flipper wrote:

Now that I've got power back I've been able to return to more fun
things like the amplifier project and I've run across some things I'd
like your learned opinion on.

To recap a bit, the design started out as a simple 12SL7 gain stage
into a conventional self bias 6V6 SE pentode configuration. Then the
idea to make it triode switchable cropped up (necessitating a CF
buffer tube to drive the 6V6) and, currently, I've been thinking to
just do the triode mode.

Btw, I'm thinking of a zener in the CF plate circuit to keep it under
300V because it's simply cheaper than a C filtered R dropper and I
wouldn't think B+ ripple should be a big problem on the plate of a CF
driver. Am I wrong?

Zeners have their uses, but I try to not use them, mainly
because they have a wide noise bandwidth, and so bypassing them with 100uF

usually makes sense to control the noise, then all the zener does is
control the DC,
when the cap has a high impedance.

When i do use zeners its usually a string of 5 watt x 75v types,
say 4 to shunt reg a voltage of about +300V.
You need to have a 350V rated 100 uF cap for the 300V supply point and if
you have
10mA of current to a tube/s from this point you should have 5mA flow in
the zeners,
so that they work ok and at a safe temperature without having to
use heavy heat sinking.



Anyway, looking at the CF buffer tube and fixed bias I wondered why
one couldn't self bias the buffer tube? No bypass cap (it's a follower
after all), just a long Rk to -50V with a variable tap at the right
spot, about -14.7V, for grid bias. A quick simulation indicates that
works but am I missing something?

I wish we had the ability to be able to place schematics attatchments to
our posts.
But if you are to drive an output tube set up for fixed bias from a CF
which is direct coupled to the output grid, and also has say 50k taken to
a
-ve supply, then the CF *must* also have a fixed bias supply to its grid
somewhat
lower voltage than the actual output tube bias voltage.

So you need a -200V supply, and a divider network, then a 470k to the
CFR grid, and then you can cap couple the CF grid to any other drive tube
you like.

Divider values have to be got just right for the correct bias current in
the output tube.




I may be overly obsessing about fixed bias but as I evaluate
variations in the power supply, like vs line variations, it sure looks
like the bias point shifts all over the place, especially with a
triode. I don't mean the voltage I generate, I mean where the bias
point should be for 34ma through the triode because plate drops as
line voltage does, unless I make a regulated B+.

The amount of B+ drop in an SET amp is negligible under musical signal
conditions.
Even when driven into class A2, the change isn't large.
Next time you test, hook up a nusic source and take the levels up until
clipping with rock and roll just starts to be visible on the CRO as little
flat spots.
Monitor the supply dc voltages; I bet they hardly vary much at all.





How do you deal with that in yours?

I only use class A1 with no grid current, so no CF buffers needed.

No shunt regulation either.

Just sensible RCRCRC filtered supplies with maybe a choke as well.
No pussy wussy caps of 20uF like they used in the silly old days of 1955.
I like large value caps in power amp
PS so I like to use 470uF caps in series/parallel arrangements to get the
hum down to
mV at the CT or other SE transformer.
Large value caps do better and sound better.






On another simulation I ran into my first real 'surprise'. I decided
to put the pieces together, in particular the power supply and amp, to
check ripple effects but the surprise came with startup.

I've got a HV delay and soft start circuit worked out but for the
first simulation cut I just used a plain 'slam it on' SS HV supply and
the original pentode amp (quicker than redrawing the whole thing).
What surprised me is that the 6V6 current spikes to over 150ma for
250ms, or so, then settles back to the design 34ma, with the culprit
being the gain stage to 6V6 RC coupling time constant. But in all the
reading I've done I've never seen anyone even mention this effect.

The current surge during turn ons shouldn't be so great.
Something is wrong about how you have your things hooked up because
in the countless radios that use a 6V6 and in all the amps and gear I
revise or
make, such things just don't happen.

From cold with an SS supply, the current always just starts with a slow
trickle then over about
a 5 second period the current increases to full idle value.

In a cathode biased amp the current trickle has to charge the cathode
caps.
Its always slow from cold, but if the tube is hot, and you turn it off,
then back on 2 seconds later, indeed sometimes current is high because the

cathode can emit as soon as power is back on, but the cathode voltage has
sagged,
to the tube is biased for a large current flow.

Fixed bias amp should be able to establish the bias voltage faster than
any
applied B+, and should sag very slowly if the power is turned off.


Is it that no one takes notice of 50 watts into a 12 watt plate for
such a 'short' duration?

if the voltage is average 150V for the 150mA, you only have 22 watts for a
short time.
The tube is able to have a current swing of up to saturation for a short
time.



Ramping HV up (the soft start circuit) makes matters worse up to a
point because while it reduces the overshoot it lengthens the over
power duration (1 sec ramp). Slow enough, say 7 seconds (0 to 330),
and the 6V6 self bias is under damped and ramps up to 34ma without
overshoot.

That's with hot heaters, of course.

This is the trouble of heating the tubes and then applying the B+.
The tubes are able to fully conduct, and when the B+ is turned on,
the grids to the output stage may initially be pulled up +ve
due to how the driver stage reacts to the B+ turn on.


It all makes sense it's just that I didn't expect it.

I think you need to more carefully measure exactly what
are the dynamic voltage movements *all* electrodes, and perhaps take
steps to limit any problems.
I just turn on the B+ fully using SS diodes and with the heaters.
Sometimes an amp will oscilate during turn on or off.
Tweets, farts, squeals are not unusual.

In that case sometimes the best way is to have a very slow rise time for
the
screen or driver voltage; such unexpected issues nearly
always arise with DIY amps.
Factory made jobs invariably don't show exactly what troubles lay behind
their development.

I had a PV2 Conrad Johnson preamp arrive this week from a guy whose
purchases always have led to a trip across town to me for a fix after any
purchase he has made on ebay.
The CJ was quite unusable.
It had been slightly modded internally in an attempt to fix the
noise at turn on problems due to the preamp stages including heavy
negative feedback loops.
An ARC amp also gave huge problems 4 mths ago for the same reasons of
instability during turn on.

The CJ has a relay which grounds its outputs for some time after turn on,
then the relay opens,
and the amp is connected to the outside world after the noise action has
subsided.
The time delay circuit was poorly designed, and wasn't working properly,
but after fixing that the amp has become usable and polite.

There is a real art, or rather a real science to analysising turn on/off
behaviour,
and finding solutions that are not damaging to anything.

Patrick Turner.

flipper wrote:

On Fri, 07 Oct 2005 11:16:31 GMT, Patrick Turner
wrote:



flipper wrote:

Now that I've got power back I've been able to return to more fun
things like the amplifier project and I've run across some things I'd
like your learned opinion on.

To recap a bit, the design started out as a simple 12SL7 gain stage
into a conventional self bias 6V6 SE pentode configuration. Then the
idea to make it triode switchable cropped up (necessitating a CF
buffer tube to drive the 6V6) and, currently, I've been thinking to
just do the triode mode.

Btw, I'm thinking of a zener in the CF plate circuit to keep it under
300V because it's simply cheaper than a C filtered R dropper and I
wouldn't think B+ ripple should be a big problem on the plate of a CF
driver. Am I wrong?

Zeners have their uses, but I try to not use them, mainly
because they have a wide noise bandwidth, and so bypassing them with 100uF

usually makes sense to control the noise, then all the zener does is
control the DC,
when the cap has a high impedance.

Ok. Point taken.


When i do use zeners its usually a string of 5 watt x 75v types,
say 4 to shunt reg a voltage of about +300V.
You need to have a 350V rated 100 uF cap for the 300V supply point and if
you have
10mA of current to a tube/s from this point you should have 5mA flow in
the zeners,
so that they work ok and at a safe temperature without having to
use heavy heat sinking.

I understand what you're saying there about a shunt reg.

However, I'm just putting a 51 volt zener in series from the plate to
B+ so there's a 51 volt drop across it.

Then any movement of the B+ is also delivered to the anode of the CF, plus some
noise.
But not much noise is transfered to the output od the CF because the
Rin to the anode circuit of a CF is very high, and the output voltage is
deterimned by the
grid voltage.

But best practice is to establish a B+ for the CF to be referenced to 0V, not
the main B+,
using shunt regulation, a string of zeners to get 300V, with a resistor down
from te main B+,
and with a 100uF cap to lower noise, and prevent transient voltages from
destroying the zeners; they are only SS devices, and are prone easy failure.
You want to use the worst type of practice.....

I may seem to be splitting hairs, but I'm just telling you my preference.



space snip

But if you are to drive an output tube set up for fixed bias from a CF
which is direct coupled to the output grid, and also has say 50k taken to
a
-ve supply, then the CF *must* also have a fixed bias supply to its grid
somewhat
lower voltage than the actual output tube bias voltage.

So you need a -200V supply, and a divider network, then a 470k to the
CFR grid, and then you can cap couple the CF grid to any other drive tube
you like.

Divider values have to be got just right for the correct bias current in
the output tube.

Yeah. But I don't follow your description and the -ve supply.

Let me give it another shot.

Working backward

6V6 plate to OPT. OPT other side to B+ 330V
screen to plate thru 180Ohm
cathode to ground
Grid to top of 12SL7 Rk connecting to 12SL7 cathode

Rk is actually a resistor divider chain
Bottom of Rk to -50V
top of Rk to 12SL7 cathode and 6V6 grid
Middle of Rk, adjustable pot to 12SL7 grid through 680k

12SL7 cathode to Rk (as just described) and 6V6 grid
grid to 680k to Rk middle pot which sets grid to -14.7
Plate to B+ 330V through 51V dropping zener.

12SL7 grid cap coupled to gain stage.

As the bias goes, the Rk divider sets 12SL7 grid to -14.7V with about
a 2.6V drop across the tube to cathode Rk which sets the 6V6 grid to
-12.1 for an idle current of 34ma.

Move the 12SL7 grid up/down with the Rk pot and the 6V6 bias follows
and idle current is adjusted.

OK, I see what you are doing.
Bu adjusting the cathode bias on the CF, ie, the grid to cathode voltage on the

SL7, the current in the SL7 is varied so the cathode voltage changes so
the output tube is biased for the wanted current by means of adjusting the pot
you have.
Ingenious.
But I would much prefer to keep the pot out of the
main dc current flow of the SL7.
I'd have a separate divider with a pot in the middle between -50V and 0V
and then have a low dc current flow in the pot and then bias the SL7 from
between
0V and -50V.
Rk for the SL7 is say a single 47k or whatever gives about the right current;
I would have a -150V supply so Rk could be as high a value as possible
to provide the CF with a high ohmic load to minimise the work the CF has to do.

A CCS between -50V and the SL7 cathode would be very good.
With 1/2 an SL7, all you'd need is about 1.5mA at idle.

For most of the easy class A action of the SL7 the load in its cathode should
be
the same as what you'd use if you had the load in the anode circuit,
so having only about 35V across a resistor in a cathode circuit isn't doing it
right.
The typical class A load for SL7 is about 150k with say 1mA,
so the drop on the load would be 150V, hence the idea that you should have a
-150V supply.




I may be overly obsessing about fixed bias but as I evaluate
variations in the power supply, like vs line variations, it sure looks
like the bias point shifts all over the place, especially with a
triode. I don't mean the voltage I generate, I mean where the bias
point should be for 34ma through the triode because plate drops as
line voltage does, unless I make a regulated B+.

The amount of B+ drop in an SET amp is negligible under musical signal
conditions.
Even when driven into class A2, the change isn't large.
Next time you test, hook up a nusic source and take the levels up until
clipping with rock and roll just starts to be visible on the CRO as little
flat spots.
Monitor the supply dc voltages; I bet they hardly vary much at all.

Yes. It isn't the drop vs signal I was talking about here, it's line
voltage variation. Mains power to you. I.E. if it's adjusted when
mains is a nominal 120V then fixed bias seems to be seriously out of
whack when mains is 105 because B+ is no longer 330 but 288.

Triodes with fixed bias are prone to becoming over biased
when the line volts go high.

The 6V6 has an Ra = about 4k, so if the B+ changes 52 v, then you get an idle
current change
of about 12.5mA, so you should set the 6V6 up for the 330V condition.
When in fact the B+ drops a bit, so to will the negative bias voltage, thus
turning on the tube a bit more
when B+ is low.
Hence there is never a need to regulate the negative fixed bias voltage; one
wants it to wander around
with the B+ voltage.
The changes of bias voltages are inaudible because usually your PS *should*
only very slowly change its voltages even when there is someone doing some arc
welding next door on an old Buick bumper bar.

In most amps I make you can switch them on and off and hear no sign
that any on/off switching is happening.

The mosfet amp I have will run switched off for 20 seconds since the PS caps
are
a pair of 100,000 uF.

Use some real capacitance, you will be so glad you did.





space snip

On another simulation I ran into my first real 'surprise'. I decided
to put the pieces together, in particular the power supply and amp, to
check ripple effects but the surprise came with startup.

I've got a HV delay and soft start circuit worked out but for the
first simulation cut I just used a plain 'slam it on' SS HV supply and
the original pentode amp (quicker than redrawing the whole thing).
What surprised me is that the 6V6 current spikes to over 150ma for
250ms, or so, then settles back to the design 34ma, with the culprit
being the gain stage to 6V6 RC coupling time constant. But in all the
reading I've done I've never seen anyone even mention this effect.

The current surge during turn ons shouldn't be so great.
Something is wrong about how you have your things hooked up because
in the countless radios that use a 6V6 and in all the amps and gear I
revise or
make, such things just don't happen.

The heaters aren't already hot in those I would imagine.

They just all switch on together, and since a 5Y3 is used for a rectifier, the
B+
comes up to its high surge voltage level, since the 5Y3 is a directly heated
diode
so conduction starts within a few seconds.
The high peak currents to get the caps charged up to a high V don't
seem to hurt the cathodes.





From cold with an SS supply, the current always just starts with a slow
trickle then over about
a 5 second period the current increases to full idle value.

In a cathode biased amp the current trickle has to charge the cathode
caps.
Its always slow from cold, but if the tube is hot, and you turn it off,
then back on 2 seconds later, indeed sometimes current is high because the

cathode can emit as soon as power is back on, but the cathode voltage has
sagged,
to the tube is biased for a large current flow.

Fixed bias amp should be able to establish the bias voltage faster than
any
applied B+, and should sag very slowly if the power is turned off.

Yes, well, 'from cold' is the thing. I was trying to take advantage of
all the wisdom posted in here and put a delayed B+ start to let the
heaters warm up first but that means it's all primed and ready to
surge when B+ comes up.

To make a B+ that switches on after 20 seconds and takes a further
20 seconds to slowly rise to full value is not mean achievement,
and really only doable with a series regulator which has a grid control
someplace that
has a slow time constant to allow the slow rise of the output voltage without
any sudden
jerks.

I have never worried about it except in the case of an SE amp with parallel
tubes which
insisted it oscillate .5 of a second after the amp was turned on.
I placed a 470uF cap to bypass the screen supply and fed with a resistor from
the B+
with hot cathodes the anode current is prevented from ever surging, and no
oscillations occured either.
But in this case the amp had cathode feedback in the output stage,
another reason why oscillation is possible.
It's like a Quad II amp cut in half, but with 4 SE output tubes.

Being SE, they can tend to oscillate because the balanced nature of a PP
circuit
tends to prevent it.


Is it that no one takes notice of 50 watts into a 12 watt plate for
such a 'short' duration?

if the voltage is average 150V for the 150mA, you only have 22 watts for a
short time.
The tube is able to have a current swing of up to saturation for a short
time.

Looks more like 225 on mine but I imagine that varies with the stage
RC coupling. Takes a good solid second to get under 50ma. too.

Ramping HV up (the soft start circuit) makes matters worse up to a
point because while it reduces the overshoot it lengthens the over
power duration (1 sec ramp). Slow enough, say 7 seconds (0 to 330),
and the 6V6 self bias is under damped and ramps up to 34ma without
overshoot.

That's with hot heaters, of course.

This is the trouble of heating the tubes and then applying the B+.
The tubes are able to fully conduct, and when the B+ is turned on,
the grids to the output stage may initially be pulled up +ve
due to how the driver stage reacts to the B+ turn on.

I think I'll skip the heater delay. Fewer parts, less work, and if it
strips the cathodes they're 2 buck tubes. Although lovely Sylvanias.

It all makes sense it's just that I didn't expect it.

I think you need to more carefully measure exactly what
are the dynamic voltage movements *all* electrodes, and perhaps take
steps to limit any problems.
I just turn on the B+ fully using SS diodes and with the heaters.
Sometimes an amp will oscilate during turn on or off.
Tweets, farts, squeals are not unusual.

This part answers my question because the entire issue goes away if
the heaters are cold and the tube can't conduct during the B+
stabilization period.

Thanks for the "Tweets, farts, squeals" tip. If I remember maybe I
won't **** in my pants from panic the first time I turn it on

In that case sometimes the best way is to have a very slow rise time for
the
screen or driver voltage; such unexpected issues nearly
always arise with DIY amps.
Factory made jobs invariably don't show exactly what troubles lay behind
their development.

Right. No one advertises the dirty laundry, especially after it's been
cleaned and folded

I had a PV2 Conrad Johnson preamp arrive this week from a guy whose
purchases always have led to a trip across town to me for a fix after any
purchase he has made on ebay.
The CJ was quite unusable.
It had been slightly modded internally in an attempt to fix the
noise at turn on problems due to the preamp stages including heavy
negative feedback loops.
An ARC amp also gave huge problems 4 mths ago for the same reasons of
instability during turn on.

The CJ has a relay which grounds its outputs for some time after turn on,
then the relay opens,
and the amp is connected to the outside world after the noise action has
subsided.
The time delay circuit was poorly designed, and wasn't working properly,
but after fixing that the amp has become usable and polite.

There is a real art, or rather a real science to analysising turn on/off
behaviour,
and finding solutions that are not damaging to anything.

Interesting that there's little, hell no, mention of these start up
noise issues in any of the amp articles, DIY or not.

Ben Duncan is the only author to have mentioned start up / shut down horrors
some amps have,
especially solid state; he wrote a great book in 1996 about the history and
development of SS
amps since the days when tubes were being phased out in 1960.

Such problems can destroy speaker drivers.

Most modern SS amps have delayed protection relays between speakers and the
amp.

And they fail to protect so often....

Patrick Turner.




Patrick Turner.

snip for space,


That was my first cut at it except I went the extra step to shunt
zener regulate the -50 to -22 for extra stability.

I would have a -150V supply so Rk could be as high a value as possible
to provide the CF with a high ohmic load to minimise the work the CF has to do.

-50V is about all I can conveniently get with the available power
components. It's voltage tripled off an 18V tranny I'm using for the
filaments (CCS, so no turn on filament surge and quiet DC).

But you could easily multiply the voltage further to make -150V just to sling a
large value resistor carrying
the low current of a mA or two.

Or hunt around for another old tranny.....

I didn't let the unavailability of parts ruin my applications of
ideas; I took steps to scrounge up whatever I could find
for breadboarding.....



A CCS between -50V and the SL7 cathode would be very good.
With 1/2 an SL7, all you'd need is about 1.5mA at idle.

I was thinking about that but I'm not sure what I'd gain for the extra
complexity.

Other than the modest Rk taking a bit more juice on the positive side,
what problem is it? Distortion?

Larger Rk, which is the RL for the SL7 means less distortion.
I split hairs again, but even though the CF distortion is low,
we shouldn't encorage it by cheating with lowering the cathode load....



As it stands, with the DC bias (not the self bias), Rk is 33k at about
1.15ma (cathode is at -12.1) and draws .84ma to 1.45ma for a 1 watt
OPT output. 'Full bore' is .45ma to 1.88ma plus 2.42ma going into the
6V6 grid.

I would of course use a 6SN7 as the driver, to get less thd since the Rout
of the SN7 as a CF would be lower perhaps.
With 1.88mA and 2.42 mA flowing in the SL7, perhaps this tube is *also*
driven into grid current; check the load lines....

A 12AT7 would be an even better driver because Ra is low
and µ is 3 times the SN7 abd about equal to the SL7, and the CF Rout
will definately be lower than the SL7, and no risk of grid current in the CF,
so the drive is all that much cleaner...
6AQ8 is also another nice tube to use, 6.3V heaters, but otherwise
almost identical to 12AT7.





Meaning 12SL7 idle is 1.15ma with peak 4.3ma. With plate at 251 (91V
dropping zener from 330V B+ and -12.1 cathode) that's a plate peak of
1.08 Watt (average comes to .38W).

For most of the easy class A action of the SL7 the load in its cathode should
be
the same as what you'd use if you had the load in the anode circuit,

What is the reason for that? I mean, you'd use a high value in the
anode because you're looking for gain. CF=1.

so having only about 35V across a resistor in a cathode circuit isn't doing it
right.
The typical class A load for SL7 is about 150k with say 1mA,
so the drop on the load would be 150V, hence the idea that you should have a
-150V supply.

Well, the ma=~1 is how I selected the 33k. Could use 39k.

OK.

you are not pushing too many volts into the 6V6.
If it was a 6550 for something with real balls, then you'd need to think
more about drive.
Best to overdesign the drive and to have a higher than needed current
ability because it usually translates to better sound dynamics.





space snip

Yes. It isn't the drop vs signal I was talking about here, it's line
voltage variation. Mains power to you. I.E. if it's adjusted when
mains is a nominal 120V then fixed bias seems to be seriously out of
whack when mains is 105 because B+ is no longer 330 but 288.

Triodes with fixed bias are prone to becoming over biased
when the line volts go high.

The 6V6 has an Ra = about 4k, so if the B+ changes 52 v, then you get an idle
current change
of about 12.5mA,

That's a pretty darn significant change for a nominal 34ma.

so you should set the 6V6 up for the 330V condition.
When in fact the B+ drops a bit, so to will the negative bias voltage, thus
turning on the tube a bit more
when B+ is low.
Hence there is never a need to regulate the negative fixed bias voltage; one
wants it to wander around
with the B+ voltage.

Well, there's my problem. I'm trying to make things 'too good' because
I'm regulating the -50 and then zener regulating the -DC bias to boot
because I'm concerned about injecting hum onto the grid.

Ordinary RC filtering should prevent hum from the B+ or B-.



High voltage caps aren't cheap but zeners and transistors are so it's
easier to regulate than throw a ton of uFs at it. Hmm. Maybe I'll look
into a SS cap multiplier arrangement so it's heavily filtered but not
regulated.

A few 100 uF won't break your bank.



My gut tells me that just letting things drift isn't optimal unless
things are ratioed properly. I need to do some thinking on that.

Letting things drift *is* ok if there is just some drift
and no LF instabilities or tendency for LF bounce.
with varying signal levels.

One instance of nil regulation and lots of direct coupling is the Loftin White
amplifier
which came into being well before WW2 and the existance of large cheap caps
or zener diodes.

Keep your iron hot...

Patrick Turner.

Hey Patrick, How's it going?

Regards
Mark

Mark Harriss wrote:

Hey Patrick, How's it going?

Ok for now.

Tomorrow is sunday, when i wonder why i am going.

If you know why your'e going, its easier to work out where your'e going,

and more about how.

The do-littles of this world spend lotsa time wondering how they are
going,
better to know only little about that and just go go go
and keep outa bother.

Patrick Turner.



Regards
Mark

flipper wrote:

On Sat, 08 Oct 2005 11:48:40 GMT, Patrick Turner
wrote:



snip for space,


That was my first cut at it except I went the extra step to shunt
zener regulate the -50 to -22 for extra stability.

I would have a -150V supply so Rk could be as high a value as possible
to provide the CF with a high ohmic load to minimise the work the CF has to do.

-50V is about all I can conveniently get with the available power
components. It's voltage tripled off an 18V tranny I'm using for the
filaments (CCS, so no turn on filament surge and quiet DC).

But you could easily multiply the voltage further to make -150V just to sling a
large value resistor carrying
the low current of a mA or two.

Well, let's say easily to -100

Ya can't just go forever ya know. Each stage has to provide enough
current to charge/discharge the next one's cap and eventually they
just play with each other. Or you make the previous stages with bigger
and bigger caps.

Voltage multiplyers with Si diodes and the cheap caps you can get
make it possible to go 10 times without too much effort; the efficiency
is not a worry when all you want is a few mA.





Anyway, I modified the thing up for -100 after the last chat.

OK,





Or hunt around for another old tranny.....

I didn't let the unavailability of parts ruin my applications of
ideas; I took steps to scrounge up whatever I could find
for breadboarding.....



A CCS between -50V and the SL7 cathode would be very good.
With 1/2 an SL7, all you'd need is about 1.5mA at idle.

I was thinking about that but I'm not sure what I'd gain for the extra
complexity.

Other than the modest Rk taking a bit more juice on the positive side,
what problem is it? Distortion?

Larger Rk, which is the RL for the SL7 means less distortion.
I split hairs again, but even though the CF distortion is low,
we shouldn't encorage it by cheating with lowering the cathode load....

OK, I don't notice any difference in the sims but I'm not sure it's as
accurate as it could be when the signal is so close to -v. I mean,
it's being driven to -35V so with -V at -50V there's not much left.

Well this is one reason why some respect to load value should be given,
and the cathode load in a CF should be worked out as if the load is in the anode
circuit
for best linearity.

Load line analysis allows a quick draw on the back of an envelope
at the kitchen table; all you have to know is the Ra line for Eg = 0V,
and then you can draw in the load line to the best position.






As it stands, with the DC bias (not the self bias), Rk is 33k at about
1.15ma (cathode is at -12.1) and draws .84ma to 1.45ma for a 1 watt
OPT output. 'Full bore' is .45ma to 1.88ma plus 2.42ma going into the
6V6 grid.

I would of course use a 6SN7 as the driver, to get less thd since the Rout
of the SN7 as a CF would be lower perhaps.
With 1.88mA and 2.42 mA flowing in the SL7, perhaps this tube is *also*
driven into grid current; check the load lines....

Yeah, I've checked. That's one reason I keep reminding myself by
posting the max plate current. Too much and it takes positive grid.

A 12AT7 would be an even better driver because Ra is low
and µ is 3 times the SN7 abd about equal to the SL7, and the CF Rout
will definately be lower than the SL7, and no risk of grid current in the CF,
so the drive is all that much cleaner...
6AQ8 is also another nice tube to use, 6.3V heaters, but otherwise
almost identical to 12AT7.

I'll keep those in mind for the next one. And there will be one.

6DJ8, 6CG7, 6H30 are also usable, the last two have fairly low gain,
the 6H30 has very low Ra, but is very favoured in high end amps now for drivers,
but its gain is the same as a 12AU7, also usable.

But say the 6V6 needs 16vrms drive in SET A2, then you'd need about
1.4V input to the gain 1/2 using 12AU7, and this is ok still if no loop NFB is to be
used.



Meaning 12SL7 idle is 1.15ma with peak 4.3ma. With plate at 251 (91V
dropping zener from 330V B+ and -12.1 cathode) that's a plate peak of
1.08 Watt (average comes to .38W).

For most of the easy class A action of the SL7 the load in its cathode should
be
the same as what you'd use if you had the load in the anode circuit,

What is the reason for that? I mean, you'd use a high value in the
anode because you're looking for gain. CF=1.

so having only about 35V across a resistor in a cathode circuit isn't doing it
right.
The typical class A load for SL7 is about 150k with say 1mA,
so the drop on the load would be 150V, hence the idea that you should have a
-150V supply.

Well, the ma=~1 is how I selected the 33k. Could use 39k.

OK.

you are not pushing too many volts into the 6V6.
If it was a 6550 for something with real balls, then you'd need to think
more about drive.
Best to overdesign the drive and to have a higher than needed current
ability because it usually translates to better sound dynamics.

Yeah. It doesn't take a ton of drive for the 6V6 and I'm measuring
that peak with the whole thing maxed out at about 4.6 watts into the
speaker where it's beyond the practical range anyway so I've got some
play room in there.

Fixed bias did get the 4.6 watts out of hard clipping though. It's a
nice fat ass sitting on ground..

So you are measuring the voltage at the output with an rms meter,
squaring the value, dividing by the load...
At the onset of clipping though maybe only 2 watts is possible in 6V6 in triode.



space snip

Well, there's my problem. I'm trying to make things 'too good' because
I'm regulating the -50 and then zener regulating the -DC bias to boot
because I'm concerned about injecting hum onto the grid.

Ordinary RC filtering should prevent hum from the B+ or B-.

Enough of it, yeah. Got a suggestion on how many volts of ripple is
acceptable?

For any SE amps I like to have Vr 10mV.
This means usually a CLC supply and for what you are doing,
a CRC supply would do, and 470uF, 220ohms, 470uF
with 50mA of total current would give 0.23vrms at C1,
and 3.55mV at C2, with 100Hz, and
considerably less if you have 120Hz ripple.
Chokes start to become cost effective when the I rises above
100mA, say for a large amp with many output tubes.
In ancient times of the 1950s, electros were unreliable,
costly and bulky and a much larger volume winding industry
existed so folks used chokes, ie, inductance rather than capacitance to
do their B+ filtering.
The problem with this was there was not a very low impedance at the point
where one connects the OPT; a 16uF cap has 200 ohms of reactive impedance at 50Hz;
470uF has 6.8 ohms, which anchors whatever is connected to it much better.
The real cost of capacitance has plumeted and the reliability has risen,
since modern electronics require electrolytics made in millions
everyday and to be able to operate over a truly wide bandwidth.
Chokes have become expensive by comparison.

The basic filter for a B+ supply could even be done with all CRC
if you want even in an amp drawing 1/2 an amp of plate current.

The rules a simple,
Choose C1 to give peak ripple voltage less than 5% of the output dc voltage.
Make sure ripple current is less than 1/4 of the max rated figure in the data for the
cap.
If a CRC filter is chosen, then the R should be at least 10 x the reactive value of C1
at Fr.
C2 can be as large as you can afford, and can be much larger than C1.
But nevertheless, C1 can be equal to C2 and both be say 470uF,
but then the peak charge currents into the cap are high, so a series R
beween the Si rectifiers and a large value C1 will reduce the peak charge currents,
so say R = 5 x Zc can reduce Ipk a lot, but only if the winding resistance is already
low.
So where you have C1 = 470uF, then maybe you'd have 22 ohms in series,
and the peak charge current would be maybe less than 1/3 of what it may otherwise
be with diodes charging straight in.
The ripple voltage won't change.

I am sure you will find more about all this at various sites around the web.

Using CRCRC is also ok, and C1 can be say 100uF, R1 = 220 ohms, C2 = 470uF,
R2 = 220ohms, C3 = 470 uF, and for 100Hz and 200mA,
the output ripple will be 1mV.

With 120Hz, its a lot lower.

At 200mA, the voltage drop in the total of 440 ohms is 88 volts,
and needs 17.6 watts of heat to be dissipated, or about as much as one of the
output tubes, so hence the idea of a choke becomes attractive.
The choke value need only be 0.33 H, to have the same effect as a 220 ohm R;
DCR will maybe = 22 ohms, so little power is lost choke is small,
like a little radio se OPT.
The only downside of the CLC or CLCLC filter is the resonance of the
LC of the final filter.
It should be less than 5 Hz for a hi-fi amp, so if C = 470uF,
the L should be 2.2H.
Fo = 5,035 / ( sq.rt of L x C ) where Fo = resonant F,
L in mH, C in uF.

CRCRC filters are not resonant.

But getting back to a CLC, say you have 200mA, and C2 = 470uF, and L = 2.2H,
then let's guess C1 = 100uF so we don't have large peak charge currents.

The formula for ripple at C1 is
Vr = 2,200 x Idc / C, where I is in amps, and C is in uF.
The constant of 2,200 is for 100Hz ripple, but if you are in a 60Hz area
use a constant of 1,833.
Notice that for each element of the filter, with 60Hz mains the the ripple is 5/6 of
the
value for 50Hz. I come from Oz, we have 50Hz,
if you are in the US, you should use 1,833 as the constant.

The results are quite accurate as long as the L or R used following C1 is = 10 x ZC1,
so if C1 = 100uF, then ZC1 = 1,600/C where C is in uF for 100Hz,
and of course for the US you would say ZC1 = 1,333 / C;
the higher the F, the lower the Z.


Then for 200mA into 100uF-2.2H-470uF we get Vr at C1 = 4.4Vrms, then the attenuation
factor of the LC
following second order filter = ZC/ZL = 3.4/1,381 = 0.00246 for 100Hz..
So ripple voltage at C2 = 4.4 x 0.00246 = 10.8mV.
For 60Hz mains, 120Hz ripple, its about 6.2mV.

For 50mA, 120Hz ripple, its 1.6mV.

And a trick you can do is to place a series CR network across the L to make the L
slightly resonant to 60Hz.
The R should be about 22 ohms, and C about 0.82uF,
so 2.2H resonates with 0.82 uF at around 120Hz.
This set up will cut the 1.6mV of ripple to around 0.4mV,
with negligible rise in ripple harmonics above 120Hz....

Most ppl don't bother gilding a lily when they have gone to the trouble
of such a CLC filter.

In my SEUL amps with a 450V x 200mA supply,
I have a voltage doubler supply using a single winding of
about 170vrms, a pair of 6 amp rated Si diodes, and a pair of 470uF caps for C1
which are in series.
Then there is a 9H choke, and two 680uF caps in series, so the
CLC is 235uF 9H 340uF, and so Vr at C1 = 1.9Vrms for 100Hz,
then Vr at C2 2mV, so because the OPT ratio is 20:1, the ripple in the output
from the PS is around 0.1mV, and considerably lower than from
other hum ingress elsewhere in the amp.
16dB of NFB reduces whatever hum I get by around 5 times to negligible levels.



High voltage caps aren't cheap but zeners and transistors are so it's
easier to regulate than throw a ton of uFs at it. Hmm. Maybe I'll look
into a SS cap multiplier arrangement so it's heavily filtered but not
regulated.

A few 100 uF won't break your bank.

Hehe. I have a very small bank.

Go and do some work to earn some $$ to pay for some parts.

( Did I say work? yes. Its the awful *w* word )



My gut tells me that just letting things drift isn't optimal unless
things are ratioed properly. I need to do some thinking on that.

Letting things drift *is* ok if there is just some drift
and no LF instabilities or tendency for LF bounce.
with varying signal levels.

Yeah. I checked that and simulating a 10% mains drop so B+ drops to
300 and -V drops to -90 I get an idle current shift from 34.07 to
32.87.

That's a dern sight better than 12 ma.

I was guessing for the worst expected figure.
But I have seen amp that were made in china with PT meant for 220Vrms,
but we get 250Vrms here many days, so the B+ is +534V instead of +470V,
and the factory bias recomendation of 50 mA per EL34 is quite wrong;
with the higher B+ the diss rises from 23.5 watts to 26.7,
and there is no margin for error; the tubes teeter on the brink of
going poof!, well, chinese tubes should **never** be run near the
rated maximum Pda if you don't loungeroom dramas.
Where you have B+ = 470V, and Ia + Ig2 = 50mA,
you have 23.5W.
Raising the B+ by 64V with fixed bias would result in
an increase in Ia of 64V / Ra = 64V / 1250 = 51mA, and the tube would overheat,
so the bias must be backed off to say 23.5 / 534 = 44mA, which is a drop of
6mA, requiring an increase of at least a volt in the Eg.
If the bias is not regged, then the bias of say -40V will increase to
about -44V, so idle current may not increase hugely.
But many amps such as Musical Reference, Stingray from Evanna Manley,
Jolida, should never be biased as the instructions insist for 230V when imported here
and operated where we often get 250Vrms on the mains.







Still, there is a shift and that's what I was alluding to. The -dc
bias is shifting a bit more than needed to compensate and I'll bet if
one were really clever they could perhaps insert some diodes or zeners
in series with the resistor divider, and recalculate of course, so
that the -dc bias tracks closer.

Most makers just don't bother with such issues, they just
say bias the tubes at such and such a current, not taking into account the real
and likely values of mains voltages in the countries where their amps may end up.

A zener could be used to transfer all the -V change at the bias supply into the divider

rather than the the divided change.

Food for thought though.
One could theoretically have a well compensated bias set up
so that where the mains went up say 15%, the bias would change
a sufficient amount to keep the Pda well within the wanted idle value.


With cathode bias, the effective Ra of a 6V6 becomes
Ra + ( µ x Rk ), so if Rk was 200ohms, then Ra' with c/bias
is about 7k? I am not sure, but the use of Rk does reduce the
rise in Ia with a rise of Ea quite a lot.
Where a pentode or beam tube has a regulated fixed screen voltage and is operating
as a pentode/beam then the Ra is the very high figure quoted in the data, and
variations in
B+ don't cause large Ia variations.
But with triode, the rise in B+ also raises the Eg2, which turns on more Ia.
Eg1 must be made more -ve to compensate.

Some tubes benefit more than others with cathode biasing;
tubes with a higher µ will give better bias regulation with
Rk.
Also amps with a high series resistance in the PS will not in fact
raise their B+ in the same proportion as the mains voltage change levels,
so if mains rise from say 220V to 250V, the B+ may only rise from
say 470V to say 510V due to the spongy nature of the PS.

With class A amps it isn't important to have a well regulated B+ voltage
since the power input to the output statge stays virtually constant
to clipping although there is a relationship between the slight increase
in Pin we measure due to the % of 2H distortion.
They used to measure the 2H by measuring the change in Pin, I don't,
I prefer to use a thd filter and meter.
Deep into overdrive the amp becomes biased like a class C jobby
because the grid current in overloaded stages charges coupling caps and biases tubes at
values which
are nearly equal to the tube being biased off if the values were used
at idle. This is how CR coupled amp stages tend to protect themselves;
output tubes in guitar amps don't mind 20dB of input voltage above that required for
clipping.
The Pin can reduce below the value measured just before clipping.
The output wave becomes squared, the tube is either on, or off, and the
average I flow in the tube is low, Pda is low, and efficiency is greater than a sine
wave condition
at onset of clip.

Patrick Turner.









One instance of nil regulation and lots of direct coupling is the Loftin White
amplifier
which came into being well before WW2 and the existance of large cheap caps
or zener diodes.

Keep your iron hot...

Patrick Turner.

Patrick Turner wrote:

Mark Harriss wrote:


Hey Patrick, How's it going?


Ok for now.

Tomorrow is sunday, when i wonder why i am going.

If you know why your'e going, its easier to work out where your'e going,

and more about how.

The do-littles of this world spend lotsa time wondering how they are
going,
better to know only little about that and just go go go
and keep outa bother.

Patrick Turner.



Regards
Mark




Sounds like you are pretty flat out at the moment, I'm
running off the principle:

"If you bite off more than you can chew, then chew like crazy"

Regards
Mark


BTW, is your email address still valid?

Mark Harriss wrote:

Patrick Turner wrote:

Mark Harriss wrote:


Hey Patrick, How's it going?


Ok for now.

Tomorrow is sunday, when i wonder why i am going.

If you know why your'e going, its easier to work out where your'e going,

and more about how.

The do-littles of this world spend lotsa time wondering how they are
going,
better to know only little about that and just go go go
and keep outa bother.

Patrick Turner.



Regards
Mark



Sounds like you are pretty flat out at the moment, I'm
running off the principle:

"If you bite off more than you can chew, then chew like crazy"

Regards
Mark

BTW, is your email address still valid?

I am getting about 100 spams a day and a few genuines so afaik the
website is working with email.

But I have two preamps before Xmas to build, then so much work I
I will be OK till about March.
Then I'd like to spend 3 mths building a few new power amps for sale
and adding much info to the website.

Patrick Turner.

Load line analysis allows a quick draw on the back of an envelope
at the kitchen table; all you have to know is the Ra line for Eg = 0V,
and then you can draw in the load line to the best position.

How do you do a load line when it's a non linear load? Grid current in
particular.

The anode load on an SET amp tube is always considered to be straight line.
The load is non linear on the driver tube due to the grid current.
So you have a load line that starts off straight and becomes a curve
upwards when the Ig1 flows.
It can still be plotted, but knowing exactly what the load is during grid current
takes some careful measuring if you wish to include it in your calculations
rather than just set a CF up which will cope without too much thd.



snip


Fixed bias did get the 4.6 watts out of hard clipping though. It's a
nice fat ass sitting on ground..

So you are measuring the voltage at the output with an rms meter,
squaring the value, dividing by the load...

Well, I'm doing it in Spice and it'll give you power in a resistive
load just by putting the 'probe' on it. Just saves manually doing the
math plus it'll handle any waveform.

At the onset of clipping though maybe only 2 watts is possible in 6V6 in triode.

2 watts looks clean as a whistle. Distortion increases at 3 but it
doesn't look like clipping kicks in till about 3.5 watts.

Well perhaps with A2.



space snip

Well, there's my problem. I'm trying to make things 'too good' because
I'm regulating the -50 and then zener regulating the -DC bias to boot
because I'm concerned about injecting hum onto the grid.

Ordinary RC filtering should prevent hum from the B+ or B-.

Enough of it, yeah. Got a suggestion on how many volts of ripple is
acceptable?

For any SE amps I like to have Vr 10mV.

That's lower than I thought you'd say was needed but about what I was
shooting for.

I saved all that for reference.

I'm going to try the cap multiplier because I like the idea and I can
always pop some more big caps in later if needed.

snip lots of great stuff

Yeah. I checked that and simulating a 10% mains drop so B+ drops to
300 and -V drops to -90 I get an idle current shift from 34.07 to
32.87.

That's a dern sight better than 12 ma.

I was guessing for the worst expected figure.

Yeah. I understand. I was just pleased to see the 'float' be rather
reasonable.

snip

Still, there is a shift and that's what I was alluding to. The -dc
bias is shifting a bit more than needed to compensate and I'll bet if
one were really clever they could perhaps insert some diodes or zeners
in series with the resistor divider, and recalculate of course, so
that the -dc bias tracks closer.

Most makers just don't bother with such issues, they just
say bias the tubes at such and such a current, not taking into account the real
and likely values of mains voltages in the countries where their amps may end up.

A zener could be used to transfer all the -V change at the bias supply into the divider

rather than the the divided change.

Food for thought though.
One could theoretically have a well compensated bias set up
so that where the mains went up say 15%, the bias would change
a sufficient amount to keep the Pda well within the wanted idle value.

I did twiddle a bit with inserting a zener just to see if it could be
balanced better and I did get it to stay within 1% with a 10% mains
drop but I don't think it's really a 'finished product' kind of thing.
Might play with it some more later.

With cathode bias, the effective Ra of a 6V6 becomes
Ra + ( µ x Rk ), so if Rk was 200ohms, then Ra' with c/bias
is about 7k? I am not sure, but the use of Rk does reduce the
rise in Ia with a rise of Ea quite a lot.
Where a pentode or beam tube has a regulated fixed screen voltage and is operating
as a pentode/beam then the Ra is the very high figure quoted in the data, and
variations in
B+ don't cause large Ia variations.
But with triode, the rise in B+ also raises the Eg2, which turns on more Ia.
Eg1 must be made more -ve to compensate.

Some tubes benefit more than others with cathode biasing;
tubes with a higher µ will give better bias regulation with
Rk.
Also amps with a high series resistance in the PS will not in fact
raise their B+ in the same proportion as the mains voltage change levels,
so if mains rise from say 220V to 250V, the B+ may only rise from
say 470V to say 510V due to the spongy nature of the PS.

That certainly complicates the matter if B+ and -V don't track each
other with a mains change.

Of course, there's doing an active servo bias too but that's a hell of
a lot more complicated than if a simple resistor network selection
would suffice and they have problems of their own.

With class A amps it isn't important to have a well regulated B+ voltage
since the power input to the output statge stays virtually constant
to clipping although there is a relationship between the slight increase
in Pin we measure due to the % of 2H distortion.
They used to measure the 2H by measuring the change in Pin, I don't,
I prefer to use a thd filter and meter.

Interesting. I hadn't though of that but it makes sense.

Deep into overdrive the amp becomes biased like a class C jobby
because the grid current in overloaded stages charges coupling caps and biases tubes at
values which
are nearly equal to the tube being biased off if the values were used
at idle. This is how CR coupled amp stages tend to protect themselves;

Wow. Hard to believe something actually works to protect itself rather
than self destruct, like most things do

Well, a sort of protection occurs unless the load value is on the low side,
then with 70 watts coming from a guitar amp
meant for 35 watts, the screens can fry....

Patrick Turner.