Introducing a New Horse to the Stable

On 16/09/2019 9:53 PM, Trevor Wilson wrote:
> On 16/09/2019 11:43 am, ~misfit~ wrote:
>> On 15/09/2019 1:58 AM, Trevor Wilson wrote:
>>> On 13/09/2019 8:02 pm, ~misfit~ wrote:
>>>> On 12/09/2019 10:18 PM, Trevor Wilson wrote:
>>>>> On 12/09/2019 12:17 am, ~misfit~ wrote:
>>>>>> On 10/09/2019 11:54 PM, Peter Wieck wrote:
>>>>>>> OK, OK, I will bite! Minor rant to follow:
>>>>>>>
>>>>>>> Tube vs. Solid State on reliability:
>>>>>>>
>>>>>>> There are not so very many 60-year old components in operation these days unmodified
>>>>>>> since-new. My oldest tube item turned 100 this year and likely works better than when it was
>>>>>>> new based on a better understanding of antenna systems, optimum tube voltages and so forth.
>>>>>>> Other than moving parts (CD player), the newest component in my office system was made in
>>>>>>> 1963. The system runs 9 hours per day, 5 days per week. Oh, and the tubes are original as well.
>>>>>>>
>>>>>>> On the other hand, and given my hobby, I see a large number of SS components that have blown
>>>>>>> transistors, exploded capacitors and much worse, irrespective of age and source. The well
>>>>>>> made, well designed stuff is serviceable, distinguishing it from the rest of the garbage out
>>>>>>> there.
>>>>>>>
>>>>>>> I would make a fairly apt comparison: A tube amplifier is much like a mid-last-century
>>>>>>> Mercedes or VW - few things were self-adjusting, and they required regular and attentive
>>>>>>> care-and-feeding. With such, they were good for several hundred thousand miles of reliable
>>>>>>> service. A contemporary Ford, Cadillac, Plymouth would be considered remarkable were it to
>>>>>>> survive 100,000 miles without heroic measures. Might run very nicely when running, but that
>>>>>>> would be your basic solid-state device in comparison.
>>>>>>>
>>>>>>> Put simply, they are different beasts designed with different things in mind, but for the
>>>>>>> same basic purpose. That one is or is not "BETTER" than the other is not relevant to the
>>>>>>> purpose in either case.
>>>>>>>
>>>>>>> Now, when I here things like "Zero global NFB" and "Critically matched components", I can
>>>>>>> smell the snake-oil from a great distance, even the 10,000 miles from here to Australia. I
>>>>>>> am sure that process also contains descriptives of "interconnects" rolled on the thighs of
>>>>>>> virgins on Walpurgis Night...
>>>>>>>
>>>>>>> Note that even "critically matched" solid-state components drift after a very short period
>>>>>>> of time in-service. All of them, such that that "less than 1%" is meaningful for perhaps 12
>>>>>>> hours or so.
>>>>>>>
>>>>>>> Being as this is a hobby for me, I get to try things that are otherwise unproductive,
>>>>>>> unprofitable or impractical. Such as shotgunning a device with single-value capacitors and
>>>>>>> then comparing it to the same device with carefully screened and matched caps. Or matching
>>>>>>> driver and output transistors and comparing to a similar device with disparate values. Guys
>>>>>>> and gals - you would be seriously shocked to discover how little difference some things make
>>>>>>> that the ALL-SEEING, ALL-KNOWING gurus will tell you are critical. Often no difference at all.
>>>>>>
>>>>>> Thanks for your input Peter. If I may ask, do you have an opinion on 'storage capacitors' on
>>>>>> an amplifier power supply? What in your opinion is 'better', a single (or few) very large
>>>>>> caps or multiple smaller caps to the same / similar capacitance?
>>>>>>
>>>>>> I have a long term project building my own amp based on PCBs taken from 100w MOSFET (two
>>>>>> pairs of J50 / K135 devices per amp) PA amps made by a New Zealand company in the 1980s.
>>>>>> (Craft, Gary Morrison's company before he went on to become head designer at Plinius until
>>>>>> 2005 when he left to set up Pure Audio). I got my hands on a rack of four of these mono amps
>>>>>> and preliminary testing using a clean source and good speakers suggest they will make a great
>>>>>> stereo amp.
>>>>>>
>>>>>> I need to put together a power supply to feed two of these and have some new 10,000uF caps
>>>>>> but was wondering if multiple smaller caps would be better. (In the PA amps they only had
>>>>>> 2,200uF but obviously weren't called on to reproduce much bass.)
>>>>>>
>>>>>> As it is I'll be using fly leads from the rectifier PCB to the caps, then to the amps and I'm
>>>>>> building my own case. I was thinking of maybe using my 10,000uF caps as well as maybe some
>>>>>> smaller ones, perhaps 1,000 in a bank, the best of both worlds. (There are also 100uF
>>>>>> electros across the rails on the amp PCBs that I'll be replacing.) That said I could also
>>>>>> just go to multiple
>>>>>>
>>>>>> Cheers,
>>>>>
>>>>> **Those old MOSFETs were pretty ordinary devices (not very linear). Evidenced by the fact that
>>>>> Plinius amps have always used BJTs. As Peter has stated, multiple small value caps will
>>>>> usually provide a superior, higher speed power supply. However, I would posit that those old
>>>>> MOSFETs are so horrible (modern MOSFETs are far superior), that it may not be worth the effort.
>>>>
>>>> I hooked a pair of them up to a preamp while still using their original power supplies and was
>>>> very pleased with the sound so decided to go ahead with the build.
>>>
>>> **I haven't listened to Craft (hi fi) amps in many years. What I heard back then was pleasing.
>>> Very wide bandwidth (ca. 1MHz), as I recall.
>>>
>>>>
>>>>> Craft amps used huge amounts of global NFB, required due to very low bias currents and the
>>>>> necessity to reduce the huge levels of distortion caused by the 'knee' at low currents (A
>>>>> Class A, or high bias MOSFET amp would have been much better). Anyway, the huge levels of
>>>>> global NFB means that PSRR (Power Supply Rejection Ratio) will be quite high, thus the
>>>>> influence of power supply changes will be relatively small.
>>>>
>>>> Unfortunately I don't own a 'scope so am unable to check a lot of stuff. When I listened to
>>>> them with the original power supplies (designed for PA use) they sounded sweet and clean at low
>>>> and moderate volume levels but seemed to run out of power at higher volumes, especially when
>>>> there was a lot of bass.
>>>
>>> **That could be due to a number of factors. Including:
>>>
>>> * Insufficient Voltage output.
>>> * Insufficient current output.
>>> * Insufficient power supply.
>>> * An unreasonable speaker impedance.
>>>
>>> Don't forget: Those meaty looking 2SJ50/2SK135 output devices are only rated for a meagre 7 Amps
>>> each and 100 Watts PDiss. By comparison, a typical output BJT of the same time period was rated
>>> at a far more respectable 20 Amps and 200 Watts PDiss (MJ15003/MJ15004). Present production
>>> variants are rated at 25 Amps and 250 Watts.
>>
>> So three pairs per side should be fine for a reasonably powerful amp?
>
> **Again: It depends on the maximum Voltage output. 3 pairs allows for a peak current ability of 21
> Amps.
>
>> I've studied the PCB and the output devices are paralleled (along with a resistor for each) so it
>> wouldn't be hard to add a third device to each (on very short flyleads - or even daughterboards -
>> mounted to the same heatsink).
>
> **Sure. However, make certain the drive circuitry can cope.

I'm not exactly sure of how to do that?

>> The speakers I'm intending to use with this are Sony SS-K90EDs.
>> Like these:
>> <https://www.stereo.net.au/forums/topic/260972-fs-sony-ss-k90ed-speakers-rare/>
>
> **OK.
>
>>
>>
>>> So, a little Ohm's Law should tell you if you are demanding more current than the output devices
>>> are capable of delivering. 14 Amps is, by high end audio standards, a relatively modest current
>>> ability for a (say) 100 Watt @ 8 Ohms amplifier. Provided the driver impedance is relatively
>>> benign, you should be OK. Fortunately, it is real hard to damage MOSFETs, by 'asking' them to
>>> deliver more current than they are rated for.
>>
>> That's one of the things I like about MOSFETs.
>
> **Well, a properly designed BJT amp should demonstrate the same robustness and reliability.
>
>>
>>>>> One more thing: Decent amounts of capacitance placed close to the output devices is far more
>>>>> influential than caps placed some distance away. In fact, long(ish) cables AFTER the main
>>>>> filter caps can be a serious limiting factor on the effectiveness of a power supply in a Class
>>>>> A/B amplifier. This is because the inductance of the wires can be a factor.
>>>>
>>>> Thanks. The fly-leads will only be 6" tops and I'll be using at least 1.5 square mm multistrand
>>>> copper conductors. If space allows I'll put a ~1,000uF cap right at the amplifier PCB as well
>>>> (or as large as I can get away with). I may end up building a wooden case as I don't have a
>>>> suitable metal one and wood's something I have experience and the tools for.
>>>
>>> **Wiring sounds good. And yeah, caps placed close to output devices is a very good thing. A
>>> wooden case, not so much. Wood is an excellent thermal insulator, which means heat may not
>>> escape too easily.
>>
>> I have a couple of big heatsinks for the amplifier modules that will sit either side of the case,
>> fins outwards in free air. They'll easily handle the power dissipation being 4x bigger than the
>> 'sinks used on the PA amp. Also I'll ventilate the top and bottom of the 'box' (if I end up going
>> with wood).
>
> **OK.
>
>>
>>>> I still haven't finalised my design yet. I might end up feeding them a few more volts than they
>>>> were getting from their original power supplies (my only suitable toroidial transformer is 10v
>>>> AC higher than original) so may parallel up a third pair of output devices onto the heatsinks
>>>> using one of the other amps as a donor. I haven't decided yet, as I said it's a long-term
>>>> project and I'm learning as I go.
>>>
>>> **Well, the MOSFETs are rated for a decent 160 Volts, so a few more rail Volts should be OK. And
>>> yes, more output devices won't hurt (refer to Ohm's Law as before). Pay attention to the drive
>>> capabilities of the preceding stages though.
>>
>> Thanks for this Trevor, I have saved it for future reference. My 300 VA toroid that I'm thinking
>> of using with this outputs 50v AC so +/- 70v DC when rectified. The original PA transformers were
>> 40v AC.
>
> **+/- 70VDC suggests a maximum power output of around 250 Watts @ 8 Ohms. If you plan on attempting
> to obtain that much power (continuously), then you will need two of those toroids.

I intend to use the system in my lounge so won't want crazy SPLs, the speakers likely wouldn't
handle that much power anyway. I actually do have two of the toroids but that would make for a big
amplifier case - and surely then I'd need to consider adding *two* more pairs of output MOSFETs per
amplifier?

I was thinking that, as I don't listen to dubstep or extremely bass-heavy music, using one toroid
and a lot of capacitance (in the region of 20,000 to 50,000 uF per rail) would be enough to handle
transients. If not then I might as well build a pair of monoblocks.

I've got a few coffee-cup sized Mepco/Electra 14,000 uF / 100v caps but they're not new... I also
have 8 new 10,000 uF / 100v Elna caps that are only about 1/4 of the size.

Cheers,
--
Shaun.

"Humans will have advanced a long, long way when religious belief has a cozy little classification
in the DSM"
David Melville

This is not an email and hasn't been checked for viruses by any half-arsed self-promoting software.

IF all you are doing is wanting to power speakers at a reasonable listening level in a reasonably sized room, and you do not want ear-bleed levels, your major task in this case is managing transients.

Your plan to use a large amount of capacitance will address that fairly nicely. And if your amp "breathes" at high volumes (you will absolutely know if that happens), then and only then consider (a) beefier power-supply(ies).

I run a pair of AR3a speakers (4-ohm, nominal) in a fairly small room (17 x 14 x 10 (feet)) with a 60 wpc/rms amp, and it does fine at any listening level I would care to use. It has about 8,000 uF/channel of capacitance.

Peter Wieck
Melrose Park, PA

On 16/09/2019 11:01 pm, ~misfit~ wrote:
> On 16/09/2019 9:53 PM, Trevor Wilson wrote:
>> On 16/09/2019 11:43 am, ~misfit~ wrote:
>>> On 15/09/2019 1:58 AM, Trevor Wilson wrote:
>>>> On 13/09/2019 8:02 pm, ~misfit~ wrote:
>>>>> On 12/09/2019 10:18 PM, Trevor Wilson wrote:
>>>>>> On 12/09/2019 12:17 am, ~misfit~ wrote:
>>>>>>> On 10/09/2019 11:54 PM, Peter Wieck wrote:
>>>>>>>> OK, OK, I will bite! Minor rant to follow:
>>>>>>>>
>>>>>>>> Tube vs. Solid State on reliability:
>>>>>>>>
>>>>>>>> There are not so very many 60-year old components in operation
>>>>>>>> these days unmodified since-new. My oldest tube item turned 100
>>>>>>>> this year and likely works better than when it was new based on
>>>>>>>> a better understanding of antenna systems, optimum tube voltages
>>>>>>>> and so forth. Other than moving parts (CD player), the newest
>>>>>>>> component in my office system was made in 1963. The system runs
>>>>>>>> 9 hours per day, 5 days per week. Oh, and the tubes are original
>>>>>>>> as well.
>>>>>>>>
>>>>>>>> On the other hand, and given my hobby, I see a large number of
>>>>>>>> SS components that have blown transistors, exploded capacitors
>>>>>>>> and much worse, irrespective of age and source. The well made,
>>>>>>>> well designed stuff is serviceable, distinguishing it from the
>>>>>>>> rest of the garbage out there.
>>>>>>>>
>>>>>>>> I would make a fairly apt comparison: A tube amplifier is much
>>>>>>>> like a mid-last-century Mercedes or VW - few things were
>>>>>>>> self-adjusting, and they required regular and attentive
>>>>>>>> care-and-feeding. With such, they were good for several hundred
>>>>>>>> thousand miles of reliable service. A contemporary Ford,
>>>>>>>> Cadillac, Plymouth would be considered remarkable were it to
>>>>>>>> survive 100,000 miles without heroic measures. Might run very
>>>>>>>> nicely when running, but that would be your basic solid-state
>>>>>>>> device in comparison.
>>>>>>>>
>>>>>>>> Put simply, they are different beasts designed with different
>>>>>>>> things in mind, but for the same basic purpose. That one is or
>>>>>>>> is not "BETTER" than the other is not relevant to the purpose in
>>>>>>>> either case.
>>>>>>>>
>>>>>>>> Now, when I here things like "Zero global NFB" and "Critically
>>>>>>>> matched components", I can smell the snake-oil from a great
>>>>>>>> distance, even the 10,000 miles from here to Australia. I am
>>>>>>>> sure that process also contains descriptives of "interconnects"
>>>>>>>> rolled on the thighs of virgins on Walpurgis Night...
>>>>>>>>
>>>>>>>> Note that even "critically matched" solid-state components drift
>>>>>>>> after a very short period of time in-service. All of them, such
>>>>>>>> that that "less than 1%" is meaningful for perhaps 12 hours or so.
>>>>>>>>
>>>>>>>> Being as this is a hobby for me, I get to try things that are
>>>>>>>> otherwise unproductive, unprofitable or impractical. Such as
>>>>>>>> shotgunning a device with single-value capacitors and then
>>>>>>>> comparing it to the same device with carefully screened and
>>>>>>>> matched caps. Or matching driver and output transistors and
>>>>>>>> comparing to a similar device with disparate values. Guys and
>>>>>>>> gals - you would be seriously shocked to discover how little
>>>>>>>> difference some things make that the ALL-SEEING, ALL-KNOWING
>>>>>>>> gurus will tell you are critical. Often no difference at all.
>>>>>>>
>>>>>>> Thanks for your input Peter. If I may ask, do you have an opinion
>>>>>>> on 'storage capacitors' on an amplifier power supply? What in
>>>>>>> your opinion is 'better', a single (or few) very large caps or
>>>>>>> multiple smaller caps to the same / similar capacitance?
>>>>>>>
>>>>>>> I have a long term project building my own amp based on PCBs
>>>>>>> taken from 100w MOSFET (two pairs of J50 / K135 devices per amp)
>>>>>>> PA amps made by a New Zealand company in the 1980s. (Craft, Gary
>>>>>>> Morrison's company before he went on to become head designer at
>>>>>>> Plinius until 2005 when he left to set up Pure Audio). I got my
>>>>>>> hands on a rack of four of these mono amps and preliminary
>>>>>>> testing using a clean source and good speakers suggest they will
>>>>>>> make a great stereo amp.
>>>>>>>
>>>>>>> I need to put together a power supply to feed two of these and
>>>>>>> have some new 10,000uF caps but was wondering if multiple smaller
>>>>>>> caps would be better. (In the PA amps they only had 2,200uF but
>>>>>>> obviously weren't called on to reproduce much bass.)
>>>>>>>
>>>>>>> As it is I'll be using fly leads from the rectifier PCB to the
>>>>>>> caps, then to the amps and I'm building my own case. I was
>>>>>>> thinking of maybe using my 10,000uF caps as well as maybe some
>>>>>>> smaller ones, perhaps 1,000 in a bank, the best of both worlds.
>>>>>>> (There are also 100uF electros across the rails on the amp PCBs
>>>>>>> that I'll be replacing.) That said I could also just go to multiple
>>>>>>>
>>>>>>> Cheers,
>>>>>>
>>>>>> **Those old MOSFETs were pretty ordinary devices (not very
>>>>>> linear). Evidenced by the fact that Plinius amps have always used
>>>>>> BJTs. As Peter has stated, multiple small value caps will usually
>>>>>> provide a superior, higher speed power supply. However, I would
>>>>>> posit that those old MOSFETs are so horrible (modern MOSFETs are
>>>>>> far superior), that it may not be worth the effort.
>>>>>
>>>>> I hooked a pair of them up to a preamp while still using their
>>>>> original power supplies and was very pleased with the sound so
>>>>> decided to go ahead with the build.
>>>>
>>>> **I haven't listened to Craft (hi fi) amps in many years. What I
>>>> heard back then was pleasing. Very wide bandwidth (ca. 1MHz), as I
>>>> recall.
>>>>
>>>>>
>>>>>> Craft amps used huge amounts of global NFB, required due to very
>>>>>> low bias currents and the necessity to reduce the huge levels of
>>>>>> distortion caused by the 'knee' at low currents (A Class A, or
>>>>>> high bias MOSFET amp would have been much better). Anyway, the
>>>>>> huge levels of global NFB means that PSRR (Power Supply Rejection
>>>>>> Ratio) will be quite high, thus the influence of power supply
>>>>>> changes will be relatively small.
>>>>>
>>>>> Unfortunately I don't own a 'scope so am unable to check a lot of
>>>>> stuff. When I listened to them with the original power supplies
>>>>> (designed for PA use) they sounded sweet and clean at low and
>>>>> moderate volume levels but seemed to run out of power at higher
>>>>> volumes, especially when there was a lot of bass.
>>>>
>>>> **That could be due to a number of factors. Including:
>>>>
>>>> * Insufficient Voltage output.
>>>> * Insufficient current output.
>>>> * Insufficient power supply.
>>>> * An unreasonable speaker impedance.
>>>>
>>>> Don't forget: Those meaty looking 2SJ50/2SK135 output devices are
>>>> only rated for a meagre 7 Amps each and 100 Watts PDiss. By
>>>> comparison, a typical output BJT of the same time period was rated
>>>> at a far more respectable 20 Amps and 200 Watts PDiss
>>>> (MJ15003/MJ15004). Present production variants are rated at 25 Amps
>>>> and 250 Watts.
>>>
>>> So three pairs per side should be fine for a reasonably powerful amp?
>>
>> **Again: It depends on the maximum Voltage output. 3 pairs allows for
>> a peak current ability of 21 Amps.
>>
>>> I've studied the PCB and the output devices are paralleled (along
>>> with a resistor for each) so it wouldn't be hard to add a third
>>> device to each (on very short flyleads - or even daughterboards -
>>> mounted to the same heatsink).
>>
>> **Sure. However, make certain the drive circuitry can cope.
>
> I'm not exactly sure of how to do that?

**You need to examine the drive circuitry, the components used and then
calculate if those components can cope with the extra load caused by
extra MOSFETs. It will PROBABLY be OK, but I don't know.

>
>>> The speakers I'm intending to use with this are Sony SS-K90EDs.
>>> Like these:
>>> <https://www.stereo.net.au/forums/topic/260972-fs-sony-ss-k90ed-speakers-rare/>
>>
>>
>> **OK.
>>
>>>
>>>
>>>> So, a little Ohm's Law should tell you if you are demanding more
>>>> current than the output devices are capable of delivering. 14 Amps
>>>> is, by high end audio standards, a relatively modest current ability
>>>> for a (say) 100 Watt @ 8 Ohms amplifier. Provided the driver
>>>> impedance is relatively benign, you should be OK. Fortunately, it is
>>>> real hard to damage MOSFETs, by 'asking' them to deliver more
>>>> current than they are rated for.
>>>
>>> That's one of the things I like about MOSFETs.
>>
>> **Well, a properly designed BJT amp should demonstrate the same
>> robustness and reliability.
>>
>>>
>>>>>> One more thing: Decent amounts of capacitance placed close to the
>>>>>> output devices is far more influential than caps placed some
>>>>>> distance away. In fact, long(ish) cables AFTER the main filter
>>>>>> caps can be a serious limiting factor on the effectiveness of a
>>>>>> power supply in a Class A/B amplifier. This is because the
>>>>>> inductance of the wires can be a factor.
>>>>>
>>>>> Thanks. The fly-leads will only be 6" tops and I'll be using at
>>>>> least 1.5 square mm multistrand copper conductors. If space allows
>>>>> I'll put a ~1,000uF cap right at the amplifier PCB as well (or as
>>>>> large as I can get away with). I may end up building a wooden case
>>>>> as I don't have a suitable metal one and wood's something I have
>>>>> experience and the tools for.
>>>>
>>>> **Wiring sounds good. And yeah, caps placed close to output devices
>>>> is a very good thing. A wooden case, not so much. Wood is an
>>>> excellent thermal insulator, which means heat may not escape too
>>>> easily.
>>>
>>> I have a couple of big heatsinks for the amplifier modules that will
>>> sit either side of the case, fins outwards in free air. They'll
>>> easily handle the power dissipation being 4x bigger than the 'sinks
>>> used on the PA amp. Also I'll ventilate the top and bottom of the
>>> 'box' (if I end up going with wood).
>>
>> **OK.
>>
>>>
>>>>> I still haven't finalised my design yet. I might end up feeding
>>>>> them a few more volts than they were getting from their original
>>>>> power supplies (my only suitable toroidial transformer is 10v AC
>>>>> higher than original) so may parallel up a third pair of output
>>>>> devices onto the heatsinks using one of the other amps as a donor.
>>>>> I haven't decided yet, as I said it's a long-term project and I'm
>>>>> learning as I go.
>>>>
>>>> **Well, the MOSFETs are rated for a decent 160 Volts, so a few more
>>>> rail Volts should be OK. And yes, more output devices won't hurt
>>>> (refer to Ohm's Law as before). Pay attention to the drive
>>>> capabilities of the preceding stages though.
>>>
>>> Thanks for this Trevor, I have saved it for future reference. My 300
>>> VA toroid that I'm thinking of using with this outputs 50v AC so +/-
>>> 70v DC when rectified. The original PA transformers were 40v AC.
>>
>> **+/- 70VDC suggests a maximum power output of around 250 Watts @ 8
>> Ohms. If you plan on attempting to obtain that much power
>> (continuously), then you will need two of those toroids.
>
> I intend to use the system in my lounge so won't want crazy SPLs, the
> speakers likely wouldn't handle that much power anyway. I actually do
> have two of the toroids but that would make for a big amplifier case -
> and surely then I'd need to consider adding *two* more pairs of output
> MOSFETs per amplifier?

**As Peter has correctly stated, provided you don't need the full
continuous power capacity of the amplifier at all times, then one
transformer will likely be plenty. From my perspective, I am a purist.
If I am presented with an amplifier rated at (say) 200 Watts/channel,
then that amplifier needs to be able to deliver 200 Watts/channel
INDEFINITELY and, possibly more importantly, it needs to be able to
deliver roughly 40% of it's maximum power without thermal distress. With
one transformer in your amplifier chassis, it would fail such a test.
But, your amplifier is not a commercial item. You can make it anything
you want.

>
> I was thinking that, as I don't listen to dubstep or extremely
> bass-heavy music, using one toroid and a lot of capacitance (in the
> region of 20,000 to 50,000 uF per rail) would be enough to handle
> transients. If not then I might as well build a pair of monoblocks.

**A worthy consideration.

>
> I've got a few coffee-cup sized Mepco/Electra 14,000 uF / 100v caps but
> they're not new... I also have 8 new 10,000 uF / 100v Elna caps that are
> only about 1/4 of the size.

**The amplifier I presently use has a 5.5kVA (yes, 5,500VA), split wound
(one winding for each channel), double C core power transformer,
followed by 92 X 3,300uF filter capacitors. The result is to ensure
that, under full power operation (at any impedance higher than 2 Ohms)
ripple is kept below 100mV. So, discussions of 10,000uF per rail doesn't
excite me. It's what I expect to see in a mass market product from
Yamaha or NAD. However, as Peter and I have both suggested, in a high
global NFB amp, such as yours, huge lumps of filter capacitance will not
be pivotal to performance. Placing a decent amount near the output
devices will be beneficial though.


--
Trevor Wilson
www.rageaudio.com.au

---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus

While much of what you say below is, strictly speaking, technically
correct under some very specific cases, it's misleading and could well
not be applicable in reality.

I've not got a lot of time to spend on this, so let me just
take on a couple of your points.

On Monday, September 16, 2019 at 5:52:54 AM UTC-4, Trevor Wilson wrote:
> On 16/09/2019 7:49 am, wrote:
> > On Saturday, September 14, 2019 at 9:58:44 AM UTC-4, Trevor Wilson wrote:
> >> So, a little Ohm's Law should tell you if you are demanding more current
> >> than the output devices are capable of delivering. 14 Amps is, by high
> >> end audio standards, a relatively modest current ability for a (say) 100
> >> Watt @ 8 Ohms amplifier. Provided the driver impedance is relatively
> >> benign, you should be OK.
> >
> > Hmm, that's not what a little Ohm's law tells me.
> >
> > 100 Watts into 8 Ohms is a tad over 3.5 amps. Let's say it's a VERY
> > robust 100 watt amplifier, delivering 200 Watts into 4 Ohms requires
> > about 7 amps, and, let's pretend it has essentially ZERO output
> > impedance and an effectively limitless power supply, you're not reaching
> > 14 amps until you're driving 400 watts into 2 ohms.
>
> **Well, no. The RMS current is certainly 3.5 Amps, but output devices
> only 'care' about PEAK currents. The peak current is, of course, 3.5 X
> 1.414 ~ 5 Amps.
>
> With a 4 Ohm load, the peak current required is 10 Amps. For 2 Ohms, it
> is 20 Amps.
>
> Assuming a 100 Watt amp. For a (say) 200 Watt amp, those peak current
> figures become 7 Amps, 14 Amps and 28 Amps respectively. WAY past the
> ability of two pairs of old Hitachi MOSFETs to deal with.

All of your calculations ASSUME several things, none of which are
likely to be true in real usage.

1. Your continuous-to-peak current calculations using a factor of
sqrt(2) ASUMMES that the excitation is pure sine. That's surely
not the case in real life, I'm sure you'd agree. Yes, the actual
crest fact may be greater than 3dB, but, again, you ASSUME that
in actual practice those peak current are REQUIRED> I suggest
they are not, in absence of any supporting evidence they are.

2. The comprehensive list of impedance curves is, yes, useful but
the interpretation of them as applied to your case ignores the
VERY important details and thus is overly simplistic and mis-
leading. Let me take just a couple of example from the list:

a. Westlake BBSM-6F
Yes, the impedance gets to 2 ohms, but look at the
broadband sensitivity 92dB/2.83V: less power is needed to
achieve a given sound pressure level, continuous, peak or
otherwise. It illustrates that you simply can't look at one
particular measurement in isolation.

And, not that it may be relevant, this is specifically
marketed towards studio use at high (deafening?:-) level.

b. The acoustat impedance curve you posted on the RageAudio
site: indeed, the impedance drops WAY down to under 1 Ohm.
BUT it does it over a VERY narrow bandwidth, and it does
it at 15 kHz. It's above 4 ohms over the entire audio range
from 10Hz to 9kHz. Exactly what kind of musical material
would require one to dump a LOT of power between 9 kHz and
20+kHz? Over the majority of the audio bandwidth, the impedance
is 6 ohms or higher. Even if you assume (quite unrealistically)
that the energy is distributed across the spectrum, equal
energy per octave, your impedance problems ar confined to about
1 octave out of 10, suggesting that if you need 100 watts in that
one octave, you'll need another 900! for everything else.

Ah, but what about short-term transients, you ask. Look at the
spectral distribution of such in actual music: sorry, you're
still NOT generating a lot of power over such a narrow bandwidth
(Sorry, Mr. Fourier, you can lay back down, we shan't be needing
you just yet).


> >
> > So, a couple of questions that Mr. Ohm may ask; what kind of loudspeaker
> > presents a broadband 2 ohm impedance or, conversely, what kind of
> > musical content would generate that kind of power requirement over
> > the pretty narrow band of frequencies where a loudspeaker has the
> > kind of pathological impedance curve that would dip to as low as
> > 2 ohms.
>
> **I have a few here that are tougher than that. Some of the Peerless
> XXLS drivers dip to the low 2 Ohm region. Most ESLs fall lower than that
> at HF.

Yes, over VERY narrow bands and at high frequencies, where your
power requirements are not anywhere near as large as at significantly
lower frequencies.

On 17/09/2019 11:54 AM, Trevor Wilson wrote:
> On 16/09/2019 11:01 pm, ~misfit~ wrote:
>> On 16/09/2019 9:53 PM, Trevor Wilson wrote:
>>>
>>> **Sure. However, make certain the drive circuitry can cope.
>>
>> I'm not exactly sure of how to do that?
>
> **You need to examine the drive circuitry, the components used and then calculate if those
> components can cope with the extra load caused by extra MOSFETs. It will PROBABLY be OK, but I
> don't know.
>
>>
>> I intend to use the system in my lounge so won't want crazy SPLs, the speakers likely wouldn't
>> handle that much power anyway. I actually do have two of the toroids but that would make for a
>> big amplifier case - and surely then I'd need to consider adding *two* more pairs of output
>> MOSFETs per amplifier?


[ Massively pruned of old quotes because your mod just
couldn't take it any more. --dsr ]



>
> **As Peter has correctly stated, provided you don't need the full continuous power capacity of the
> amplifier at all times, then one transformer will likely be plenty. From my perspective, I am a
> purist. If I am presented with an amplifier rated at (say) 200 Watts/channel, then that amplifier
> needs to be able to deliver 200 Watts/channel INDEFINITELY and, possibly more importantly, it needs
> to be able to deliver roughly 40% of it's maximum power without thermal distress. With one
> transformer in your amplifier chassis, it would fail such a test. But, your amplifier is not a
> commercial item. You can make it anything you want.
>
>>
>> I was thinking that, as I don't listen to dubstep or extremely bass-heavy music, using one toroid
>> and a lot of capacitance (in the region of 20,000 to 50,000 uF per rail) would be enough to
>> handle transients. If not then I might as well build a pair of monoblocks.
>
> **A worthy consideration.
>
>>
>> I've got a few coffee-cup sized Mepco/Electra 14,000 uF / 100v caps but they're not new... I also
>> have 8 new 10,000 uF / 100v Elna caps that are only about 1/4 of the size.
>
> **The amplifier I presently use has a 5.5kVA (yes, 5,500VA), split wound (one winding for each
> channel), double C core power transformer, followed by 92 X 3,300uF filter capacitors.

Wow! Decades ago I used to work with a touring band doing stage lighting and (some) sound mixing
and that's a more capable amplifier power supply than were in some amps we used at medium-sized gigs.

> The result
> is to ensure that, under full power operation (at any impedance higher than 2 Ohms) ripple is kept
> below 100mV. So, discussions of 10,000uF per rail doesn't excite me. It's what I expect to see in a
> mass market product from Yamaha or NAD. However, as Peter and I have both suggested, in a high
> global NFB amp, such as yours, huge lumps of filter capacitance will not be pivotal to performance.
> Placing a decent amount near the output devices will be beneficial though.

Thanks for your input Trevor. It helps me to decide how to go about building my 'franken-amp'.
--
Shaun.

"Humans will have advanced a long, long way when religious belief has a cozy little classification
in the DSM"
David Melville

This is not an email and hasn't been checked for viruses by any half-arsed self-promoting software.

On 19/09/2019 8:00 pm, wrote:
> While much of what you say below is, strictly speaking, technically
> correct under some very specific cases, it's misleading and could well
> not be applicable in reality.
>
> I've not got a lot of time to spend on this, so let me just
> take on a couple of your points.
>
> On Monday, September 16, 2019 at 5:52:54 AM UTC-4, Trevor Wilson wrote:
>> On 16/09/2019 7:49 am, wrote:
>>> On Saturday, September 14, 2019 at 9:58:44 AM UTC-4, Trevor Wilson wrote:
>>>> So, a little Ohm's Law should tell you if you are demanding more current
>>>> than the output devices are capable of delivering. 14 Amps is, by high
>>>> end audio standards, a relatively modest current ability for a (say) 100
>>>> Watt @ 8 Ohms amplifier. Provided the driver impedance is relatively
>>>> benign, you should be OK.
>>>
>>> Hmm, that's not what a little Ohm's law tells me.
>>>
>>> 100 Watts into 8 Ohms is a tad over 3.5 amps. Let's say it's a VERY
>>> robust 100 watt amplifier, delivering 200 Watts into 4 Ohms requires
>>> about 7 amps, and, let's pretend it has essentially ZERO output
>>> impedance and an effectively limitless power supply, you're not reaching
>>> 14 amps until you're driving 400 watts into 2 ohms.
>>
>> **Well, no. The RMS current is certainly 3.5 Amps, but output devices
>> only 'care' about PEAK currents. The peak current is, of course, 3.5 X
>> 1.414 ~ 5 Amps.
>>
>> With a 4 Ohm load, the peak current required is 10 Amps. For 2 Ohms, it
>> is 20 Amps.
>>
>> Assuming a 100 Watt amp. For a (say) 200 Watt amp, those peak current
>> figures become 7 Amps, 14 Amps and 28 Amps respectively. WAY past the
>> ability of two pairs of old Hitachi MOSFETs to deal with.
>
> All of your calculations ASSUME several things, none of which are
> likely to be true in real usage.
>
> 1. Your continuous-to-peak current calculations using a factor of
> sqrt(2) ASUMMES that the excitation is pure sine. That's surely
> not the case in real life, I'm sure you'd agree. Yes, the actual
> crest fact may be greater than 3dB, but, again, you ASSUME that
> in actual practice those peak current are REQUIRED> I suggest
> they are not, in absence of any supporting evidence they are.

**Whoaa. Hang on a sec. Let's get down to basics. The peak to average
figure available in music is not the issue here and is certainly not
what I am talking about.

Let us assume, for the sake of simplicity, that the music played IS
close to pure sine waves (maybe the listener is a pipe organ lover). Now
I certainly recognise that 99.9999% of music is not sine waves
(apologies to Fourier), but I need to make an important point. Let's say
we have (again, for simplicity and ignoring SOA considerations for a
moment) an amplifier that uses the most excellent Toshiba
2SC5200/2SA1943 output devices. Let's say an RMS output current of 15
Amps is required. This would not be healthy for the output devices, as
their maximum current rating is 15 Amps (peak). Yes, it can
(theoretically, SOA considerations aside) deliver a DC current of 15
Amps. However, a sine wave of 15 Amps means that the peak current will
be in excess of 21 Amps. Examine the SOA curve of the 2SC5200:


https://toshiba.semicon-storage.com/...l.2SC5200.html

You may care to note that even if the current flowing through the device
exceeds 15 Amps for as little as ONE MILLISECOND (in truth it is far
less), the device will be destroyed.

It is for this reason that peak current MUST be taken into account when
designing any solid state amp.

>
> 2. The comprehensive list of impedance curves is, yes, useful but
> the interpretation of them as applied to your case ignores the
> VERY important details and thus is overly simplistic and mis-
> leading. Let me take just a couple of example from the list:
>
> a. Westlake BBSM-6F
> Yes, the impedance gets to 2 ohms, but look at the
> broadband sensitivity 92dB/2.83V: less power is needed to
> achieve a given sound pressure level, continuous, peak or
> otherwise. It illustrates that you simply can't look at one
> particular measurement in isolation.

**Indeed. I can assure you, however, having owned a pair for some years,
that they are utterly ruthless in exposing an amplifier's ability to
deal with tough loads. Even at modest listening levels.

>
> And, not that it may be relevant, this is specifically
> marketed towards studio use at high (deafening?:-) level.

**I can assure you that I do not and have not, for many years listened
to music at silly levels. Back when I was in my teens and early 20s,
yes. But not now and not when I owned the Westlakes.

>
> b. The acoustat impedance curve you posted on the RageAudio
> site: indeed, the impedance drops WAY down to under 1 Ohm.
> BUT it does it over a VERY narrow bandwidth, and it does
> it at 15 kHz. It's above 4 ohms over the entire audio range
> from 10Hz to 9kHz. Exactly what kind of musical material
> would require one to dump a LOT of power between 9 kHz and
> 20+kHz? Over the majority of the audio bandwidth, the impedance
> is 6 ohms or higher. Even if you assume (quite unrealistically)
> that the energy is distributed across the spectrum, equal
> energy per octave, your impedance problems ar confined to about
> 1 octave out of 10, suggesting that if you need 100 watts in that
> one octave, you'll need another 900! for everything else.
>
> Ah, but what about short-term transients, you ask. Look at the
> spectral distribution of such in actual music: sorry, you're
> still NOT generating a lot of power over such a narrow bandwidth
> (Sorry, Mr. Fourier, you can lay back down, we shan't be needing
> you just yet).

**Oh, I agree, you won't be generating much power. However, VI limiting
systems in most amplifiers don't care about such things. They just
operate anyway. Mostly. And to the detriment of sound quality. And, more
seriously, when using almost any valve amplifier with the Accustats,
will guarantee a suck-out at the impedance dip. That is likely to be
audible.

>
>
>>>
>>> So, a couple of questions that Mr. Ohm may ask; what kind of loudspeaker
>>> presents a broadband 2 ohm impedance or, conversely, what kind of
>>> musical content would generate that kind of power requirement over
>>> the pretty narrow band of frequencies where a loudspeaker has the
>>> kind of pathological impedance curve that would dip to as low as
>>> 2 ohms.
>>
>> **I have a few here that are tougher than that. Some of the Peerless
>> XXLS drivers dip to the low 2 Ohm region. Most ESLs fall lower than that
>> at HF.
>
> Yes, over VERY narrow bands and at high frequencies, where your
> power requirements are not anywhere near as large as at significantly
> lower frequencies.

**The Peerless XXLS drivers are LF drivers, requiring prodigious currents:

http://www.loudspeakerdatabase.com/P...LS-P835017.pdf

I have passive radiator based subwoofer in the workshop right now. I
will run a real world impedance plot for the system when I get some time.


--
Trevor Wilson
www.rageaudio.com.au

---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus

On 20/09/2019 10:52 PM, ~misfit~ wrote:
> On 17/09/2019 11:54 AM, Trevor Wilson wrote:
>> On 16/09/2019 11:01 pm, ~misfit~ wrote:
>>> On 16/09/2019 9:53 PM, Trevor Wilson wrote:
>>>>
>>>> **Sure. However, make certain the drive circuitry can cope.
>>>
>>> I'm not exactly sure of how to do that?
>>
>> **You need to examine the drive circuitry, the components used and then calculate if those
>> components can cope with the extra load caused by extra MOSFETs. It will PROBABLY be OK, but I
>> don't know.
>>
>>>
>>> I intend to use the system in my lounge so won't want crazy SPLs, the speakers likely wouldn't
>>> handle that much power anyway. I actually do have two of the toroids but that would make for a
>>> big amplifier case - and surely then I'd need to consider adding *two* more pairs of output
>>> MOSFETs per amplifier?
>
>
> Â*Â* [ Massively pruned of old quotes because your mod just
> Â*Â*Â*Â* couldn't take it any more. --dsrÂ*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* ]

Thanks. I actually seriously considered doing the same but, as a relative newbie here didn't want
to upset any pie-carts.

[ Always feel free to trim quotes in your reply down to
the points you are actively addressing. Usenet is
archived all over the world, it's really easy to
reconstruct threads. --dsr ]



--
Shaun.

"Humans will have advanced a long, long way when religious belief has a cozy little classification
in the DSM"
David Melville

This is not an email and hasn't been checked for viruses by any half-arsed self-promoting software.

On 12 Sep 2019 20:28:33 GMT, Trevor Wilson wrote:

<snip>
>
>**Let me be very clear about several things:
>
>* NFB is fine. In fact, NO audio amplifier can work without it.
>* GLOBAL NFB is also fine. When properly applied.
>* I have a personal preference for the amplifiers I use, which employ
>lots of local NFB and no global NFB. Others may have a different opinion.
>* As part of my education into the world of zero global NFB amplifiers,
>I subjected myself to a couple of single (unfortunately) blind tests,
>between two, otherwise identical, amplifiers. One employed zero GNFB and
>one employed a modest amount of GNFB. I preferred the zero GNFB one.
>Since that time, I subjected several (10) of my clients to the same test
>(DBT). The zero GNFB models was preferred every time. Except one.
>* Once mo I would posit that part of the reason why some listeners
>prefer valve amplifiers, is due to the fact that global NFB levels are
>very low, or non-existent.
>
>
> But, again, in the real world, negative feedback, done properly, has
>many more advantages than disadvantages.
>
>**Again: No issue with NFB. In fact, no issue with GNFB, when done well.

How do modern switching amps (class D) stack up for HiFi use? Aren't
most PA systems now fully digital? Do they actually use FB? If I look
at the spec sheet of the TDA7492 it doesn't look like it. Do they
sound worse than a good analog amp?

The class-D amps typically have a series inductance between the
switching elements and the speakers, does that influence transients?
Even a tweeter has already 15-20 microHenry of inductance.

Mat Nieuwenhoven

On 14/10/2019 8:00 am, Mat Nieuwenhoven wrote:
> On 12 Sep 2019 20:28:33 GMT, Trevor Wilson wrote:
>
> <snip>
>>
>> **Let me be very clear about several things:
>>
>> * NFB is fine. In fact, NO audio amplifier can work without it.
>> * GLOBAL NFB is also fine. When properly applied.
>> * I have a personal preference for the amplifiers I use, which employ
>> lots of local NFB and no global NFB. Others may have a different opinion.
>> * As part of my education into the world of zero global NFB amplifiers,
>> I subjected myself to a couple of single (unfortunately) blind tests,
>> between two, otherwise identical, amplifiers. One employed zero GNFB and
>> one employed a modest amount of GNFB. I preferred the zero GNFB one.
>> Since that time, I subjected several (10) of my clients to the same test
>> (DBT). The zero GNFB models was preferred every time. Except one.
>> * Once mo I would posit that part of the reason why some listeners
>> prefer valve amplifiers, is due to the fact that global NFB levels are
>> very low, or non-existent.
>>
>>
>> But, again, in the real world, negative feedback, done properly, has
>> many more advantages than disadvantages.
>>
>> **Again: No issue with NFB. In fact, no issue with GNFB, when done well.
>
> How do modern switching amps (class D) stack up for HiFi use?

**Provided the switching frequency is high enough and the load impedance
is benign, then they should work well. I was particularly impressed with
the interesting design from Devialet. A small Class A stage is used in
much the same way that Quad did several decades ago for their Current
Dumpingâ„¢ products. They used a Class A stage, combined with a Class C
power stage.

Aren't
> most PA systems now fully digital?

**No such thing.

Do they actually use FB?

**EVERY amplifier uses NFB. Every single one. Regardless of technology
or claims from manufacturers.

If I look
> at the spec sheet of the TDA7492 it doesn't look like it. Do they
> sound worse than a good analog amp?

**I see a loop feedback mechanism in the block diagram. I see some
audibly significant problems with the amplifier. Max THD is cited aas
0.4% and the frequency response is poor, compared to even modest Class
A/B amplifiers. The low switching frequency ensures that low impedance
(<4 Ohms) loads are not well catered for.

>
> The class-D amps typically have a series inductance between the
> switching elements and the speakers, does that influence transients?

**Of course.

> Even a tweeter has already 15-20 microHenry of inductance.
>
> Mat Nieuwenhoven
>
>
>
>
>
>


--
Trevor Wilson
www.rageaudio.com.au

--
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus

On 14/10/2019 8:00 am, Mat Nieuwenhoven wrote:

> Even a tweeter has already 15-20 microHenry of inductance.

**Not so. I haven't measured one in quite some time, but the EMIT HF
drivers, used in many Infinity speakers exhibit far lower inductance
figures than that. If I had to guess, I'd estimate the inductance figure
to be around 5 X 10^-6H. I'll see if I can locate one to measure. Then,
of course, is the sadly deleted Audax HD-3P Piezo HF driver. And any
number of ELS HF drivers.


--
Trevor Wilson
www.rageaudio.com.au

--
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus