Process to AB test op-amps

[davmol]

I recently picked up two Fosi V3 monos and I have some op-amps that I received a year ago for a different device but never used.

I've watched with interest the videos and measurements from Amir on this subject, alongside YouTubers extolling the many benefits of these op-amp swaps, so I'd like to conduct some listening tests if my own.

On that basis I'd like to plan how I might do an AB test with a reasonable degree of control.

Here are my initial thoughts:
1. Use a single speaker, to avoid speaker unit variation. Will this require mono test tracks?

2. Use a single (left or right) channel from WiiM ultra, out to RCA switching box. There shouldn't be channel imbalance from the ultra but will assume this takes that factor out of the loop.

3. Take same single channel out from the switching box and into each V3 mono using XLR (identical rca to trs cable will be used). Using XLR input eliminates one of the op-amps used with RCA input, and I think Fosi states that the remaining two op-amps are the ones that have an effect on the tone / "sound signature".

4. Use 1khz test tone and multimeter to check voltage / output level match on each V3 mono. If there is a difference at this stage then I'll have to review options for correction, such as putting the amps on separate L/R channels and using balance adjustments on the wiim, which removes benefits of step 2, but I can't think of an easier alternative yet.

5. Connect speaker output from each V3 mono into ET30 splitter / vu meter box, as separate amplifiers, again using single channel only.

6. Connect the ET30 speaker output to the single speaker.

7. Complete an AB test with the original op-amps still in each V3 mono, to check for any audible differences in the amps themselves (not expecting unit variation but just checking).

8. Assuming all ok then swap out both op-amps in one of the V3 monos.

9. Repeat output voltage measurements, record any change and level match if necessary.

10. Repeat AB test to listen for audible differences between the different op-amps.

11. If audible differences are noted and not explainable, then proceed with blind test (process tbc later, as only if required).


For clarity, as with my previous test using DACs, my only objective here is to satisfy my own curiosity, and hopefully a side benefit is it helps others who might want to do the same.

I don't intend to execute any testing until August as I've got a lot of other stuff going on, so that gives plenty of time to get the process clear, and I welcome any input / recommendations to improve it. Thanks!

 

[DVDdoug]

I wouldn't even bother... If there's a difference it's likely be noise but noise should be inaudible unless the op-amp is used in a high gain microphone preamp or a phono preamp (where any noise is greatly amplified). There might be measurable differences but line-level audio is about the easiest thing in world to make better than human hearing, especially with op-amps.

But anyway... If you hear an obvious noise difference, you're done.

If there's not some gross defect or problem, sighted listening tests are pretty-much useless for this kind of thing. Even Amir admits to fooling himself. What is a blind ABX test? ABX testing will require assistance from someone else. And since the results need to be statistically valid, they have to be repeated. It's time consuming and tedious, especially if your assistant/partner is not as interested as you.

It might also be worthwhile to read Audiophoolery which describes the FEW REAL characteristics of "sound quality", which for electronics are noise, distortion, and frequency response. (Speakers and room acoustics get more complex.) That will help to avoid listening for all of the undefined nonsense you might be reading about.

and level match if necessary.

Because the in-circuit ("closed loop") gain depends on resistors, not the op-amp, you shouldn't have to level match.* It "doesn't hurt" to level-match with a test-tone and a meter, and some people will demand it if you claim to hear a difference. ...I wont "demand" anything, but I won't believe you and I'll ignore you unless you say something like, "I'm hearing a buzz", or "the bass goes-away completely, or something like that. Of course, don't touch the volume control. It's also a dead give-away and useless if "A" is on the left and "B" is in the right...

Or if there is a level difference, that IS a valid audible difference. The main reasons for level matching is that sometimes level differences are heard-identified as something else... "more dynamic", etc. And, if there are audible level differences it's a dead give-away and you can always identify "X" in an ABX test.



* P.S. it probably is a good idea to make sure both amps match.

 

[IngMishael]

Don't. It's a fruitless venture. Moving your speakers 3cm towards or away from the front wall has a bigger effect than OP Amp changes

 

[wwenze]

Head position results in actual differences, just ask anyone with a measurement mic.

That's why any and every of your 11 steps are susceptible to false positive.

For op-amp testing. I'd suggest running two amps in parallel (with a series resistor added before the output/feedback node), each amp can be removed using a switch that does not click or make any noise. Have a person switch the op-amps on the fly without telling you. At the end of the testing, answer how many times the op-amps were switched.

 

[solderdude]

I recently picked up two Fosi V3 monos and I have some op-amps that I received a year ago for a different device but never used.

I've watched with interest the videos and measurements from Amir on this subject, alongside YouTubers extolling the many benefits of these op-amp swaps, so I'd like to conduct some listening tests if my own.

On that basis I'd like to plan how I might do an AB test with a reasonable degree of control.

Here are my initial thoughts:
1. Use a single speaker, to avoid speaker unit variation. Will this require mono test tracks?

2. Use a single (left or right) channel from WiiM ultra, out to RCA switching box. There shouldn't be channel imbalance from the ultra but will assume this takes that factor out of the loop.

3. Take same single channel out from the switching box and into each V3 mono using XLR (identical rca to trs cable will be used). Using XLR input eliminates one of the op-amps used with RCA input, and I think Fosi states that the remaining two op-amps are the ones that have an effect on the tone / "sound signature".

4. Use 1khz test tone and multimeter to check voltage / output level match on each V3 mono. If there is a difference at this stage then I'll have to review options for correction, such as putting the amps on separate L/R channels and using balance adjustments on the wiim, which removes benefits of step 2, but I can't think of an easier alternative yet.

5. Connect speaker output from each V3 mono into ET30 splitter / vu meter box, as separate amplifiers, again using single channel only.

6. Connect the ET30 speaker output to the single speaker.

7. Complete an AB test with the original op-amps still in each V3 mono, to check for any audible differences in the amps themselves (not expecting unit variation but just checking).

8. Assuming all ok then swap out both op-amps in one of the V3 monos.

9. Repeat output voltage measurements, record any change and level match if necessary.

10. Repeat AB test to listen for audible differences between the different op-amps.

11. If audible differences are noted and not explainable, then proceed with blind test (process tbc later, as only if required).


For clarity, as with my previous test using DACs, my only objective here is to satisfy my own curiosity, and hopefully a side benefit is it helps others who might want to do the same.

I don't intend to execute any testing until August as I've got a lot of other stuff going on, so that gives plenty of time to get the process clear, and I welcome any input / recommendations to improve it. Thanks!

Distortion and frequency response is solely determined by the output stage that comes after the input op-amp buffer stage.
The performance of as good as any op-amp (with the exception of a few totally NOT suited op-amps) is vastly superior of that of the output stage.
So ... in all cases you will be listening to the output stage only and that does not change.

This means any form of controlled testing will (when done in a proper way) will never reveal any audible differences.
Any uncontrolled and 'sighted' testing is highly likely to show audible differences ranging from 'barely audible' to 'night and day' depending on who does the 'testing'.

 

[DVDdoug]

That's why any and every of your 11 steps are susceptible to false positive.

That's where statistical blind tests help. i.e. If you can identify "X" 9 out of 10 times in an ABX test, there's only about a 1% chance that you got it right by guessing or "dumb luck". If you get it right 10 out of 10 times the odds of guessing correctly are about 0.1%.

...Or electrical measurements. But we can measure things that we can't hear so you need some "feel" for when things become audible. I'm NOT an expert on that and different people have different hearing ability.... within normal human limits. I've got a 100MHz oscilloscope on my bench and I could build a preamp that goes into the MHz range (at least to 1MHz easily) but so what?

 

[davmol]

I understand about blind and ABX etc but if I can't hear any difference in the first place when sighted then there is no point in a blind test, as I would just be guessing. These steps are just to do a simple AB first with at least some basic rules to minimise external factors, and then if I were to hear a difference then I proceed with an ABX blind test under stricter rules.

 

[wwenze]

I understand about blind and ABX etc but if I can't hear any difference in the first place when sighted then there is no point in a blind test, as I would just be guessing. These steps are just to do a simple AB first with at least some basic rules to minimise external factors, and then if I were to hear a difference then I proceed with an ABX blind test under stricter rules.

Our point is that the variations that are already present while simply doing nothing outweighs any minimization of external factors you are trying to do with your 11 steps.

It will help if you understand how much this variation is. Then you can save the 11 steps.

 

[davmol]

I'm still not sure I understand the variations being discussed, but are you saying that it simply isn't possible to do an AB test between two amps of the same architecture and find that they sound the same (due to these variations being mentioned)? Again, I'm not looking to write a paper or prove anything to anyone, I'm just doing it for my own enjoyment.

 

[Speedskater]

If there are differences, either measurable or audible, it will come down to:
a) Op-amps that are poorly chosen for the circuit at hand.
b) Ringing or oscillation (maybe well outside the audio range).