Try opa132P instead of opa134.You'll be amazed.
You can use low offset bipolar op amps like opa2228P , but you can decouple the negative input anytime with a good quality bipolar cap .It's not the input current that causes the offset in your case.I used the inverting topology instead and believe me: I have no problem with lm6172...
The key is to have high openloop gain at 10kHz (high GBW is needed) and also high slew rate. I did work with OPA132 but it is not enough here. Same for OPA228/2228, they have only 11V/us and 33MHz, this is useless in this amp. It will certainly work and still would be acceptable at 1kHz, but no way at 10kHz and with heavy load of the amp.
R3 x Ib x Gain is the issue, input bias current x input resistance x gain, as there is a C1 coupling cap. Yes DC FB blocking cap from R4 to GND could be used but why. There is no advantage of such solution and with the BJT opamp.
There is no reason to change the selected opamps as the parameters are excellent. The only option from the suggestions here would be ADA4637, which has JFET input, GBW = 79MHz, SR = 170V/us, very low distortion and openloop gain high enough at 10kHz. This would give similar results as OPA637. BTW I also prefer JFET opamps as they are much more immune against EMI.
Remember that parameters quickly deteriorate when it moves from class A to AB. You need to stay mostly <2 x Iq (2 x idle current) with the output peak current. You can go a bit higher, but then the O/P stage increases input current and decreases its input impedance and starts to be a significant load for the driving circuit. If you push it even harder, Q1 and Q2 start to work as a sort of output current limiter. So best to stay in class A. This would not work as a typical class AB with quite low idle current and high power above the 2xIq.
There must be smth wrong with your assessement unless there's smth catastrophically wrong about the power transistors. Operating output trz in class A gives maximum slew rate so I'd rule out that in the first place.
There's sufficient slew rate even with opa2227 in your circuit .The slew rate depends also on the sheer power of the op-amp output stage and both op-amps that I recommended are pretty capable.The op-amp needs to deliver its current all the way upor down to the rails and even if it had 1 billion v/us slew rate if it could only swing a few miliamps its useless.The op-amp is driving a common base circuit , op-amp's bandwidth is no problem.I had no bandwidth problem driving 60's germanium transistors...The main ingredients you're looking up for in your op-amps are output voltage swing /output current , not bandwidth.
Last edited:
That DC blocking works especially well with electrolitic capacitors and bipolar input op amps and it rules out voltage offset which improves your headroom also.With jfet opams bipolar electrolitics work better than unipolar.You might need selecting them for lower leakage, but try the inverting topology with cheaper lower slew rate op-amps first instead
What I'm really convinced in a lot of talks about high vs low slew rate op amps is that there is a coincidence of two different factors that leads to false premises. Usually high slew rate op amps are used in driving 50 ohms loads when supplied below +-5V so naturally they have high output capabilities even rail to rail 100mA outputs while the slower op-amps aren't usually optimized for driving tough loads near the rails.Thus if you're going to use lm6172 of course you're going to drive almost anything down to 32 ohms headphones, but that does not mean you're going to drive them at 3200V/us because it is impossible at any frequency.
Driving a common base circuit with 4.6mA apparently needs just a bit of current, but a lot of headroom which usually video or adsl op-amps have, but the are a few audio op amps that also have good headroom and output capabilities and I feel that a simple njm4556 or a common lm4562 would be a tough nut in a blind test against the highest speed op amp used in this circuit if the high speed op amp is stable and not ringing in your circuit.
What I see in clear need for optimisation in your circuit is q1, q2 , q7, q9, q8, q10 which would better be low sat zetex trz or at least good old 2n3904/06 .You don't need bd139/140 .At 4.6mA and max 60v swing you can still go with bc337/327 or 2n3904/06 at low voltage supplies but ztx, fmmt and fzt trz are widely available at higher voltages and even better specs.
I think your op-amps are actually fighting with reduced headroom!
Of course if you have the EMI in mind you'd preffer j-fet op amps, but I'm pretty sure that you don't need especially high slew rate op amp.
Driving a common base circuit with 4.6mA apparently needs just a bit of current, but a lot of headroom which usually video or adsl op-amps have, but the are a few audio op amps that also have good headroom and output capabilities and I feel that a simple njm4556 or a common lm4562 would be a tough nut in a blind test against the highest speed op amp used in this circuit if the high speed op amp is stable and not ringing in your circuit.
What I see in clear need for optimisation in your circuit is q1, q2 , q7, q9, q8, q10 which would better be low sat zetex trz or at least good old 2n3904/06 .You don't need bd139/140 .At 4.6mA and max 60v swing you can still go with bc337/327 or 2n3904/06 at low voltage supplies but ztx, fmmt and fzt trz are widely available at higher voltages and even better specs.
I think your op-amps are actually fighting with reduced headroom!
Of course if you have the EMI in mind you'd preffer j-fet op amps, but I'm pretty sure that you don't need especially high slew rate op amp.
Last edited:
PMA
I'm also interested in this class A amplifier, and wondering if you will release the gerbers for this amplifier?
Have you compared the sound from this amplifier with your PM-AB2 that also are an inreresting amp.
Regards
Last edited:
I will consider releasing the Gerber files, but I need to make small changes that I did during years by "cut track/ paste wire" method
.AB2 is made just in one piece, one channel, as a 1st sample, so no listening yet. From my previous experience with A/B level matched tests with well measuring amplifiers, we may have a feeling that we hear some difference, but it is then very difficult or impossible to verify it in a double blind, ABX test.
So, I am a bit sceptical. Low distortion amplifiers with wide and flat frequency response and reasonably low Zout tend to sound very similar or same. Double blind test is the only reliable method to evaluate this. Otherwise expectation bias tends to dominate in a sighted test, even if we are not willing to realize it and admit it.
What you mean as "proof"? The old thread is here
https://www.diyaudio.com/community/threads/new-error-correction-amp.40543/
https://www.diyaudio.com/community/threads/new-error-correction-amp.40543/
This is not true. Slew rate in voltage feedback amplifier (VFA) depend on input current (LTP) and the compensation.
Bob Cordell and Douglas Self wrote it in their book. And I can confirm it in simulation or in measurement.
Definitely, you do not know how to design amplifier.
Thank God I don't have to! Ltspice does it for me!
Slew rate is defined as V/us, but you have no damn slew rate if you can't push your current through the damn 2...8 ohm INDUCTIVE load.
You could have a billion V/femto second in the op amp...if the final stage can't deliver it to the load you have 0V / 1 billion years slew rate.Try listen BA 5218 on headphones and you'll see what a 2...3v/us op amp can do just because it can deliver enough current to the load.There's no slew rate without the ability to push a current through a load and no op amp can do that on 8 ohms that is why you have a buffer...the op-amp itself has an internsl buffer too.Why are you biasing for class A to make 20khz crossover distortion as low as possible and why are you looking for low cob transistors in the final stage when you could just use your billion volt slew rate op-amp and you're done...!?
As for slew rate requirements of an amp...now that it's clear you're the true engineer here, maybe you could tell me how a 9v/us vfb op amp can drive a qsc1700 with 21 v/us at 600 watts of power into 4 ohms if only the input ltp of ne5532 is the one responsible with slew rate and basically you can only go down below 9v/us with any added compensation, not up...Apparently there's no need for more than 21V/us either ...
