Can you hear a difference between OPA2134 and LME49720 in the PRE-TC10?

[pma]

The op-amps OPA2134 and LME49720 were swapped in the Audiophonics PRE-TC10 preamp and the files (matched and aligned) were recorded and uploaded to:

2134-49720.zip

drive.google.com

Can you hear any difference? This is how they measure:

And this is the preamp:

 

[pma]

BTW, this is what I got from Foobar ABX DBT, no comment:

foo_abx 2.1 report
foobar2000 v2.0
2025-04-05 15:41:11

File A: 2134.wav
SHA1: be3c81c4a4dd185262c511bcb0447d7f8c6da138
File B: 49720.wav
SHA1: ec38050c91e3d36050a8a4d580c51324ca3beb4f

Output:
ASIO : Topping USB Audio Device
Crossfading: NO

15:41:11 : Test started.
15:42:15 : 01/01
15:42:44 : 01/02
15:42:54 : 02/03
15:43:04 : 02/04
15:43:22 : 02/05
15:43:33 : 03/06
15:43:52 : 04/07
15:44:04 : 05/08
15:44:21 : 06/09
15:44:33 : 06/10
15:44:45 : 07/11
15:45:01 : 08/12
15:45:13 : 08/13
15:45:23 : 09/14
15:45:46 : 10/15
15:46:07 : 11/16
15:46:07 : Test finished.

 ----------
Total: 11/16
p-value: 0.1051 (10.51%)

 -- signature --
040e38d97b904f8d1b2dd365b0edd44c46a963dc

 

[DanielT]

Good job PMA. I don't hear any difference. By the way, are you supposed to hear any difference? In that case, I blame my results on my substandard HiFi.
If the outcome is instead meant to show no audible difference, then I have superb HiFi instead.
(I actually have ok/mediocre HiFi)

 

[pma]

Good job PMA. I don't hear any difference. By the way, are you supposed to hear any difference? In that case, I blame my results on my substandard HiFi.
If the outcome is to show no difference, then I have superb HiFi stuff instead.
(I actually have ok/mediocre HiFi)

My mind is open . I "think" I can hear differences in woman voice sibilants . You also get them as a Deltawave delta waveform. Paul? (@pkane )
BTW, Deltawave says this:

DeltaWave v2.0.13, 2025-04-05T15:53:05.4871763+02:00
Reference: 2134.wav[L] 3155000 samples 48000Hz 24bits, stereo, MD5=00
Comparison: 49720.wav[L] 3155000 samples 48000Hz 24bits, stereo, MD5=00
Settings:
Gain:True, Remove DC:True
Non-linear Gain EQ:False Non-linear Phase EQ: False
EQ FFT Size:65536, EQ Frequency Cut: 0Hz - 0Hz, EQ Threshold: -160dB
Correct Non-linearity: False
Correct Drift:True, Precision:30, Subsample Align:True
Non-Linear drift Correction:True
Upsample:False, Window:Hann
Spectrum Window:Kaiser10, Spectrum Size:262144
Spectrogram Window:Hann, Spectrogram Size:4096, Spectrogram Steps:2048
Filter Type:FIR, window:Hann, taps:8192, minimum phase=False
Dither:False bits=0
Trim Silence:True
Enable Simple Waveform Measurement: False

Discarding Reference: Start=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -4.941dB Comparison: -4.967dB
Initial RMS values Reference: -26.747dB Comparison: -26.747dB

Null Depth=56.996dB
Trimming 0 samples at start and 0 samples at the end that are below -90.31dB level

X-Correlation offset: 0 samples
Trimming 0 samples at start and 0 samples at the end that are below -90.31dB level

Drift computation quality, #1: Excellent (0.09μs)


Trimmed 3122 samples ( 65.041667ms) front, 74 samples ( 1.541667ms end)


Final peak values Reference: -4.941dB Comparison: -4.94dB
Final RMS values Reference: -26.746dB Comparison: -26.746dB

Gain= 0dB (1x) DC=0 Phase offset=-0.001712ms (-0.082 samples)
Difference (rms) = -82.92dB [-84.58dBA]
Correlated Null Depth=89.24dB [84.84dBA]
Clock drift: -0.01 ppm


Files are NOT a bit-perfect match (match=53.03%) at 16 bits
Files are NOT a bit-perfect match (match=0.38%) at 24 bits
Files match @ 49.9857% when reduced to 16.2 bits


---- Phase difference (full bandwidth): 0.573095621379185°
0-10kHz: 0.21°
0-20kHz: 0.45°
0-24kHz: 0.57°
Timing error (rms jitter): 10.6ns
PK Metric (step=400ms, overlap=50%):
RMS=-110.7dBr
Median=-112.8
Max=-104.4

99%: -104.73
75%: -109.02
50%: -112.82
25%: -117.38
1%: -122.37

gn=1.00000435592333, dc=5.23198169699686E-08, dr=-5.92159560203277E-09, of=-0.0821710331450599

DONE!

Signature: 2206d5e5bd9772a4321ce98d7fd85954

RMS of the difference of spectra: -141.080276006535dB
DF Metric (step=400ms, overlap=0%):
Median=-69.4dB
Max=-48.3dB Min=-84.6dB

1% > -83.68dB
10% > -76.85dB
25% > -72.92dB
50% > -69.41dB
75% > -64.44dB
90% > -56.42dB
99% > -46.71dB

Linearity 27.7bits @ 0.5dB error

10ns jitter, 27.7bits linearity/0.5dB. That's not bad matching. No manipulation was done except for cutting the files to the same start and length.

And Pkmetric:



rms 110.7 dBr .......

 

[audio_tony]

Now here's a thought. If like me, you don't expect to hear any difference (and you don't) is that expectation bias as well?

Because expectation bias can work both ways I guess.

I think for tests like this, the best way to gather opinions is simply to publish the two wav files and then ask if anyone can hear any difference, and then reveal what you are comparing once results are in.

I seem to recall that's how folk did this on DIY audio a few years back.

 

[DanielT]

PMA,

When do you think you could hear clear differences when swapping op amps in HiFi gear? If it is even possible under any circumstances?
Not if you pop in op amp X in Y that is not technically designed for it and it goes to hell. That is an exception, a poorly thought out op amp swap.

I think more in terms of for example old type of op amp in some phono pre amp swap to a new type of op amp. If you then can measure such large differences that it EVENTUALLY becomes audible? That is, where there is at least a potential that there can be audible differences?

 

[pkane]

My mind is open . I "think" I can hear differences in woman voice sibilants . You also get them as a Deltawave delta waveform. Paul? (@pkane )
BTW, Deltawave says this:



10ns jitter, 27.7bits linearity/0.5dB. That's not bad matching. No manipulation was done except for cutting the files to the same start and length.

And Pkmetric:

View attachment 442098

rms 110.7 dBr .......

Unlikely that any of this is audible. The highest levels in the delta waveform are around 40-41sec mark at -53dBFS. These contain mostly energy above 10kHz, and are masked by the actual signal that's at about -8dBFS peak:

On a LUFS scale:

And almost invisible on PKMetric scale because it is a high-frequency burst that's weighted down by the equal loudness curves and the masking by the main signal:

 

[pma]

Unlikely that any of this is audible. The highest levels in the delta waveform are around 40-41sec mark at -53dBFS. These contain mostly energy above 10kHz, and are masked by the actual signal that's at about -8dBFS peak:

I can make the same plots with YOUR software. But rather than telling what would be or not be likely/unlikely audible, shall we have more ABX results? They are always asked to be posted, but if there are properly matched and aligned test files, there is scarcely someone willing to try. Instead, we can read guesses if it would or would not be audible. On the other hand, videos with elementary recording mistakes attract attention of many readers.

Note: the AA mixdown confirms that the biggest difference is in energy of the sibilants, exactly as is observed by listening.

 

[pkane]

I can make the same plots with YOUR software. But rather than telling what would be or not be likely/unlikely audible, shall we have more ABX results? They are always asked to be posted, but if there are properly matched and aligned test files, there is scarcely someone willing to try. Instead, we can read guesses if it would or would not be audible. On the other hand, videos with elementary recording mistakes attract attention of many readers.

Note: the AA mixdown confirms that the biggest difference is in energy of the sibilants, exactly as is observed by listening.

View attachment 442211

My apologies, since you invited me to post here, I had assumed you wanted more DeltaWave results

 

[Noricha]

I can make the same plots with YOUR software. But rather than telling what would be or not be likely/unlikely audible, shall we have more ABX results? They are always asked to be posted, but if there are properly matched and aligned test files, there is scarcely someone willing to try. Instead, we can read guesses if it would or would not be audible. On the other hand, videos with elementary recording mistakes attract attention of many readers.

Note: the AA mixdown confirms that the biggest difference is in energy of the sibilants, exactly as is observed by listening.

View attachment 442211

Is the blue chart for OPA2134?

 

[amirm]

Can you hear any difference? This is how they measure:

I didn't get an answer to my question about your measurements. Why is there a step down in distortion for the one in green at the high frequency end? Did you not use the same measurement bandwidth for both?

 

[KL....]

With there being so many repeatable conclusions (thousands was indicated which is a large dataset, isn't it?), that Anticipation/Expectation Bias does occur, then lets say, of those repeatable conclusions, that it is 50/50 (that is 50% are genuine/objective repeatable conclusions and 50% are genuine/subjective Anticipation/Expectation Bias).

50% of thousands is still a lot, isn't it? What does this mean, well, it raises the question that needs to be asked (@pma @solderdude @amirm), is FFT (a very useful tool, isn't it?) 100% accurate in all situations and beyond any doubt or is it not quite 100% (perhaps 80% or 90%) accurate? At this time, the 100% accurate in all situations and beyond any doubt, is the accepted assumption/practice, so lets ask the question/s for the not quite 100% accurate....

• FFT is Interpolation of analytic measurement/algorithms, isn't it?
• who created FFT?
• how did they decide that it was 100% accurate in all situations and beyond any doubt?
  • If no it was not 100% (perhaps 80% or 90%) then what assumptions have/were made and what are/were the assumptions and how do they affect Interpretation of the Interpolation of analytic measurement/algorithms?.... this is reasonable, isn't it?
  • If yes, 100%, then it (still) needs to be shown that the Interpretation of the Interpolation was a reasonable and a repeatable conclusion. This is to determine the level of Trust and where Trust can be considered as Known/Accepted, beyond any doubt?.... this is reasonable, isn't it?
• when they created FFT what setup did they use to determine that it is 100% accurate in all situations and beyond any doubt?
• are there any other considerations that can (need to) be clarifyed that FFT is 100% accurate in all situations and beyond any doubt?

 

[solderdude]

Fast Fourier transform - Wikipedia

en.wikipedia.org

Guide to FFT Analysis (Fast Fourier Transform)

This is the ultimate guide to FFT analysis. Learn what FFT is, how to use it, the equipment needed, and what are some standard FFT analyzer settings.

dewesoft.com

But here's a question for you.
Pavel went to the trouble of creating a blind test using 3 different op-amps and all else being equal.
He (as well as Amir) went to the trouble of measuring and analyzing the results and posting those.

Why don't you simply do a blind test using Foobar abx plugin and post the results ?
This, at least, will tell you if your hearing is sensitive/good enough to hear the measured differences.
See if you can get 100% or even 80% accuracy using your hearing only.

 

[KL....]

Fast Fourier transform - Wikipedia

en.wikipedia.org

Guide to FFT Analysis (Fast Fourier Transform)

This is the ultimate guide to FFT analysis. Learn what FFT is, how to use it, the equipment needed, and what are some standard FFT analyzer settings.

dewesoft.com

But here's a question for you.
Pavel went to the trouble of creating a blind test using 3 different op-amps and all else being equal.
He (as well as Amir) went to the trouble of measuring and analyzing the results and posting those.

Why don't you simply do a blind test using Foobar abx plugin and post the results ?
This, at least, will tell you if your hearing is sensitive/good enough to hear the measured differences.
See if you can get 100% or even 80% accuracy using your hearing only.

Thank you, let's have a read to see if they suggest that FFT is100% accurate in all situations and beyond any doubt....

Know that I am not suggesting that ASR patrons MUST follow FFT nor that they MUST follow blind test via Foobar abx plugin but I and they are being asked to (even blindly so, 100%), aren't we? So, best to establish that this is a reasonable request to do so, isn't it?

Also, know that I respect and appreciate yours, @pma, @amirm advice and efforts (very much) and due diligence/process is always important, isn't it? FFT is a very useful Tool, isn't it?, and measurements are very important, use and setup/application/implementation are even more important, aren't they?

 

[solderdude]

Note: it is math and the final result is dependent on the used ADC quality, recording length (amount of available data) and sample frequency as well as the many possible settings.

I still don't get the '100 percent accuracy' and 'doubt' part though.

 

[Scrivs]

Thank you, let's have a read to see if they suggest that FFT is100% accurate in all situations and beyond any doubt....

Know that I am not suggesting that ASR patrons MUST follow FFT nor that they MUST follow blind test via Foobar abx plugin but I and they are being asked to (even blindly so, 100%), aren't we? So, best to establish that this is a reasonable request to do so, isn't it?

Also, know that I respect and appreciate yours, @pma, @amirm advice and efforts (very much) and due diligence/process is always important, isn't it? FFT is a very useful Tool, isn't it?, and measurements are very important, use and setup/application/implementation are even more important, aren't they?

Solderdude,
Why do you continue to ket this donkey troll you?

Edit: I mean, the more lures he throws out there, the more Burley you ladle out which sustains his argumentative idiot-cycle.

 

[KL....]

Solderdude,
Why do you continue to ket this donkey troll you?

So, the insults begin.... unfortunate?

Perhaps, you can answer this question.... is FFT is100% accurate in all situations and beyond any doubt?

 

[KL....]

Note: it is math and the final result is dependent on the used ADC quality, recording length (amount of available data) and sample frequency as well as the many possible settings.

I still don't get the '100 percent accuracy' and 'doubt' part though.

this

It is very clear that there are differences between op-amps in specs and actual performance in certain circuits.
Measurements and data sheets show this. That is not under debate (or should not be).

The question is would there be differences that reach audible levels in any particular circuit when performance is not limited by the differences between the used (and suitable fairly high performance) op-amps.

Can different op-amps, that have better performance of the rest of the device it is used in, lead to the usual flowery sound descriptions ?
So far I don't think that is nor will ever be possible to be 'proven' without a shadow of a doubt.

and asking.... reasonable isn't it?
edit: Let's endeavour to make it possible (without a shadow of a doubt), if possible, with yours, @pma, @amirm expertise/s :=) Otherwise the same question/s will be asked over, over, and over again, won't they?

With there being so many repeatable conclusions (thousands was indicated which is a large dataset, isn't it?), that Anticipation/Expectation Bias does occur, then lets say, of those repeatable conclusions, that it is 50/50 (that is 50% are genuine/objective repeatable conclusions and 50% are genuine/subjective Anticipation/Expectation Bias).

50% of thousands is still a lot, isn't it?

 

[Scrivs]

I won't pretend to know that answer with 100% certainty.

However, no matter how succinctly, and in how ever many ways your questions are answered, you insist on regurgitating essentially what sounds like LLM generated responses replete with the fails, in an attempt to muddy up whatever information thats presented to you.

Good luck wasting other people's time going forward. I, for one won't be one of 'em.

 

[amirm]

• FFT is Interpolation of analytic measurement/algorithms, isn't it?
• who created FFT?
• how did they decide that it was 100% accurate in all situations and beyond any doubt?
  • If no it was not 100% (perhaps 80% or 90%) then what assumptions have/were made and what are/were the assumptions and how do they affect Interpretation of the Interpolation of analytic measurement/algorithms?.... this is reasonable, isn't it?
  • If yes, 100%, then it (still) needs to be shown that the Interpretation of the Interpolation was a reasonable and a repeatable conclusion. This is to determine the level of Trust and where Trust can be considered as Known/Accepted, beyond any doubt?.... this is reasonable, isn't it?
• when they created FFT what setup did they use to determine that it is 100% accurate in all situations and beyond any doubt?
• are there any other considerations that can (need to) be clarifyed that FFT is 100% accurate in all situations and beyond any doubt?

None of these comments belong here as they are generic questions. Please post them in dedicated thread: https://www.audiosciencereview.com/...are-measurements-everything-or-nothing.29062/