composite amplifiers

My source for this circuit is almost 20 years old*, so a forum and web search didn't turn up anything about it. I'm soliciting comments, opinions, and possible mods using other op amps or more powerful chip amps. I'm attracted to its elegant simplicity. And the numbers ain't bad either:
compamp.jpg

10 watts RMS into 8 ohms ____________ >0.003% THD @ 1kHz

* Charles Kitchin, Scott Wurcer, and Jeff Smith. Composite Audio Power Amplifiers. Electronics Now, Nov 1992: 38-44
 
Hi,

Unless I'm mistaken very low input impedance circa 1K and very poor gain, x2, I
cannot see the point at all, cannot see it being remotely stable either, very poor.

rgds, sreten.

The below avoids bipolar electrolytics in the signal path, gain is x20 :
 

Attachments

  • Untitled (1).png
    Untitled (1).png
    21.2 KB · Views: 2,735
Last edited:

jcx

Member
2003-02-17 7:38 pm
..
the OP circuit "throws away" ~36 dB of gain in the 4k/65 divider - rather inefficient use of potential loop gain - and it still has a region of 2nd order loop gain slope giving ~ 0 phase margin

composite/multiloop amps with op amp added inside global loop with audio power chip amp is difficult given the limitations of achieving composite amp global stability when you may have to assume worst case loop intercept for the chip amp as low as 200 kHz you have limited added loop gain @ 20 kHz

if you stick to Bode's Maximum Loop Gain you only get ~10 dB added gain at 20 kHz, adding more loop gain requires allowing conditionally stable compensation and consequent possible clipping recovery sticking/oscillation

an interesting option is to use Black’s Feedforward to improve higher frequencies while relying on added loop gain at lower frequencies

the resulting complexity may make you wonder why you aren’t just building a discrete amp with 30+ MHz output Q from Cordell, Self books
 
Last edited:
There have been a few small discussions on this type of design and they are very hard to find even after doing some major digging on the subject.

I have the original articles published back in that era somewhere.

It is funny that you should bring it up as I was just thinking about such a design today.

Yesterday I had finaly found my copy of a Randall RG300 300watt power amplifier that I have had for 30 years,I had thought that I had lost it forever as web searches had turned up nothing.

I have seen a DIY stereo version of this amp in action as a small PA amp for a band.
It is super simple and very heavy duty as it uses 6 X 2N3773's in the output stage and is rated at 400 watts into 1 ohm.
There are only four transistor types in the whole circuit,TIS97,TIP31c,TIP32c and 2N3773.
So my idea was to use a Composite input setup in order to bring its performance up to date although it performed nicely the way it is from what I remember.

This type of configuration is supposed to make the overall performance of the amplifier as the same as the opamp in the first stage.

jer :)
 
sreten, what is the source of the circuit you have there? It's roughly the same thing my circuit shows. This article also shows a non-inverting version with the input and feedback resistors removed and the source brought to the + input on a 1M resistor to ground.
There's also 33, 40, and 70 watt versions, repeating this basic "module". I agree that it probably needs a strong input signal and attention to grounding and layout, but with just 5 passives and some decoupling caps this 10W amp just seems too tempting not to give it a shot. If I can't get it to behave, a little more soldering will give me that 3-stage circuit.
 

jcx

Member
2003-02-17 7:38 pm
..
sreten's circuit is simply inverting mode for the power chip amp - the only other "improvement" from the op amps is the DC servo

buffering, inversion gives a hi Z positive input

this is essentially the same as Cordell's "Super Gain Clone" - page 540 of his book

I consider it a very feeble attempt at really improving the chip amp audio performance

the only chip amp distortion mechanism addressed is the nonlinear common mode input Z

the bigger thermal coupling distortion is better fixed by the OP circuit with the input op amp inside the global loop but physically separate from the chip amp and its high power, large temperature excursions

also the OP circuit input op amp is adding to the global loop gain below ~50 kHz - this added loop gain reduces all distortions of the chip amp

the input op amp is driving a relatively high 4k Ohm load, and only to a fraction of the output V at audio frequencies - its not working very hard at all so should have low distortion
 
Last edited:
The AD711 is part of the magic with this circuit. According to the article, it seems to be particularly well-suited for this application. Since Mr. Wurcer designed the opamp, if I'm not mistaken, I figure he ought to know.
The output is meant to mirror the performance of the opamp, so the LM1875's datasheet THD of 0.015% gets reduced significantly. My motivation is simply that I have the chip amps on hand and am looking to use 'em. If this puppy needs a good preamp, I have a LME49710-based diy unit ready and waiting.
Thank you for your responses, which I will continue to consider. But don't blow your credibility with talk of low-gain inefficiencies and thermal coupling, please.
 
Hi,

I cannot see why the gain is so low. If x10ing the values of R2 and R4
it would make more sense to to me. The LM1875 is stable with with
closed loop gains of 10 or greater, x2 = oscillation guaranteed AFAICS.

The usual 1R/0.22uF output loading is missing. No reason the chip-amp
should be run off +/- 18V, +/-25V is better, but too high for the op-amp.

rgds, sreten.

I really can't see the "magic" of the circuit at all .....
 
Last edited:
Don't get me wrong. I am very wary of the circuit's stability. But I'll probably give it a try anyway. The network on the output of the opamp is intended to stabilize the response, and a table is provided showing different values and their effect on phase margin and bandwidth. If things go well, I may try higher ps voltage for the LM1875, but I'd be happy with 10W.
 

jcx

Member
2003-02-17 7:38 pm
..
interesting method of encouraging "...comments, opinions, and possible mods"

nothing "magic" about the op amp - the AD711 is barley an improvement over LF411, TL070

OPA134 beats it by >2x on noise, GBW

if you want to talk about "magic" jfet op amps for audio you should be thinking OPA627 and the last decade's devices aimed at it like AD8610, ADA4627, OPA827


the OP composite amp circuit isn't particularly sophisticated in compensation, use of the 2 amps worth of gain - Scott Wurcer's name doesn't help its performance at all as far as I can see

Walt Jung, Gerald Graeme, Bob Pease, Jim Williams have all written about various op amp deficiencies and ways to help out



but since SofaSpud can dismiss their ideas so easily as having poor credibility I guess all we can do sit back and await enlightenment
 
The last post showed up at the same time as mine, so I missed it.
I hadn't said anything at all about "magic" jfet op amps.
I mentioned Mr. Wurcer only because I believe he created the AD711, the composite circuit and variations, and the article I was using for reference. Sounds like credibility to me. YMMV.
But I won't suggest what anyone can do. Even though I'm strongly tempted.
I went back to this thread today because last night I was doing some light reading before bed. I happened upon some very relevant info in Jerald Graeme's book, Optimizing Op Amp Performance. Specifically, oscillation conditions in composite amps and how to prevent them. The conditions, not the amps :)
Not a bad book. It could find a place in the forum book recommendations.
 
Last edited:

jcx

Member
2003-02-17 7:38 pm
..
The AD711 is part of the magic with this circuit. According to the article, it seems to be particularly well-suited for this application. Since Mr. Wurcer designed the opamp, if I'm not mistaken, I figure he ought to know.

I see the word "magic" - and I see an appeal to authority - a fine one for op amp design but still needs to be tempered with knowledge of what other experts have written, Walt Jung's name ring a bell there? – I have spent a few decades designing composite op amp circuits in precision instrumentation and may know of a few more experts, articles, and have experience from my own analysis, design, debugging

...Thank you for your responses, which I will continue to consider. But don't blow your credibility with talk of low-gain inefficiencies and thermal coupling, please.

a little condensation is always a good way to encourage others to explain, be helpful - particularly when it reveals underlying ignorance coupled to the arrogance

The last post showed up at the same time as mine, so I missed it.
I hadn't said anything at all about "magic" jfet op amps.
I mentioned Mr. Wurcer only because I believe he created the AD711, the composite circuit and variations, and the article I was using for reference. Sounds like credibility to me. YMMV.
But I won't suggest what anyone can do. Even though I'm strongly tempted.

I do have a suggestion - read carefully, ask for clarification if you don't recognize something

I went back to this thread today because last night I was doing some light reading before bed. I happened upon some very relevant info in Jerald Graeme's book, Optimizing Op Amp Performance. Specifically, oscillation conditions in composite amps and how to prevent them. The conditions, not the amps :)
Not a bad book. It could find a place in the forum book recommendations.

Graeme is one of the authors I suggested - maybe there's hope - but you need to think about your posting style if you want to engage others in a dialog
 
I see the word "magic" - and I see an appeal to authority - a fine one for op amp design but still needs to be tempered with knowledge of what other experts have written, Walt Jung's name ring a bell there?
The word "magic" is often used as slang for performance. In that context, I've never known it to cause fits in others. Until now.
It was not an appeal to authority; it was backing up the qualified person who built the thing and wrote it up for RE. And you'll find Wurcer credited in many of Jung's composite amp papers.
– I have spent a few decades designing composite op amp circuits in precision instrumentation and may know of a few more experts, articles, and have experience from my own analysis, design, debugging
I've become more aware of that fact... which makes this all the more disappointing.
a little condensation is always a good way to encourage others to explain, be helpful - particularly when it reveals underlying ignorance coupled to the arrogance
Is this for me or the responders? If someone wants to dispute the output figures, that's fine. Otherwise, low gain is irrelevant. And I'm sorry, but I just can't imagine thermal coupling with the op amp being an issue. That would be one pitiful layout IMO.
I do have a suggestion - read carefully, ask for clarification if you don't recognize something
I do recognize something, but it's quite hideous.
Graeme is one of the authors I suggested - maybe there's hope
I like to think so. (Was that too brash?)
but you need to think about your posting style if you want to engage others in a dialog
Certainly a black kettle comment.
I was attracted to the 20 year old circuit in post #1 because I had parts in stock to build it, and the "dirty work" had mostly been done already. BTW, from the pictures, it looks like the working prototypes were built on perfboard.
I was also interested in information in regards to use of the technique with newer/higher power chip amps and/or other/newer FET op amps. I didn't realize that was too much to ask.
 

jcx

Member
2003-02-17 7:38 pm
..
it certainly isn't too much to Ask - but given the opportunity to ask for my reasoning when you didn't understand some of the points I presented you chose to offer ridicule

now instead of just asking for my technical reasoning you are escalating with the
I do recognize something, but it's quite hideous.
line

my use of the terms ignorance and arrogance are descriptions of you posts' tone with respect to my volunteered technical comments, not sly insinuations about your character, motives

drop the self justification, recognize you put your foot in it and move on to the technical discussion
 
Quoting the 1st reply:
cannot see the point at all,
Quoting the 2nd reply:
the resulting complexity may make you wonder why you aren’t just building a discrete amp
Quoting the 3rd reply:
It is funny that you should bring it up as I was just thinking about such a design today.
Quoting the 4th reply:
I consider it a very feeble attempt at really improving the chip amp audio performance
Quoting the 5th reply:
I really can't see the "magic" of the circuit at all .....
Quoting 6th reply:
Walt Jung, Gerald Graeme, Bob Pease, Jim Williams have all written about various op amp deficiencies and ways to help out
but since SofaSpud can dismiss their ideas so easily as having poor credibility I guess all we can do sit back and await enlightenment
Now I should "move on to the technical discussion." I'm the OP, remember? Thanks, but I'll fly this one solo.
 
The schematic in post #1 doesn't insist on running the AD711 on its own private clean regulated power, unaffected by the high current noise striking at the LM1875/LM675. However, please don't miss the opportunity to run pre sections cleanly on hi-fi amplifiers.

STK456 and TDA7294 can be run that way, with the pre on regs (shuntie, etc. . .) and not more than 220uF, while the output section's separate power taps can have large capacitance unregulated supply. They, and similar discrete amplifiers, have something very similar to your application, which is really quite like a pre with an output buffer.

This reminds me that one of the possibilities of regulated power, uses a couple of extra parts to insure a clean ground, and that's the version you're after for the pre (AD711), since the majority of the high current noise that you're trying to avoid is the big hammering coming back down the speaker return line. Contrary to legend, 100% of that noise doesn't flow into the power supply board voluntarily. No, not really. Some of that noise makes it up the ground leg of the gain divider, gets gain applied, and is the main cause of the typical hall of mirrors effect, that you have the opportunity to avoid.
 
Last edited:
They, and similar discrete amplifiers, have something very similar to your application, which is really quite like a pre with an output buffer.
Thanks, that is along the same lines as I was thinking. I hadn't thought about the separate supplies. As I mentioned, a big part of the appeal was the few passives and some decoupling caps.
This is really just a junkbox project of mine. Nothing's concrete yet. This circuit may well be too unstable for a production line, but to me that just gives it more diy appeal. Please tell me more about your clean ground concept. I could probably use some education on the C1/R1/R2 compensation network also, if you'd be so kind.
 
Thanks, that is along the same lines as I was thinking. I hadn't thought about the separate supplies. . . . Please tell me more about your clean ground concept.

The earth itself isn't a rock steady reference during an earthquake. Speaker return forces upon an amplifier board are the same as that.
The loud noise broadcast doesn't necessarily, automatically, entirely, flow into a power supply board. No. Although textbooks may insist that is the case, its just not true of AC, so I'd ask you to review what happens if you open key a multi-kilowatt transmitter in a crowded neighborhood. :) That's simply not a flow of any sort--it is an omnidirectional blast.

Nesting/composite amplifiers give the opportunity to have the pre run from regs, if you so choose, and since the pre is controlling global feedback, it can also reduce/cancel power circuit noise of the output stage (which is usually jolted/fluctuted from operating the speaker).

The concept seems to be similar to chips with 4 power pins and likewise discrete amplifiers, which all provide the same opportunity (run the pre clean). Your circuit seems to add that feature to a chip-amp that otherwise wouldn't have such a benefit due to lacking connections for it.

There are many cases where features like that aren't used, for instance the ST (TDA7294) and STK (a discrete amp hidden in a little black box) datasheets show hooking V+ small signal power to V+ output stage power and likewise hooking V- small signal power to V- output stage power, thus wasting the high resolution facility of the amplifier. That approach is unfortunately common practice with many discrete amplifiers as well. As usual, the only reason to do it wrong is for reduced costs.

In the recent newsletter, Jan Didden wrote an article on LM317 Shunt regulator and it seems that if you were going to use a regulator at small signal (your AD711 "pre section") then that regulator might as well get more work done, in the form of a clean signal reference. After all, a voltage is only as important as its reference. And, this isn't a balanced input amplifier. Therefore every bit of input is 100% reliant on small signal ground--its not possible for any bit of input to be more important than small signal ground.

Newletter 8, Jan Didden, Shunt Regulator

I really hope that link works because I can't explain the topic.
Also, I found a reference that "Jung Regulator" can substitute for the shunt regulator illustrated in the article, but I've honestly no idea if that's true. . . except that any reg is better than wiring the pre section to the exact same unregulated power as the output stage uses to shove the speaker, which shoves back too. Considering that big difference, minor differences amongst regulators might be less important or not important. That sort of judgement on the quality of regulators probably follows in direct proportion to the gain applied, whereby more gain makes minor differences more obvious.

I could probably use some education on the C1/R1/R2 compensation network also, if you'd be so kind.
I don't have the education to thoroughly answer your question.
What I see there is that the components you specify are a Local negative feedback loop. Usually nested amplifiers have 1 local feedback loop per each op-amp. And this was going well until I noticed that not all of your feedback is negative. The AD711 is running both R3 and R4 in an adder function. Since both op-amps aren't inverting, R4 is running forward, and after spotting that, I'm lost because I'm just not familiar with nested design. It does appear to be not much different than an inverting discrete amplifier. Perhaps it is merely comparable to the wide range of functions available to ordinary inverting mode amplifiers. But, its got me bluffed. A simulator might estimate some functional component values.