This is a summary of my approach to designing discrete-transistor op-amps for audio. These op-amps favor open-loop linearity, bandwidth, and PSRR over gain.
IC op-amps are invariably designed for the highest possible open-loop gain (>=100dB) so that the closed-loop gain is almost entirely determined by the feedback network. IC op-amps are well-suited for precision applications up to frequencies of about 100Hz. Above that, the gain must be rolled-off and performance deteriorates.
In addition to audio being fast by op-amp standards, the feedback networks used in audio are always linear and often time-invariant. Audio can tolerate errors of 1% as long as the errors are linear and time-invariant.
The special properties of audio give rise to a new (old) approach - design an amplifier with good open-loop linearity, bandwidth, and PSRR, and apply moderate feedback. An amplifier with 60dB of open-loop gain can come out ahead of an amplifier with 100dB of open-loop gain if the starting point is 40dB better. Keep in mind that we are not trying to beat IC op-amps on all metrics, just the ones that matter.
Naturally, I have omitted all details. This approach is applicable to class A amplifiers where distortion can be made arbitrarily low through an appropriate choice of operating point.
Ed
ETA: I added a graph in post 63 which illustrates the idea.
IC op-amps are invariably designed for the highest possible open-loop gain (>=100dB) so that the closed-loop gain is almost entirely determined by the feedback network. IC op-amps are well-suited for precision applications up to frequencies of about 100Hz. Above that, the gain must be rolled-off and performance deteriorates.
In addition to audio being fast by op-amp standards, the feedback networks used in audio are always linear and often time-invariant. Audio can tolerate errors of 1% as long as the errors are linear and time-invariant.
The special properties of audio give rise to a new (old) approach - design an amplifier with good open-loop linearity, bandwidth, and PSRR, and apply moderate feedback. An amplifier with 60dB of open-loop gain can come out ahead of an amplifier with 100dB of open-loop gain if the starting point is 40dB better. Keep in mind that we are not trying to beat IC op-amps on all metrics, just the ones that matter.
Naturally, I have omitted all details. This approach is applicable to class A amplifiers where distortion can be made arbitrarily low through an appropriate choice of operating point.
Ed
ETA: I added a graph in post 63 which illustrates the idea.
Last edited:
Oh, the JE-990. Shouldn't that be a LM394H at the input instead of the LM394N? Guess I still have the parts to make a few.
What metrics are "the ones that matter"? What aspects of performance "deteriorate"? What closed loop gains are you interested in, for example?
Some clarity would be appreciated.
I posted this thread to avoid derailing another thread. https://www.diyaudio.com/community/...om-opa1656-thread.377331/page-20#post-7940257 I was not planning to post more.
Single-digit PPM THD can be achieved. The closed-loop gain cannot be too large.
Ed
Single-digit PPM THD can be achieved. The closed-loop gain cannot be too large.
Ed
Is this thread intended to talk about opamps with lower OLG, with examples of these opamps, their use, and suggestions for implementation? That would be a beneficial topic for this forum, and to further illustrate your ideas and point of view. I’m sure people will be interested in that.
Indeed. Here is a much simplified FET input opamp design from 40+ years ago.
Very similar to the discretes that are shown.
Single digit PPM distortion over the audio band could be achieved with this in unity gain mode and remember this was 40+ years ago, and it was not even intended for audio purposes. Modern opamps, optimized for audio, can be significantly better.
@6L6: Yes, the topic is moderate open-loop gain and how OLG is only one of many parameters.
@wynpalmer2: Unfortunately, I am not up-to-date on IC op-amps.
Ed
@wynpalmer2: Unfortunately, I am not up-to-date on IC op-amps.
Ed
@Chris Hornbeck, A perfectly high output impedance transconductance input stage, followed by lossy RIAA, followed by perfectly high input impedance second stage, has frequency response independent of stage gains. This isn't something easily done with the usual monolythic opamps, but is very interesting for discrete designs.
When I correctly understand, you are describing a CFA like the AD844.
@Hierfi made an interesting design with this opamp.
@EdGr,
What can a discrete design improve to an integrated design using the same elements ? Even companies like Mark Levinson that went on a long time using discrete designs when others switched, are now also using opamps in preamps and main amps.
Hans
When I correctly understand, you are describing a CFA like the AD844.
@Hierfi made an interesting design with this opamp.
@EdGr,
What can a discrete design improve to an integrated design using the same elements ? Even companies like Mark Levinson that went on a long time using discrete designs when others switched, are now also using opamps in preamps and main amps.
Hans
Well, thanks for your, well, opinion, I guess...
In the future, you’ll have more people interested and engaged if you actually have something to show your ideas rather than just telling everyone how right you are. This is a forum for discussion, learning, illumination, and education through diverse interests. You are welcome to come here and announce your beliefs if you like, but to do so without any support is akin to shouting at the clouds like an old man sitting on a park bench… The passers by will shake their heads in disbelief, and mutter to themselves that when they are in their dotage that they will hopefully purport themselves more coherently.
We were hoping to see these ideas discussed and you to show your reasoning and results as to why this is a good idea. There may be much merit to your point. It’s a disappointment to all that it’s not happening.
ADI produced two CFAs in the 80's- although they didn't call them CFAs, as that was a trademarked term at the time, they called them transimpedance amplifiers, then later rebranded them to be CFAs.
These were the AD844 and the AD846. The 844 was, in many ways, a simpler form of the 846, but was targeting a somewhat different market, and, as it turns out, a more long lasting one as the 846 is now obsolete.
The 846 was a precision version with almost ideal current conveyance from the inverting node to the compensation node, with extremely high output impedance, and even lower distortion than the 844.
@Mark Johnson may remember that it had an ISSCC paper about it in the day.
It could indeed be used as an RIAA stage with a suitable passive interstage filter and was used as such at least once.
Indeed, I was a party to demonstrations of its use in this way.
These were the AD844 and the AD846. The 844 was, in many ways, a simpler form of the 846, but was targeting a somewhat different market, and, as it turns out, a more long lasting one as the 846 is now obsolete.
The 846 was a precision version with almost ideal current conveyance from the inverting node to the compensation node, with extremely high output impedance, and even lower distortion than the 844.
@Mark Johnson may remember that it had an ISSCC paper about it in the day.
It could indeed be used as an RIAA stage with a suitable passive interstage filter and was used as such at least once.
Indeed, I was a party to demonstrations of its use in this way.
@wynpalmer2 Is there a difference in OLG when comparing current feedback opamps to voltage feedback? I'm not particularly familiar with the differences...
Yes. However, this is not a good forum to discuss the differences.
I would refer you to various sources.
The 846 datasheet.
https://www.analog.com/media/en/technical-documentation/obsolete-data-sheets/270284ad846.pdf
The 844 datasheet.
https://www.analog.com/media/en/technical-documentation/data-sheets/ad844.pdf
The JFE150 datasheet. The composite amplifier section.
https://www.ti.com/lit/ds/symlink/j...JEuBm4IBuTyZw4W53iuQGpdk8gAaQf7BoCa54QAvD_BwE
I would refer you to various sources.
The 846 datasheet.
https://www.analog.com/media/en/technical-documentation/obsolete-data-sheets/270284ad846.pdf
The 844 datasheet.
https://www.analog.com/media/en/technical-documentation/data-sheets/ad844.pdf
The JFE150 datasheet. The composite amplifier section.
https://www.ti.com/lit/ds/symlink/j...JEuBm4IBuTyZw4W53iuQGpdk8gAaQf7BoCa54QAvD_BwE