This setup worked great for me. Of course, you use other chips, but they belong to equivalent families and can be interchanged (AD823 with OPA627 and OPA541 with OPA549).
It played very nice nice but I dismantled it soon because I needed the chips for another project which sounded even better.
It played very nice nice but I dismantled it soon because I needed the chips for another project which sounded even better.
The AD823/OPA549 lashup has the some shortcoming as the Clown - the amp inside the loop (the 549) is too slow - even slower than the 541. Having said that, the Clown is doing duty in my living room, and it doesn't sound too bad. I was watching some Battlestar DVDs last night, and the amp is good at rendering the little low level sounds- maybe because its no-signal output noise is so low. Having the Clown around is giving me some breathing space to do some mods on my screen driven SE amp - a choke in the B+ supply to squelch some output noise that looks like B+ ripple feedthrough, and another look at the screen supply for the input stage. My "Almost Blameless" class AB amp (heavy on the A) is just about ready, too. I expect It'll kick the Clown's sorry rump.
At nay rate, the Clown was just a short foray into chip-amp land. I'll be going back to tubes and discrete sand state projects, especially as I have 5-6 projects on the bench in various state of disarray.
At nay rate, the Clown was just a short foray into chip-amp land. I'll be going back to tubes and discrete sand state projects, especially as I have 5-6 projects on the bench in various state of disarray.
The "Almost Blameless" discrete amp is up and running, and I'm in the process of tweaking in the bias for thermal stability, as there is a heavy dose of Class A. Even at this early date. I'd have to say that it kicks the Clown's spotty behind... I'll post a separate thread in the Solid State section on the "AB" when I'm more or less sure it's complete.
Look at the beginning of this thread - I used an OPA627 and OPA541. I have very mixed feelings about this amp, as it is not a proper design for a composite amplifier. A composite amplifier consists of a slower, more accurate amp with a faster amp included in its feedback loop. This one is the opposite way, with the slow! 541 included in the feedback loop of the much faster 627. The 627 needs to be slowed down quite a bit via inner loop compensation so that the combination does not oscillate.
The square wave response shows some well-damped overshoot (I don't like to see any) plus a very noticeable crossover notch at the zero crossing of the square wave, indicating the the 541 is under-biased, and though the 627 tries its damndest to speed the 541 through the zero crossing, it's not successful for a square wave with 200 ns edges.
A fast square wave is a pretty severe test for an audio amplifier. I suspect that for slower signals, the 627 helps to smooth out the crossover notch from the 541. Also, due to the low offset of the 627, there is little or no DC at the output of the amp (I couldn't measure any).
Summary - Though you can build this amp and it will make what appear to be fairly decent sounds, it is basically flawed in concept. I may take the trouble to hook it up to an analyzer to look at the harmonic/intermodulation distortion spectrum at some point, or I may just gut it and use the chassis and power supply for another project. I'm having much better results with my discrete and vacuum-state amps. I was going to slam all the big amp-in-one products as too slow and under-biased, but some of the National amps look a bit better in that respect. I'm not curious enough to buy some and try them in this kind of a circuit, however, as I have way too many other (more succulent) fish to fry. The exercise is left to someone else with more motivation.
The square wave response shows some well-damped overshoot (I don't like to see any) plus a very noticeable crossover notch at the zero crossing of the square wave, indicating the the 541 is under-biased, and though the 627 tries its damndest to speed the 541 through the zero crossing, it's not successful for a square wave with 200 ns edges.
A fast square wave is a pretty severe test for an audio amplifier. I suspect that for slower signals, the 627 helps to smooth out the crossover notch from the 541. Also, due to the low offset of the 627, there is little or no DC at the output of the amp (I couldn't measure any).
Summary - Though you can build this amp and it will make what appear to be fairly decent sounds, it is basically flawed in concept. I may take the trouble to hook it up to an analyzer to look at the harmonic/intermodulation distortion spectrum at some point, or I may just gut it and use the chassis and power supply for another project. I'm having much better results with my discrete and vacuum-state amps. I was going to slam all the big amp-in-one products as too slow and under-biased, but some of the National amps look a bit better in that respect. I'm not curious enough to buy some and try them in this kind of a circuit, however, as I have way too many other (more succulent) fish to fry. The exercise is left to someone else with more motivation.
Last edited:
200nS edge sq wave isn't a "audio" test signal
it may be useful to learn about some amp properties but shouldn't really be accepted as a clear indication of "poor" audio performance
likewize damped +100KHz ringing of a few 10% needn't be evidence of bad audio amplification - and isn't always an indication of poor amplifier stability/phase margin
in the case of a slow power amp with a fast outer feedback amp you may have to accept that critical or under damped response is too restrictive given the contraints you have to work within
add a prefilter if you're really bothered by the appearance of the sq wave - cutting input signal slew rate cuts slew rate error
power chip amp output could be biased as discussed in the http://www.diyaudio.com/forums/chip-amps/164634-chip-amp-rival-hi-end-design-advice.html thread - should give better low signal level linearity per the "first watt" principle
it may be useful to learn about some amp properties but shouldn't really be accepted as a clear indication of "poor" audio performance
likewize damped +100KHz ringing of a few 10% needn't be evidence of bad audio amplification - and isn't always an indication of poor amplifier stability/phase margin
in the case of a slow power amp with a fast outer feedback amp you may have to accept that critical or under damped response is too restrictive given the contraints you have to work within
add a prefilter if you're really bothered by the appearance of the sq wave - cutting input signal slew rate cuts slew rate error
power chip amp output could be biased as discussed in the http://www.diyaudio.com/forums/chip-amps/164634-chip-amp-rival-hi-end-design-advice.html thread - should give better low signal level linearity per the "first watt" principle
I don't have to accept any of these sort of warts from my other designs, because they aren't there. I could probably cut down the overshoot with more inner loop compensation around the 627. The crossover notch will stay, as it's the 541's fault - it does run rather lean
I originally threw this together on a whim because I had both the 627s and the 541s on hand, with a view to getting around the low bandwidth of the 541. The concept may work better with something like an LM3886, as it is faster and runs with higher bias current. Paired with the 627 or some other high quality JFET-input amp you get the DC accuracy and the ability to run easily with an AC-coupled feedback loop, something I have insisted on for over 30 years. I refuse to put an electrolytic cap in my feedback loop.
To be honest, I like the sound of my other amps better, which is another reason I have moved on. I'm having a lot of fun marrying solid state with tubes, but that's another set of threads altogether.
I originally threw this together on a whim because I had both the 627s and the 541s on hand, with a view to getting around the low bandwidth of the 541. The concept may work better with something like an LM3886, as it is faster and runs with higher bias current. Paired with the 627 or some other high quality JFET-input amp you get the DC accuracy and the ability to run easily with an AC-coupled feedback loop, something I have insisted on for over 30 years. I refuse to put an electrolytic cap in my feedback loop.
To be honest, I like the sound of my other amps better, which is another reason I have moved on. I'm having a lot of fun marrying solid state with tubes, but that's another set of threads altogether.
the 200ns edge sq wave test is like asking if a +/-200 V ps direct coupled amp is "better" than +/-80 V ps amp with a gain of 20 and a max source of 2 Vrms - standard "consumer" desktop CD/SACD/DVD DAC output level
if the output V never reaches the limit most of us don't agonize over missing the "improvement" offered by the +/-200 Vsupply amp - despite it being clearly seen with a 14 Vrms source from our function generator
if the output V never reaches the limit most of us don't agonize over missing the "improvement" offered by the +/-200 Vsupply amp - despite it being clearly seen with a 14 Vrms source from our function generator
Last edited:
That's not the way I look at it. If an amplifier is presented with "out of band" signal that doesn't saturate it, it should simply roll it off instead of trying to respond to it and screwing up while trying to do so. The problem in my circuit is that there is a very slow opamp inside the feedback loop of a very fast one. This is begging for trouble. The overshoot I mentioned can (and should be) fixed by compensating the inner loop. The cap I used (C2 in the schematic) straightened out the small signal response (too small a value causes pretty drastic peaking in the frequency response), but was still too small to fix the large signal. Making C2 larger would probably nail the overshoot, but do nothing for the crossover distortion. Looking at the harmonics with "audio" frequencies would be revealing. I would expect higher harmonics to show up at high frequencies as the 627 is less able to correct the 541 crossover distortion. A LM3886 would possibly fare better in this respect, but I don't have the time or the inclination to try it.
I always hit my amps with a square wave that is large enough to cause the amplifier to slew, but small enough not to force it into saturation. It shakes out a lot of problems.All my other solid state amps and all my tube amps are a lot better behaved than this one, with the same square wave excitation.
I always hit my amps with a square wave that is large enough to cause the amplifier to slew, but small enough not to force it into saturation. It shakes out a lot of problems.All my other solid state amps and all my tube amps are a lot better behaved than this one, with the same square wave excitation.
Dear Wrenchone,
I think you gave up a little bit to fast. I am agree with you about the fact that the OPA541/549's are to slow compared to most drive opamps in this configuration. However I achieved breath taking results in such a neted feedback configuration. I managed it to get near perfect square wave reproduction till 20Khz. with a combination of a cap direct from the driver opamp to to inverted input, and a second compensation cap from the output chip to the negative input of the driver chip. This way I was able to tune the overshoot, and roll of perfectly from the square-wave. After this tuning was done, the amplifier was stable, and dead silent (one of the biggest advantages of this configuration). The sound was 3D, dark, black and musical, and the perception with percussion enz. was that this is a lightning "Fast" amp even though the slew-rate is poor. I will take a look if I can still find my schematics.
Two questions for those who built those or comparable nested feedback chip amps.
1: What about clipping behavior? My experience is clipping behavior is really bad, because the driver opamp tries to correct fo the clipped wave, which it can't and the result is under and overhsoot.
2: Thermal protection. What if the amplifier chip get into thermal protection, the driver chip start to get infinite gain.
Any of you thought about this?
With kind regards,
Bas
I think you gave up a little bit to fast. I am agree with you about the fact that the OPA541/549's are to slow compared to most drive opamps in this configuration. However I achieved breath taking results in such a neted feedback configuration. I managed it to get near perfect square wave reproduction till 20Khz. with a combination of a cap direct from the driver opamp to to inverted input, and a second compensation cap from the output chip to the negative input of the driver chip. This way I was able to tune the overshoot, and roll of perfectly from the square-wave. After this tuning was done, the amplifier was stable, and dead silent (one of the biggest advantages of this configuration). The sound was 3D, dark, black and musical, and the perception with percussion enz. was that this is a lightning "Fast" amp even though the slew-rate is poor. I will take a look if I can still find my schematics.
Two questions for those who built those or comparable nested feedback chip amps.
1: What about clipping behavior? My experience is clipping behavior is really bad, because the driver opamp tries to correct fo the clipped wave, which it can't and the result is under and overhsoot.
2: Thermal protection. What if the amplifier chip get into thermal protection, the driver chip start to get infinite gain.
Any of you thought about this?
With kind regards,
Bas
I did drag the amp out and have a listen to it last night, and it is still as I remember it. Soundstage in absence of signal is utterly quiet - better than any of my other amps. I suspect that power supply rejection may play a part in that. I also have 50,000 uF filtering each rail, plus 1000 uF local bypass. Bass extension is very good, and bass is tightly controlled - in fact, low bass is so good that I have had problems with acoustic feedback from my turntable if I don't watch the volume. Little details are apparent. I find the presentation to be more "dry" than that of my other amps.
I may look at this amp again after I'm finished with some other projects. As I mentioned before, taking a look at the harmonic spectrum would tell me some things. As I also said before, a combination of OPA627 and LM3886 may have better synergy. I used the OPA541 because I had it on hand, as this thing started out as a lark.
I have a small subwoofer amp that has been in the works for a while, using the OPA541, plus a OPA4134 - based crossover filter. I have an extra opamp from the 4134 floating around (actually hooked up as a follower so it doesn't cause any mischief) - it may be interesting to use that in a composite with the LM3886.
I may look at this amp again after I'm finished with some other projects. As I mentioned before, taking a look at the harmonic spectrum would tell me some things. As I also said before, a combination of OPA627 and LM3886 may have better synergy. I used the OPA541 because I had it on hand, as this thing started out as a lark.
I have a small subwoofer amp that has been in the works for a while, using the OPA541, plus a OPA4134 - based crossover filter. I have an extra opamp from the 4134 floating around (actually hooked up as a follower so it doesn't cause any mischief) - it may be interesting to use that in a composite with the LM3886.
Dear Wrenchclone,
Thanks for your comments. I share your experience. The damping-factor is insane high because of the nested feedback's low Z-out. I will search tomorrow again for my schematics and post it.
I prefer a configuration with the LM3886's as well, the reason I choose the OPA549 is because this has a E/S pin that gives a logic level when the chip goes into thermal shutdown. This is needed to activate a relays to short the input opamp, to prevent it from self-destruction when the amplifier chip shut's off.
My two question's remains open,
1: How to improve clipping behavior?
2: How to prevent the drive op-amp from self destruction when the amplifier chip get's in thermal shutdown.
Ps. I am working on a same approach with a discrete output-stage in nested feedback with an opamp.
With kind regards,
Bas
I have always admired the composite configuration since I first saw it in an electronics magazine (radio electronics I think it was).
I believe it was a 70watt BPA using lm1875's.
Has any one considered trying this method using lm3886's in a BPA-300 or larger using 4 to 8 chips per leg or even just a plain ol' parallel configuration?
I believe it was a 70watt BPA using lm1875's.
Has any one considered trying this method using lm3886's in a BPA-300 or larger using 4 to 8 chips per leg or even just a plain ol' parallel configuration?
Yes I do bridged/parallel with total 12 OPA549's. And I go even one step further, but I will publish here once it is finished.
Do you have any print from it? Wonder how they handle with the thermal shutdown issue. If the outer chip goes in shutdown but there is still signal on the inner opamp, it will get an infinite gain, because it keep trying to get the output equal to the input.
With kind regards,
Bas
Thank you in advance! It would be amazing if you could find it.
with kind regards,
Bas
Ps. I build my first with the so hated but still amazing NE5534. With this (still good) opamp, I achieved insane low noise. Second unique feature of the NE5534 is the compensation pins which will help to tune the speed to the follower stage. All in all the NE5534 is already not a very fast opamp, and therefor a perfect candidate for such a project with slow chip amps.
With kind regards,
Bas
Here is the 33watt version.
I am still looking for the 70watt version.
I remember the paralleled lm1875's for outputs but the inputs each may have been in some sorted nested composite configuration.
I believe there were two articles.
The first one showed this circuit and some single ended ones to descirbe the concept of reducing the higher distortion and noise of the power amp to that of the lower noise and distortion of the driving opamp.
The second article was the BPA 70watt amp itself.
This performance was amazing for the early 90's and for what was reasonable cost at the time.
Nowdays it is even better and cheaper in cost, even, with the higher power output chips and even lower noise and distortion opamps than that of day. jer
I am still looking for the 70watt version.
I remember the paralleled lm1875's for outputs but the inputs each may have been in some sorted nested composite configuration.
I believe there were two articles.
The first one showed this circuit and some single ended ones to descirbe the concept of reducing the higher distortion and noise of the power amp to that of the lower noise and distortion of the driving opamp.
The second article was the BPA 70watt amp itself.
This performance was amazing for the early 90's and for what was reasonable cost at the time.
Nowdays it is even better and cheaper in cost, even, with the higher power output chips and even lower noise and distortion opamps than that of day. jer
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.