Does Op-amp Rolling Work? [Video]

[DualTriode]

PEAQ is used to evaluate an audio codec vs the original signal, what does that have to do with testing an amplifier?

Think of the amplifier (op-amp) as the codec and compare output to input.
The PEAQ algorithm uses filters to model the average human ear and FFT does the calculations.

AI, not.
MATLAB Machine Learning maybe.

 

[levimax]

Think of the amplifier (op-amp) as the codec and compare output to input.
The PEAQ algorithm uses filters to model the average human ear and FFT does the calculations.

AI, not.
MATLAB Machine Learning maybe.

That doesn't make sense to me because the purpose of an op-amp amp is to accurately amplify the original signal, if the signal is the same as the original except louder it would be perceived the same as the original signal. There is no need that I can see to evaluate an op-amp like a Codec where parts of the original signal are discarded to "save space".

 

[amirm]

The PEAQ algorithm uses filters to model the average human ear and FFT does the calculations.

Audio objective metrics are rather poor as far as correlation with hearing. My team would only use them to test to make sure nothing was broken. Otherwise, all evaluation of our signal processing/codecs was done by ear.

 

[amirm]

The SINAD application in the APx500 software can use some help. As previously stated, a APx555 audio analyzer with a GRAS microphone is not a human ear. Upgrade as in get the add ons.

You sure you know what thread you are in? We are discussing the performance of an amplifier, not headphone or speaker. No GRAS fixture was used or would hav been appropriate (its noise floor would dwarf amplifier's).

Measurements show identical performance as far as the signal coming out of the amplifier. This confirms no audible difference.

 

[DualTriode]

You sure you know what thread you are in? We are discussing the performance of an amplifier, not headphone or speaker. No GRAS fixture was used or would hav been appropriate (its noise floor would dwarf amplifier's).

Measurements show identical performance as far as the signal coming out of the amplifier. This confirms no audible difference.

This does not confirm no audible difference.

Amplifier performance is important. There is much more to it. When your FFT zooms in on harmonics we can see that there are visible differences in the harmonic sequences. The SINAD numbers may calculate the same. Also think modulation side bands.

There was no GRAS headphone measurement fixture mentioned. I have one here on my shelf.
If you read my comment it was about a (calibrated Laboratory Grade) GRAS measurement microphone, it could be B&K as well. Think GRAS model number 46AG. I have a pair, plus others.

What op-amp rolling is about, are there reliable perceived sound differences among a handful of op-amps as they are swapped in and out of a amplifier?

Amplifier SINAD and FFT measurements alone can not tell you that there are any op-amp caused human perceptible sound differences coming from of the speaker or not. PEAQ in the basic form uses filters (think transfer function) and FFT to model the function of a average human ear.

Audio Precision has it's own version(s) of PEAQ.

 

[solderdude]

When your FFT zooms in on harmonics we can see that there are visible differences in the harmonic sequences

Look at the levels these differences are at. Then prove that you can actually hear a few -105dB tones together with a 105dB louder fundamental.
It is mentioned in the plot by Amir. Also some of the measured differences might also come from tolerances between channels or layout differences.

Best to try to get your point across on the actual measurements thread. Comments in this thread should be about the video ?

 

[antcollinet]

This does not confirm no audible difference.

Amplifier performance is important. There is much more to it. When your FFT zooms in on harmonics we can see that there are visible differences in the harmonic sequences. The SINAD numbers may calculate the same. Also think modulation side bands.

There was no GRAS headphone measurement fixture mentioned. I have one here on my shelf.
If you read my comment it was about a (calibrated Laboratory Grade) GRAS measurement microphone, it could be B&K as well. Think GRAS model number 46AG. I have a pair, plus others.

What op-amp rolling is about, are there reliable perceived sound differences among a handful of op-amps as they are swapped in and out of a amplifier?

Amplifier SINAD and FFT measurements alone can not tell you that there are any op-amp caused human perceptible sound differences coming from of the speaker or not. PEAQ in the basic form uses filters (think transfer function) and FFT to model the function of a average human ear.

Audio Precision has it's own version(s) of PEAQ.

So you seem to be suggesting that changing the op amp can (EDIT : audibly) change the harmonic content of the distortion while somehow - almost miraculously - keeping the THD essentially the same, and even more miraculously doing so across the power range of the amp, AND the frequency range of the amp.

And this despite the fact that the engineering know how tell's us that swapping two op amps (where both are appropriately specced for the application) is exceptionally unlikely to have any significant impact on distortion at all - let alone managing this near miraculous co-incidence.

We need to be careful of "all things are possible" thinking while missing the fact that most of those possibilities are so improbable, they can be safely ignored.

The test shows that there has been no change to disotortion charcteristis of the amp. Not to THD and neither to the harmonic content.

 

[DualTriode]

See the notes on the image below image below:

Douk Audio A5 1 kHz Test Tone
-Stock op-amp has higher , high order distortion
-But levels are near or below threshold of hearing

Now turn up the the volume and the higher, high order distortion of the stock op-amp is more than audible even without the likely added modulation distortion.

Point made.

 

[SIY]

Now turn up the the volume and the higher, high order distortion of the stock op-amp is more than audible

Really? With a full-scale fundamental? That makes absolutely zero sense.

 

[antcollinet]

Really? With a full-scale fundamental? That makes absolutely zero sense.

Exactly.

And @Dual triode Ive added the work "audibly" to more accurately reflect my intent.

 

[ssashton]

I have an amp with -90dB noise floor. I can bearly hear any hiss even with my ear next to the speaker and with no music playing. How would anybody hear a noise -105dB being masked by music cranked out at maximum volume so loud the neighbours are calling the cops?

Now I like good measuring and performing gear as much as the next graph sniffer, but let us not pretend we can hear such small signals.

Edit: Though I think Dual Triode hinted at sum and difference tests. Indeed a null test where the two signals are mixed out of phase to leave only the difference can be quite revealing.

 

[antcollinet]

Edit: Though I think Dual Triode hinted at sum and difference tests. Indeed a null test where the two signals are mixed out of phase to leave only the difference can be quite revealing.

Difficult to get right though unless you use specialised software to align and level match accurately.

And in this case, still completely pointless.

 

[amirm]

Difficult to get right though unless you use specialised software to align and level match accurately.

And in this case, still completely pointless.

Even harder as the performance of these amp is impacted by heat and ground loops. Even if you put the same opamp back in, the null would show differences. To wit, someone noticed that SINAD was higher here than the original A5 review.

 

[solderdude]

Now turn up the the volume and the higher, high order distortion of the stock op-amp is more than audible even without the likely added modulation distortion.

Suppose you reproduce the 1kHz tone at a VERY uncomfortable 110dB SPL... the higher order harmonics would be at 0dB SPL. So at the audible threshold around 1kHz in an anechoic room where one is acclimated for quite a while in order to detect constant tones at these levels... while your ears are being blasted at 110dB at 1kHz.
Let me tell you a constant 1kHz tone at 110dB is no fun... now if that were 10Hz that would not be as deafening but hearing things at the absolute hearing treshold would even then be a tall order.
Nah... there is nothing audible about harmonics at that level.

 

[DualTriode]

Really? With a full-scale fundamental? That makes absolutely zero sense.

Full scale on the dBrA scale?
Is that 1volt. 5watts or what?

Looking at the usual ASR dashboard FFT in the review video, full scale on the dashboard FFT dBrA scale is 5 watts.

So, turning up the volume above 5 watts does make sense.

 

[The_Fall]

As a layman, I see op-amp rolling as an extension of the broader audio hobby—very similar to tube rolling in valve amps. The idea of swapping out components to "tune" the sound feels like a natural evolution of the enthusiast mindset. My gut tells me this entire concept was likely created—or at least amplified—by clever marketing to capitalize on the FOMO of Class D amp owners. It also reminds me of how some companies push capacitor upgrades as a way to "optimize" performance.

Rolling tubes can make pretty major changes... different but not better so its a matter of tastes for some hobbyists.

Rolling is more about personal preference (or placebo) than a guaranteed improvement. Much like boutique capacitors, and high-end fuses, the appeal is often in the tweaking itself rather than a fundamental change in performance. And at the level you all are explaining in terms of Class D, feels pretty suspect on the motivations of these companies.

That said, I get why people enjoy experimenting—it’s part of the fun of the hobby. But I also can’t shake the feeling that a lot of this is just another way for companies to keep enthusiasts engaged and spending.

 

[antcollinet]

So, turning up the volume above 5 watts does make sense.

That 5W at typical listening difference might represent about 95dB average SPL. Lets say you take it up to 110dB SPL - +15dB or around 160W

Suppose you reproduce the 1kHz tone at a VERY uncomfortable 110dB SPL... the higher order harmonics would be at 0dB SPL. So at the audible threshold around 1kHz in an anechoic room where one is acclimated for quite a while in order to detect constant tones at these levels... while your ears are being blasted at 110dB at 1kHz.
Let me tell you a constant 1kHz tone at 110dB is no fun... now if that were 10Hz that would not be as deafening but hearing things at the absolute hearing treshold would even then be a tall order.
Nah... there is nothing audible about harmonics at that level.

EDITED TO ADD. Thinking you can hear distortion products at -110dB is like being in a nightclub - you know, one of those where your mate has to shout one inch from your ear to make himself heard - and you still can't make out what he is saying, and then seeing across the room, a down feather drop onto the back of a kitten, and thinking you can hear the impact.

 

[craigenputtock]

It wasn't actually. It was genuinely motivated by the many comments in my Douk A5 review. I thought my usual audience knows that these things don't make a difference. Then realized that I had last tested these a long time ago so decided to do a fresh one in the context of what was asked (the A5).

Have you heard the clips posted by New Record Day (and Danny of GR) purporting to prove the differences? If so, any comments?

 

[JustJones]

Have you heard the clips posted by New Record Day (and Danny of GR) purporting to prove the differences? If so, any comments?

You can start at the linked comment and read on through. Looks like the files weren't level matched which will make them sound different. Looks like one is clipping.

Post in thread 'Serious Question: How can Op Amp rolling improve the sound of an already well measuring device? Are that many confused? (Master Thread)' https://audiosciencereview.com/foru...any-confused-master-thread.61010/post-2256982

 

[DanTheMan]

Let's give Danny the benefit of the doubt and say he didn't deliberately put his finger on the scale and make his recommendations progressively louder, then something is unusual with those 'artisanal' op-amps as they are louder in the recordings... Maybe someone would know how that could be possible without a deliberate cheat? The most recommended one is actually 2dB louder and clipping the recording. This is now 2 of these op-amp comparisons by true believers that are tainted in the same manner in favor of the desired conclusion. A 2dB boost over the entire audible band should be so easy to distinguish, even an audiophile's wife could hear it.