Dear Rodents,
Since my 35W P/P EL34s UL won't cut the cake anymore with my Infinity RS
IIIbs, I'm looking at paralleling output tubes 2 pair, 3 pair, or 4 pair.
I'm somewhat familiar with 6550s & EL34s. I read that there can be many
problems with this configuration. Are there any success stories out there?
BTW: does anyone have experience with OTLs such as ATma-Spheres? I need more
power but I don't want to loose that sweet sound of tubes. I really could
use your opinions and thank you very much in advance.
Cordially,
west

Parallel tubes are a common feature in many amps, our Aussie fellow Mr.
Turner has a 12+12*6550s beast listed on his site.
Nevertheless with "only" 4x KT88/6550 per channel You can get 100-120W,
which is usually more than enough for normal listening levels; I wouldn't
risk my spine handling a bigger one.
If You don't mind to wire 2 more sockets and prefer the EL34s, just use 6 of
them to get the same results.
This way, the amp remains relatively simple. Voltages are in the
"customary" range, parts are available off the shelf and do not cost an arm
and a leg.
I definitely don't like the OTLs for a "conceptual" reason: "normal" tubed
amps are so good 'cause they're so simple that - even when they distort -
they don't fall apart in unpredictable ways bursting out any kind of nasty
overtones. In this field, less is mo this is all the SE DHT magic.
The other option is - bigger toobs, like SV811s or even 845s, but
voltages... yuk! Not worth unless You want DHTs.
The OTLs need to use tons of feedback, complicated topologies and often the
tubes used are questionable in terms of linearity. The problem saved by
eliminating the transformer is transferred to the gain/driving stages that
need to swing very large voltages. Last but not least, to get down to 4
Ohms, OTLs need to use A LOT of paralleled tubes, ending up being a sort of
electric room heater. Here, in Summer we have 100°F with 90% humidity... I
wouldn't turn on one of these.

BTW, I'm a chemical engineer. I usually think that a project is well done
when I can no longer wipe off a single pipe, a single valve or a single
transmitter without stopping the whole thing, just 'cos the only valve that
never locks or leaks is the one that has been left in the warehouse.

Ciao

Fabio



"west" ha scritto nel messaggio
...
Dear Rodents,
Since my 35W P/P EL34s UL won't cut the cake anymore with my Infinity RS
IIIbs, I'm looking at paralleling output tubes 2 pair, 3 pair, or 4 pair.
I'm somewhat familiar with 6550s & EL34s. I read that there can be many
problems with this configuration. Are there any success stories out there?
BTW: does anyone have experience with OTLs such as ATma-Spheres? I need
more
power but I don't want to loose that sweet sound of tubes. I really could
use your opinions and thank you very much in advance.
Cordially,
west

west wrote:

Dear Rodents,
Since my 35W P/P EL34s UL won't cut the cake anymore with my Infinity RS
IIIbs, I'm looking at paralleling output tubes 2 pair, 3 pair, or 4 pair.

Using a six pack of EL34/6CA7 will easily give 100 watts.

With good load matching, you should have no worries. See
http://www.turneraudio.com.au/htmlwe...00monobloc.htm

I sold a pair of these mono amps about 12 mths ago.

But they sure did the business with some ESL speakers.

4 x 6550 will give around 100 watts easily.


I'm somewhat familiar with 6550s & EL34s. I read that there can be many
problems with this configuration. Are there any success stories out there?

I routinely parallel tubes to get a high power to allow high dynamic levels.

BTW: does anyone have experience with OTLs such as ATma-Spheres? I need more
power but I don't want to loose that sweet sound of tubes. I really could
use your opinions and thank you very much in advance.

If you have a 100 watt amp using a quad of 6550 in UL, fixed bias,
you should be able to get the first 25 watts in class A and the
sound should be that creamy accuracy of tubes.


but first you should borrow a high power SS amp,
and measure and plot the impedance graph, and then
do a rough calculation of the power max needed with a peak level
and hold dvm with some busy music at a loud level you must prefer.

That will tell you how many maximum volts of output you need
regardless of the impedance.

Say you need a max of 20 peak volts, and the lowest dip in Z is
say 3 ohms, then you want an amp at least capable of
14 vrms into 3 ohms which is 65 watts.

A pair of KT88/6550 with Ea = 560v and RL about 5k will give you
the 65 watts; even a pair of EL34 will, but both won't give you
more than a few watts of class A.

So you should use a six pack of EL34 with comfortable Ea = +430v,
a total Ia = 6 x 47 mA = 282 mA, so Pd per tube is 20 watts.

A quad of KT88/6550 with Ea = 430v and the same total Ia will
have 70 mA / tube for 30 watts of Pd each.

The ATma-spheres may not give you as good a sound as the folks
writing reviews bestow on these amps.
Often there are no technical details.

Many PP OTLs are actually class AN with a tiny amount of class A
and a shirtload of NFB to try to linearize the mess of an amp
that one has without NFB.


When you find out how much pure class A they make before
going into class AB operation, and what the Rout for these amps is,
let me know.

But as far as I am concerned, the ATma-spheres would be a darn sight better
with an OPT, since there are *zero* tube OTLs with any sort of a decent
load match.

Patrick Turner.





Cordially,
west

What sounds nicer to you. The El34 6 pack or KT88 4 pack? Does one have any
advantage over the other? Thanks.
west

"Patrick Turner" wrote in message
...


west wrote:

Dear Rodents,
Since my 35W P/P EL34s UL won't cut the cake anymore with my Infinity RS
IIIbs, I'm looking at paralleling output tubes 2 pair, 3 pair, or 4
pair.

Using a six pack of EL34/6CA7 will easily give 100 watts.

With good load matching, you should have no worries. See
http://www.turneraudio.com.au/htmlwe...00monobloc.htm

I sold a pair of these mono amps about 12 mths ago.

But they sure did the business with some ESL speakers.

4 x 6550 will give around 100 watts easily.


I'm somewhat familiar with 6550s & EL34s. I read that there can be many
problems with this configuration. Are there any success stories out
there?

I routinely parallel tubes to get a high power to allow high dynamic
levels.

BTW: does anyone have experience with OTLs such as ATma-Spheres? I need
more
power but I don't want to loose that sweet sound of tubes. I really
could
use your opinions and thank you very much in advance.

If you have a 100 watt amp using a quad of 6550 in UL, fixed bias,
you should be able to get the first 25 watts in class A and the
sound should be that creamy accuracy of tubes.


but first you should borrow a high power SS amp,
and measure and plot the impedance graph, and then
do a rough calculation of the power max needed with a peak level
and hold dvm with some busy music at a loud level you must prefer.

That will tell you how many maximum volts of output you need
regardless of the impedance.

Say you need a max of 20 peak volts, and the lowest dip in Z is
say 3 ohms, then you want an amp at least capable of
14 vrms into 3 ohms which is 65 watts.

A pair of KT88/6550 with Ea = 560v and RL about 5k will give you
the 65 watts; even a pair of EL34 will, but both won't give you
more than a few watts of class A.

So you should use a six pack of EL34 with comfortable Ea = +430v,
a total Ia = 6 x 47 mA = 282 mA, so Pd per tube is 20 watts.

A quad of KT88/6550 with Ea = 430v and the same total Ia will
have 70 mA / tube for 30 watts of Pd each.

The ATma-spheres may not give you as good a sound as the folks
writing reviews bestow on these amps.
Often there are no technical details.

Many PP OTLs are actually class AN with a tiny amount of class A
and a shirtload of NFB to try to linearize the mess of an amp
that one has without NFB.


When you find out how much pure class A they make before
going into class AB operation, and what the Rout for these amps is,
let me know.

But as far as I am concerned, the ATma-spheres would be a darn sight
better
with an OPT, since there are *zero* tube OTLs with any sort of a decent
load match.

Patrick Turner.





Cordially,
west

west wrote:

What sounds nicer to you. The El34 6 pack or KT88 4 pack? Does one have any
advantage over the other? Thanks.
west

I don't find any astonishing differences between powerful UL amps.

The last 8585 PP amp had a quad of 6550 per channel
with 12.5% CFB applied like Quad II.
About 95 watts max is possible into 3 ohms.
Only 55 watts is possible into 8 ohms because the amp is set up
to provide a high amount of pure class A without having a huge
dissipation in the tubes, only about 25 watts each at idle.
The drive amp and OPT and general topology made this amp
exceptionally hard to better; the detail and dynamics and timbre
of music seemed the best my client and his friends have ever heard,
and he has tried a lot of amps.
So at present while I still think a normal UL PP amp with splendid OPT
and nice driver amp and about 16 dB of global FB is potentially
extremely hard to better, the CFB version using a fixed screen voltage
sounds just that much better, despite measuring virtually the same,
ie, with less than 0.05% thd at all listening levels.
ARC and CJ and all thse fancy brands will wax lyrical about their
complex and ambitious creations full of gee whiz techniques and part brands,
but believe me there is only so much goodness available from a
given amount of iron, copper, glass and vacuum, and its all available
for you to make for yourself if you try hard enough and learn enough detail.
If you use a quad of 6550 conservatively for a lot of class A
and 50 watts is plenty, then you won't be dissapointed.
You must establish what power you really need, as i described below.
Its no good building an amp if your speakers are
so insensitive that you will run into clipping easily.

Patrick Turner




"Patrick Turner" wrote in message
...


west wrote:

Dear Rodents,
Since my 35W P/P EL34s UL won't cut the cake anymore with my Infinity RS
IIIbs, I'm looking at paralleling output tubes 2 pair, 3 pair, or 4
pair.

Using a six pack of EL34/6CA7 will easily give 100 watts.

With good load matching, you should have no worries. See
http://www.turneraudio.com.au/htmlwe...00monobloc.htm

I sold a pair of these mono amps about 12 mths ago.

But they sure did the business with some ESL speakers.

4 x 6550 will give around 100 watts easily.


I'm somewhat familiar with 6550s & EL34s. I read that there can be many
problems with this configuration. Are there any success stories out
there?

I routinely parallel tubes to get a high power to allow high dynamic
levels.

BTW: does anyone have experience with OTLs such as ATma-Spheres? I need
more
power but I don't want to loose that sweet sound of tubes. I really
could
use your opinions and thank you very much in advance.

If you have a 100 watt amp using a quad of 6550 in UL, fixed bias,
you should be able to get the first 25 watts in class A and the
sound should be that creamy accuracy of tubes.


but first you should borrow a high power SS amp,
and measure and plot the impedance graph, and then
do a rough calculation of the power max needed with a peak level
and hold dvm with some busy music at a loud level you must prefer.

That will tell you how many maximum volts of output you need
regardless of the impedance.

Say you need a max of 20 peak volts, and the lowest dip in Z is
say 3 ohms, then you want an amp at least capable of
14 vrms into 3 ohms which is 65 watts.

A pair of KT88/6550 with Ea = 560v and RL about 5k will give you
the 65 watts; even a pair of EL34 will, but both won't give you
more than a few watts of class A.

So you should use a six pack of EL34 with comfortable Ea = +430v,
a total Ia = 6 x 47 mA = 282 mA, so Pd per tube is 20 watts.

A quad of KT88/6550 with Ea = 430v and the same total Ia will
have 70 mA / tube for 30 watts of Pd each.

The ATma-spheres may not give you as good a sound as the folks
writing reviews bestow on these amps.
Often there are no technical details.

Many PP OTLs are actually class AN with a tiny amount of class A
and a shirtload of NFB to try to linearize the mess of an amp
that one has without NFB.


When you find out how much pure class A they make before
going into class AB operation, and what the Rout for these amps is,
let me know.

But as far as I am concerned, the ATma-spheres would be a darn sight
better
with an OPT, since there are *zero* tube OTLs with any sort of a decent
load match.

Patrick Turner.





Cordially,
west

"Patrick Turner" wrote in message
...


west wrote:

What sounds nicer to you. The El34 6 pack or KT88 4 pack? Does one have
any
advantage over the other? Thanks.
west

I don't find any astonishing differences between powerful UL amps.

The last 8585 PP amp had a quad of 6550 per channel
with 12.5% CFB applied like Quad II.
About 95 watts max is possible into 3 ohms.
Only 55 watts is possible into 8 ohms because the amp is set up
to provide a high amount of pure class A without having a huge
dissipation in the tubes, only about 25 watts each at idle.
The drive amp and OPT and general topology made this amp
exceptionally hard to better; the detail and dynamics and timbre
of music seemed the best my client and his friends have ever heard,
and he has tried a lot of amps.
So at present while I still think a normal UL PP amp with splendid OPT
and nice driver amp and about 16 dB of global FB is potentially
extremely hard to better, the CFB version using a fixed screen voltage
sounds just that much better, despite measuring virtually the same,
ie, with less than 0.05% thd at all listening levels.
ARC and CJ and all thse fancy brands will wax lyrical about their
complex and ambitious creations full of gee whiz techniques and part
brands,
but believe me there is only so much goodness available from a
given amount of iron, copper, glass and vacuum, and its all available
for you to make for yourself if you try hard enough and learn enough
detail.
If you use a quad of 6550 conservatively for a lot of class A
and 50 watts is plenty, then you won't be dissapointed.
You must establish what power you really need, as i described below.
Its no good building an amp if your speakers are
so insensitive that you will run into clipping easily.

Patrick Turner

I agree with the measurement. I just have to get hold of a peak & hold DVM.
If I knew some model #s, I could look in eBay.
west


"Patrick Turner" wrote in message
...


west wrote:

Dear Rodents,
Since my 35W P/P EL34s UL won't cut the cake anymore with my
Infinity RS
IIIbs, I'm looking at paralleling output tubes 2 pair, 3 pair, or 4
pair.

Using a six pack of EL34/6CA7 will easily give 100 watts.

With good load matching, you should have no worries. See
http://www.turneraudio.com.au/htmlwe...00monobloc.htm

I sold a pair of these mono amps about 12 mths ago.

But they sure did the business with some ESL speakers.

4 x 6550 will give around 100 watts easily.


I'm somewhat familiar with 6550s & EL34s. I read that there can be
many
problems with this configuration. Are there any success stories out
there?

I routinely parallel tubes to get a high power to allow high dynamic
levels.

BTW: does anyone have experience with OTLs such as ATma-Spheres? I
need
more
power but I don't want to loose that sweet sound of tubes. I really
could
use your opinions and thank you very much in advance.

If you have a 100 watt amp using a quad of 6550 in UL, fixed bias,
you should be able to get the first 25 watts in class A and the
sound should be that creamy accuracy of tubes.


but first you should borrow a high power SS amp,
and measure and plot the impedance graph, and then
do a rough calculation of the power max needed with a peak level
and hold dvm with some busy music at a loud level you must prefer.

That will tell you how many maximum volts of output you need
regardless of the impedance.

Say you need a max of 20 peak volts, and the lowest dip in Z is
say 3 ohms, then you want an amp at least capable of
14 vrms into 3 ohms which is 65 watts.

A pair of KT88/6550 with Ea = 560v and RL about 5k will give you
the 65 watts; even a pair of EL34 will, but both won't give you
more than a few watts of class A.

So you should use a six pack of EL34 with comfortable Ea = +430v,
a total Ia = 6 x 47 mA = 282 mA, so Pd per tube is 20 watts.

A quad of KT88/6550 with Ea = 430v and the same total Ia will
have 70 mA / tube for 30 watts of Pd each.

The ATma-spheres may not give you as good a sound as the folks
writing reviews bestow on these amps.
Often there are no technical details.

Many PP OTLs are actually class AN with a tiny amount of class A
and a shirtload of NFB to try to linearize the mess of an amp
that one has without NFB.


When you find out how much pure class A they make before
going into class AB operation, and what the Rout for these amps is,
let me know.

But as far as I am concerned, the ATma-spheres would be a darn sight
better
with an OPT, since there are *zero* tube OTLs with any sort of a
decent
load match.

Patrick Turner.





Cordially,
west

"west" wrote

Since my 35W P/P EL34s UL won't cut the cake anymore with my
Infinity RS
IIIbs, I'm looking at paralleling output tubes 2 pair, 3 pair, or
4 pair.
I'm somewhat familiar with 6550s & EL34s. I read that there can be
many
problems with this configuration....

What problems?

If you use UL taps then the lower the impedance of the output stage,
the more the inclination to ring, I think. You will probably need to
damp this, most directly with zobels between anode and screen
windings.

I don't know of any other problems associated with parallel output
valves. There are several advantages.

cheers, Ian

west wrote:

"Patrick Turner" wrote in message
...


west wrote:

What sounds nicer to you. The El34 6 pack or KT88 4 pack? Does one have
any
advantage over the other? Thanks.
west

I don't find any astonishing differences between powerful UL amps.

The last 8585 PP amp had a quad of 6550 per channel
with 12.5% CFB applied like Quad II.
About 95 watts max is possible into 3 ohms.
Only 55 watts is possible into 8 ohms because the amp is set up
to provide a high amount of pure class A without having a huge
dissipation in the tubes, only about 25 watts each at idle.
The drive amp and OPT and general topology made this amp
exceptionally hard to better; the detail and dynamics and timbre
of music seemed the best my client and his friends have ever heard,
and he has tried a lot of amps.
So at present while I still think a normal UL PP amp with splendid OPT
and nice driver amp and about 16 dB of global FB is potentially
extremely hard to better, the CFB version using a fixed screen voltage
sounds just that much better, despite measuring virtually the same,
ie, with less than 0.05% thd at all listening levels.
ARC and CJ and all thse fancy brands will wax lyrical about their
complex and ambitious creations full of gee whiz techniques and part
brands,
but believe me there is only so much goodness available from a
given amount of iron, copper, glass and vacuum, and its all available
for you to make for yourself if you try hard enough and learn enough
detail.
If you use a quad of 6550 conservatively for a lot of class A
and 50 watts is plenty, then you won't be dissapointed.
You must establish what power you really need, as i described below.
Its no good building an amp if your speakers are
so insensitive that you will run into clipping easily.

Patrick Turner

I agree with the measurement. I just have to get hold of a peak & hold DVM.
If I knew some model #s, I could look in eBay.
west

Radio Shack, Tandy should have a cheap model of dvm with a peak&hold function.
it quickly measures the peak voltage of a signal.
leave it clipped to a circuit and the highest voltage over a period is stored.

Patrick Turner.




"Patrick Turner" wrote in message
...


west wrote:

Dear Rodents,
Since my 35W P/P EL34s UL won't cut the cake anymore with my
Infinity RS
IIIbs, I'm looking at paralleling output tubes 2 pair, 3 pair, or 4
pair.

Using a six pack of EL34/6CA7 will easily give 100 watts.

With good load matching, you should have no worries. See
http://www.turneraudio.com.au/htmlwe...00monobloc.htm

I sold a pair of these mono amps about 12 mths ago.

But they sure did the business with some ESL speakers.

4 x 6550 will give around 100 watts easily.


I'm somewhat familiar with 6550s & EL34s. I read that there can be
many
problems with this configuration. Are there any success stories out
there?

I routinely parallel tubes to get a high power to allow high dynamic
levels.

BTW: does anyone have experience with OTLs such as ATma-Spheres? I
need
more
power but I don't want to loose that sweet sound of tubes. I really
could
use your opinions and thank you very much in advance.

If you have a 100 watt amp using a quad of 6550 in UL, fixed bias,
you should be able to get the first 25 watts in class A and the
sound should be that creamy accuracy of tubes.


but first you should borrow a high power SS amp,
and measure and plot the impedance graph, and then
do a rough calculation of the power max needed with a peak level
and hold dvm with some busy music at a loud level you must prefer.

That will tell you how many maximum volts of output you need
regardless of the impedance.

Say you need a max of 20 peak volts, and the lowest dip in Z is
say 3 ohms, then you want an amp at least capable of
14 vrms into 3 ohms which is 65 watts.

A pair of KT88/6550 with Ea = 560v and RL about 5k will give you
the 65 watts; even a pair of EL34 will, but both won't give you
more than a few watts of class A.

So you should use a six pack of EL34 with comfortable Ea = +430v,
a total Ia = 6 x 47 mA = 282 mA, so Pd per tube is 20 watts.

A quad of KT88/6550 with Ea = 430v and the same total Ia will
have 70 mA / tube for 30 watts of Pd each.

The ATma-spheres may not give you as good a sound as the folks
writing reviews bestow on these amps.
Often there are no technical details.

Many PP OTLs are actually class AN with a tiny amount of class A
and a shirtload of NFB to try to linearize the mess of an amp
that one has without NFB.


When you find out how much pure class A they make before
going into class AB operation, and what the Rout for these amps is,
let me know.

But as far as I am concerned, the ATma-spheres would be a darn sight
better
with an OPT, since there are *zero* tube OTLs with any sort of a
decent
load match.

Patrick Turner.





Cordially,
west

west wrote:

Dear Rodents,
Since my 35W P/P EL34s UL won't cut the cake anymore with my Infinity RS
IIIbs, I'm looking at paralleling output tubes 2 pair, 3 pair, or 4 pair.
I'm somewhat familiar with 6550s & EL34s. I read that there can be many
problems with this configuration. Are there any success stories out there?
BTW: does anyone have experience with OTLs such as ATma-Spheres? I need more
power but I don't want to loose that sweet sound of tubes. I really could
use your opinions and thank you very much in advance.
Cordially,
west

A lot will depend on what kind of output hookup used, triode, pentode or UL.

Pentodes are current devices, so have no problem sharing the load. However the
same tubes connected as triodes are voltage sources. Paralleled voltage sources
do not share the load well at all. The result is higher than expected distortion
at all levels.

So if you don't intend to use any NFB, the result may suprise you!

The sharing of load & distortion problems are a real problem working into a
complex load such as a loudspeaker.

JLS

"John Stewart" wrote

...Pentodes are current devices, so have no problem sharing the
load. However the
same tubes connected as triodes are voltage sources. Paralleled
voltage sources
do not share the load well at all. The result is higher than
expected distortion
at all levels....

I'm surprised. You are usually one for posting supporting evidence,
John. I can see the idea, but do you have actual data on this?

cheers, Ian

Ian Iveson wrote:

"John Stewart" wrote

...Pentodes are current devices, so have no problem sharing the
load. However the
same tubes connected as triodes are voltage sources. Paralleled
voltage sources
do not share the load well at all. The result is higher than
expected distortion
at all levels....

I'm surprised. You are usually one for posting supporting evidence,
John. I can see the idea, but do you have actual data on this?

cheers, Ian

It is here somewhere. I will dig it out & post in the next day or
two. JLS

John Stewart wrote:

Ian Iveson wrote:

"John Stewart" wrote

...Pentodes are current devices, so have no problem sharing the
load. However the
same tubes connected as triodes are voltage sources. Paralleled
voltage sources
do not share the load well at all. The result is higher than
expected distortion
at all levels....

I'm surprised. You are usually one for posting supporting evidence,
John. I can see the idea, but do you have actual data on this?

cheers, Ian

It is here somewhere. I will dig it out & post in the next day or
two. JLS

I must apologize on this one Ian. I can't find those results anywhere in
my file. I did that work sometime more than 40 years ago before going
into sales with HP. Perhaps I will have another look again when I get
time to do some tests since it is a rather interesting subject. Or
perhaps someone out there will take a shot at it!

However, in the meantime we can use a little network theory to illustrate
what is happening. Imagine in a perfect world we have a pair of identical
voltage sources connected to a common load, both thru identical
impedances. The voltage source polarities are connected so that +ve is to
+ve & -ve is to -ve. The simplest form this would take would be a pair of
identical batteries driving a resistive load thru identical resistors.
The resultant current in the load would be the simple sum of the currents
provided by the two batteries, those currents being equal to each other.

But as usual the world is far from perfect as are this pair of batteries
& the resistors. One or the other battery will manage to sink or source a
current thru the other battery while at the same time push some current
thru the load. The two currents in the load are no longer identical &
never will be.

For the audio case circumstances get much worse. If we are lucky enough
to have say a pair of WE300B's, carefully matched on an AVO tester it
will help a lot, but there are still problems & lots of them. At how many
points can we match? It is unlikely we can get a match for all conditions
of operation on the plate family of curves.

Lets replace the batteries & resistors above with the WE300B's.

The WE300B's (And any other triode) are voltages sources as ordinarily
connected & pretty good ones at that. If they are driven by a common
source into a resistive load will they each provide the same current to
the load? It seems rather unlikely no matter how careful we are.

And the loadline presented to them will not be a nice straight line as
you often see in a text book. That is the simplest possible condition &
seldom happens in real life. The loadline if driven by a single frequency
might be at best a straight line but more than likely elliptical. That is
because the load is not a simple resistor but a loudspeaker. And there is
an OPT on the way to the loudspeaker to further confuse things. The
WE300B's have some non-linearites & so do the OPT's, all generating some
distortion of their own.

Things get really crazy with the program material we would like to use
whether it be Bach or The Rolling Stones (I prefer the Stones). That is
because we now have a complex signal of many frequencies running into a
complex load (the loudspeaker).

Here again we have a condition where one or the other of the WE300B's can
sink or source some current into the other while at the same time driving
part of the load current! That means that each of the tubes are at times
driving the other. Some of the power developed by each tube is simply
dissipated in the other & never does show up in your loudspeaker load.

So if you parallel a pair of triodes you won't get double the power at a
particular D. But with great care you might come close.

Pentodes are current sources & can work well paralleled. But they still
have all the objectionable distortion characteristics typical of their
family. A way out of that is loop NFB & we see that often used to lower D
& get a reasonable DF.

You could also run the triodes with cathode resistors to move them closer
to CC operation. But that lowers the DF.

When running tubes in parallel one usually needs to included swamping
resistors on all grid leads & sometimes plate leads as well in order to
suppress parasitic RF oscillations.

The circuit I did my tests on was a simple stereo amp using a pair of
cathode biased 6AQ5's into Hammond 125D OPT's in each channel. Not the
best amp at all but it did illustrate the principle well. Since each amp
of the pair had loop NFB it was in a sense a voltage source. If my memory
is still in gear the DF was about 3 so in that respect very much like an
SET circuit. But there the similarity ends.

Seems we still can't get something for nothing!

Cheers, John Stewart

John Stewart wrote:

John Stewart wrote:

Ian Iveson wrote:

"John Stewart" wrote

...Pentodes are current devices, so have no problem sharing the
load. However the
same tubes connected as triodes are voltage sources. Paralleled
voltage sources
do not share the load well at all. The result is higher than
expected distortion
at all levels....

I'm surprised. You are usually one for posting supporting evidence,
John. I can see the idea, but do you have actual data on this?

cheers, Ian

It is here somewhere. I will dig it out & post in the next day or
two. JLS

I must apologize on this one Ian. I can't find those results anywhere in
my file. I did that work sometime more than 40 years ago before going
into sales with HP. Perhaps I will have another look again when I get
time to do some tests since it is a rather interesting subject. Or
perhaps someone out there will take a shot at it!

However, in the meantime we can use a little network theory to illustrate
what is happening. Imagine in a perfect world we have a pair of identical
voltage sources connected to a common load, both thru identical
impedances. The voltage source polarities are connected so that +ve is to
+ve & -ve is to -ve. The simplest form this would take would be a pair of
identical batteries driving a resistive load thru identical resistors.
The resultant current in the load would be the simple sum of the currents
provided by the two batteries, those currents being equal to each other.

I have a colleague in Sydney who built a pair of 80 watt SET amps
using 10 x 300B in parallel.
The load was about 400 ohms for the lot and
they all shared the load pretty darn well.
The ratio between any one tube's Ra of 800 ohms to the
rest of the parallel tubes' Ra which is 88 0hms and the 400 ohms RL
is 800 : 72, so each tube works as a current source
to the total resistance it is connected to, since one tube's Ra
is large compared to all the the external resistances.

In all the experiments I have made with triode connected tubes
connected in parallel I have found only very slight differences
between the signal cathode currents in the parallel tubes.




But as usual the world is far from perfect as are this pair of batteries
& the resistors. One or the other battery will manage to sink or source a
current thru the other battery while at the same time push some current
thru the load. The two currents in the load are no longer identical &
never will be.

For the audio case circumstances get much worse.

Nope.

If we are lucky enough
to have say a pair of WE300B's, carefully matched on an AVO tester it
will help a lot, but there are still problems & lots of them.

Nope.

At how many
points can we match? It is unlikely we can get a match for all conditions
of operation on the plate family of curves.

The variations don't make much difference...



Lets replace the batteries & resistors above with the WE300B's.

The WE300B's (And any other triode) are voltages sources as ordinarily
connected & pretty good ones at that.

They are very weak voltage sources.
Ra of 800 ohms is not that much lower than the 4k one might load the tube
with.
It seems like a lot, but it isn't.
If the Ra was 8 ohms, there might be a slight problem



If they are driven by a common
source into a resistive load will they each provide the same current to
the load? It seems rather unlikely no matter how careful we are.

It is in practice very likely they will share the current fairly evenly.

One could have 625 EL34 in triode all hooked up to 8 ohms
The average load seen by each EL34 would be 5,000 ohms, and a reasonable load
match for each tube.
Since no two tubes are exactly matched and they have different gm,
then although the anode voltage swing of 160vrms would be
the same for each tube, the current in each tube would differ according to
each tubes's
gm, and since u would be so close for all it is of negligible concern, since
u
depends on the physical dimensions between electrodes.
Therefore it could be shown that each tube would itself experience a
different load to its neighbour,
one might think its load is 5,300 ohms, the next 4,900, next 5,100 and so on.

Not all horses pull the carriage with the same force, but they all do what
suits them.



And the loadline presented to them will not be a nice straight line as
you often see in a text book.

The load line in PP class A amps for each tube is actually a bent line,
and very bent in PP class AB amps.
while in class A, each tube of a pair sees say 5k, to make up the total
10k a-a RL.
But when one has cut off, the other tube sees just 2.5k.
Changes of gain during the cycle cause distortion.

That is the simplest possible condition &
seldom happens in real life.

The SE amp will have the straightest load line.

The loadline if driven by a single frequency
might be at best a straight line but more than likely elliptical. That is
because the load is not a simple resistor but a loudspeaker. And there is
an OPT on the way to the loudspeaker to further confuse things.

The same elliptical load line is shared between all parallel tubes.

The
WE300B's have some non-linearites & so do the OPT's, all generating some
distortion of their own.

And the individual distortion currents sum to make but one average distortion
current.

In PP amps with a quad or more OP tubes in parallel we have the opportunity
to swap tubes around so the total signal currents of each 1/2 of the
PP output circuit are closely matched, thus cancelling the effect of the
mismatch between tubes, and getting almost perfect cancellation
of 2H between the two PP halves of the circuit.
Within each 1/2 I have found the currents near equal between 2 or 3 or 6
output tubes
in parallel with the same phase of signal.



Things get really crazy with the program material we would like to use
whether it be Bach or The Rolling Stones (I prefer the Stones). That is
because we now have a complex signal of many frequencies running into a
complex load (the loudspeaker).

Signal complexity makes no difference to the sharing ability of tube
behaviour.
A signal is a signal is a signal. Just voltage current changes.



Here again we have a condition where one or the other of the WE300B's can
sink or source some current into the other while at the same time driving
part of the load current! That means that each of the tubes are at times
driving the other.

I have never seen this. One could say that if you pull one tube out, the
other 5
have to work harder.
But the direction of current flow is never into one tube when its out of
all the rest.
What simply happens is that one tube experiences a different load to the
other.
You may say that if the difference in current is 2 mA between a pair
of parallel tubes then the net 2 mA is flowing from one tube to the other,
some how creating power in a tube to reduce efficiency.
Its over stating what's going on.



Some of the power developed by each tube is simply
dissipated in the other & never does show up in your loudspeaker load.

You mean one tube dissipates more heat than the other.
If a weak gm tube is paired with a much stronger gm tube then
of course the weaker one might see 7k and the stronger one 4k,
and the actual load is 2.54 k.
The one that acts like it sees 4k will have more dissipation
than the one that has the 7k, and the efficiencies of each will vary as you
calculate.
In practice it isn't a big deal.

One could make the contribution of power near equal if we had
current FB as one is forced to have in SS amps with multiple
bjt or mosfets in parallel. Especially bjts, because if one bjt conducts more
than the others it heats up
and conducts even more, hence the use of the 0.33 ohm emitter resistors
seen in many SS amps with multiple devices.
Mosfets and tubes back off a bit when they get hotter, since they have
negative temp
characteristics, ie, hotter means less gm.

Tubes when paralleled can tend to stay matched, because any that work harder
than the rest
wear out faster, and they gm reduces......



So if you parallel a pair of triodes you won't get double the power at a
particular D. But with great care you might come close.

I routinely double the power with parallel tubes.
I sometimes quadruple it, sixtuple it, or twelvopple it.

Colleagues have the same experiences, even with triodes.

Even when no special efforts are made to match, multiples of power
are produced.



Pentodes are current sources & can work well paralleled. But they still
have all the objectionable distortion characteristics typical of their
family. A way out of that is loop NFB & we see that often used to lower D
& get a reasonable DF.

You could also run the triodes with cathode resistors to move them closer
to CC operation. But that lowers the DF.

50 ohms in the cathode circuit of an EL34 in triode
makes the Ra rise from 1,250 ohms to about 1,750 ohms,
and it wastes 0.05 watts when there is 160 vrms across the 5k anode load, (
5.12 watts of output power )
So one could have 200 ohms in the '34 cathode circuit, Ra would rise to
3,250 ohms or about equal to a 50% UL connected '34.
With 160vrms across the 5k load, the same 5.12 watts is produced, but
0.2 watts is wasted in the cathode R, and this is a very small wastage.
The applied current FB will reduce the gain and distortion like all NFB does,

but only marginally, but it won't do any good in a PP circuit where
individual unbypassed cathode resistors in a class AB circuit are not too
good a thing
where one wants to have a fixed bias operation.
Quite OK in any SEW circuit though.

To reduce the Ro of the amp after increasing it with unbypassed Rk,
external loops of voltage NFB must be used, and to get a circuit
with 3 times the usual Ra back down to the original Ra
might take at least 8 dB of global or other series voltage NFB, or perhaps a
little
positive current FB which makes distortion worse after the betterment from
the negative series current FB in a given output triode cathode circuit.

Its all rather bloomin messy.


When running tubes in parallel one usually needs to included swamping
resistors on all grid leads & sometimes plate leads as well in order to
suppress parasitic RF oscillations.

In my 300 watt amps with a dozen 6550 all I needed was the rather low value
2.2k g1 resistor stoppers.
There are 220 ohms in series with the screens and their supply.
Most important are the zobel neworks across each 1/2 of the primary
on the OPT.
This is placed half way along the line up of six tubes on each side of the PP
circuit.
Not a sign of any RF instability, and I have an OPT with 270 khz of
open loop bandwidth.

Should a tube decide to go beserko and go into thermal
runaway, then sometimes it will also oscillate at RF,
as well as conduct a huge cathode current.
The active protection I have in all my amps detects
the excess DC always associated with a hot tube.
If one or more tubes sustains twice the idle current for more than
4 seconds, the amp shuts down.

I once repaired a PA amp twice which had an octet of EL34, with
one single fixed bias voltage, Ea = 900v, Eg2 = 450v.
After the second repair I insisted I inspect his two
bins with 6 x 12" speakers in each.
He maintained he had 8 ohm speakers.
Sure, he was dead right, but all were wired in parallel,
and the most he got was 50 watts in the speakers, 25 watts in the cables,
and a shirtload of heat in the tubes, which had been pooping out regularly
for 15 years.
Once the amp was re-loaded correctly, biased with 8 pots,
had zobels connected right, and some stability added with the
12 dB of GNFB I added as well as a regulated screen supply, he never phoned
me again,
and the sound was truly glorious.



Instead of using unbypassed cathode resistors and triode connection in my 300
watt amps
to only marginally assist load sharing, I have 20% of the primary
signal voltage applied to the cathodes as series voltage NFB from the OPT.
The screens are taken to +330v, Ea = 470v, and individual cathode bias is
applied
to each tube to regulate the bias.
Some fixed bias is used to keep the value of Rk to a minimum
and keep the DC Pd in the RK low, yet regulate bias well enough.
I also use a couple of power bjts to dynamically shunt the
excess 2H currents in the cathode circuits thus maintaining thre voltage
across the total of 12,000 uF in each bypass cap on each half of the PP
circuit.
So with dynamic stabilizing of the cathode bias the tubes tend to share their
load
pretty darn well.
They are all subject to the same cathode series voltage FB, not current FB.
But since they are tetrode connected but with voltage FB there seems to be no

tendency for the tubes to differ much in their current donation to the load
power.

With many rather poorly conceived commercially made amps such as the Baird
with a six pack of EL84 to give the rather large 60 watts of power per
channel,
there is but one sole value of fixed bias.
Baird didn't want to load down customers with the responsibility to adjust
12 bias pots, something many users would invariably get wrong.
Therefore as all these little tube age their bias drift is large, and some
would
fade faster than others, and conduct differently, but the drift wouldn't be
any worse than
having just a pair of output 6550 with the same grid bias applied.
3 x EL84 can do about what one 6550 can do.
10 x EL84 each with its own cathode bias circuit would be more effective
for 50 watts of class AB with a larger amount of class A,
and since the power used won't be high, the tubes would last and stay
well enough matched to provide the power wanted.




The circuit I did my tests on was a simple stereo amp using a pair of
cathode biased 6AQ5's into Hammond 125D OPT's in each channel. Not the
best amp at all but it did illustrate the principle well. Since each amp
of the pair had loop NFB it was in a sense a voltage source. If my memory
is still in gear the DF was about 3 so in that respect very much like an
SET circuit. But there the similarity ends.

Paralleling amps which have two separate FB loops which of course that will
means that the Or of one amp is say 2 ohms, and the other might be
2.2 ohms, due to vagaries of tubes and matching, and if the common load is
8 ohms then one amp sees nearly the same Ro as the other.
You can do the math with 2 ohms from a true zero ohms voltage source,
and another voltage source feeding 2.2 ohms, and tell me what the
mismatch does, but I'd say not much.

With a pair of solid state amps with very low Ro of say 100 milliohms each,
the effects of different amp gains could have serious load sharing problems.
But not when Ro is over 1 ohm.

With very low Ro SS amps one could happily power
a common load with 4 amps in parallel, but each with 1 ohm as a series
R in the output, and this would give the effect of having Rout = about 0.25
ohms.
Its not hard to have 1% accurately matched feedback networks in each
amp to ensure the voltage gain is the same.
Such close matching of voltage gain is far easier to attain with
modern metal film resistors than rely on device matching,
or the fact that we have current sources.
if the Ro of an amp = 0.1 ohms or less like most SS amps, the
1 ohm series R will convert the amp to a current source from the amp's
point of view, since it sees at least 1 ohm of load.
The voltage NFB ensures that the actual load value seen by each of multiple
amps
will stay close when the series 1 ohm "safety resistors" are in place
Such R can be used to sense current overloading.



Seems we still can't get something for nothing!

Well, be prepared to pay the price.

Youse will never ever hear a gal say
" jus lie down here luvvy; this won't costyer anyfink "


I think you overstate a problem.

Triodes are current sources for the reasons I gave above
in the top paragraph. One tube sees all the parallel Ra and RL
of the other tubes.
All the rowers in a boat don't make the same power,
some row harder than others; none make it more difficult
for any other to row towards that medal.
Unless one falls out of the boat and tangles her oars in a daze of over
exertion,
and this happened in one of the Oz boats with a girls crew in the last
Olympics.
One girl just passed out, ****ed right up, and
stuffed any chance of a medal. The other girls were not happy......

Anyway, next time you are in your workshop, plug a 6L6
and KT88 into each side of a PP circuit and wire them as triodes
and bias them both for 22 watts, and do a test in class A,
then parallel some tubes in triode in an SE circuit, and get back to us
with your findings.

Forget the theorizing, see what happens in reality land.

BTW, with 625 trioded EL34 in parallel, a single EL34
sees a load and Ra resistance = approximately 1.6 ohms.

Output power at 5.1 watts average each = 3,187 watts.

The 625 tubes are needed for a very nice OTL load match for SE class A.

In many OTL amps, paralel output triodes are very prone to
individually thermaling out since the load match is usually
so damn attrocious.

Patrick Turner.



Cheers, John Stewart

Interesting. I'm generally an opponent of paralleled tubes for absolutely
no technical reason, just 'cause in my job I always faced a f%#!!ing mess to
make two compressors or two pumps to work together in parallel. I was
surprised sometimes by the fact that it was easier to share load among 3 or
4 units than just 2, but now I got the "intuitive" point: it depends from
the perturbation that the single unit causes to the whole. If 10
compressors are blowing into a single pipe and one fails, the 9 still
running do not even notice, but if it's one out of two...
In the same way (I suppose) goes tubes behavior: in a parallel of 4 units,
the differences in the Gm of any single one will affect in a minor way the
working point of all other ones.
But when only 2 tubes are paralleled, "matched pairs" are mandatory;.
Triode/pentode mode has the same meaning: what happens to the second tube if
the first conducts more/less?
Adjustable bias will fix only a part of the problem (DC current) but still
any single tube will operate in a different point of its curve and reach
saturation at different signal levels.
My "less is more" approach, here, is not a good one (it's never useful to be
too much devoted to a rule, without putting it in a context).

(Nevertheless, I still think that 2 tubes only are a bit of a trouble: I
remember I saw a paralleled ECC83 on a 'scope, 1K sine wave. One half had a
higher gain. The output graph was a real "seno" (in Italian the sine wave
and the feminine breast share the same word): a beautiful sine wave, with a
clearly visible "nipple" on its top, where the "strongest" half tube got
in..)

Ciao

Fabio


"Patrick Turner" ha scritto nel messaggio
...


John Stewart wrote:

John Stewart wrote:

Ian Iveson wrote:

"John Stewart" wrote

...Pentodes are current devices, so have no problem sharing the
load. However the
same tubes connected as triodes are voltage sources. Paralleled
voltage sources
do not share the load well at all. The result is higher than
expected distortion
at all levels....

I'm surprised. You are usually one for posting supporting evidence,
John. I can see the idea, but do you have actual data on this?

cheers, Ian

It is here somewhere. I will dig it out & post in the next day or
two. JLS

I must apologize on this one Ian. I can't find those results anywhere in
my file. I did that work sometime more than 40 years ago before going
into sales with HP. Perhaps I will have another look again when I get
time to do some tests since it is a rather interesting subject. Or
perhaps someone out there will take a shot at it!

However, in the meantime we can use a little network theory to illustrate
what is happening. Imagine in a perfect world we have a pair of identical
voltage sources connected to a common load, both thru identical
impedances. The voltage source polarities are connected so that +ve is to
+ve & -ve is to -ve. The simplest form this would take would be a pair of
identical batteries driving a resistive load thru identical resistors.
The resultant current in the load would be the simple sum of the currents
provided by the two batteries, those currents being equal to each other.

I have a colleague in Sydney who built a pair of 80 watt SET amps
using 10 x 300B in parallel.
The load was about 400 ohms for the lot and
they all shared the load pretty darn well.
The ratio between any one tube's Ra of 800 ohms to the
rest of the parallel tubes' Ra which is 88 0hms and the 400 ohms RL
is 800 : 72, so each tube works as a current source
to the total resistance it is connected to, since one tube's Ra
is large compared to all the the external resistances.

In all the experiments I have made with triode connected tubes
connected in parallel I have found only very slight differences
between the signal cathode currents in the parallel tubes.




But as usual the world is far from perfect as are this pair of batteries
& the resistors. One or the other battery will manage to sink or source a
current thru the other battery while at the same time push some current
thru the load. The two currents in the load are no longer identical &
never will be.

For the audio case circumstances get much worse.

Nope.

If we are lucky enough
to have say a pair of WE300B's, carefully matched on an AVO tester it
will help a lot, but there are still problems & lots of them.

Nope.

At how many
points can we match? It is unlikely we can get a match for all conditions
of operation on the plate family of curves.

The variations don't make much difference...



Lets replace the batteries & resistors above with the WE300B's.

The WE300B's (And any other triode) are voltages sources as ordinarily
connected & pretty good ones at that.

They are very weak voltage sources.
Ra of 800 ohms is not that much lower than the 4k one might load the tube
with.
It seems like a lot, but it isn't.
If the Ra was 8 ohms, there might be a slight problem



If they are driven by a common
source into a resistive load will they each provide the same current to
the load? It seems rather unlikely no matter how careful we are.

It is in practice very likely they will share the current fairly evenly.

One could have 625 EL34 in triode all hooked up to 8 ohms
The average load seen by each EL34 would be 5,000 ohms, and a reasonable
load
match for each tube.
Since no two tubes are exactly matched and they have different gm,
then although the anode voltage swing of 160vrms would be
the same for each tube, the current in each tube would differ according to
each tubes's
gm, and since u would be so close for all it is of negligible concern,
since
u
depends on the physical dimensions between electrodes.
Therefore it could be shown that each tube would itself experience a
different load to its neighbour,
one might think its load is 5,300 ohms, the next 4,900, next 5,100 and so
on.

Not all horses pull the carriage with the same force, but they all do what
suits them.



And the loadline presented to them will not be a nice straight line as
you often see in a text book.

The load line in PP class A amps for each tube is actually a bent line,
and very bent in PP class AB amps.
while in class A, each tube of a pair sees say 5k, to make up the total
10k a-a RL.
But when one has cut off, the other tube sees just 2.5k.
Changes of gain during the cycle cause distortion.

That is the simplest possible condition &
seldom happens in real life.

The SE amp will have the straightest load line.

The loadline if driven by a single frequency
might be at best a straight line but more than likely elliptical. That is
because the load is not a simple resistor but a loudspeaker. And there is
an OPT on the way to the loudspeaker to further confuse things.

The same elliptical load line is shared between all parallel tubes.

The
WE300B's have some non-linearites & so do the OPT's, all generating some
distortion of their own.

And the individual distortion currents sum to make but one average
distortion
current.

In PP amps with a quad or more OP tubes in parallel we have the
opportunity
to swap tubes around so the total signal currents of each 1/2 of the
PP output circuit are closely matched, thus cancelling the effect of the
mismatch between tubes, and getting almost perfect cancellation
of 2H between the two PP halves of the circuit.
Within each 1/2 I have found the currents near equal between 2 or 3 or 6
output tubes
in parallel with the same phase of signal.



Things get really crazy with the program material we would like to use
whether it be Bach or The Rolling Stones (I prefer the Stones). That is
because we now have a complex signal of many frequencies running into a
complex load (the loudspeaker).

Signal complexity makes no difference to the sharing ability of tube
behaviour.
A signal is a signal is a signal. Just voltage current changes.



Here again we have a condition where one or the other of the WE300B's can
sink or source some current into the other while at the same time driving
part of the load current! That means that each of the tubes are at times
driving the other.

I have never seen this. One could say that if you pull one tube out, the
other 5
have to work harder.
But the direction of current flow is never into one tube when its out of
all the rest.
What simply happens is that one tube experiences a different load to the
other.
You may say that if the difference in current is 2 mA between a pair
of parallel tubes then the net 2 mA is flowing from one tube to the other,
some how creating power in a tube to reduce efficiency.
Its over stating what's going on.



Some of the power developed by each tube is simply
dissipated in the other & never does show up in your loudspeaker load.

You mean one tube dissipates more heat than the other.
If a weak gm tube is paired with a much stronger gm tube then
of course the weaker one might see 7k and the stronger one 4k,
and the actual load is 2.54 k.
The one that acts like it sees 4k will have more dissipation
than the one that has the 7k, and the efficiencies of each will vary as
you
calculate.
In practice it isn't a big deal.

One could make the contribution of power near equal if we had
current FB as one is forced to have in SS amps with multiple
bjt or mosfets in parallel. Especially bjts, because if one bjt conducts
more
than the others it heats up
and conducts even more, hence the use of the 0.33 ohm emitter resistors
seen in many SS amps with multiple devices.
Mosfets and tubes back off a bit when they get hotter, since they have
negative temp
characteristics, ie, hotter means less gm.

Tubes when paralleled can tend to stay matched, because any that work
harder
than the rest
wear out faster, and they gm reduces......



So if you parallel a pair of triodes you won't get double the power at a
particular D. But with great care you might come close.

I routinely double the power with parallel tubes.
I sometimes quadruple it, sixtuple it, or twelvopple it.

Colleagues have the same experiences, even with triodes.

Even when no special efforts are made to match, multiples of power
are produced.



Pentodes are current sources & can work well paralleled. But they still
have all the objectionable distortion characteristics typical of their
family. A way out of that is loop NFB & we see that often used to lower D
& get a reasonable DF.

You could also run the triodes with cathode resistors to move them closer
to CC operation. But that lowers the DF.

50 ohms in the cathode circuit of an EL34 in triode
makes the Ra rise from 1,250 ohms to about 1,750 ohms,
and it wastes 0.05 watts when there is 160 vrms across the 5k anode load,
(
5.12 watts of output power )
So one could have 200 ohms in the '34 cathode circuit, Ra would rise to
3,250 ohms or about equal to a 50% UL connected '34.
With 160vrms across the 5k load, the same 5.12 watts is produced, but
0.2 watts is wasted in the cathode R, and this is a very small wastage.
The applied current FB will reduce the gain and distortion like all NFB
does,

but only marginally, but it won't do any good in a PP circuit where
individual unbypassed cathode resistors in a class AB circuit are not too
good a thing
where one wants to have a fixed bias operation.
Quite OK in any SEW circuit though.

To reduce the Ro of the amp after increasing it with unbypassed Rk,
external loops of voltage NFB must be used, and to get a circuit
with 3 times the usual Ra back down to the original Ra
might take at least 8 dB of global or other series voltage NFB, or perhaps
a
little
positive current FB which makes distortion worse after the betterment from
the negative series current FB in a given output triode cathode circuit.

Its all rather bloomin messy.


When running tubes in parallel one usually needs to included swamping
resistors on all grid leads & sometimes plate leads as well in order to
suppress parasitic RF oscillations.

In my 300 watt amps with a dozen 6550 all I needed was the rather low
value
2.2k g1 resistor stoppers.
There are 220 ohms in series with the screens and their supply.
Most important are the zobel neworks across each 1/2 of the primary
on the OPT.
This is placed half way along the line up of six tubes on each side of the
PP
circuit.
Not a sign of any RF instability, and I have an OPT with 270 khz of
open loop bandwidth.

Should a tube decide to go beserko and go into thermal
runaway, then sometimes it will also oscillate at RF,
as well as conduct a huge cathode current.
The active protection I have in all my amps detects
the excess DC always associated with a hot tube.
If one or more tubes sustains twice the idle current for more than
4 seconds, the amp shuts down.

I once repaired a PA amp twice which had an octet of EL34, with
one single fixed bias voltage, Ea = 900v, Eg2 = 450v.
After the second repair I insisted I inspect his two
bins with 6 x 12" speakers in each.
He maintained he had 8 ohm speakers.
Sure, he was dead right, but all were wired in parallel,
and the most he got was 50 watts in the speakers, 25 watts in the cables,
and a shirtload of heat in the tubes, which had been pooping out regularly
for 15 years.
Once the amp was re-loaded correctly, biased with 8 pots,
had zobels connected right, and some stability added with the
12 dB of GNFB I added as well as a regulated screen supply, he never
phoned
me again,
and the sound was truly glorious.



Instead of using unbypassed cathode resistors and triode connection in my
300
watt amps
to only marginally assist load sharing, I have 20% of the primary
signal voltage applied to the cathodes as series voltage NFB from the OPT.
The screens are taken to +330v, Ea = 470v, and individual cathode bias is
applied
to each tube to regulate the bias.
Some fixed bias is used to keep the value of Rk to a minimum
and keep the DC Pd in the RK low, yet regulate bias well enough.
I also use a couple of power bjts to dynamically shunt the
excess 2H currents in the cathode circuits thus maintaining thre voltage
across the total of 12,000 uF in each bypass cap on each half of the PP
circuit.
So with dynamic stabilizing of the cathode bias the tubes tend to share
their
load
pretty darn well.
They are all subject to the same cathode series voltage FB, not current
FB.
But since they are tetrode connected but with voltage FB there seems to be
no

tendency for the tubes to differ much in their current donation to the
load
power.

With many rather poorly conceived commercially made amps such as the Baird
with a six pack of EL84 to give the rather large 60 watts of power per
channel,
there is but one sole value of fixed bias.
Baird didn't want to load down customers with the responsibility to adjust
12 bias pots, something many users would invariably get wrong.
Therefore as all these little tube age their bias drift is large, and some
would
fade faster than others, and conduct differently, but the drift wouldn't
be
any worse than
having just a pair of output 6550 with the same grid bias applied.
3 x EL84 can do about what one 6550 can do.
10 x EL84 each with its own cathode bias circuit would be more effective
for 50 watts of class AB with a larger amount of class A,
and since the power used won't be high, the tubes would last and stay
well enough matched to provide the power wanted.




The circuit I did my tests on was a simple stereo amp using a pair of
cathode biased 6AQ5's into Hammond 125D OPT's in each channel. Not the
best amp at all but it did illustrate the principle well. Since each amp
of the pair had loop NFB it was in a sense a voltage source. If my memory
is still in gear the DF was about 3 so in that respect very much like an
SET circuit. But there the similarity ends.

Paralleling amps which have two separate FB loops which of course that
will
means that the Or of one amp is say 2 ohms, and the other might be
2.2 ohms, due to vagaries of tubes and matching, and if the common load is
8 ohms then one amp sees nearly the same Ro as the other.
You can do the math with 2 ohms from a true zero ohms voltage source,
and another voltage source feeding 2.2 ohms, and tell me what the
mismatch does, but I'd say not much.

With a pair of solid state amps with very low Ro of say 100 milliohms
each,
the effects of different amp gains could have serious load sharing
problems.
But not when Ro is over 1 ohm.

With very low Ro SS amps one could happily power
a common load with 4 amps in parallel, but each with 1 ohm as a series
R in the output, and this would give the effect of having Rout = about
0.25
ohms.
Its not hard to have 1% accurately matched feedback networks in each
amp to ensure the voltage gain is the same.
Such close matching of voltage gain is far easier to attain with
modern metal film resistors than rely on device matching,
or the fact that we have current sources.
if the Ro of an amp = 0.1 ohms or less like most SS amps, the
1 ohm series R will convert the amp to a current source from the amp's
point of view, since it sees at least 1 ohm of load.
The voltage NFB ensures that the actual load value seen by each of
multiple
amps
will stay close when the series 1 ohm "safety resistors" are in place
Such R can be used to sense current overloading.



Seems we still can't get something for nothing!

Well, be prepared to pay the price.

Youse will never ever hear a gal say
" jus lie down here luvvy; this won't costyer anyfink "


I think you overstate a problem.

Triodes are current sources for the reasons I gave above
in the top paragraph. One tube sees all the parallel Ra and RL
of the other tubes.
All the rowers in a boat don't make the same power,
some row harder than others; none make it more difficult
for any other to row towards that medal.
Unless one falls out of the boat and tangles her oars in a daze of over
exertion,
and this happened in one of the Oz boats with a girls crew in the last
Olympics.
One girl just passed out, ****ed right up, and
stuffed any chance of a medal. The other girls were not happy......

Anyway, next time you are in your workshop, plug a 6L6
and KT88 into each side of a PP circuit and wire them as triodes
and bias them both for 22 watts, and do a test in class A,
then parallel some tubes in triode in an SE circuit, and get back to us
with your findings.

Forget the theorizing, see what happens in reality land.

BTW, with 625 trioded EL34 in parallel, a single EL34
sees a load and Ra resistance = approximately 1.6 ohms.

Output power at 5.1 watts average each = 3,187 watts.

The 625 tubes are needed for a very nice OTL load match for SE class A.

In many OTL amps, paralel output triodes are very prone to
individually thermaling out since the load match is usually
so damn attrocious.

Patrick Turner.



Cheers, John Stewart

Fabio Berutti wrote:

Interesting. I'm generally an opponent of paralleled tubes for absolutely
no technical reason, just 'cause in my job I always faced a f%#!!ing mess to
make two compressors or two pumps to work together in parallel. I was
surprised sometimes by the fact that it was easier to share load among 3 or
4 units than just 2, but now I got the "intuitive" point: it depends from
the perturbation that the single unit causes to the whole. If 10
compressors are blowing into a single pipe and one fails, the 9 still
running do not even notice, but if it's one out of two...
In the same way (I suppose) goes tubes behavior: in a parallel of 4 units,
the differences in the Gm of any single one will affect in a minor way the
working point of all other ones.
But when only 2 tubes are paralleled, "matched pairs" are mandatory;.

I have made a lot of amps with a quad of tubes and had no trouble
with the parallel pairs each side of the PP circuit, and in some the tubes were
not matched.


Triode/pentode mode has the same meaning: what happens to the second tube if
the first conducts more/less?

With PP hi-fi amps the tubes rarely work hard.
In most amps with a quad of tubes you can pull out one tube and the amp
still works OK for the next 20 years. The thd is a good bit higher, more 2H.


Adjustable bias will fix only a part of the problem (DC current) but still
any single tube will operate in a different point of its curve and reach
saturation at different signal levels.

Its an insignificant problem.


My "less is more" approach, here, is not a good one (it's never useful to be
too much devoted to a rule, without putting it in a context).

(Nevertheless, I still think that 2 tubes only are a bit of a trouble: I
remember I saw a paralleled ECC83 on a 'scope, 1K sine wave. One half had a
higher gain. The output graph was a real "seno" (in Italian the sine wave
and the feminine breast share the same word): a beautiful sine wave, with a
clearly visible "nipple" on its top, where the "strongest" half tube got
in..)

I routinely parallel both halves of a 6SN7, 12AT7, like many other makers.

I have never ever seen nipples on a breast wave yet.

Every time I have ever paralleled triode tube halves the thd is reduced.
Its a good reason to parallel tubes.

The strongest share this Earth with its weakest, and we hope it is with
tolerance and compassion
that they work together harmoniously, and multiple triodes seem to show us what
mankind and womenkind could so easily do, but won't.

Patrick Turner.





Ciao

Fabio

"Patrick Turner" ha scritto nel messaggio
...


John Stewart wrote:

John Stewart wrote:

Ian Iveson wrote:

"John Stewart" wrote

...Pentodes are current devices, so have no problem sharing the
load. However the
same tubes connected as triodes are voltage sources. Paralleled
voltage sources
do not share the load well at all. The result is higher than
expected distortion
at all levels....

I'm surprised. You are usually one for posting supporting evidence,
John. I can see the idea, but do you have actual data on this?

cheers, Ian

It is here somewhere. I will dig it out & post in the next day or
two. JLS

I must apologize on this one Ian. I can't find those results anywhere in
my file. I did that work sometime more than 40 years ago before going
into sales with HP. Perhaps I will have another look again when I get
time to do some tests since it is a rather interesting subject. Or
perhaps someone out there will take a shot at it!

However, in the meantime we can use a little network theory to illustrate
what is happening. Imagine in a perfect world we have a pair of identical
voltage sources connected to a common load, both thru identical
impedances. The voltage source polarities are connected so that +ve is to
+ve & -ve is to -ve. The simplest form this would take would be a pair of
identical batteries driving a resistive load thru identical resistors.
The resultant current in the load would be the simple sum of the currents
provided by the two batteries, those currents being equal to each other.

I have a colleague in Sydney who built a pair of 80 watt SET amps
using 10 x 300B in parallel.
The load was about 400 ohms for the lot and
they all shared the load pretty darn well.
The ratio between any one tube's Ra of 800 ohms to the
rest of the parallel tubes' Ra which is 88 0hms and the 400 ohms RL
is 800 : 72, so each tube works as a current source
to the total resistance it is connected to, since one tube's Ra
is large compared to all the the external resistances.

In all the experiments I have made with triode connected tubes
connected in parallel I have found only very slight differences
between the signal cathode currents in the parallel tubes.




But as usual the world is far from perfect as are this pair of batteries
& the resistors. One or the other battery will manage to sink or source a
current thru the other battery while at the same time push some current
thru the load. The two currents in the load are no longer identical &
never will be.

For the audio case circumstances get much worse.

Nope.

If we are lucky enough
to have say a pair of WE300B's, carefully matched on an AVO tester it
will help a lot, but there are still problems & lots of them.

Nope.

At how many
points can we match? It is unlikely we can get a match for all conditions
of operation on the plate family of curves.

The variations don't make much difference...



Lets replace the batteries & resistors above with the WE300B's.

The WE300B's (And any other triode) are voltages sources as ordinarily
connected & pretty good ones at that.

They are very weak voltage sources.
Ra of 800 ohms is not that much lower than the 4k one might load the tube
with.
It seems like a lot, but it isn't.
If the Ra was 8 ohms, there might be a slight problem



If they are driven by a common
source into a resistive load will they each provide the same current to
the load? It seems rather unlikely no matter how careful we are.

It is in practice very likely they will share the current fairly evenly.

One could have 625 EL34 in triode all hooked up to 8 ohms
The average load seen by each EL34 would be 5,000 ohms, and a reasonable
load
match for each tube.
Since no two tubes are exactly matched and they have different gm,
then although the anode voltage swing of 160vrms would be
the same for each tube, the current in each tube would differ according to
each tubes's
gm, and since u would be so close for all it is of negligible concern,
since
u
depends on the physical dimensions between electrodes.
Therefore it could be shown that each tube would itself experience a
different load to its neighbour,
one might think its load is 5,300 ohms, the next 4,900, next 5,100 and so
on.

Not all horses pull the carriage with the same force, but they all do what
suits them.



And the loadline presented to them will not be a nice straight line as
you often see in a text book.

The load line in PP class A amps for each tube is actually a bent line,
and very bent in PP class AB amps.
while in class A, each tube of a pair sees say 5k, to make up the total
10k a-a RL.
But when one has cut off, the other tube sees just 2.5k.
Changes of gain during the cycle cause distortion.

That is the simplest possible condition &
seldom happens in real life.

The SE amp will have the straightest load line.

The loadline if driven by a single frequency
might be at best a straight line but more than likely elliptical. That is
because the load is not a simple resistor but a loudspeaker. And there is
an OPT on the way to the loudspeaker to further confuse things.

The same elliptical load line is shared between all parallel tubes.

The
WE300B's have some non-linearites & so do the OPT's, all generating some
distortion of their own.

And the individual distortion currents sum to make but one average
distortion
current.

In PP amps with a quad or more OP tubes in parallel we have the
opportunity
to swap tubes around so the total signal currents of each 1/2 of the
PP output circuit are closely matched, thus cancelling the effect of the
mismatch between tubes, and getting almost perfect cancellation
of 2H between the two PP halves of the circuit.
Within each 1/2 I have found the currents near equal between 2 or 3 or 6
output tubes
in parallel with the same phase of signal.



Things get really crazy with the program material we would like to use
whether it be Bach or The Rolling Stones (I prefer the Stones). That is
because we now have a complex signal of many frequencies running into a
complex load (the loudspeaker).

Signal complexity makes no difference to the sharing ability of tube
behaviour.
A signal is a signal is a signal. Just voltage current changes.



Here again we have a condition where one or the other of the WE300B's can
sink or source some current into the other while at the same time driving
part of the load current! That means that each of the tubes are at times
driving the other.

I have never seen this. One could say that if you pull one tube out, the
other 5
have to work harder.
But the direction of current flow is never into one tube when its out of
all the rest.
What simply happens is that one tube experiences a different load to the
other.
You may say that if the difference in current is 2 mA between a pair
of parallel tubes then the net 2 mA is flowing from one tube to the other,
some how creating power in a tube to reduce efficiency.
Its over stating what's going on.



Some of the power developed by each tube is simply
dissipated in the other & never does show up in your loudspeaker load.

You mean one tube dissipates more heat than the other.
If a weak gm tube is paired with a much stronger gm tube then
of course the weaker one might see 7k and the stronger one 4k,
and the actual load is 2.54 k.
The one that acts like it sees 4k will have more dissipation
than the one that has the 7k, and the efficiencies of each will vary as
you
calculate.
In practice it isn't a big deal.

One could make the contribution of power near equal if we had
current FB as one is forced to have in SS amps with multiple
bjt or mosfets in parallel. Especially bjts, because if one bjt conducts
more
than the others it heats up
and conducts even more, hence the use of the 0.33 ohm emitter resistors
seen in many SS amps with multiple devices.
Mosfets and tubes back off a bit when they get hotter, since they have
negative temp
characteristics, ie, hotter means less gm.

Tubes when paralleled can tend to stay matched, because any that work
harder
than the rest
wear out faster, and they gm reduces......



So if you parallel a pair of triodes you won't get double the power at a
particular D. But with great care you might come close.

I routinely double the power with parallel tubes.
I sometimes quadruple it, sixtuple it, or twelvopple it.

Colleagues have the same experiences, even with triodes.

Even when no special efforts are made to match, multiples of power
are produced.



Pentodes are current sources & can work well paralleled. But they still
have all the objectionable distortion characteristics typical of their
family. A way out of that is loop NFB & we see that often used to lower D
& get a reasonable DF.

You could also run the triodes with cathode resistors to move them closer
to CC operation. But that lowers the DF.

50 ohms in the cathode circuit of an EL34 in triode
makes the Ra rise from 1,250 ohms to about 1,750 ohms,
and it wastes 0.05 watts when there is 160 vrms across the 5k anode load,
(
5.12 watts of output power )
So one could have 200 ohms in the '34 cathode circuit, Ra would rise to
3,250 ohms or about equal to a 50% UL connected '34.
With 160vrms across the 5k load, the same 5.12 watts is produced, but
0.2 watts is wasted in the cathode R, and this is a very small wastage.
The applied current FB will reduce the gain and distortion like all NFB
does,

but only marginally, but it won't do any good in a PP circuit where
individual unbypassed cathode resistors in a class AB circuit are not too
good a thing
where one wants to have a fixed bias operation.
Quite OK in any SEW circuit though.

To reduce the Ro of the amp after increasing it with unbypassed Rk,
external loops of voltage NFB must be used, and to get a circuit
with 3 times the usual Ra back down to the original Ra
might take at least 8 dB of global or other series voltage NFB, or perhaps
a
little
positive current FB which makes distortion worse after the betterment from
the negative series current FB in a given output triode cathode circuit.

Its all rather bloomin messy.


When running tubes in parallel one usually needs to included swamping
resistors on all grid leads & sometimes plate leads as well in order to
suppress parasitic RF oscillations.

In my 300 watt amps with a dozen 6550 all I needed was the rather low
value
2.2k g1 resistor stoppers.
There are 220 ohms in series with the screens and their supply.
Most important are the zobel neworks across each 1/2 of the primary
on the OPT.
This is placed half way along the line up of six tubes on each side of the
PP
circuit.
Not a sign of any RF instability, and I have an OPT with 270 khz of
open loop bandwidth.

Should a tube decide to go beserko and go into thermal
runaway, then sometimes it will also oscillate at RF,
as well as conduct a huge cathode current.
The active protection I have in all my amps detects
the excess DC always associated with a hot tube.
If one or more tubes sustains twice the idle current for more than
4 seconds, the amp shuts down.

I once repaired a PA amp twice which had an octet of EL34, with
one single fixed bias voltage, Ea = 900v, Eg2 = 450v.
After the second repair I insisted I inspect his two
bins with 6 x 12" speakers in each.
He maintained he had 8 ohm speakers.
Sure, he was dead right, but all were wired in parallel,
and the most he got was 50 watts in the speakers, 25 watts in the cables,
and a shirtload of heat in the tubes, which had been pooping out regularly
for 15 years.
Once the amp was re-loaded correctly, biased with 8 pots,
had zobels connected right, and some stability added with the
12 dB of GNFB I added as well as a regulated screen supply, he never
phoned
me again,
and the sound was truly glorious.



Instead of using unbypassed cathode resistors and triode connection in my
300
watt amps
to only marginally assist load sharing, I have 20% of the primary
signal voltage applied to the cathodes as series voltage NFB from the OPT.
The screens are taken to +330v, Ea = 470v, and individual cathode bias is
applied
to each tube to regulate the bias.
Some fixed bias is used to keep the value of Rk to a minimum
and keep the DC Pd in the RK low, yet regulate bias well enough.
I also use a couple of power bjts to dynamically shunt the
excess 2H currents in the cathode circuits thus maintaining thre voltage
across the total of 12,000 uF in each bypass cap on each half of the PP
circuit.
So with dynamic stabilizing of the cathode bias the tubes tend to share
their
load
pretty darn well.
They are all subject to the same cathode series voltage FB, not current
FB.
But since they are tetrode connected but with voltage FB there seems to be
no

tendency for the tubes to differ much in their current donation to the
load
power.

With many rather poorly conceived commercially made amps such as the Baird
with a six pack of EL84 to give the rather large 60 watts of power per
channel,
there is but one sole value of fixed bias.
Baird didn't want to load down customers with the responsibility to adjust
12 bias pots, something many users would invariably get wrong.
Therefore as all these little tube age their bias drift is large, and some
would
fade faster than others, and conduct differently, but the drift wouldn't
be
any worse than
having just a pair of output 6550 with the same grid bias applied.
3 x EL84 can do about what one 6550 can do.
10 x EL84 each with its own cathode bias circuit would be more effective
for 50 watts of class AB with a larger amount of class A,
and since the power used won't be high, the tubes would last and stay
well enough matched to provide the power wanted.




The circuit I did my tests on was a simple stereo amp using a pair of
cathode biased 6AQ5's into Hammond 125D OPT's in each channel. Not the
best amp at all but it did illustrate the principle well. Since each amp
of the pair had loop NFB it was in a sense a voltage source. If my memory
is still in gear the DF was about 3 so in that respect very much like an
SET circuit. But there the similarity ends.

Paralleling amps which have two separate FB loops which of course that
will
means that the Or of one amp is say 2 ohms, and the other might be
2.2 ohms, due to vagaries of tubes and matching, and if the common load is
8 ohms then one amp sees nearly the same Ro as the other.
You can do the math with 2 ohms from a true zero ohms voltage source,
and another voltage source feeding 2.2 ohms, and tell me what the
mismatch does, but I'd say not much.

With a pair of solid state amps with very low Ro of say 100 milliohms
each,
the effects of different amp gains could have serious load sharing
problems.
But not when Ro is over 1 ohm.

With very low Ro SS amps one could happily power
a common load with 4 amps in parallel, but each with 1 ohm as a series
R in the output, and this would give the effect of having Rout = about
0.25
ohms.
Its not hard to have 1% accurately matched feedback networks in each
amp to ensure the voltage gain is the same.
Such close matching of voltage gain is far easier to attain with
modern metal film resistors than rely on device matching,
or the fact that we have current sources.
if the Ro of an amp = 0.1 ohms or less like most SS amps, the
1 ohm series R will convert the amp to a current source from the amp's
point of view, since it sees at least 1 ohm of load.
The voltage NFB ensures that the actual load value seen by each of
multiple
amps
will stay close when the series 1 ohm "safety resistors" are in place
Such R can be used to sense current overloading.



Seems we still can't get something for nothing!

Well, be prepared to pay the price.

Youse will never ever hear a gal say
" jus lie down here luvvy; this won't costyer anyfink "


I think you overstate a problem.

Triodes are current sources for the reasons I gave above
in the top paragraph. One tube sees all the parallel Ra and RL
of the other tubes.
All the rowers in a boat don't make the same power,
some row harder than others; none make it more difficult
for any other to row towards that medal.
Unless one falls out of the boat and tangles her oars in a daze of over
exertion,
and this happened in one of the Oz boats with a girls crew in the last
Olympics.
One girl just passed out, ****ed right up, and
stuffed any chance of a medal. The other girls were not happy......

Anyway, next time you are in your workshop, plug a 6L6
and KT88 into each side of a PP circuit and wire them as triodes
and bias them both for 22 watts, and do a test in class A,
then parallel some tubes in triode in an SE circuit, and get back to us
with your findings.

Forget the theorizing, see what happens in reality land.

BTW, with 625 trioded EL34 in parallel, a single EL34
sees a load and Ra resistance = approximately 1.6 ohms.

Output power at 5.1 watts average each = 3,187 watts.

The 625 tubes are needed for a very nice OTL load match for SE class A.

In many OTL amps, paralel output triodes are very prone to
individually thermaling out since the load match is usually
so damn attrocious.

Patrick Turner.



Cheers, John Stewart

Patrick,
In your monster amp of 12 KT88s, have you ever thought of banking them? For
instance, flip a switch and have only 1 pair/side running, another 2
pairs/side, etc. I suppose you would do it with turning the B+ on & off. Not
always will you need 12 tubes running. This will save on tube life and will
not force you to kill a mosquito with a cannon. Thoughts?
west

"Patrick Turner" wrote in message
...


Fabio Berutti wrote:

Interesting. I'm generally an opponent of paralleled tubes for
absolutely
no technical reason, just 'cause in my job I always faced a f%#!!ing
mess to
make two compressors or two pumps to work together in parallel. I was
surprised sometimes by the fact that it was easier to share load among 3
or
4 units than just 2, but now I got the "intuitive" point: it depends
from
the perturbation that the single unit causes to the whole. If 10
compressors are blowing into a single pipe and one fails, the 9 still
running do not even notice, but if it's one out of two...
In the same way (I suppose) goes tubes behavior: in a parallel of 4
units,
the differences in the Gm of any single one will affect in a minor way
the
working point of all other ones.
But when only 2 tubes are paralleled, "matched pairs" are mandatory;.

I have made a lot of amps with a quad of tubes and had no trouble
with the parallel pairs each side of the PP circuit, and in some the tubes
were
not matched.


Triode/pentode mode has the same meaning: what happens to the second
tube if
the first conducts more/less?

With PP hi-fi amps the tubes rarely work hard.
In most amps with a quad of tubes you can pull out one tube and the amp
still works OK for the next 20 years. The thd is a good bit higher, more
2H.


Adjustable bias will fix only a part of the problem (DC current) but
still
any single tube will operate in a different point of its curve and reach
saturation at different signal levels.

Its an insignificant problem.


My "less is more" approach, here, is not a good one (it's never useful
to be
too much devoted to a rule, without putting it in a context).

(Nevertheless, I still think that 2 tubes only are a bit of a trouble: I
remember I saw a paralleled ECC83 on a 'scope, 1K sine wave. One half
had a
higher gain. The output graph was a real "seno" (in Italian the sine
wave
and the feminine breast share the same word): a beautiful sine wave,
with a
clearly visible "nipple" on its top, where the "strongest" half tube got
in..)

I routinely parallel both halves of a 6SN7, 12AT7, like many other makers.

I have never ever seen nipples on a breast wave yet.

Every time I have ever paralleled triode tube halves the thd is reduced.
Its a good reason to parallel tubes.

The strongest share this Earth with its weakest, and we hope it is with
tolerance and compassion
that they work together harmoniously, and multiple triodes seem to show us
what
mankind and womenkind could so easily do, but won't.

Patrick Turner.





Ciao

Fabio

"Patrick Turner" ha scritto nel messaggio
...


John Stewart wrote:

John Stewart wrote:

Ian Iveson wrote:

"John Stewart" wrote

...Pentodes are current devices, so have no problem sharing the
load. However the
same tubes connected as triodes are voltage sources. Paralleled
voltage sources
do not share the load well at all. The result is higher than
expected distortion
at all levels....

I'm surprised. You are usually one for posting supporting
evidence,
John. I can see the idea, but do you have actual data on this?

cheers, Ian

It is here somewhere. I will dig it out & post in the next day or
two. JLS

I must apologize on this one Ian. I can't find those results anywhere
in
my file. I did that work sometime more than 40 years ago before going
into sales with HP. Perhaps I will have another look again when I get
time to do some tests since it is a rather interesting subject. Or
perhaps someone out there will take a shot at it!

However, in the meantime we can use a little network theory to
illustrate
what is happening. Imagine in a perfect world we have a pair of
identical
voltage sources connected to a common load, both thru identical
impedances. The voltage source polarities are connected so that +ve
is to
+ve & -ve is to -ve. The simplest form this would take would be a
pair of
identical batteries driving a resistive load thru identical
resistors.
The resultant current in the load would be the simple sum of the
currents
provided by the two batteries, those currents being equal to each
other.

I have a colleague in Sydney who built a pair of 80 watt SET amps
using 10 x 300B in parallel.
The load was about 400 ohms for the lot and
they all shared the load pretty darn well.
The ratio between any one tube's Ra of 800 ohms to the
rest of the parallel tubes' Ra which is 88 0hms and the 400 ohms RL
is 800 : 72, so each tube works as a current source
to the total resistance it is connected to, since one tube's Ra
is large compared to all the the external resistances.

In all the experiments I have made with triode connected tubes
connected in parallel I have found only very slight differences
between the signal cathode currents in the parallel tubes.




But as usual the world is far from perfect as are this pair of
batteries
& the resistors. One or the other battery will manage to sink or
source a
current thru the other battery while at the same time push some
current
thru the load. The two currents in the load are no longer identical &
never will be.

For the audio case circumstances get much worse.

Nope.

If we are lucky enough
to have say a pair of WE300B's, carefully matched on an AVO tester it
will help a lot, but there are still problems & lots of them.

Nope.

At how many
points can we match? It is unlikely we can get a match for all
conditions
of operation on the plate family of curves.

The variations don't make much difference...



Lets replace the batteries & resistors above with the WE300B's.

The WE300B's (And any other triode) are voltages sources as
ordinarily
connected & pretty good ones at that.

They are very weak voltage sources.
Ra of 800 ohms is not that much lower than the 4k one might load the
tube
with.
It seems like a lot, but it isn't.
If the Ra was 8 ohms, there might be a slight problem



If they are driven by a common
source into a resistive load will they each provide the same current
to
the load? It seems rather unlikely no matter how careful we are.

It is in practice very likely they will share the current fairly
evenly.

One could have 625 EL34 in triode all hooked up to 8 ohms
The average load seen by each EL34 would be 5,000 ohms, and a
reasonable
load
match for each tube.
Since no two tubes are exactly matched and they have different gm,
then although the anode voltage swing of 160vrms would be
the same for each tube, the current in each tube would differ
according to
each tubes's
gm, and since u would be so close for all it is of negligible concern,
since
u
depends on the physical dimensions between electrodes.
Therefore it could be shown that each tube would itself experience a
different load to its neighbour,
one might think its load is 5,300 ohms, the next 4,900, next 5,100 and
so
on.

Not all horses pull the carriage with the same force, but they all do
what
suits them.



And the loadline presented to them will not be a nice straight line
as
you often see in a text book.

The load line in PP class A amps for each tube is actually a bent
line,
and very bent in PP class AB amps.
while in class A, each tube of a pair sees say 5k, to make up the
total
10k a-a RL.
But when one has cut off, the other tube sees just 2.5k.
Changes of gain during the cycle cause distortion.

That is the simplest possible condition &
seldom happens in real life.

The SE amp will have the straightest load line.

The loadline if driven by a single frequency
might be at best a straight line but more than likely elliptical.
That is
because the load is not a simple resistor but a loudspeaker. And
there is
an OPT on the way to the loudspeaker to further confuse things.

The same elliptical load line is shared between all parallel tubes.

The
WE300B's have some non-linearites & so do the OPT's, all generating
some
distortion of their own.

And the individual distortion currents sum to make but one average
distortion
current.

In PP amps with a quad or more OP tubes in parallel we have the
opportunity
to swap tubes around so the total signal currents of each 1/2 of the
PP output circuit are closely matched, thus cancelling the effect of
the
mismatch between tubes, and getting almost perfect cancellation
of 2H between the two PP halves of the circuit.
Within each 1/2 I have found the currents near equal between 2 or 3 or
6
output tubes
in parallel with the same phase of signal.



Things get really crazy with the program material we would like to
use
whether it be Bach or The Rolling Stones (I prefer the Stones). That
is
because we now have a complex signal of many frequencies running into
a
complex load (the loudspeaker).

Signal complexity makes no difference to the sharing ability of tube
behaviour.
A signal is a signal is a signal. Just voltage current changes.



Here again we have a condition where one or the other of the WE300B's
can
sink or source some current into the other while at the same time
driving
part of the load current! That means that each of the tubes are at
times
driving the other.

I have never seen this. One could say that if you pull one tube out,
the
other 5
have to work harder.
But the direction of current flow is never into one tube when its out
of
all the rest.
What simply happens is that one tube experiences a different load to
the
other.
You may say that if the difference in current is 2 mA between a pair
of parallel tubes then the net 2 mA is flowing from one tube to the
other,
some how creating power in a tube to reduce efficiency.
Its over stating what's going on.



Some of the power developed by each tube is simply
dissipated in the other & never does show up in your loudspeaker
load.

You mean one tube dissipates more heat than the other.
If a weak gm tube is paired with a much stronger gm tube then
of course the weaker one might see 7k and the stronger one 4k,
and the actual load is 2.54 k.
The one that acts like it sees 4k will have more dissipation
than the one that has the 7k, and the efficiencies of each will vary
as
you
calculate.
In practice it isn't a big deal.

One could make the contribution of power near equal if we had
current FB as one is forced to have in SS amps with multiple
bjt or mosfets in parallel. Especially bjts, because if one bjt
conducts
more
than the others it heats up
and conducts even more, hence the use of the 0.33 ohm emitter
resistors
seen in many SS amps with multiple devices.
Mosfets and tubes back off a bit when they get hotter, since they have
negative temp
characteristics, ie, hotter means less gm.

Tubes when paralleled can tend to stay matched, because any that work
harder
than the rest
wear out faster, and they gm reduces......



So if you parallel a pair of triodes you won't get double the power
at a
particular D. But with great care you might come close.

I routinely double the power with parallel tubes.
I sometimes quadruple it, sixtuple it, or twelvopple it.

Colleagues have the same experiences, even with triodes.

Even when no special efforts are made to match, multiples of power
are produced.



Pentodes are current sources & can work well paralleled. But they
still
have all the objectionable distortion characteristics typical of
their
family. A way out of that is loop NFB & we see that often used to
lower D
& get a reasonable DF.

You could also run the triodes with cathode resistors to move them
closer
to CC operation. But that lowers the DF.

50 ohms in the cathode circuit of an EL34 in triode
makes the Ra rise from 1,250 ohms to about 1,750 ohms,
and it wastes 0.05 watts when there is 160 vrms across the 5k anode
load,
(
5.12 watts of output power )
So one could have 200 ohms in the '34 cathode circuit, Ra would rise
to
3,250 ohms or about equal to a 50% UL connected '34.
With 160vrms across the 5k load, the same 5.12 watts is produced, but
0.2 watts is wasted in the cathode R, and this is a very small
wastage.
The applied current FB will reduce the gain and distortion like all
NFB
does,

but only marginally, but it won't do any good in a PP circuit where
individual unbypassed cathode resistors in a class AB circuit are not
too
good a thing
where one wants to have a fixed bias operation.
Quite OK in any SEW circuit though.

To reduce the Ro of the amp after increasing it with unbypassed Rk,
external loops of voltage NFB must be used, and to get a circuit
with 3 times the usual Ra back down to the original Ra
might take at least 8 dB of global or other series voltage NFB, or
perhaps
a
little
positive current FB which makes distortion worse after the betterment
from
the negative series current FB in a given output triode cathode
circuit.

Its all rather bloomin messy.


When running tubes in parallel one usually needs to included swamping
resistors on all grid leads & sometimes plate leads as well in order
to
suppress parasitic RF oscillations.

In my 300 watt amps with a dozen 6550 all I needed was the rather low
value
2.2k g1 resistor stoppers.
There are 220 ohms in series with the screens and their supply.
Most important are the zobel neworks across each 1/2 of the primary
on the OPT.
This is placed half way along the line up of six tubes on each side of
the
PP
circuit.
Not a sign of any RF instability, and I have an OPT with 270 khz of
open loop bandwidth.

Should a tube decide to go beserko and go into thermal
runaway, then sometimes it will also oscillate at RF,
as well as conduct a huge cathode current.
The active protection I have in all my amps detects
the excess DC always associated with a hot tube.
If one or more tubes sustains twice the idle current for more than
4 seconds, the amp shuts down.

I once repaired a PA amp twice which had an octet of EL34, with
one single fixed bias voltage, Ea = 900v, Eg2 = 450v.
After the second repair I insisted I inspect his two
bins with 6 x 12" speakers in each.
He maintained he had 8 ohm speakers.
Sure, he was dead right, but all were wired in parallel,
and the most he got was 50 watts in the speakers, 25 watts in the
cables,
and a shirtload of heat in the tubes, which had been pooping out
regularly
for 15 years.
Once the amp was re-loaded correctly, biased with 8 pots,
had zobels connected right, and some stability added with the
12 dB of GNFB I added as well as a regulated screen supply, he never
phoned
me again,
and the sound was truly glorious.



Instead of using unbypassed cathode resistors and triode connection in
my
300
watt amps
to only marginally assist load sharing, I have 20% of the primary
signal voltage applied to the cathodes as series voltage NFB from the
OPT.
The screens are taken to +330v, Ea = 470v, and individual cathode bias
is
applied
to each tube to regulate the bias.
Some fixed bias is used to keep the value of Rk to a minimum
and keep the DC Pd in the RK low, yet regulate bias well enough.
I also use a couple of power bjts to dynamically shunt the
excess 2H currents in the cathode circuits thus maintaining thre
voltage
across the total of 12,000 uF in each bypass cap on each half of the
PP
circuit.
So with dynamic stabilizing of the cathode bias the tubes tend to
share
their
load
pretty darn well.
They are all subject to the same cathode series voltage FB, not
current
FB.
But since they are tetrode connected but with voltage FB there seems
to be
no

tendency for the tubes to differ much in their current donation to the
load
power.

With many rather poorly conceived commercially made amps such as the
Baird
with a six pack of EL84 to give the rather large 60 watts of power per
channel,
there is but one sole value of fixed bias.
Baird didn't want to load down customers with the responsibility to
adjust
12 bias pots, something many users would invariably get wrong.
Therefore as all these little tube age their bias drift is large, and
some
would
fade faster than others, and conduct differently, but the drift
wouldn't
be
any worse than
having just a pair of output 6550 with the same grid bias applied.
3 x EL84 can do about what one 6550 can do.
10 x EL84 each with its own cathode bias circuit would be more
effective
for 50 watts of class AB with a larger amount of class A,
and since the power used won't be high, the tubes would last and stay
well enough matched to provide the power wanted.




The circuit I did my tests on was a simple stereo amp using a pair of
cathode biased 6AQ5's into Hammond 125D OPT's in each channel. Not
the
best amp at all but it did illustrate the principle well. Since each
amp
of the pair had loop NFB it was in a sense a voltage source. If my
memory
is still in gear the DF was about 3 so in that respect very much like
an
SET circuit. But there the similarity ends.

Paralleling amps which have two separate FB loops which of course that
will
means that the Or of one amp is say 2 ohms, and the other might be
2.2 ohms, due to vagaries of tubes and matching, and if the common
load is
8 ohms then one amp sees nearly the same Ro as the other.
You can do the math with 2 ohms from a true zero ohms voltage source,
and another voltage source feeding 2.2 ohms, and tell me what the
mismatch does, but I'd say not much.

With a pair of solid state amps with very low Ro of say 100 milliohms
each,
the effects of different amp gains could have serious load sharing
problems.
But not when Ro is over 1 ohm.

With very low Ro SS amps one could happily power
a common load with 4 amps in parallel, but each with 1 ohm as a series
R in the output, and this would give the effect of having Rout = about
0.25
ohms.
Its not hard to have 1% accurately matched feedback networks in each
amp to ensure the voltage gain is the same.
Such close matching of voltage gain is far easier to attain with
modern metal film resistors than rely on device matching,
or the fact that we have current sources.
if the Ro of an amp = 0.1 ohms or less like most SS amps, the
1 ohm series R will convert the amp to a current source from the amp's
point of view, since it sees at least 1 ohm of load.
The voltage NFB ensures that the actual load value seen by each of
multiple
amps
will stay close when the series 1 ohm "safety resistors" are in place
Such R can be used to sense current overloading.



Seems we still can't get something for nothing!

Well, be prepared to pay the price.

Youse will never ever hear a gal say
" jus lie down here luvvy; this won't costyer anyfink "


I think you overstate a problem.

Triodes are current sources for the reasons I gave above
in the top paragraph. One tube sees all the parallel Ra and RL
of the other tubes.
All the rowers in a boat don't make the same power,
some row harder than others; none make it more difficult
for any other to row towards that medal.
Unless one falls out of the boat and tangles her oars in a daze of
over
exertion,
and this happened in one of the Oz boats with a girls crew in the last
Olympics.
One girl just passed out, ****ed right up, and
stuffed any chance of a medal. The other girls were not happy......

Anyway, next time you are in your workshop, plug a 6L6
and KT88 into each side of a PP circuit and wire them as triodes
and bias them both for 22 watts, and do a test in class A,
then parallel some tubes in triode in an SE circuit, and get back to
us
with your findings.

Forget the theorizing, see what happens in reality land.

BTW, with 625 trioded EL34 in parallel, a single EL34
sees a load and Ra resistance = approximately 1.6 ohms.

Output power at 5.1 watts average each = 3,187 watts.

The 625 tubes are needed for a very nice OTL load match for SE class
A.

In many OTL amps, paralel output triodes are very prone to
individually thermaling out since the load match is usually
so damn attrocious.

Patrick Turner.



Cheers, John Stewart

Patrick Turner wrote:

Thus Spake The Oracle (as follows) again leading us into a land covered by a 'Fog
of Bull****'. Patrick T will now repeal Kirchhoff's Laws. Later this year we will
have a new Law of Gravitation as well!!!! JLS

John Stewart wrote:

John Stewart wrote:

Ian Iveson wrote:

"John Stewart" wrote

...Pentodes are current devices, so have no problem sharing the
load. However the
same tubes connected as triodes are voltage sources. Paralleled
voltage sources
do not share the load well at all. The result is higher than
expected distortion
at all levels....

I'm surprised. You are usually one for posting supporting evidence,
John. I can see the idea, but do you have actual data on this?

cheers, Ian

It is here somewhere. I will dig it out & post in the next day or
two. JLS

I must apologize on this one Ian. I can't find those results anywhere in
my file. I did that work sometime more than 40 years ago before going
into sales with HP. Perhaps I will have another look again when I get
time to do some tests since it is a rather interesting subject. Or
perhaps someone out there will take a shot at it!

However, in the meantime we can use a little network theory to illustrate
what is happening. Imagine in a perfect world we have a pair of identical
voltage sources connected to a common load, both thru identical
impedances. The voltage source polarities are connected so that +ve is to
+ve & -ve is to -ve. The simplest form this would take would be a pair of
identical batteries driving a resistive load thru identical resistors.
The resultant current in the load would be the simple sum of the currents
provided by the two batteries, those currents being equal to each other.

I have a colleague in Sydney who built a pair of 80 watt SET amps
using 10 x 300B in parallel.
The load was about 400 ohms for the lot and
they all shared the load pretty darn well.
The ratio between any one tube's Ra of 800 ohms to the
rest of the parallel tubes' Ra which is 88 0hms and the 400 ohms RL
is 800 : 72, so each tube works as a current source
to the total resistance it is connected to, since one tube's Ra
is large compared to all the the external resistances.

In all the experiments I have made with triode connected tubes
connected in parallel I have found only very slight differences
between the signal cathode currents in the parallel tubes.



But as usual the world is far from perfect as are this pair of batteries
& the resistors. One or the other battery will manage to sink or source a
current thru the other battery while at the same time push some current
thru the load. The two currents in the load are no longer identical &
never will be.

For the audio case circumstances get much worse.

Nope.

If we are lucky enough
to have say a pair of WE300B's, carefully matched on an AVO tester it
will help a lot, but there are still problems & lots of them.

Nope.

At how many
points can we match? It is unlikely we can get a match for all conditions
of operation on the plate family of curves.

The variations don't make much difference...



Lets replace the batteries & resistors above with the WE300B's.

The WE300B's (And any other triode) are voltages sources as ordinarily
connected & pretty good ones at that.

They are very weak voltage sources.
Ra of 800 ohms is not that much lower than the 4k one might load the tube
with.
It seems like a lot, but it isn't.
If the Ra was 8 ohms, there might be a slight problem

If they are driven by a common
source into a resistive load will they each provide the same current to
the load? It seems rather unlikely no matter how careful we are.

It is in practice very likely they will share the current fairly evenly.

One could have 625 EL34 in triode all hooked up to 8 ohms
The average load seen by each EL34 would be 5,000 ohms, and a reasonable load
match for each tube.
Since no two tubes are exactly matched and they have different gm,
then although the anode voltage swing of 160vrms would be
the same for each tube, the current in each tube would differ according to
each tubes's
gm, and since u would be so close for all it is of negligible concern, since
u
depends on the physical dimensions between electrodes.
Therefore it could be shown that each tube would itself experience a
different load to its neighbour,
one might think its load is 5,300 ohms, the next 4,900, next 5,100 and so on.

Not all horses pull the carriage with the same force, but they all do what
suits them.



And the loadline presented to them will not be a nice straight line as
you often see in a text book.

The load line in PP class A amps for each tube is actually a bent line,
and very bent in PP class AB amps.
while in class A, each tube of a pair sees say 5k, to make up the total
10k a-a RL.
But when one has cut off, the other tube sees just 2.5k.
Changes of gain during the cycle cause distortion.

That is the simplest possible condition &
seldom happens in real life.

The SE amp will have the straightest load line.

The loadline if driven by a single frequency
might be at best a straight line but more than likely elliptical. That is
because the load is not a simple resistor but a loudspeaker. And there is
an OPT on the way to the loudspeaker to further confuse things.

The same elliptical load line is shared between all parallel tubes.

The
WE300B's have some non-linearites & so do the OPT's, all generating some
distortion of their own.

And the individual distortion currents sum to make but one average distortion
current.

In PP amps with a quad or more OP tubes in parallel we have the opportunity
to swap tubes around so the total signal currents of each 1/2 of the
PP output circuit are closely matched, thus cancelling the effect of the
mismatch between tubes, and getting almost perfect cancellation
of 2H between the two PP halves of the circuit.
Within each 1/2 I have found the currents near equal between 2 or 3 or 6
output tubes
in parallel with the same phase of signal.



Things get really crazy with the program material we would like to use
whether it be Bach or The Rolling Stones (I prefer the Stones). That is
because we now have a complex signal of many frequencies running into a
complex load (the loudspeaker).

Signal complexity makes no difference to the sharing ability of tube
behaviour.
A signal is a signal is a signal. Just voltage current changes.



Here again we have a condition where one or the other of the WE300B's can
sink or source some current into the other while at the same time driving
part of the load current! That means that each of the tubes are at times
driving the other.

I have never seen this. One could say that if you pull one tube out, the
other 5
have to work harder.
But the direction of current flow is never into one tube when its out of
all the rest.
What simply happens is that one tube experiences a different load to the
other.
You may say that if the difference in current is 2 mA between a pair
of parallel tubes then the net 2 mA is flowing from one tube to the other,
some how creating power in a tube to reduce efficiency.
Its over stating what's going on.

Some of the power developed by each tube is simply
dissipated in the other & never does show up in your loudspeaker load.

You mean one tube dissipates more heat than the other.
If a weak gm tube is paired with a much stronger gm tube then
of course the weaker one might see 7k and the stronger one 4k,
and the actual load is 2.54 k.
The one that acts like it sees 4k will have more dissipation
than the one that has the 7k, and the efficiencies of each will vary as you
calculate.
In practice it isn't a big deal.

One could make the contribution of power near equal if we had
current FB as one is forced to have in SS amps with multiple
bjt or mosfets in parallel. Especially bjts, because if one bjt conducts more
than the others it heats up
and conducts even more, hence the use of the 0.33 ohm emitter resistors
seen in many SS amps with multiple devices.
Mosfets and tubes back off a bit when they get hotter, since they have
negative temp
characteristics, ie, hotter means less gm.

Tubes when paralleled can tend to stay matched, because any that work harder
than the rest
wear out faster, and they gm reduces......



So if you parallel a pair of triodes you won't get double the power at a
particular D. But with great care you might come close.

I routinely double the power with parallel tubes.
I sometimes quadruple it, sixtuple it, or twelvopple it.

Colleagues have the same experiences, even with triodes.

Even when no special efforts are made to match, multiples of power
are produced.



Pentodes are current sources & can work well paralleled. But they still
have all the objectionable distortion characteristics typical of their
family. A way out of that is loop NFB & we see that often used to lower D
& get a reasonable DF.

You could also run the triodes with cathode resistors to move them closer
to CC operation. But that lowers the DF.

50 ohms in the cathode circuit of an EL34 in triode
makes the Ra rise from 1,250 ohms to about 1,750 ohms,
and it wastes 0.05 watts when there is 160 vrms across the 5k anode load, (
5.12 watts of output power )
So one could have 200 ohms in the '34 cathode circuit, Ra would rise to
3,250 ohms or about equal to a 50% UL connected '34.
With 160vrms across the 5k load, the same 5.12 watts is produced, but
0.2 watts is wasted in the cathode R, and this is a very small wastage.
The applied current FB will reduce the gain and distortion like all NFB does,

but only marginally, but it won't do any good in a PP circuit where
individual unbypassed cathode resistors in a class AB circuit are not too
good a thing
where one wants to have a fixed bias operation.
Quite OK in any SEW circuit though.

To reduce the Ro of the amp after increasing it with unbypassed Rk,
external loops of voltage NFB must be used, and to get a circuit
with 3 times the usual Ra back down to the original Ra
might take at least 8 dB of global or other series voltage NFB, or perhaps a
little
positive current FB which makes distortion worse after the betterment from
the negative series current FB in a given output triode cathode circuit.

Its all rather bloomin messy.


When running tubes in parallel one usually needs to included swamping
resistors on all grid leads & sometimes plate leads as well in order to
suppress parasitic RF oscillations.

In my 300 watt amps with a dozen 6550 all I needed was the rather low value
2.2k g1 resistor stoppers.
There are 220 ohms in series with the screens and their supply.
Most important are the zobel neworks across each 1/2 of the primary
on the OPT.
This is placed half way along the line up of six tubes on each side of the PP
circuit.
Not a sign of any RF instability, and I have an OPT with 270 khz of
open loop bandwidth.

Should a tube decide to go beserko and go into thermal
runaway, then sometimes it will also oscillate at RF,
as well as conduct a huge cathode current.
The active protection I have in all my amps detects
the excess DC always associated with a hot tube.
If one or more tubes sustains twice the idle current for more than
4 seconds, the amp shuts down.

I once repaired a PA amp twice which had an octet of EL34, with
one single fixed bias voltage, Ea = 900v, Eg2 = 450v.
After the second repair I insisted I inspect his two
bins with 6 x 12" speakers in each.
He maintained he had 8 ohm speakers.
Sure, he was dead right, but all were wired in parallel,
and the most he got was 50 watts in the speakers, 25 watts in the cables,
and a shirtload of heat in the tubes, which had been pooping out regularly
for 15 years.
Once the amp was re-loaded correctly, biased with 8 pots,
had zobels connected right, and some stability added with the
12 dB of GNFB I added as well as a regulated screen supply, he never phoned
me again,
and the sound was truly glorious.

Instead of using unbypassed cathode resistors and triode connection in my 300
watt amps
to only marginally assist load sharing, I have 20% of the primary
signal voltage applied to the cathodes as series voltage NFB from the OPT.
The screens are taken to +330v, Ea = 470v, and individual cathode bias is
applied
to each tube to regulate the bias.
Some fixed bias is used to keep the value of Rk to a minimum
and keep the DC Pd in the RK low, yet regulate bias well enough.
I also use a couple of power bjts to dynamically shunt the
excess 2H currents in the cathode circuits thus maintaining thre voltage
across the total of 12,000 uF in each bypass cap on each half of the PP
circuit.
So with dynamic stabilizing of the cathode bias the tubes tend to share their
load
pretty darn well.
They are all subject to the same cathode series voltage FB, not current FB.
But since they are tetrode connected but with voltage FB there seems to be no

tendency for the tubes to differ much in their current donation to the load
power.

With many rather poorly conceived commercially made amps such as the Baird
with a six pack of EL84 to give the rather large 60 watts of power per
channel,
there is but one sole value of fixed bias.
Baird didn't want to load down customers with the responsibility to adjust
12 bias pots, something many users would invariably get wrong.
Therefore as all these little tube age their bias drift is large, and some
would
fade faster than others, and conduct differently, but the drift wouldn't be
any worse than
having just a pair of output 6550 with the same grid bias applied.
3 x EL84 can do about what one 6550 can do.
10 x EL84 each with its own cathode bias circuit would be more effective
for 50 watts of class AB with a larger amount of class A,
and since the power used won't be high, the tubes would last and stay
well enough matched to provide the power wanted.


The circuit I did my tests on was a simple stereo amp using a pair of
cathode biased 6AQ5's into Hammond 125D OPT's in each channel. Not the
best amp at all but it did illustrate the principle well. Since each amp
of the pair had loop NFB it was in a sense a voltage source. If my memory
is still in gear the DF was about 3 so in that respect very much like an
SET circuit. But there the similarity ends.

Paralleling amps which have two separate FB loops which of course that will
means that the Or of one amp is say 2 ohms, and the other might be
2.2 ohms, due to vagaries of tubes and matching, and if the common load is
8 ohms then one amp sees nearly the same Ro as the other.
You can do the math with 2 ohms from a true zero ohms voltage source,
and another voltage source feeding 2.2 ohms, and tell me what the
mismatch does, but I'd say not much.

With a pair of solid state amps with very low Ro of say 100 milliohms each,
the effects of different amp gains could have serious load sharing problems.
But not when Ro is over 1 ohm.

With very low Ro SS amps one could happily power
a common load with 4 amps in parallel, but each with 1 ohm as a series
R in the output, and this would give the effect of having Rout = about 0.25
ohms.
Its not hard to have 1% accurately matched feedback networks in each
amp to ensure the voltage gain is the same.
Such close matching of voltage gain is far easier to attain with
modern metal film resistors than rely on device matching,
or the fact that we have current sources.
if the Ro of an amp = 0.1 ohms or less like most SS amps, the
1 ohm series R will convert the amp to a current source from the amp's
point of view, since it sees at least 1 ohm of load.
The voltage NFB ensures that the actual load value seen by each of multiple
amps
will stay close when the series 1 ohm "safety resistors" are in place
Such R can be used to sense current overloading.



Seems we still can't get something for nothing!

Well, be prepared to pay the price.

Youse will never ever hear a gal say
" jus lie down here luvvy; this won't costyer anyfink "

I think you overstate a problem.

Triodes are current sources for the reasons I gave above
in the top paragraph. One tube sees all the parallel Ra and RL
of the other tubes.
All the rowers in a boat don't make the same power,
some row harder than others; none make it more difficult
for any other to row towards that medal.
Unless one falls out of the boat and tangles her oars in a daze of over
exertion,
and this happened in one of the Oz boats with a girls crew in the last
Olympics.
One girl just passed out, ****ed right up, and
stuffed any chance of a medal. The other girls were not happy......

Anyway, next time you are in your workshop, plug a 6L6
and KT88 into each side of a PP circuit and wire them as triodes
and bias them both for 22 watts, and do a test in class A,
then parallel some tubes in triode in an SE circuit, and get back to us
with your findings.

Forget the theorizing, see what happens in reality land.

BTW, with 625 trioded EL34 in parallel, a single EL34
sees a load and Ra resistance = approximately 1.6 ohms.

Output power at 5.1 watts average each = 3,187 watts.

The 625 tubes are needed for a very nice OTL load match for SE class A.

In many OTL amps, paralel output triodes are very prone to
individually thermaling out since the load match is usually
so damn attrocious.

Patrick Turner.



Cheers, John Stewart

west wrote:

Patrick,
In your monster amp of 12 KT88s, have you ever thought of banking them? For
instance, flip a switch and have only 1 pair/side running, another 2
pairs/side, etc. I suppose you would do it with turning the B+ on & off. Not
always will you need 12 tubes running. This will save on tube life and will
not force you to kill a mosquito with a cannon. Thoughts?
west

If somebody wants me to sell them an amp with the option you mention,
then fine, I'll do it.

But then the load on the 6 active tubes is half what it really should be.
Still acceptable, but not as much class A.......
The 300 watters make 250 watts into 8 ohms,
and the 6550 tubes are run very cool at about 25 watts each
so about 140 watts of pure class A is possible without raising a sweat....

Plenty of power for low 80's dB sensitive speakers with an awkward impedance
and a huge room and loud levels.

When they sell you a Mazzeratti, they don't sell you a special spanner
to unbolt half the engine for driving around town to do the shopping.

Patrick Turner.




"Patrick Turner" wrote in message
...


Fabio Berutti wrote:

Interesting. I'm generally an opponent of paralleled tubes for
absolutely
no technical reason, just 'cause in my job I always faced a f%#!!ing
mess to
make two compressors or two pumps to work together in parallel. I was
surprised sometimes by the fact that it was easier to share load among 3
or
4 units than just 2, but now I got the "intuitive" point: it depends
from
the perturbation that the single unit causes to the whole. If 10
compressors are blowing into a single pipe and one fails, the 9 still
running do not even notice, but if it's one out of two...
In the same way (I suppose) goes tubes behavior: in a parallel of 4
units,
the differences in the Gm of any single one will affect in a minor way
the
working point of all other ones.
But when only 2 tubes are paralleled, "matched pairs" are mandatory;.

I have made a lot of amps with a quad of tubes and had no trouble
with the parallel pairs each side of the PP circuit, and in some the tubes
were
not matched.


Triode/pentode mode has the same meaning: what happens to the second
tube if
the first conducts more/less?

With PP hi-fi amps the tubes rarely work hard.
In most amps with a quad of tubes you can pull out one tube and the amp
still works OK for the next 20 years. The thd is a good bit higher, more
2H.


Adjustable bias will fix only a part of the problem (DC current) but
still
any single tube will operate in a different point of its curve and reach
saturation at different signal levels.

Its an insignificant problem.


My "less is more" approach, here, is not a good one (it's never useful
to be
too much devoted to a rule, without putting it in a context).

(Nevertheless, I still think that 2 tubes only are a bit of a trouble: I
remember I saw a paralleled ECC83 on a 'scope, 1K sine wave. One half
had a
higher gain. The output graph was a real "seno" (in Italian the sine
wave
and the feminine breast share the same word): a beautiful sine wave,
with a
clearly visible "nipple" on its top, where the "strongest" half tube got
in..)

I routinely parallel both halves of a 6SN7, 12AT7, like many other makers.

I have never ever seen nipples on a breast wave yet.

Every time I have ever paralleled triode tube halves the thd is reduced.
Its a good reason to parallel tubes.

The strongest share this Earth with its weakest, and we hope it is with
tolerance and compassion
that they work together harmoniously, and multiple triodes seem to show us
what
mankind and womenkind could so easily do, but won't.

Patrick Turner.





Ciao

Fabio

"Patrick Turner" ha scritto nel messaggio
...


John Stewart wrote:

John Stewart wrote:

Ian Iveson wrote:

"John Stewart" wrote

...Pentodes are current devices, so have no problem sharing the
load. However the
same tubes connected as triodes are voltage sources. Paralleled
voltage sources
do not share the load well at all. The result is higher than
expected distortion
at all levels....

I'm surprised. You are usually one for posting supporting
evidence,
John. I can see the idea, but do you have actual data on this?

cheers, Ian

It is here somewhere. I will dig it out & post in the next day or
two. JLS

I must apologize on this one Ian. I can't find those results anywhere
in
my file. I did that work sometime more than 40 years ago before going
into sales with HP. Perhaps I will have another look again when I get
time to do some tests since it is a rather interesting subject. Or
perhaps someone out there will take a shot at it!

However, in the meantime we can use a little network theory to
illustrate
what is happening. Imagine in a perfect world we have a pair of
identical
voltage sources connected to a common load, both thru identical
impedances. The voltage source polarities are connected so that +ve
is to
+ve & -ve is to -ve. The simplest form this would take would be a
pair of
identical batteries driving a resistive load thru identical
resistors.
The resultant current in the load would be the simple sum of the
currents
provided by the two batteries, those currents being equal to each
other.

I have a colleague in Sydney who built a pair of 80 watt SET amps
using 10 x 300B in parallel.
The load was about 400 ohms for the lot and
they all shared the load pretty darn well.
The ratio between any one tube's Ra of 800 ohms to the
rest of the parallel tubes' Ra which is 88 0hms and the 400 ohms RL
is 800 : 72, so each tube works as a current source
to the total resistance it is connected to, since one tube's Ra
is large compared to all the the external resistances.

In all the experiments I have made with triode connected tubes
connected in parallel I have found only very slight differences
between the signal cathode currents in the parallel tubes.




But as usual the world is far from perfect as are this pair of
batteries
& the resistors. One or the other battery will manage to sink or
source a
current thru the other battery while at the same time push some
current
thru the load. The two currents in the load are no longer identical &
never will be.

For the audio case circumstances get much worse.

Nope.

If we are lucky enough
to have say a pair of WE300B's, carefully matched on an AVO tester it
will help a lot, but there are still problems & lots of them.

Nope.

At how many
points can we match? It is unlikely we can get a match for all
conditions
of operation on the plate family of curves.

The variations don't make much difference...



Lets replace the batteries & resistors above with the WE300B's.

The WE300B's (And any other triode) are voltages sources as
ordinarily
connected & pretty good ones at that.

They are very weak voltage sources.
Ra of 800 ohms is not that much lower than the 4k one might load the
tube
with.
It seems like a lot, but it isn't.
If the Ra was 8 ohms, there might be a slight problem



If they are driven by a common
source into a resistive load will they each provide the same current
to
the load? It seems rather unlikely no matter how careful we are.

It is in practice very likely they will share the current fairly
evenly.

One could have 625 EL34 in triode all hooked up to 8 ohms
The average load seen by each EL34 would be 5,000 ohms, and a
reasonable
load
match for each tube.
Since no two tubes are exactly matched and they have different gm,
then although the anode voltage swing of 160vrms would be
the same for each tube, the current in each tube would differ
according to
each tubes's
gm, and since u would be so close for all it is of negligible concern,
since
u
depends on the physical dimensions between electrodes.
Therefore it could be shown that each tube would itself experience a
different load to its neighbour,
one might think its load is 5,300 ohms, the next 4,900, next 5,100 and
so
on.

Not all horses pull the carriage with the same force, but they all do
what
suits them.



And the loadline presented to them will not be a nice straight line
as
you often see in a text book.

The load line in PP class A amps for each tube is actually a bent
line,
and very bent in PP class AB amps.
while in class A, each tube of a pair sees say 5k, to make up the
total
10k a-a RL.
But when one has cut off, the other tube sees just 2.5k.
Changes of gain during the cycle cause distortion.

That is the simplest possible condition &
seldom happens in real life.

The SE amp will have the straightest load line.

The loadline if driven by a single frequency
might be at best a straight line but more than likely elliptical.
That is
because the load is not a simple resistor but a loudspeaker. And
there is
an OPT on the way to the loudspeaker to further confuse things.

The same elliptical load line is shared between all parallel tubes.

The
WE300B's have some non-linearites & so do the OPT's, all generating
some
distortion of their own.

And the individual distortion currents sum to make but one average
distortion
current.

In PP amps with a quad or more OP tubes in parallel we have the
opportunity
to swap tubes around so the total signal currents of each 1/2 of the
PP output circuit are closely matched, thus cancelling the effect of
the
mismatch between tubes, and getting almost perfect cancellation
of 2H between the two PP halves of the circuit.
Within each 1/2 I have found the currents near equal between 2 or 3 or
6
output tubes
in parallel with the same phase of signal.



Things get really crazy with the program material we would like to
use
whether it be Bach or The Rolling Stones (I prefer the Stones). That
is
because we now have a complex signal of many frequencies running into
a
complex load (the loudspeaker).

Signal complexity makes no difference to the sharing ability of tube
behaviour.
A signal is a signal is a signal. Just voltage current changes.



Here again we have a condition where one or the other of the WE300B's
can
sink or source some current into the other while at the same time
driving
part of the load current! That means that each of the tubes are at
times
driving the other.

I have never seen this. One could say that if you pull one tube out,
the
other 5
have to work harder.
But the direction of current flow is never into one tube when its out
of
all the rest.
What simply happens is that one tube experiences a different load to
the
other.
You may say that if the difference in current is 2 mA between a pair
of parallel tubes then the net 2 mA is flowing from one tube to the
other,
some how creating power in a tube to reduce efficiency.
Its over stating what's going on.



Some of the power developed by each tube is simply
dissipated in the other & never does show up in your loudspeaker
load.

You mean one tube dissipates more heat than the other.
If a weak gm tube is paired with a much stronger gm tube then
of course the weaker one might see 7k and the stronger one 4k,
and the actual load is 2.54 k.
The one that acts like it sees 4k will have more dissipation
than the one that has the 7k, and the efficiencies of each will vary
as
you
calculate.
In practice it isn't a big deal.

One could make the contribution of power near equal if we had
current FB as one is forced to have in SS amps with multiple
bjt or mosfets in parallel. Especially bjts, because if one bjt
conducts
more
than the others it heats up
and conducts even more, hence the use of the 0.33 ohm emitter
resistors
seen in many SS amps with multiple devices.
Mosfets and tubes back off a bit when they get hotter, since they have
negative temp
characteristics, ie, hotter means less gm.

Tubes when paralleled can tend to stay matched, because any that work
harder
than the rest
wear out faster, and they gm reduces......



So if you parallel a pair of triodes you won't get double the power
at a
particular D. But with great care you might come close.

I routinely double the power with parallel tubes.
I sometimes quadruple it, sixtuple it, or twelvopple it.

Colleagues have the same experiences, even with triodes.

Even when no special efforts are made to match, multiples of power
are produced.



Pentodes are current sources & can work well paralleled. But they
still
have all the objectionable distortion characteristics typical of
their
family. A way out of that is loop NFB & we see that often used to
lower D
& get a reasonable DF.

You could also run the triodes with cathode resistors to move them
closer
to CC operation. But that lowers the DF.

50 ohms in the cathode circuit of an EL34 in triode
makes the Ra rise from 1,250 ohms to about 1,750 ohms,
and it wastes 0.05 watts when there is 160 vrms across the 5k anode
load,
(
5.12 watts of output power )
So one could have 200 ohms in the '34 cathode circuit, Ra would rise
to
3,250 ohms or about equal to a 50% UL connected '34.
With 160vrms across the 5k load, the same 5.12 watts is produced, but
0.2 watts is wasted in the cathode R, and this is a very small
wastage.
The applied current FB will reduce the gain and distortion like all
NFB
does,

but only marginally, but it won't do any good in a PP circuit where
individual unbypassed cathode resistors in a class AB circuit are not
too
good a thing
where one wants to have a fixed bias operation.
Quite OK in any SEW circuit though.

To reduce the Ro of the amp after increasing it with unbypassed Rk,
external loops of voltage NFB must be used, and to get a circuit
with 3 times the usual Ra back down to the original Ra
might take at least 8 dB of global or other series voltage NFB, or
perhaps
a
little
positive current FB which makes distortion worse after the betterment
from
the negative series current FB in a given output triode cathode
circuit.

Its all rather bloomin messy.


When running tubes in parallel one usually needs to included swamping
resistors on all grid leads & sometimes plate leads as well in order
to
suppress parasitic RF oscillations.

In my 300 watt amps with a dozen 6550 all I needed was the rather low
value
2.2k g1 resistor stoppers.
There are 220 ohms in series with the screens and their supply.
Most important are the zobel neworks across each 1/2 of the primary
on the OPT.
This is placed half way along the line up of six tubes on each side of
the
PP
circuit.
Not a sign of any RF instability, and I have an OPT with 270 khz of
open loop bandwidth.

Should a tube decide to go beserko and go into thermal
runaway, then sometimes it will also oscillate at RF,
as well as conduct a huge cathode current.
The active protection I have in all my amps detects
the excess DC always associated with a hot tube.
If one or more tubes sustains twice the idle current for more than
4 seconds, the amp shuts down.

I once repaired a PA amp twice which had an octet of EL34, with
one single fixed bias voltage, Ea = 900v, Eg2 = 450v.
After the second repair I insisted I inspect his two
bins with 6 x 12" speakers in each.
He maintained he had 8 ohm speakers.
Sure, he was dead right, but all were wired in parallel,
and the most he got was 50 watts in the speakers, 25 watts in the
cables,
and a shirtload of heat in the tubes, which had been pooping out
regularly
for 15 years.
Once the amp was re-loaded correctly, biased with 8 pots,
had zobels connected right, and some stability added with the
12 dB of GNFB I added as well as a regulated screen supply, he never
phoned
me again,
and the sound was truly glorious.



Instead of using unbypassed cathode resistors and triode connection in
my
300
watt amps
to only marginally assist load sharing, I have 20% of the primary
signal voltage applied to the cathodes as series voltage NFB from the
OPT.
The screens are taken to +330v, Ea = 470v, and individual cathode bias
is
applied
to each tube to regulate the bias.
Some fixed bias is used to keep the value of Rk to a minimum
and keep the DC Pd in the RK low, yet regulate bias well enough.
I also use a couple of power bjts to dynamically shunt the
excess 2H currents in the cathode circuits thus maintaining thre
voltage
across the total of 12,000 uF in each bypass cap on each half of the
PP
circuit.
So with dynamic stabilizing of the cathode bias the tubes tend to
share
their
load
pretty darn well.
They are all subject to the same cathode series voltage FB, not
current
FB.
But since they are tetrode connected but with voltage FB there seems
to be
no

tendency for the tubes to differ much in their current donation to the
load
power.

With many rather poorly conceived commercially made amps such as the
Baird
with a six pack of EL84 to give the rather large 60 watts of power per
channel,
there is but one sole value of fixed bias.
Baird didn't want to load down customers with the responsibility to
adjust
12 bias pots, something many users would invariably get wrong.
Therefore as all these little tube age their bias drift is large, and
some
would
fade faster than others, and conduct differently, but the drift
wouldn't
be
any worse than
having just a pair of output 6550 with the same grid bias applied.
3 x EL84 can do about what one 6550 can do.
10 x EL84 each with its own cathode bias circuit would be more
effective
for 50 watts of class AB with a larger amount of class A,
and since the power used won't be high, the tubes would last and stay
well enough matched to provide the power wanted.




The circuit I did my tests on was a simple stereo amp using a pair of
cathode biased 6AQ5's into Hammond 125D OPT's in each channel. Not
the
best amp at all but it did illustrate the principle well. Since each
amp
of the pair had loop NFB it was in a sense a voltage source. If my
memory
is still in gear the DF was about 3 so in that respect very much like
an
SET circuit. But there the similarity ends.

Paralleling amps which have two separate FB loops which of course that
will
means that the Or of one amp is say 2 ohms, and the other might be
2.2 ohms, due to vagaries of tubes and matching, and if the common
load is
8 ohms then one amp sees nearly the same Ro as the other.
You can do the math with 2 ohms from a true zero ohms voltage source,
and another voltage source feeding 2.2 ohms, and tell me what the
mismatch does, but I'd say not much.

With a pair of solid state amps with very low Ro of say 100 milliohms
each,
the effects of different amp gains could have serious load sharing
problems.
But not when Ro is over 1 ohm.

With very low Ro SS amps one could happily power
a common load with 4 amps in parallel, but each with 1 ohm as a series
R in the output, and this would give the effect of having Rout = about
0.25
ohms.
Its not hard to have 1% accurately matched feedback networks in each
amp to ensure the voltage gain is the same.
Such close matching of voltage gain is far easier to attain with
modern metal film resistors than rely on device matching,
or the fact that we have current sources.
if the Ro of an amp = 0.1 ohms or less like most SS amps, the
1 ohm series R will convert the amp to a current source from the amp's
point of view, since it sees at least 1 ohm of load.
The voltage NFB ensures that the actual load value seen by each of
multiple
amps
will stay close when the series 1 ohm "safety resistors" are in place
Such R can be used to sense current overloading.



Seems we still can't get something for nothing!

Well, be prepared to pay the price.

Youse will never ever hear a gal say
" jus lie down here luvvy; this won't costyer anyfink "


I think you overstate a problem.

Triodes are current sources for the reasons I gave above
in the top paragraph. One tube sees all the parallel Ra and RL
of the other tubes.
All the rowers in a boat don't make the same power,
some row harder than others; none make it more difficult
for any other to row towards that medal.
Unless one falls out of the boat and tangles her oars in a daze of
over
exertion,
and this happened in one of the Oz boats with a girls crew in the last
Olympics.
One girl just passed out, ****ed right up, and
stuffed any chance of a medal. The other girls were not happy......

Anyway, next time you are in your workshop, plug a 6L6
and KT88 into each side of a PP circuit and wire them as triodes
and bias them both for 22 watts, and do a test in class A,
then parallel some tubes in triode in an SE circuit, and get back to
us
with your findings.

Forget the theorizing, see what happens in reality land.

BTW, with 625 trioded EL34 in parallel, a single EL34
sees a load and Ra resistance = approximately 1.6 ohms.

Output power at 5.1 watts average each = 3,187 watts.

The 625 tubes are needed for a very nice OTL load match for SE class
A.

In many OTL amps, paralel output triodes are very prone to
individually thermaling out since the load match is usually
so damn attrocious.

Patrick Turner.



Cheers, John Stewart

John Stewart wrote:

Patrick Turner wrote:

Thus Spake The Oracle (as follows) again leading us into a land covered by a 'Fog
of Bull****'. Patrick T will now repeal Kirchhoff's Laws. Later this year we will
have a new Law of Gravitation as well!!!! JLS

But you have not said why I lead us all unto ungodly foggy doom.
I just offer an oppinion, take it or leave it.

So maybe my own described experience of what is possible and real with parallel
tubes will be the guide to folks in the group wanting a bit higher power ceiling,
more class A,
less thd, better sound without resorting to mainly class B and shirtfulls of NFB.

The Law of Levitation is still intact, and has not been disproved yet
to the believers, allied to the Flat Earth Society.

And it is damn flat when you drive out into the middle of OZ.
There's a dip after Western Australia, and its full of sea water.
Bloody Not Always Flat.
There's some theory about it in some book.........

Patrick Turner.





John Stewart wrote:

John Stewart wrote:

Ian Iveson wrote:

"John Stewart" wrote

...Pentodes are current devices, so have no problem sharing the
load. However the
same tubes connected as triodes are voltage sources. Paralleled
voltage sources
do not share the load well at all. The result is higher than
expected distortion
at all levels....

I'm surprised. You are usually one for posting supporting evidence,
John. I can see the idea, but do you have actual data on this?

cheers, Ian

It is here somewhere. I will dig it out & post in the next day or
two. JLS

I must apologize on this one Ian. I can't find those results anywhere in
my file. I did that work sometime more than 40 years ago before going
into sales with HP. Perhaps I will have another look again when I get
time to do some tests since it is a rather interesting subject. Or
perhaps someone out there will take a shot at it!

However, in the meantime we can use a little network theory to illustrate
what is happening. Imagine in a perfect world we have a pair of identical
voltage sources connected to a common load, both thru identical
impedances. The voltage source polarities are connected so that +ve is to
+ve & -ve is to -ve. The simplest form this would take would be a pair of
identical batteries driving a resistive load thru identical resistors.
The resultant current in the load would be the simple sum of the currents
provided by the two batteries, those currents being equal to each other.

I have a colleague in Sydney who built a pair of 80 watt SET amps
using 10 x 300B in parallel.
The load was about 400 ohms for the lot and
they all shared the load pretty darn well.
The ratio between any one tube's Ra of 800 ohms to the
rest of the parallel tubes' Ra which is 88 0hms and the 400 ohms RL
is 800 : 72, so each tube works as a current source
to the total resistance it is connected to, since one tube's Ra
is large compared to all the the external resistances.

In all the experiments I have made with triode connected tubes
connected in parallel I have found only very slight differences
between the signal cathode currents in the parallel tubes.



But as usual the world is far from perfect as are this pair of batteries
& the resistors. One or the other battery will manage to sink or source a
current thru the other battery while at the same time push some current
thru the load. The two currents in the load are no longer identical &
never will be.

For the audio case circumstances get much worse.

Nope.

If we are lucky enough
to have say a pair of WE300B's, carefully matched on an AVO tester it
will help a lot, but there are still problems & lots of them.

Nope.

At how many
points can we match? It is unlikely we can get a match for all conditions
of operation on the plate family of curves.

The variations don't make much difference...



Lets replace the batteries & resistors above with the WE300B's.

The WE300B's (And any other triode) are voltages sources as ordinarily
connected & pretty good ones at that.

They are very weak voltage sources.
Ra of 800 ohms is not that much lower than the 4k one might load the tube
with.
It seems like a lot, but it isn't.
If the Ra was 8 ohms, there might be a slight problem

If they are driven by a common
source into a resistive load will they each provide the same current to
the load? It seems rather unlikely no matter how careful we are.

It is in practice very likely they will share the current fairly evenly.

One could have 625 EL34 in triode all hooked up to 8 ohms
The average load seen by each EL34 would be 5,000 ohms, and a reasonable load
match for each tube.
Since no two tubes are exactly matched and they have different gm,
then although the anode voltage swing of 160vrms would be
the same for each tube, the current in each tube would differ according to
each tubes's
gm, and since u would be so close for all it is of negligible concern, since
u
depends on the physical dimensions between electrodes.
Therefore it could be shown that each tube would itself experience a
different load to its neighbour,
one might think its load is 5,300 ohms, the next 4,900, next 5,100 and so on.

Not all horses pull the carriage with the same force, but they all do what
suits them.



And the loadline presented to them will not be a nice straight line as
you often see in a text book.

The load line in PP class A amps for each tube is actually a bent line,
and very bent in PP class AB amps.
while in class A, each tube of a pair sees say 5k, to make up the total
10k a-a RL.
But when one has cut off, the other tube sees just 2.5k.
Changes of gain during the cycle cause distortion.

That is the simplest possible condition &
seldom happens in real life.

The SE amp will have the straightest load line.

The loadline if driven by a single frequency
might be at best a straight line but more than likely elliptical. That is
because the load is not a simple resistor but a loudspeaker. And there is
an OPT on the way to the loudspeaker to further confuse things.

The same elliptical load line is shared between all parallel tubes.

The
WE300B's have some non-linearites & so do the OPT's, all generating some
distortion of their own.

And the individual distortion currents sum to make but one average distortion
current.

In PP amps with a quad or more OP tubes in parallel we have the opportunity
to swap tubes around so the total signal currents of each 1/2 of the
PP output circuit are closely matched, thus cancelling the effect of the
mismatch between tubes, and getting almost perfect cancellation
of 2H between the two PP halves of the circuit.
Within each 1/2 I have found the currents near equal between 2 or 3 or 6
output tubes
in parallel with the same phase of signal.



Things get really crazy with the program material we would like to use
whether it be Bach or The Rolling Stones (I prefer the Stones). That is
because we now have a complex signal of many frequencies running into a
complex load (the loudspeaker).

Signal complexity makes no difference to the sharing ability of tube
behaviour.
A signal is a signal is a signal. Just voltage current changes.



Here again we have a condition where one or the other of the WE300B's can
sink or source some current into the other while at the same time driving
part of the load current! That means that each of the tubes are at times
driving the other.

I have never seen this. One could say that if you pull one tube out, the
other 5
have to work harder.
But the direction of current flow is never into one tube when its out of
all the rest.
What simply happens is that one tube experiences a different load to the
other.
You may say that if the difference in current is 2 mA between a pair
of parallel tubes then the net 2 mA is flowing from one tube to the other,
some how creating power in a tube to reduce efficiency.
Its over stating what's going on.

Some of the power developed by each tube is simply
dissipated in the other & never does show up in your loudspeaker load.

You mean one tube dissipates more heat than the other.
If a weak gm tube is paired with a much stronger gm tube then
of course the weaker one might see 7k and the stronger one 4k,
and the actual load is 2.54 k.
The one that acts like it sees 4k will have more dissipation
than the one that has the 7k, and the efficiencies of each will vary as you
calculate.
In practice it isn't a big deal.

One could make the contribution of power near equal if we had
current FB as one is forced to have in SS amps with multiple
bjt or mosfets in parallel. Especially bjts, because if one bjt conducts more
than the others it heats up
and conducts even more, hence the use of the 0.33 ohm emitter resistors
seen in many SS amps with multiple devices.
Mosfets and tubes back off a bit when they get hotter, since they have
negative temp
characteristics, ie, hotter means less gm.

Tubes when paralleled can tend to stay matched, because any that work harder
than the rest
wear out faster, and they gm reduces......



So if you parallel a pair of triodes you won't get double the power at a
particular D. But with great care you might come close.

I routinely double the power with parallel tubes.
I sometimes quadruple it, sixtuple it, or twelvopple it.

Colleagues have the same experiences, even with triodes.

Even when no special efforts are made to match, multiples of power
are produced.



Pentodes are current sources & can work well paralleled. But they still
have all the objectionable distortion characteristics typical of their
family. A way out of that is loop NFB & we see that often used to lower D
& get a reasonable DF.

You could also run the triodes with cathode resistors to move them closer
to CC operation. But that lowers the DF.

50 ohms in the cathode circuit of an EL34 in triode
makes the Ra rise from 1,250 ohms to about 1,750 ohms,
and it wastes 0.05 watts when there is 160 vrms across the 5k anode load, (
5.12 watts of output power )
So one could have 200 ohms in the '34 cathode circuit, Ra would rise to
3,250 ohms or about equal to a 50% UL connected '34.
With 160vrms across the 5k load, the same 5.12 watts is produced, but
0.2 watts is wasted in the cathode R, and this is a very small wastage.
The applied current FB will reduce the gain and distortion like all NFB does,

but only marginally, but it won't do any good in a PP circuit where
individual unbypassed cathode resistors in a class AB circuit are not too
good a thing
where one wants to have a fixed bias operation.
Quite OK in any SEW circuit though.

To reduce the Ro of the amp after increasing it with unbypassed Rk,
external loops of voltage NFB must be used, and to get a circuit
with 3 times the usual Ra back down to the original Ra
might take at least 8 dB of global or other series voltage NFB, or perhaps a
little
positive current FB which makes distortion worse after the betterment from
the negative series current FB in a given output triode cathode circuit.

Its all rather bloomin messy.


When running tubes in parallel one usually needs to included swamping
resistors on all grid leads & sometimes plate leads as well in order to
suppress parasitic RF oscillations.

In my 300 watt amps with a dozen 6550 all I needed was the rather low value
2.2k g1 resistor stoppers.
There are 220 ohms in series with the screens and their supply.
Most important are the zobel neworks across each 1/2 of the primary
on the OPT.
This is placed half way along the line up of six tubes on each side of the PP
circuit.
Not a sign of any RF instability, and I have an OPT with 270 khz of
open loop bandwidth.

Should a tube decide to go beserko and go into thermal
runaway, then sometimes it will also oscillate at RF,
as well as conduct a huge cathode current.
The active protection I have in all my amps detects
the excess DC always associated with a hot tube.
If one or more tubes sustains twice the idle current for more than
4 seconds, the amp shuts down.

I once repaired a PA amp twice which had an octet of EL34, with
one single fixed bias voltage, Ea = 900v, Eg2 = 450v.
After the second repair I insisted I inspect his two
bins with 6 x 12" speakers in each.
He maintained he had 8 ohm speakers.
Sure, he was dead right, but all were wired in parallel,
and the most he got was 50 watts in the speakers, 25 watts in the cables,
and a shirtload of heat in the tubes, which had been pooping out regularly
for 15 years.
Once the amp was re-loaded correctly, biased with 8 pots,
had zobels connected right, and some stability added with the
12 dB of GNFB I added as well as a regulated screen supply, he never phoned
me again,
and the sound was truly glorious.

Instead of using unbypassed cathode resistors and triode connection in my 300
watt amps
to only marginally assist load sharing, I have 20% of the primary
signal voltage applied to the cathodes as series voltage NFB from the OPT.
The screens are taken to +330v, Ea = 470v, and individual cathode bias is
applied
to each tube to regulate the bias.
Some fixed bias is used to keep the value of Rk to a minimum
and keep the DC Pd in the RK low, yet regulate bias well enough.
I also use a couple of power bjts to dynamically shunt the
excess 2H currents in the cathode circuits thus maintaining thre voltage
across the total of 12,000 uF in each bypass cap on each half of the PP
circuit.
So with dynamic stabilizing of the cathode bias the tubes tend to share their
load
pretty darn well.
They are all subject to the same cathode series voltage FB, not current FB.
But since they are tetrode connected but with voltage FB there seems to be no

tendency for the tubes to differ much in their current donation to the load
power.

With many rather poorly conceived commercially made amps such as the Baird
with a six pack of EL84 to give the rather large 60 watts of power per
channel,
there is but one sole value of fixed bias.
Baird didn't want to load down customers with the responsibility to adjust
12 bias pots, something many users would invariably get wrong.
Therefore as all these little tube age their bias drift is large, and some
would
fade faster than others, and conduct differently, but the drift wouldn't be
any worse than
having just a pair of output 6550 with the same grid bias applied.
3 x EL84 can do about what one 6550 can do.
10 x EL84 each with its own cathode bias circuit would be more effective
for 50 watts of class AB with a larger amount of class A,
and since the power used won't be high, the tubes would last and stay
well enough matched to provide the power wanted.


The circuit I did my tests on was a simple stereo amp using a pair of
cathode biased 6AQ5's into Hammond 125D OPT's in each channel. Not the
best amp at all but it did illustrate the principle well. Since each amp
of the pair had loop NFB it was in a sense a voltage source. If my memory
is still in gear the DF was about 3 so in that respect very much like an
SET circuit. But there the similarity ends.

Paralleling amps which have two separate FB loops which of course that will
means that the Or of one amp is say 2 ohms, and the other might be
2.2 ohms, due to vagaries of tubes and matching, and if the common load is
8 ohms then one amp sees nearly the same Ro as the other.
You can do the math with 2 ohms from a true zero ohms voltage source,
and another voltage source feeding 2.2 ohms, and tell me what the
mismatch does, but I'd say not much.

With a pair of solid state amps with very low Ro of say 100 milliohms each,
the effects of different amp gains could have serious load sharing problems.
But not when Ro is over 1 ohm.

With very low Ro SS amps one could happily power
a common load with 4 amps in parallel, but each with 1 ohm as a series
R in the output, and this would give the effect of having Rout = about 0.25
ohms.
Its not hard to have 1% accurately matched feedback networks in each
amp to ensure the voltage gain is the same.
Such close matching of voltage gain is far easier to attain with
modern metal film resistors than rely on device matching,
or the fact that we have current sources.
if the Ro of an amp = 0.1 ohms or less like most SS amps, the
1 ohm series R will convert the amp to a current source from the amp's
point of view, since it sees at least 1 ohm of load.
The voltage NFB ensures that the actual load value seen by each of multiple
amps
will stay close when the series 1 ohm "safety resistors" are in place
Such R can be used to sense current overloading.



Seems we still can't get something for nothing!

Well, be prepared to pay the price.

Youse will never ever hear a gal say
" jus lie down here luvvy; this won't costyer anyfink "

I think you overstate a problem.

Triodes are current sources for the reasons I gave above
in the top paragraph. One tube sees all the parallel Ra and RL
of the other tubes.
All the rowers in a boat don't make the same power,
some row harder than others; none make it more difficult
for any other to row towards that medal.
Unless one falls out of the boat and tangles her oars in a daze of over
exertion,
and this happened in one of the Oz boats with a girls crew in the last
Olympics.
One girl just passed out, ****ed right up, and
stuffed any chance of a medal. The other girls were not happy......

Anyway, next time you are in your workshop, plug a 6L6
and KT88 into each side of a PP circuit and wire them as triodes
and bias them both for 22 watts, and do a test in class A,
then parallel some tubes in triode in an SE circuit, and get back to us
with your findings.

Forget the theorizing, see what happens in reality land.

BTW, with 625 trioded EL34 in parallel, a single EL34
sees a load and Ra resistance = approximately 1.6 ohms.

Output power at 5.1 watts average each = 3,187 watts.

The 625 tubes are needed for a very nice OTL load match for SE class A.

In many OTL amps, paralel output triodes are very prone to
individually thermaling out since the load match is usually
so damn attrocious.

Patrick Turner.



Cheers, John Stewart

"John Stewart" wrote

[below]

Hmm, er, well...

The voltage source model may not be the best in this case. Really
you can take your pick of model, considering triodes are normally
loaded with RL of the same order of magnitude as Ra, so they are
somewhere between a voltage and current source.

I can see where you are coming from, but not quite where you are
going. I suspect some may have difficulty with the principle of
superposition, and wonder how you get current going back through one
of the valves!

After much grappling with thoughts, I decided the maths of the
triode is far too complicated to illustrate the point. So I will
simulate and compare when I get time to figure out how to simulate
the difference between two examples of the same valve. Would you
happen to know which of the parameters varies? Most convenient would
be to alter the perveance, which would be like stretching or
squeezing the current axis of the anode characteristic curves.

Considering a resistance model, parallel resistors sum to A.B/(A+B),
of course. If it were legitimate to express the transfer function of
the pair by similarly combining that of each valve, then you may end
up with an average distortion, plus a higher order content arising
from the interaction between the two, since the product A.B would be
distortion of distortion, as if the valves were in series. That is a
pretty loose thought at the moment...but it is what you might
expect. It would be interesting to put some typical practical
figures on the effect.

Just a further note. If the valves were perfectly matched, then each
would see the other as an infinite resistance, ie a current source.
That is why it is awkward to choose a voltage source model. As soon
as they are not matched, each is seen by the other as an active
finite load, proportional to the degree of the mismatch. If you had
two separate but equal loads on the two triodes, and then shorted
the anodes together, you would change both the voltage and the
current for each.

Someone should measure this. It will presumably be particularly
problematic in the common practice of doubling input valves in high
gain preamps.

cheers, Ian

in message ...

I must apologize on this one Ian. I can't find those results
anywhere in
my file. I did that work sometime more than 40 years ago before
going
into sales with HP. Perhaps I will have another look again when I
get
time to do some tests since it is a rather interesting subject. Or
perhaps someone out there will take a shot at it!

However, in the meantime we can use a little network theory to
illustrate
what is happening. Imagine in a perfect world we have a pair of
identical
voltage sources connected to a common load, both thru identical
impedances. The voltage source polarities are connected so that
+ve is to
+ve & -ve is to -ve. The simplest form this would take would be a
pair of
identical batteries driving a resistive load thru identical
resistors.
The resultant current in the load would be the simple sum of the
currents
provided by the two batteries, those currents being equal to each
other.

But as usual the world is far from perfect as are this pair of
batteries
& the resistors. One or the other battery will manage to sink or
source a
current thru the other battery while at the same time push some
current
thru the load. The two currents in the load are no longer
identical &
never will be.

For the audio case circumstances get much worse. If we are lucky
enough
to have say a pair of WE300B's, carefully matched on an AVO tester
it
will help a lot, but there are still problems & lots of them. At
how many
points can we match? It is unlikely we can get a match for all
conditions
of operation on the plate family of curves.

Lets replace the batteries & resistors above with the WE300B's.

The WE300B's (And any other triode) are voltages sources as
ordinarily
connected & pretty good ones at that. If they are driven by a
common
source into a resistive load will they each provide the same
current to
the load? It seems rather unlikely no matter how careful we are.

And the loadline presented to them will not be a nice straight
line as
you often see in a text book. That is the simplest possible
condition &
seldom happens in real life. The loadline if driven by a single
frequency
might be at best a straight line but more than likely elliptical.
That is
because the load is not a simple resistor but a loudspeaker. And
there is
an OPT on the way to the loudspeaker to further confuse things.
The
WE300B's have some non-linearites & so do the OPT's, all
generating some
distortion of their own.

Things get really crazy with the program material we would like to
use
whether it be Bach or The Rolling Stones (I prefer the Stones).
That is
because we now have a complex signal of many frequencies running
into a
complex load (the loudspeaker).

Here again we have a condition where one or the other of the
WE300B's can
sink or source some current into the other while at the same time
driving
part of the load current! That means that each of the tubes are at
times
driving the other. Some of the power developed by each tube is
simply
dissipated in the other & never does show up in your loudspeaker
load.

So if you parallel a pair of triodes you won't get double the
power at a
particular D. But with great care you might come close.

Pentodes are current sources & can work well paralleled. But they
still
have all the objectionable distortion characteristics typical of
their
family. A way out of that is loop NFB & we see that often used to
lower D
& get a reasonable DF.

You could also run the triodes with cathode resistors to move them
closer
to CC operation. But that lowers the DF.

When running tubes in parallel one usually needs to included
swamping
resistors on all grid leads & sometimes plate leads as well in
order to
suppress parasitic RF oscillations.

The circuit I did my tests on was a simple stereo amp using a pair
of
cathode biased 6AQ5's into Hammond 125D OPT's in each channel. Not
the
best amp at all but it did illustrate the principle well. Since
each amp
of the pair had loop NFB it was in a sense a voltage source. If my
memory
is still in gear the DF was about 3 so in that respect very much
like an
SET circuit. But there the similarity ends.

Seems we still can't get something for nothing!

Cheers, John Stewart

Ian Iveson wrote:

"John Stewart" wrote

[below]

Hmm, er, well...

The voltage source model may not be the best in this case. Really
you can take your pick of model, considering triodes are normally
loaded with RL of the same order of magnitude as Ra, so they are
somewhere between a voltage and current source.

I can see where you are coming from, but not quite where you are
going. I suspect some may have difficulty with the principle of
superposition, and wonder how you get current going back through one
of the valves!

After much grappling with thoughts, I decided the maths of the
triode is far too complicated to illustrate the point. So I will
simulate and compare when I get time to figure out how to simulate
the difference between two examples of the same valve. Would you
happen to know which of the parameters varies? Most convenient would
be to alter the perveance, which would be like stretching or
squeezing the current axis of the anode characteristic curves.

Considering a resistance model, parallel resistors sum to A.B/(A+B),
of course. If it were legitimate to express the transfer function of
the pair by similarly combining that of each valve, then you may end
up with an average distortion, plus a higher order content arising
from the interaction between the two, since the product A.B would be
distortion of distortion, as if the valves were in series. That is a
pretty loose thought at the moment...but it is what you might
expect. It would be interesting to put some typical practical
figures on the effect.

Just a further note. If the valves were perfectly matched, then each
would see the other as an infinite resistance, ie a current source.
That is why it is awkward to choose a voltage source model. As soon
as they are not matched, each is seen by the other as an active
finite load, proportional to the degree of the mismatch. If you had
two separate but equal loads on the two triodes, and then shorted
the anodes together, you would change both the voltage and the
current for each.

Someone should measure this. It will presumably be particularly
problematic in the common practice of doubling input valves in high
gain preamps.

Indeed its well that you ask about two triodes loaded the same way,
same load and input signal.
Usually the gains will be slightly different as no two tubes are
perfectly matched.
When you short the anodes, one tube will do slightly more work than the
other.
so maybe one tube sees 5.200k and the other is 4.8146k,
based on observations of the anode swing voltage, same for both tubes,
and the anode current swing, different for each tube.
The load after two 5k loads are paralleled is 2.5k.

There is no dreadful sudden increase in tube dissipation in one of the
tubes
resulting in a 10% difference in gM.
Usually tube circuits have a large tolerance to slight if not
large load mismatches, and usually unmatched samples of tubes bought
together
are well within 10% of each other for Gm and Ra.


Paralleling triodes is OK.

Some parts of a single anode structure in a given tube get hotter than
other parts.
So its as if numerous slightly different sections of the tube are in
parallel.
Paralleling two or 10 tubes merely extends the plate, grid, screen and
cathode structure.
And if one tube cops out and goes beserko, pull it out and replace it,
not a drama at all
and you avoid having to replace a large expensive tube that would do the
same job as the multiple
small ones.

Patrick Turner.







cheers, Ian

in message ...

I must apologize on this one Ian. I can't find those results
anywhere in
my file. I did that work sometime more than 40 years ago before
going
into sales with HP. Perhaps I will have another look again when I
get
time to do some tests since it is a rather interesting subject. Or
perhaps someone out there will take a shot at it!

However, in the meantime we can use a little network theory to
illustrate
what is happening. Imagine in a perfect world we have a pair of
identical
voltage sources connected to a common load, both thru identical
impedances. The voltage source polarities are connected so that
+ve is to
+ve & -ve is to -ve. The simplest form this would take would be a
pair of
identical batteries driving a resistive load thru identical
resistors.
The resultant current in the load would be the simple sum of the
currents
provided by the two batteries, those currents being equal to each
other.

But as usual the world is far from perfect as are this pair of
batteries
& the resistors. One or the other battery will manage to sink or
source a
current thru the other battery while at the same time push some
current
thru the load. The two currents in the load are no longer
identical &
never will be.

For the audio case circumstances get much worse. If we are lucky
enough
to have say a pair of WE300B's, carefully matched on an AVO tester
it
will help a lot, but there are still problems & lots of them. At
how many
points can we match? It is unlikely we can get a match for all
conditions
of operation on the plate family of curves.

Lets replace the batteries & resistors above with the WE300B's.

The WE300B's (And any other triode) are voltages sources as
ordinarily
connected & pretty good ones at that. If they are driven by a
common
source into a resistive load will they each provide the same
current to
the load? It seems rather unlikely no matter how careful we are.

And the loadline presented to them will not be a nice straight
line as
you often see in a text book. That is the simplest possible
condition &
seldom happens in real life. The loadline if driven by a single
frequency
might be at best a straight line but more than likely elliptical.
That is
because the load is not a simple resistor but a loudspeaker. And
there is
an OPT on the way to the loudspeaker to further confuse things.
The
WE300B's have some non-linearites & so do the OPT's, all
generating some
distortion of their own.

Things get really crazy with the program material we would like to
use
whether it be Bach or The Rolling Stones (I prefer the Stones).
That is
because we now have a complex signal of many frequencies running
into a
complex load (the loudspeaker).

Here again we have a condition where one or the other of the
WE300B's can
sink or source some current into the other while at the same time
driving
part of the load current! That means that each of the tubes are at
times
driving the other. Some of the power developed by each tube is
simply
dissipated in the other & never does show up in your loudspeaker
load.

So if you parallel a pair of triodes you won't get double the
power at a
particular D. But with great care you might come close.

Pentodes are current sources & can work well paralleled. But they
still
have all the objectionable distortion characteristics typical of
their
family. A way out of that is loop NFB & we see that often used to
lower D
& get a reasonable DF.

You could also run the triodes with cathode resistors to move them
closer
to CC operation. But that lowers the DF.

When running tubes in parallel one usually needs to included
swamping
resistors on all grid leads & sometimes plate leads as well in
order to
suppress parasitic RF oscillations.

The circuit I did my tests on was a simple stereo amp using a pair
of
cathode biased 6AQ5's into Hammond 125D OPT's in each channel. Not
the
best amp at all but it did illustrate the principle well. Since
each amp
of the pair had loop NFB it was in a sense a voltage source. If my
memory
is still in gear the DF was about 3 so in that respect very much
like an
SET circuit. But there the similarity ends.

Seems we still can't get something for nothing!

Cheers, John Stewart

"John Stewart" wrote

[below]

Hmm, er, well...

The voltage source model may not be the best in this case. Really
you can take your pick of model, considering triodes are normally
loaded with RL of the same order of magnitude as Ra, so they are
somewhere between a voltage and current source.

I can see where you are coming from, but not quite where you are
going. I suspect some may have difficulty with the principle of
superposition, and wonder how you get current going back through one
of the valves!

After much grappling with thoughts, I decided the maths of the
triode is far too complicated to illustrate the point. So I will
simulate and compare when I get time to figure out how to simulate
the difference between two examples of the same valve. Would you
happen to know which of the parameters varies? Most convenient would
be to alter the perveance, which would be like stretching or
squeezing the current axis of the anode characteristic curves.

Considering a resistance model, parallel resistors sum to A.B/(A+B),
of course. If it were legitimate to express the transfer function of
the pair by similarly combining that of each valve, then you may end
up with an average distortion, plus a higher order content arising
from the interaction between the two, since the product A.B would be
distortion of distortion, as if the valves were in series. That is a
pretty loose thought at the moment...but it is what you might
expect. It would be interesting to put some typical practical
figures on the effect.

Just a further note. If the valves were perfectly matched, then each
would see the other as an infinite resistance, ie a current source.
That is why it is awkward to choose a voltage source model. As soon
as they are not matched, each is seen by the other as an active
finite load, proportional to the degree of the mismatch. If you had
two separate but equal loads on the two triodes, and then shorted
the anodes together, you would change both the voltage and the
current for each.

Someone should measure this. It will presumably be particularly
problematic in the common practice of doubling input valves in high
gain preamps.

cheers, Ian

in message ...

I must apologize on this one Ian. I can't find those results
anywhere in
my file. I did that work sometime more than 40 years ago before
going
into sales with HP. Perhaps I will have another look again when I
get
time to do some tests since it is a rather interesting subject. Or
perhaps someone out there will take a shot at it!

However, in the meantime we can use a little network theory to
illustrate
what is happening. Imagine in a perfect world we have a pair of
identical
voltage sources connected to a common load, both thru identical
impedances. The voltage source polarities are connected so that
+ve is to
+ve & -ve is to -ve. The simplest form this would take would be a
pair of
identical batteries driving a resistive load thru identical
resistors.
The resultant current in the load would be the simple sum of the
currents
provided by the two batteries, those currents being equal to each
other.

But as usual the world is far from perfect as are this pair of
batteries
& the resistors. One or the other battery will manage to sink or
source a
current thru the other battery while at the same time push some
current
thru the load. The two currents in the load are no longer
identical &
never will be.

For the audio case circumstances get much worse. If we are lucky
enough
to have say a pair of WE300B's, carefully matched on an AVO tester
it
will help a lot, but there are still problems & lots of them. At
how many
points can we match? It is unlikely we can get a match for all
conditions
of operation on the plate family of curves.

Lets replace the batteries & resistors above with the WE300B's.

The WE300B's (And any other triode) are voltages sources as
ordinarily
connected & pretty good ones at that. If they are driven by a
common
source into a resistive load will they each provide the same
current to
the load? It seems rather unlikely no matter how careful we are.

And the loadline presented to them will not be a nice straight
line as
you often see in a text book. That is the simplest possible
condition &
seldom happens in real life. The loadline if driven by a single
frequency
might be at best a straight line but more than likely elliptical.
That is
because the load is not a simple resistor but a loudspeaker. And
there is
an OPT on the way to the loudspeaker to further confuse things.
The
WE300B's have some non-linearites & so do the OPT's, all
generating some
distortion of their own.

Things get really crazy with the program material we would like to
use
whether it be Bach or The Rolling Stones (I prefer the Stones).
That is
because we now have a complex signal of many frequencies running
into a
complex load (the loudspeaker).

Here again we have a condition where one or the other of the
WE300B's can
sink or source some current into the other while at the same time
driving
part of the load current! That means that each of the tubes are at
times
driving the other. Some of the power developed by each tube is
simply
dissipated in the other & never does show up in your loudspeaker
load.

So if you parallel a pair of triodes you won't get double the
power at a
particular D. But with great care you might come close.

Pentodes are current sources & can work well paralleled. But they
still
have all the objectionable distortion characteristics typical of
their
family. A way out of that is loop NFB & we see that often used to
lower D
& get a reasonable DF.

You could also run the triodes with cathode resistors to move them
closer
to CC operation. But that lowers the DF.

When running tubes in parallel one usually needs to included
swamping
resistors on all grid leads & sometimes plate leads as well in
order to
suppress parasitic RF oscillations.

The circuit I did my tests on was a simple stereo amp using a pair
of
cathode biased 6AQ5's into Hammond 125D OPT's in each channel. Not
the
best amp at all but it did illustrate the principle well. Since
each amp
of the pair had loop NFB it was in a sense a voltage source. If my
memory
is still in gear the DF was about 3 so in that respect very much
like an
SET circuit. But there the similarity ends.

Seems we still can't get something for nothing!

Cheers, John Stewart

On Fri, 22 Apr 2005 01:26:53 GMT, "Ian Iveson"
wrote:

how to simulate
the difference between two examples of the same valve. Would you
happen to know which of the parameters varies? Most convenient would
be to alter the perveance, which would be like stretching or
squeezing the current axis of the anode characteristic curves.

Probably a good choice to describe aging in initially similar
valves, but how about mu? Grid alignment and pitch are pretty
fiddly as a production problem, and different plants make
wildly different assumptions. Just look at what passes for
12AX7's for example.

Anyway, not my first thought, which was along the line of:
why, with your excellent modeling intuition, not just go
back to first principles? Triodes are just transconductance
engines; like that. Piece of cake, not?


Considering a resistance model, parallel resistors sum to A.B/(A+B),
of course. If it were legitimate to express the transfer function of
the pair by similarly combining that of each valve, then you may end
up with an average distortion, plus a higher order content arising
from the interaction between the two, since the product A.B would be
distortion of distortion, as if the valves were in series. That is a
pretty loose thought at the moment...but it is what you might
expect.

I fear a slippery slope here, where the "separate" outputs of the
paralleled valves are to be considered as somehow "separate".
Dangerous ground and, sadly, well trod.


Just a further note. If the valves were perfectly matched, then each
would see the other as an infinite resistance, ie a current source.

Brilliant. But then, in the words of Conrad Weiss (Major Strasse) in
_Casablanca_ "I expected no less."

Chris Hornbeck

John Stewart wrote:

Ian Iveson wrote:

"John Stewart" wrote

[below]

Hmm, er, well...

The voltage source model may not be the best in this case. Really
you can take your pick of model, considering triodes are normally
loaded with RL of the same order of magnitude as Ra, so they are
somewhere between a voltage and current source.

What we see most often in the text books are two kinds of models. First
of all there is the small signal equivalent circuits using a linear
voltage or current generator connected series or parallel with a linear
resistor. The voltage model most often used with triodes while the
current model mostly with pentodes, but can be used either way freely.
They are valid only for small signals & don't work worth a damn in the
case of a power amplifier stage.

The equivalent pure voltage generator with an output voltage of
µ x Eg in series with a resistance representing the Ra works fine when
modelling
any type of tube. The model of a current gene of gm x Eg with an R strapped
across the output to
model Ra is also OK with any tube. OK with power tubes.
Usually we think of using small signal to avoid distortion affecting
gain considerations.
The current or voltage generator model for a tube is useful with power tubes
imho.
And all power tubes, not just a few.





All tubes are non-linear & to a first order include the 3/2's power law.

Unless they happen to be a triode operating as a pure voltage device with no
current change,
ie, with a CCS load.
Then the triode is remarkably linear compared to most other common devices.


But in fact that is only first order & to predict the results accurately
we would needed to go to a power series. So to get better results from
the models I referenced above either the voltage (current) generator or
it's associated resistor would need to be non-linear to account for the
development of distortion products when passing a large signal. The
small signal models do not do that at all.

Hmm, I'm having trouble seeing how all this relates to paralleling tubes....



I can see where you are coming from, but not quite where you are
going. I suspect some may have difficulty with the principle of
superposition, and wonder how you get current going back through one
of the valves!

You are certainly correct there with the business of Superposition.

Just as long as we don't lurch into Superstition....

But
it is a useful tool provided one knows what is going on. It does
seperate the signals from the biasing potentials for any kind of circuit
we may choose to deal with, whether it be SS or vacuum. But it is the
signal (AC) current going back thru the tubes, not the supply current.
You than have one non-linear device sourcing or sinking current from
another. If that don't make for some distortion I sure don't know what
does!

What's the net result of having dissimilar triodes driving the same load?
One just works hader than the other; OK the weaker does provoke
the stronger to pass more current, but the mismatch has to be large to cause
a bother
in the form of the stronger tube doing so much more work than its neighbour
that
its load is lower, so the thd is higher.

Most unmatched tubes are within 10%, and paralleling isn't a problem.





After much grappling with thoughts, I decided the maths of the
triode is far too complicated to illustrate the point. So I will
simulate and compare when I get time to figure out how to simulate
the difference between two examples of the same valve. Would you
happen to know which of the parameters varies? Most convenient would
be to alter the perveance, which would be like stretching or
squeezing the current axis of the anode characteristic curves.

Considering a resistance model, parallel resistors sum to A.B/(A+B),
of course. If it were legitimate to express the transfer function of
the pair by similarly combining that of each valve, then you may end
up with an average distortion, plus a higher order content arising
from the interaction between the two, since the product A.B would be
distortion of distortion, as if the valves were in series. That is a
pretty loose thought at the moment...but it is what you might
expect. It would be interesting to put some typical practical
figures on the effect.

Just a further note. If the valves were perfectly matched, then each
would see the other as an infinite resistance, ie a current source.

The small signal equivalent circuits when modeled as current sources
(typically pentode models, but could be just as easily a triode) always
include a resistor in parallel with the source generator. For a triode
that is usually a low resistance. So as a generator each tube no longer
sees the other as infinite resistance. In any case, that is only small
signal & does not account for even the 3/2's power law.

That is why it is awkward to choose a voltage source model. As soon
as they are not matched, each is seen by the other as an active
finite load, proportional to the degree of the mismatch. If you had
two separate but equal loads on the two triodes, and then shorted
the anodes together, you would change both the voltage and the
current for each.

Someone should measure this. It will presumably be particularly
problematic in the common practice of doubling input valves in high
gain preamps.

cheers, Ian



I'm not so sure that simulation IS the easy way out for the
determination of results. It depends too much on very accurate modeling,
something not always available. The results I measured many years ago
were made on a pair of supposedly identical amplifiers, both
individually & then paralleled. I was surprised at the results & that is
why I remember all that so well. But my only real claim earlier in this
thread was that if one paralleled 'n' triodes, don't expect to get 'n'
times the power at the same distortion level.

Probably the easy way out is just do the experiment, but with great care
& be objective. Don't cook the results. If I had some time now I would
try it again. But Spring is finally here & there is much to do outside.
We had a foot of snow here on the 2nd & 3rd of April. There is still
some in the bush. And the graphics are much better there than on this
display!

Its simple to create a model of a bunch of triodes as voltage generators
each one puting an output of µ x Eg at its output and powering a common load

via its own separate Ra resistance.

The simulator should tell you what's going on with such a simple easy thing
to set up.
But a human can churn the number by changing Ra a bit to see what effect it
has on
gain and thd.

Paralleling tubes isn't a flip flop situation where if one tube conducts
slightly more
signal load current then the weaker tube suddently quits and the stronger
burns up from over exertion.

In guitar amps using a six pack of tetrodes, regularly over driven heavily,
there will always be a tube which will
fail first, but this can be due to a number of reasons, not just because
things are paralleled.


Patrick Turner.



Cheers, John Stewart

It's been a while, but I do seem to recall finding that parallel power
tubes can give problems with stability. Seems you need may need to do
things like putting damping resistors in _series_ with the signal grids,
possibly in the plate circuits, stuff like that. I dunno, maybe it was all
that extra wire.

And keep in mind, doubling the power is only 3db...

/kenw

"west" wrote:

Dear Rodents,
Since my 35W P/P EL34s UL won't cut the cake anymore with my Infinity RS
IIIbs, I'm looking at paralleling output tubes 2 pair, 3 pair, or 4 pair.
I'm somewhat familiar with 6550s & EL34s. I read that there can be many
problems with this configuration. Are there any success stories out there?
BTW: does anyone have experience with OTLs such as ATma-Spheres? I need more
power but I don't want to loose that sweet sound of tubes. I really could
use your opinions and thank you very much in advance.
Cordially,
west


Ken Wallewein
K&M Systems Integration
Phone (403)274-7848
Fax (403)275-4535

www.kmsi.net

Patrick Turner wrote:

John Stewart wrote:

Ian Iveson wrote:

"John Stewart" wrote

[below]

Hmm, er, well...

The voltage source model may not be the best in this case. Really
you can take your pick of model, considering triodes are normally
loaded with RL of the same order of magnitude as Ra, so they are
somewhere between a voltage and current source.

What we see most often in the text books are two kinds of models. Fir=
st
of all there is the small signal equivalent circuits using a linear
voltage or current generator connected series or parallel with a line=
ar
resistor. The voltage model most often used with triodes while the
current model mostly with pentodes, but can be used either way freely=
=2E
They are valid only for small signals & don't work worth a damn in th=
e
case of a power amplifier stage.

The equivalent pure voltage generator with an output voltage of
=B5 x Eg in series with a resistance representing the Ra works fine whe=
n
modelling
any type of tube. The model of a current gene of gm x Eg with an R stra=
pped
across the output to
model Ra is also OK with any tube. OK with power tubes.
Usually we think of using small signal to avoid distortion affecting
gain considerations.
The current or voltage generator model for a tube is useful with power =
tubes
imho.
And all power tubes, not just a few.

True enough, but this thread among other things is a discussion of distor=
tion in
amplifiers where tubes are paralleled. What reason other than increased p=
ower
would anyone use paralleling? In the case of more power, any analysis usi=
ng
linear models of the generator & it's associated resistance fails at high=
er
levels. One needs to allow for the non-linearites in order to predict the=

distortions which result naturally.

All tubes are non-linear & to a first order include the 3/2's power l=
aw.

Unless they happen to be a triode operating as a pure voltage device wi=
th no
current change,
ie, with a CCS load.
Then the triode is remarkably linear compared to most other common devi=
ces.

Yes again, but a triode working into a constant current delivers no power=
to
it's load (the CC Source). In order to deliver power to the load we must
interpose a power amplifier & then we return to the problem of accounting=
for
the non-linearites. In a common MU-Follower, that is part of the function=
of the
top triode. So the top triode operates as a CF & sees a load of whatever
resistor we choose to go between the tubes to develop drive for the upper=
triode
of the MU-Follower. It is well able to do that & drive a following grid a=
t low
distortion. But then comes the power amp, singly or paralleled.

But in fact that is only first order & to predict the results accurat=
ely
we would needed to go to a power series. So to get better results fro=
m
the models I referenced above either the voltage (current) generator =
or
it's associated resistor would need to be non-linear to account for t=
he
development of distortion products when passing a large signal. The
small signal models do not do that at all.

Hmm, I'm having trouble seeing how all this relates to paralleling tube=
s....

Look harder. It little investigation will reveal all!

I can see where you are coming from, but not quite where you are
going. I suspect some may have difficulty with the principle of
superposition, and wonder how you get current going back through on=
e
of the valves!

You are certainly correct there with the business of Superposition.

Just as long as we don't lurch into Superstition....

But
it is a useful tool provided one knows what is going on. It does
seperate the signals from the biasing potentials for any kind of circ=
uit
we may choose to deal with, whether it be SS or vacuum. But it is the=

signal (AC) current going back thru the tubes, not the supply current=
=2E
You than have one non-linear device sourcing or sinking current from
another. If that don't make for some distortion I sure don't know wha=
t
does!

What's the net result of having dissimilar triodes driving the same loa=
d?
One just works hader than the other; OK the weaker does provoke
the stronger to pass more current, but the mismatch has to be large to =
cause
a bother
in the form of the stronger tube doing so much more work than its neigh=
bour
that
its load is lower, so the thd is higher.

Most unmatched tubes are within 10%, and paralleling isn't a problem.

Paralleling is no big deal with pentodes, a current source. We see that i=
n many
amplifiers, including yours. Triodes are quite a different matter, altho =
they
will work. I've tried to point out some warts which seem not obvious to o=
thers.
But you can be sure they are there.

After much grappling with thoughts, I decided the maths of the
triode is far too complicated to illustrate the point. So I will
simulate and compare when I get time to figure out how to simulate
the difference between two examples of the same valve. Would you
happen to know which of the parameters varies? Most convenient woul=
d
be to alter the perveance, which would be like stretching or
squeezing the current axis of the anode characteristic curves.

Considering a resistance model, parallel resistors sum to A.B/(A+B)=
,
of course. If it were legitimate to express the transfer function o=
f
the pair by similarly combining that of each valve, then you may en=
d
up with an average distortion, plus a higher order content arising
from the interaction between the two, since the product A.B would b=
e
distortion of distortion, as if the valves were in series. That is =
a
pretty loose thought at the moment...but it is what you might
expect. It would be interesting to put some typical practical
figures on the effect.

Just a further note. If the valves were perfectly matched, then eac=
h
would see the other as an infinite resistance, ie a current source.=


The small signal equivalent circuits when modeled as current sources
(typically pentode models, but could be just as easily a triode) alwa=
ys
include a resistor in parallel with the source generator. For a triod=
e
that is usually a low resistance. So as a generator each tube no long=
er
sees the other as infinite resistance. In any case, that is only smal=
l
signal & does not account for even the 3/2's power law.

That is why it is awkward to choose a voltage source model. As soon=

as they are not matched, each is seen by the other as an active
finite load, proportional to the degree of the mismatch. If you had=

two separate but equal loads on the two triodes, and then shorted
the anodes together, you would change both the voltage and the
current for each.

Someone should measure this. It will presumably be particularly
problematic in the common practice of doubling input valves in high=

gain preamps.

cheers, Ian



I'm not so sure that simulation IS the easy way out for the
determination of results. It depends too much on very accurate modeli=
ng,
something not always available. The results I measured many years ago=

were made on a pair of supposedly identical amplifiers, both
individually & then paralleled. I was surprised at the results & that=
is
why I remember all that so well. But my only real claim earlier in th=
is
thread was that if one paralleled 'n' triodes, don't expect to get 'n=
'
times the power at the same distortion level.

Probably the easy way out is just do the experiment, but with great c=
are
& be objective. Don't cook the results. If I had some time now I woul=
d
try it again. But Spring is finally here & there is much to do outsid=
e.
We had a foot of snow here on the 2nd & 3rd of April. There is still
some in the bush. And the graphics are much better there than on this=

display!

Its simple to create a model of a bunch of triodes as voltage generator=
s
each one puting an output of =B5 x Eg at its output and powering a comm=
on load

via its own separate Ra resistance.

The simulator should tell you what's going on with such a simple easy t=
hing
to set up.
But a human can churn the number by changing Ra a bit to see what effec=
t it
has on
gain and thd.

Paralleling tubes isn't a flip flop situation where if one tube conduct=
s
slightly more
signal load current then the weaker tube suddently quits and the strong=
er
burns up from over exertion.

In guitar amps using a six pack of tetrodes, regularly over driven heav=
ily,
there will always be a tube which will
fail first, but this can be due to a number of reasons, not just becaus=
e
things are paralleled.

Patrick Turner.



Cheers, John Stewart

If amplifiers are joined to increase available power, a good alternative =
is
bridging, if that is possible. By simply paralleling the outputs their
respective internal resistances get in the way, which causes current hogg=
ing by
one amplifier or the other. That is typical of any pair of voltage source=
s,
including triodes. Whether it be in audio amplifiers or large power syste=
ms,
some means of sharing the load must be provided. =
JLS