AES Paper Digest: Do Audio Op-amps Sound Different?

[Krunok]

@amirm My English is really being stretched to its limits here, so please feel free to help if additional explanation will be needed.

 

[oivavoi]

Made perfect sense. Thanks!

 

[Krunok]

Made perfect sense. Thanks!

My pleasure.

 

[Arnold Krueger]

The most obvious red herring in the test is the fact that the op amps are being run with a closed loop gain of 100 (40 dB) and approximately flat response. IOW response less than 3 dB down at 20 kHz. Audio op amps are usually run with a gain of 10 (20 dB) or less. Unity gain ( 0 dB) is not unusual. The high gain operation vastly reduces the inverse feedback available for reduction of distortion.

 

[tomelex]

Amir pointed out:
Notice that they ran the signal through two of the fixtures shown before. So the total picture is the op-amp being in the signal path four times. It is not explained why they had to do this.


Frankly, this test was very amateurish for pro audio engineers. THD+N should never be used as a test when spectrum analysis and multitone (or IM) as mentioned by others can show what is really coming out of a system. The fact that differences could be heard between op amps is nothing new. And folks seemingly preferring distortion, well, just play a sample of 1 KHz from your digital source (that would be a nearly perfect waveform) and play a 1Khz tone from a test LP, and, wait for it, a lot of folks would prefer that LP distorted version. And this applies to a whole lot of tones as in music. Now, however, if one were asked which tone sounded the "cleanest" then the digital would win every time.

Is this what the AES has to offer these days?

 

[amirm]

Is this what the AES has to offer these days?

This is a conference paper so it means it has less rigor and usually a lot shorter as compared to papers for the Journal of AES.

 

[Blumlein 88]

It doesn't look mysterious to me. All save one of the significant results had some form of distortion in the -40 db range. Most of those had at least one op amp with IMD in the -40 db range. Also it isn't clear are the measurements of one pass thru the devices or two. If its one, then all save one of the significant results have distortion somewhere above 1 %. It appears IMD is more noticeable more often, and distortion up into the 40 db range is heard.

Result #10 is an outlier in that it doesn't show any higher level of distortion. Results #13 and #14 had IMD in the -40 db range, but weren't detected.

Not sure what their thinking is behind doing the circuits this way. Looks like they should have studied Jung's op amp book more thoroughly. It doesn't appear to mimic how the devices normally get used.

So distortion is audible well above 1%, who knew????

 

[Arnold Krueger]

Amir pointed out:
Notice that they ran the signal through two of the fixtures shown before. So the total picture is the op-amp being in the signal path four times. It is not explained why they had to do this.


Frankly, this test was very amateurish for pro audio engineers. THD+N should never be used as a test when spectrum analysis and multitone (or IM) as mentioned by others can show what is really coming out of a system. The fact that differences could be heard between op amps is nothing new. And folks seemingly preferring distortion, well, just play a sample of 1 KHz from your digital source (that would be a nearly perfect waveform) and play a 1Khz tone from a test LP, and, wait for it, a lot of folks would prefer that LP distorted version. And this applies to a whole lot of tones as in music. Now, however, if one were asked which tone sounded the "cleanest" then the digital would win every time.

Is this what the AES has to offer these days?

The McGill U audio group has a reputation among many of the AES members around here as being willing to do a wide range of things in order to court interest from audio's high end. :-(

I took advantage of my AES membership to post a comment as follows:

"
The paper seems to be testing Op amps in circuits where closed loop gains are far higher (100 or 40 dB) than are commonly used in consumer and professional gear.

Most consumer and professional gear uses regular op amps (as opposed to special high gain differential op amps such as those designed for mic preamps) with closed loop gains in the range of unity gain to a gain of 10 or so.

One apparent exception to op amp applications typically having closed loop gains of less than 10 is RIAA preamplifiers, but their closed loop gain often takes the approximate form of a low pass filter, and so their feedback at high frequencies is fallling in a similar fashion as the open loop gain of the op amp. In the same application there are a few RIAA preamps that use passive equalization, but they generally use separate op amps for buffering the inputs and outputs, and so the burden of providing gain is essentially split among two separate devices.

Furthermore, today most audio gear today is either a digital converter or based on digital processing, and so there is only one op amp stage at the input for buffering an ADC, and at most 2 (if there is a separate V/I converter and buffer) at the ouput. This paper seems to be using a test circuit with 4 stages of op amps.
"

The AES discussion page for this paper is: https://secure.aes.org/forum/pubs/conventions/?ID=717 . I haven't had a lot of luck getting my comments posted there lately, as it appears that criticism is not exactly honored.

 

[SIY]

This is a very interesting paper, thanks for bringing it up. A couple small flaws I might have pointed out if I were a referee. Several have noted that the frequency responses were not maintained equal- and the test subjects were relatively young. Several have also noted that the distortion was relatively high, which should perhaps also have been dealt with. The conclusion that the detected differences were due to the distortion is not supported by the data- it might be true, but there's no evidence for it. Their decision to not disclose the important detail of which opamps were used is, to me, a major flaw- full disclosure is needed when making claims of this sort.

But the other missing factor takes me back to a simple experiment I mentioned in an article I did in Linear Audio a few years back where I daisy-chained some opamps configured as buffers to see how many it took before I could reliably distinguish whether they were in or out of circuit. If memory serves, it was about five, and I invited the readers to guess what the difference was that I was able to perceive. As far as I'm aware, not one guessed correctly, and at least one more mendacious sort of fellow was highly bothered by the results. I will reveal it here, especially because it relates back to disclosing opamp identities so that a good correlation could be drawn: noise floor.

 

[Krunok]

This is a very interesting paper, thanks for bringing it up. A couple small flaws I might have pointed out if I were a referee. Several have noted that the frequency responses were not maintained equal- and the test subjects were relatively young. Several have also noted that the distortion was relatively high, which should perhaps also have been dealt with. The conclusion that the detected differences were due to the distortion is not supported by the data- it might be true, but there's no evidence for it. Their decision to not disclose the important detail of which opamps were used is, to me, a major flaw- full disclosure is needed when making claims of this sort.

But the other missing factor takes me back to a simple experiment I mentioned in an article I did in Linear Audio a few years back where I daisy-chained some opamps configured as buffers to see how many it took before I could reliably distinguish whether they were in or out of circuit. If memory serves, it was about five, and I invited the readers to guess what the difference was that I was able to perceive. As far as I'm aware, not one guessed correctly, and at least one more mendacious sort of fellow was highly bothered by the results. I will reveal it here, especially because it relates back to disclosing opamp identities so that a good correlation could be drawn: noise floor.

But as those opamps were configured as buffers there gain was 1, right? In that scenario THD of the amp is really small so there's nothing much to add up in that chain beside noise. Or..?

 

[SIY]

That's right, though I did not measure distortion in my experiment (it was really meant as a demonstration of a controlled listening test method). It was indeed likely low, but it was not measured nor did I (at least consciously) perceive any distortion.

Like I said, the relatively high distortion in the McGill preprint might be the audible factor, but that's not demonstrated by the data. It could be frequency response and, I'm throwing in noise as another possible variable that the authors should have addressed by disclosing opamp identities.

 

[Krunok]

That's right, though I did not measure distortion in my experiment (it was really meant as a demonstration of a controlled listening test method). It was indeed likely low, but it was not measured nor did I (at least consciously) perceive any distortion.

Like I said, the relatively high distortion in the McGill preprint might be the audible factor, but that's not demonstrated by the data. It could be frequency response and, I'm throwing in noise as another possible variable that the authors should have addressed by disclosing opamp identities.

Sure, it may be noise, but I actually believe distortion was the audible factor in their experiment, but only because they used opamps in a way that emphasized distortion (as has already being pointed out). Your experiment, on the other hand, emphasized adding noise. As the truth is pretty much always somewhere in the middle (and out there! ), I believe that opamps, when used properly don't show perceptible distortion nor noise, nor they can be differentiated.

 

[SIY]

Again, that could indeed be right, but the authors got to that conclusion without demonstrating this it was right. The frequency response factor could also be part of it. So could noise. That's the part where, to me, the preprint went wrong, not looking at all the factors that could lead to positive identification or at the minimum, providing enough disclosure so the reader can balance the probabilities..

 

[Krunok]

Again, that could indeed be right, but the authors got to that conclusion without demonstrating this it was right. The frequency response factor could also be part of it. So could noise. That's the part where, to me, the preprint went wrong, not looking at all the factors that could lead to positive identification or at the minimum, providing enough disclosure so the reader can balance the probabilities..

You are absolutely right - they made a conclusion but no demonstrated facts are actually supporting it. And yes, frequency response and noise may also be the factor, as it has not been proven that they weren't. All in all, not much of a scientific values can be found in that paper.

 

[Arnold Krueger]

Again, that could indeed be right, but the authors got to that conclusion without demonstrating this it was right. The frequency response factor could also be part of it. So could noise. That's the part where, to me, the preprint went wrong, not looking at all the factors that could lead to positive identification or at the minimum, providing enough disclosure so the reader can balance the probabilities..

As far as frequency response goes, these are the FR plots from the article:

and this is a graph of frequency response tolerances necessary for the absence of intrusion on a listening test:

I'll interpret the curves from the McGill paper as showing a 1/3 wide variation at 20 KHz on the order of 2 to 5 dB. According to the graph above it would take more than 4 dB variation (gain or loss) to fail the criteria for matching which is conservative - it is much less than actual expected audibility. It's near or on the edge, but I would expect it to not be a strong effect. OTOH the threshold for audibility of isolated artifacts is around 70 dB (again conservative) and a number of the tests seem to fail that criteria. As others have mentioned numbers for distortion from IM were given, I don't recall the nature of the IM test being mentioned. THD is pretty standard but there are at least two mainstream tests for IM being SMPTE and CCIR. and they are IMO very different.

 

[SIY]

Extrapolating particulars from the general (especially a "general" which is pretty old) is not something I'd want to hang my hat on. These guys have some work to do before they can get this past a referee, assuming they ever bother to try.

 

[oivavoi]

Extrapolating particulars from the general (especially a "general" which is pretty old) is not something I'd want to hang my hat on. These guys have some work to do before they can get this past a referee, assuming they ever bother to try.

My understanding is that the Gaskell brothers have finished their PhDs, have left academia and now work in the audio industry (making graphene transducers and some analog equipment). So I doubt that they will put much effort into getting this old preprint published. If they do publish anything, I expect it to be on the acoustical properties of graphene, and not opamps, as the former seems to hold much more promise for the future.

 

[Krunok]

My idea of testing if audio op-amps sound different or not wouldn't be pushing op-amps to 40dB gain to cause high distortion nor to chain buffers until noise start to appear but to take some standard circuitry (like that one in D10 DAC/preamp where 2 op-amps are doing I-V amplification and the 3rd is doing low pass filtering (I guess) and than you start exchanging op-amps and test if there is a difference in sound. My guess is nobody could hear the difference between op-amps. My guess is also that nobody will make such test any time soon as such test would also imply that there may be no noticeable difference between $100 DAC/preamp, $1500 DAC/preamp and $5000 DAC/preamp, and I'm pretty sure nobody would like to hear that, not in audio manufacturing industry but also not the consumers, who like to think they can "improve" the sound not only by changing the op-amp but also by changing the power cable.

 

[Arnold Krueger]

My idea of testing if audio op-amps sound different or not wouldn't be pushing op-amps to 40dB gain to cause high distortion nor to chain buffers until noise start to appear but to take some standard circuitry (like that one in D10 DAC/preamp where 2 op-amps are doing I-V amplification and the 3rd is doing low pass filtering (I guess) and then you start exchanging op-amps and test if there is a difference in sound. My guess is nobody could hear the difference between op-amps. My guess is also that nobody will make such test any time soon as such test would also imply that there may be no noticeable difference between $100 DAC/preamp, $1500 DAC/preamp and $5000 DAC/preamp, and I'm pretty sure nobody would like to hear that, not in audio manufacturing industry but also not the consumers, who like to think they can "improve" the sound not only by changing the op-amp but also by changing the power cable.

Let me put it this way.

I have a D10 on order (alleged to arrive next week) and I do have a stash of fairly typical and modern audio op-amps on hand. I have test gear with far lower residuals than what I've posted results for, thus far. I have a stash of DIP sockets that sometimes get soldered into existing gear. I also have a suite of files of music that has historically been known for its diagnostic properties...

 

[Krunok]

Let me put it this way.

I have a D10 on order (alleged to arrive next week) and I do have a stash of fairly typical and modern audio op-amps on hand. I have test gear with far lower residuals than what I've posted results for, thus far. I have a stash of DIP sockets that sometimes get soldered into existing gear. I also have a suite of files of music that has historically been known for its diagnostic properties...

Sounds perfect - all I can think of you'll be needing in addition to that is a good bottle of wine, but that can be easily arranged!