I have heard much derision of op-amps over the years and have yet to hear a detailed, technical analysis of what is wrong with them. Usually all I hear is "they sound bad", and some stuff about negative feedback, but no real technical analysis.
If they "sound" bad, there must be some identifiable deficiency in performance- either distortion, noise, frequency response, or dynamic response.
Can anyone in this forum steer me toward any good engineering-based analysis of discrete vs. integrated circuits and how measured peformance relates to perceived "sound"?
Thanks,
MR
If they "sound" bad, there must be some identifiable deficiency in performance- either distortion, noise, frequency response, or dynamic response.
Can anyone in this forum steer me toward any good engineering-based analysis of discrete vs. integrated circuits and how measured peformance relates to perceived "sound"?
Thanks,
MR
Not all op-amps suck.
Some people just feel that discreet designs have much more flexibiliy about what transistors to use, what topology to use, how much feedback ect. It is also more fun than just picking feedback resistors for op-amp design. If op amps are your cup tea there are dozens of very good ones. The question reminds of something I read once about one THE best opamp designers (I have listened to several of his opamp designs and they a very good) having a preamp made with discreet Jfets in his stereo at home. Kind of funny huh..... plenty of us roll your own types have used opamps, in all fairness I must admit that some opamps are quite good. Do a search on the forum.
http://www.dself.demon.co.uk/webbop/opamp.htm
http://www.passdiy.com/pdf/diyopamp.pdf
H.H.
Some people just feel that discreet designs have much more flexibiliy about what transistors to use, what topology to use, how much feedback ect. It is also more fun than just picking feedback resistors for op-amp design. If op amps are your cup tea there are dozens of very good ones. The question reminds of something I read once about one THE best opamp designers (I have listened to several of his opamp designs and they a very good) having a preamp made with discreet Jfets in his stereo at home. Kind of funny huh..... plenty of us roll your own types have used opamps, in all fairness I must admit that some opamps are quite good. Do a search on the forum.
http://www.dself.demon.co.uk/webbop/opamp.htm
http://www.passdiy.com/pdf/diyopamp.pdf
H.H.
Op-Amp Blues!
Well.
1) Insufficient bias in output stage for some applications and switches to class B. Sound better with a buffer attached so as not to load the output stage down.
2) Transient thermal distortion. Since the device is built on the same substrate when the output stage heats up it thermally modulates the other parts of the amp.
3) Output stage is usuall quasi-comp. problem of making matching p and n channel devices in the manufacturing process.
4) Istability problems and require some form of external compensation applied at the WRONG place. Compensated devices usually don't sound good.
5)Headroom limited by the rail voltage.
6)Duals usually have poor seperation detween devices
The list goes on but my main gripe is lack of control of the circuits operaing parameters.
Jam
P.S. Harry feel free to chime in at any time.....
Well.
1) Insufficient bias in output stage for some applications and switches to class B. Sound better with a buffer attached so as not to load the output stage down.
2) Transient thermal distortion. Since the device is built on the same substrate when the output stage heats up it thermally modulates the other parts of the amp.
3) Output stage is usuall quasi-comp. problem of making matching p and n channel devices in the manufacturing process.
4) Istability problems and require some form of external compensation applied at the WRONG place. Compensated devices usually don't sound good.
5)Headroom limited by the rail voltage.
6)Duals usually have poor seperation detween devices
The list goes on but my main gripe is lack of control of the circuits operaing parameters.
Jam
P.S. Harry feel free to chime in at any time.....
Well, if you like them, good for you. I don't mind th sound of opamps, some designs with them sound really good. But they just aren't any fun. Like harry said picking a feedback resistor isn't any fun. Also with discrete designs you can change whatever you want and see how it affects the sound. With opamps you can't change anything, so you can't improve upon it in any way.
Engineers......
Man I have had these discussions over and over and over again.
Like I said if op-amps float your boat, go for it. I you want to do the work and go listen and build some discreet transistor circuits there are plenty links and references on the forum. I am a degreed BSEE and I don't get bogged down in the objective verses sujective debates anymore. After a few years of building and comparing different circuits you will form your own opinions and come by them honestly and the hard way which is the best way to learn. Check out the Borbely website since he is both a very technically astute engineer and also designs by listening. Psssst....... don't tell anyone but the two approaches are not mutually exclusive. Good audio designs are made to be listened to, not measured. The ear is the final arbitor......
http://www.borbelyaudio.com/
H.H.
Man I have had these discussions over and over and over again.
Like I said if op-amps float your boat, go for it. I you want to do the work and go listen and build some discreet transistor circuits there are plenty links and references on the forum. I am a degreed BSEE and I don't get bogged down in the objective verses sujective debates anymore. After a few years of building and comparing different circuits you will form your own opinions and come by them honestly and the hard way which is the best way to learn. Check out the Borbely website since he is both a very technically astute engineer and also designs by listening. Psssst....... don't tell anyone but the two approaches are not mutually exclusive. Good audio designs are made to be listened to, not measured. The ear is the final arbitor......
http://www.borbelyaudio.com/
H.H.
In addition to all the points mentioned, maybe the bottom line is that IC op-amps are designed to be multi-application by design. Regardless of what it is, generally speaking, something that is designed to do multiple things, rarely does one thing especially well. Without getting too technical about the IC’s design, it basically means that the IC may have extra stages and components on the substrate that are superfluous to our needs. Many feel these “extras” detract from the ultimate possible performance.
Personally I’m over the “challenge” of designing, it’s just that I feel discrete components are able to do the job better for most applications in which we may otherwise use an IC op-amp.
Cheers,
Pete
Personally I’m over the “challenge” of designing, it’s just that I feel discrete components are able to do the job better for most applications in which we may otherwise use an IC op-amp.
Cheers,
Pete
Opamps *DO NOT SUCK*... there're EXCELLENT opamps for audio, and if you use them well, they'll surely sound great. As JoeBob and Harry said, i tend to pick discrete circuits because they're more fun from the building/designing point of view. And tend to sound better, but this is only my opinion.
I want a discrete preamp for my next project (well... a preamp
) but the new preamp Rod Elliot proposed, using the burr brown chip looks yummy...
I want a discrete preamp for my next project (well... a preamp
) but the new preamp Rod Elliot proposed, using the burr brown chip looks yummy...Well, no, some op amps are specifically meant to amplify audio, e.g. OPA2134.Pete Fleming said:
I remember Thagard's phono preamp design article in Audio Express, where he said it was very difficult to surpass op-amp quality using discrete components. I would bet that at this level of performance, the layout is critical and probably more important than the choice of discrete vs. opamp.
Having said all that, I agree that discrete is probably more fun. But if I have a little bit of room left in my preamp box and need a quick, high-quality headphone amp for my Senns, I'll throw in an opamp.
Op Amps: Do they suck??
After reading most of posts, I think it is ill-advised to take a definitive stand. Like all things in engineering, there are good ones and bad ones.
I'm more interested in the reasons why some are good and some are not so good.
I believe that most sonic problems in SS amplification can be boiled down to four important pressure points in the most common designs. In my opinion (and nothing is more dangerous than one man's opinion, read any history book!!) they are, in order of significance:
1. The introduction of insidious and accumulating phase errors across the voltage amplifier with increasing frequency. This is the direct result of insertion of a lag compensation capacitor across the collector/base of the voltage amp to secure stability. Transistors do voltage amplification imperfectly because of their highly variable parasitics, particularly Miller capacitance.
2. Obsessive use of current sources, particularly on voltage amplifiers, which have intrinsically good frequency response and thus make compensation of the amplifier much more difficult. Of course, this is related to 1./ above.
3. The appalling performance of the feedback shunt electrolytic capacitor in its dual role of AC ground zero and DC block. In my experience this need to perform two functions at the feedback node is a very tall order and fraught with sonic pitfalls.
4. The extraordinarily sharp and transient-rich behavior of most large signal solid state devices during the crossover event.
Since modern opamps are, by and large, redolent with CCS and electronic gimcrackery of an extremely clever but often non-linear kind, it follows that often they do require electronic handcuffs to make them stable. These handcuffs tame the prisoner but mute his voice utterly, and this, I believe, is often the problem. The size and quality of the compensation capacitor is crucial.
I have not used a lot of opamps, so I can't credibly attack them in a public place like this, and I prefer to do discrete because I have the habit after many years of tinkering. Up to about six or seven years ago I never found an opamp I really liked. I also have a preference for minimal active devices in the signal chain in order to minimise group delay and phase anomalies, and most opamps I study seem to have at least double the number of active devices over most simple discrete designs. Recently I did a four transistor fully Class A gainblock using simple, clean design, and the performance sonically - where it matters - was extraordinary. It had a gain of 13dB, 3dB points at 10Hz and 280KHz, and Zout of just 35R! It sounded sensational, with a full width, full depth sound stage, and wonderful layering of complex passages. Why would one use an opamp in this application, when the cost of the parts was about two dollars (including the LED and all caps!) and most trick opamps are around this price and often call for SMD mounting anyway!
I guess we shouldn't make categoric statements; some opamps are fantastic, and some functions, such as balanced line ICs, are very expensive to do any other way. I love ICs for servo control, comparator functions; oscillators; instrumentation; VU meters; microprocessor control, etc. I think there is a role at the very low levels in phono preamps too. In these distinctive areas they are invaluable. But at line levels I suspect the full-on PP output stage is unnecessary, overkill, and you really need single end output to maintain musicality. (There's a controversial statement thrown in!!).
Cheers,
Hugh R. Dean
www.printedelectronics.com
After reading most of posts, I think it is ill-advised to take a definitive stand. Like all things in engineering, there are good ones and bad ones.
I'm more interested in the reasons why some are good and some are not so good.
I believe that most sonic problems in SS amplification can be boiled down to four important pressure points in the most common designs. In my opinion (and nothing is more dangerous than one man's opinion, read any history book!!) they are, in order of significance:
1. The introduction of insidious and accumulating phase errors across the voltage amplifier with increasing frequency. This is the direct result of insertion of a lag compensation capacitor across the collector/base of the voltage amp to secure stability. Transistors do voltage amplification imperfectly because of their highly variable parasitics, particularly Miller capacitance.
2. Obsessive use of current sources, particularly on voltage amplifiers, which have intrinsically good frequency response and thus make compensation of the amplifier much more difficult. Of course, this is related to 1./ above.
3. The appalling performance of the feedback shunt electrolytic capacitor in its dual role of AC ground zero and DC block. In my experience this need to perform two functions at the feedback node is a very tall order and fraught with sonic pitfalls.
4. The extraordinarily sharp and transient-rich behavior of most large signal solid state devices during the crossover event.
Since modern opamps are, by and large, redolent with CCS and electronic gimcrackery of an extremely clever but often non-linear kind, it follows that often they do require electronic handcuffs to make them stable. These handcuffs tame the prisoner but mute his voice utterly, and this, I believe, is often the problem. The size and quality of the compensation capacitor is crucial.
I have not used a lot of opamps, so I can't credibly attack them in a public place like this, and I prefer to do discrete because I have the habit after many years of tinkering. Up to about six or seven years ago I never found an opamp I really liked. I also have a preference for minimal active devices in the signal chain in order to minimise group delay and phase anomalies, and most opamps I study seem to have at least double the number of active devices over most simple discrete designs. Recently I did a four transistor fully Class A gainblock using simple, clean design, and the performance sonically - where it matters - was extraordinary. It had a gain of 13dB, 3dB points at 10Hz and 280KHz, and Zout of just 35R! It sounded sensational, with a full width, full depth sound stage, and wonderful layering of complex passages. Why would one use an opamp in this application, when the cost of the parts was about two dollars (including the LED and all caps!) and most trick opamps are around this price and often call for SMD mounting anyway!
I guess we shouldn't make categoric statements; some opamps are fantastic, and some functions, such as balanced line ICs, are very expensive to do any other way. I love ICs for servo control, comparator functions; oscillators; instrumentation; VU meters; microprocessor control, etc. I think there is a role at the very low levels in phono preamps too. In these distinctive areas they are invaluable. But at line levels I suspect the full-on PP output stage is unnecessary, overkill, and you really need single end output to maintain musicality. (There's a controversial statement thrown in!!).
Cheers,
Hugh R. Dean
www.printedelectronics.com
open loop linearity
MRehorst,
Both the passive and active devices on the OP-amp substrate are much more non-linear than their discrete counterparts. E.g. an OP-amp internal resistor can be quite non-linear. All these shortcomings result in a violation of one of the basics for top class audio performance - open loop linearity.
Huge amounts of NFB restores the figures but subjectively affects the dynamics. The big crescendo of a full orchestra lose some of its overwhelming power. The same piece played on a top-class discrete design will nail you to the wall. When feedback is reduced too much on the discete design the sound shifts toward being non-dynamic. Each design has its optimum amound of NFB, and the problem with OP-amps is that we are at least some 50 dB off the optimum.
In specific applications OP-amps can in fact work well. I once listened to a Chord DSC900 DAC that had excellent bass, really excellent. But the sonic problems started to show up in the mid´s and upwards, which is kind of typical for OP-amp sound. From this listening session I can at least conclude that good bass is possible with OP-amps. If the target design is a low-pass crossover for a sub then this OP-amp would work just great (sorry, don´t know the OP type).
Regards
Syl
MRehorst,
Both the passive and active devices on the OP-amp substrate are much more non-linear than their discrete counterparts. E.g. an OP-amp internal resistor can be quite non-linear. All these shortcomings result in a violation of one of the basics for top class audio performance - open loop linearity.
Huge amounts of NFB restores the figures but subjectively affects the dynamics. The big crescendo of a full orchestra lose some of its overwhelming power. The same piece played on a top-class discrete design will nail you to the wall. When feedback is reduced too much on the discete design the sound shifts toward being non-dynamic. Each design has its optimum amound of NFB, and the problem with OP-amps is that we are at least some 50 dB off the optimum.
In specific applications OP-amps can in fact work well. I once listened to a Chord DSC900 DAC that had excellent bass, really excellent. But the sonic problems started to show up in the mid´s and upwards, which is kind of typical for OP-amp sound. From this listening session I can at least conclude that good bass is possible with OP-amps. If the target design is a low-pass crossover for a sub then this OP-amp would work just great (sorry, don´t know the OP type).
Regards
Syl
Opamp vs. Discrete
Well Hugh,
I almost completely disagree with your post.
1) Folded cascode designs like AD817, AD797 do not use this cap and have only one gain stage.
2) In my experience current sources improve the sound.
3) Electrolytic in the feedback loop I only use in the poweramp. Because I find this the better sounding solution compared to servo's. My preamp discrete ""opamp"" is completely DC coupled from input to output.
4) Output stages not starved for bias current show good behaviour on transients even low level ones.
I am curious what kind of opamps sound fantastic as I have not yet met these in this life.
Well Hugh,
I almost completely disagree with your post.
1) Folded cascode designs like AD817, AD797 do not use this cap and have only one gain stage.
2) In my experience current sources improve the sound.
3) Electrolytic in the feedback loop I only use in the poweramp. Because I find this the better sounding solution compared to servo's. My preamp discrete ""opamp"" is completely DC coupled from input to output.
4) Output stages not starved for bias current show good behaviour on transients even low level ones.
I am curious what kind of opamps sound fantastic as I have not yet met these in this life.
Hi Elso,
I have read your posts over some time with great interest, and am pleased to finally meet you in this forum!
I am delighted that you disagree with me. Hell, I'm probably wrong, and might actually learn something from you! Let me answer your points briefly:
1) Folded cascode designs like AD817, AD797 do not use this cap and have only one gain stage.
I wasn't aware of this, as I don't use opamps for voltage amplification at all. But I was aware the 797 is a very good sounding opamp, and pleased that it has only one gain stage. Elso, do you have a schematic handy I might look at?
2) In my experience current sources improve the sound.
Yes, and I have used them despite my prejudices, in a 28W SE mosfet design some years ago, and recently in a direct coupled gain block. I only use them where I have to, but let's just say I don't quite agree they improve the sound. Rarely can circuits with CCS and without be directly compared unless bootstraps or transformers are used, so it really apples and pears.
3) Electrolytic in the feedback loop I only use in the poweramp. Because I find this the better sounding solution compared to servo's. My preamp discrete ""opamp"" is completely DC coupled from input to output.
Yes, I think I agree with this. But there may be ways of wiring the electrolytic so it supports two AC paths through it; this I know improves the sound considerably.
4) Output stages not starved for bias current show good behaviour on transients even low level ones.
Yes, can't disagree with this, but I'm concerned more about the crossover disjunction, which happens so quickly the NFB loop can't properly correct. I actually would say that in comparisons between well biased Class AB and Class A, the Class AB seems to me to have the edge for dynamics, all other things being equal.
Thank you for your post. I welcome such input!
Cheers,
Hugh
www.printedelectronics.com
I have read your posts over some time with great interest, and am pleased to finally meet you in this forum!
I am delighted that you disagree with me. Hell, I'm probably wrong, and might actually learn something from you! Let me answer your points briefly:
1) Folded cascode designs like AD817, AD797 do not use this cap and have only one gain stage.
I wasn't aware of this, as I don't use opamps for voltage amplification at all. But I was aware the 797 is a very good sounding opamp, and pleased that it has only one gain stage. Elso, do you have a schematic handy I might look at?
2) In my experience current sources improve the sound.
Yes, and I have used them despite my prejudices, in a 28W SE mosfet design some years ago, and recently in a direct coupled gain block. I only use them where I have to, but let's just say I don't quite agree they improve the sound. Rarely can circuits with CCS and without be directly compared unless bootstraps or transformers are used, so it really apples and pears.
3) Electrolytic in the feedback loop I only use in the poweramp. Because I find this the better sounding solution compared to servo's. My preamp discrete ""opamp"" is completely DC coupled from input to output.
Yes, I think I agree with this. But there may be ways of wiring the electrolytic so it supports two AC paths through it; this I know improves the sound considerably.
4) Output stages not starved for bias current show good behaviour on transients even low level ones.
Yes, can't disagree with this, but I'm concerned more about the crossover disjunction, which happens so quickly the NFB loop can't properly correct. I actually would say that in comparisons between well biased Class AB and Class A, the Class AB seems to me to have the edge for dynamics, all other things being equal.
Thank you for your post. I welcome such input!
Cheers,
Hugh
www.printedelectronics.com
Guys, this discussion is at some parts out in the blue. Virtually EVERY CD-player has an opamp somewhere in the audio chain and the discussion wasn't about opamp-less CD-players.
I must though point out that CS4328 was a DAC with no need for an opamp. It had an excellent output stage which could perform DC! The chip was powered with +5 V digital and +-5 V analog. Very expensive but. If you check my DAC project I have an "audiophile" output, strait from the DAC chip! I think the CS4328 was the best 16,18-bit chip around (not longer available).
If anyone want datasheets, send me a message but you have to wait until monday.
I must though point out that CS4328 was a DAC with no need for an opamp. It had an excellent output stage which could perform DC! The chip was powered with +5 V digital and +-5 V analog. Very expensive but. If you check my DAC project I have an "audiophile" output, strait from the DAC chip! I think the CS4328 was the best 16,18-bit chip around (not longer available).
If anyone want datasheets, send me a message but you have to wait until monday.
the AD797 datasheet can be found here .
Look at page 8.
For the current source :
A badly made currentsource will performe badly. The currentsource is subjected to voltagemodulation. So higher "ro" you can get the lower this modulation will be. Both JFET and BJT currentsource are exposed to this problem. I think some times this is overlooked!? But i would not trade an currentsource for a resistor as common emitter resistor in a diff pair or in the VAS stage.
Sonny
Look at page 8.
For the current source :
A badly made currentsource will performe badly. The currentsource is subjected to voltagemodulation. So higher "ro" you can get the lower this modulation will be. Both JFET and BJT currentsource are exposed to this problem. I think some times this is overlooked!? But i would not trade an currentsource for a resistor as common emitter resistor in a diff pair or in the VAS stage.
Sonny
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.