Fosi Audio V3 op-amp rolling, has anyone tried it? Snake oil? Or are there actual differences?

[antcollinet]

To throw in something here that might muddle the discussion, consider "blind sight" research in neuroscience. There are people who, due to specific damage to the visual cortex, have areas of vision on which they cannot consciously report. Yet they are able to respond to stimuli in those areas. So if you're talking to them about what the see in these visual areas, their report is uniformly "nothing." Yet their behavior shows that they nonetheless at some other level are aware of what's before their eyes in these blind areas.

The potential complication for A/B tests in audio is that, just as there can be nonreportable differences in what we see even though our behavior demonstrates that somehow unconsciously we're still seeing, and acting accordingly, there may possibly be nonreportable difference in what we hear, which in A/B tests we cannot report, but which nonetheless have significant effects on our behavior -- where in this case behavior includes level of enjoyment of music. Given the far-from-complete theories about why music even works, that some of the work may be significantly by communication across unconscious channels can't be safely discounted.

The feeling we get from music may be affected by nonreportable channels of musical perception, a "blind" sort of perception that won't show up in "blind" A/B tests. For my part, I've noticed that as I mix and match individual speakers as well as amplifiers, that new speakers or new amplifiers on several channels of a multi-channel AV setup will sound fine to me in the overall context, at first. But within a couple of weeks I develop a clear feeling of a mismatch, where the older, lesser speaker or amp no longer fits so well. What immediate conscious impression didn't reveal, awareness of seems to percolate up to the conscious level given more time.

I like A/B tests. The concept is clear. When they do show a difference, that's something quite real. But not showing an A/B difference may be less conclusive, just if there are unconscious levels of hearing just as there are of seeing.

Any studies you can point to (such as the blind sight ones you reference) that show this is actually happening?

Because if not you are just speculating/guessing. Which is pointless.

In any case people are not claiming "unreportable" impressions or behaviours related to them. They are claiming they can hear it. Typically "night and day"

 

[antcollinet]

Exactly what I did with a little Aiyima, which allowed OPamp rolling on both its I/V and LPF functions. I tested several combinations, running 44 measurements per channel each time (THD, THD+N, IMD, bandwidth, phase, jitter, SNR, linearity, crosstalk, etc…). A Motu Ultralite Mk5 was used as the capturing device to perform measurements with REW.
A friend of mine did the same exercice using an old Yamaha CDX-390 CD player and using a Focusrite for measurements.

Here below are some results (with the Aiyima only), restricted to few measurements showing tangible differences (bandwidth and phase were always the same for instance). Note the relatively low resolution of the device anyways and fact I used 16bits/44.1kHz dithered test files as we initially wanted to test CD players only.

I’m currently repeating the same kind of tests with a DAC of much higher resolution than the Aiyima. Conclusions shared below are the same, that said.

A * next the the OPamp indicates the original one in its original slot.

View attachment 384959
View attachment 384960
View attachment 384961
View attachment 384962
View attachment 384963

In addition, we also organized ABX tests, including random people as well as audiophiles. We used two tracks of our choice (including the famous Tracy Chapman - Fast Track) and accepted any track from our audiophiles friends. We recorded the output of the DAC with high resolution audio interface, and gave the anonymized files to our testers.

From the above, you can see that the LPF slot is a little more demanding to our OPamps.

Only one of our audiophiles was able to spot the worst case scenario (here 0.4% THD) using a recording of his choice: it was a 24bits/192kHz file of the "Symphony No. 3 Op.55 Eroica IV. Finale“ and therefore slightly deviating from our initial idea to limit our tests to 16bits/44.1kHz.
I failed replicating his performance though (I tried hard). We then continued to include his file in the ongoing tests.

Conclusions have already been shared in this thread, nothing new under the sun. Don’t mess with the original design, and avoid randomly swapping OPamps, it’s useless.

We wanted to put all of that in a paper, but it’s still work in progress as we’re not done enjoying ourselves with higher end DACs and some CD players. It takes time.

Great work.

 

[MaxwellsEq]

Exactly what I did with a little Aiyima, which allowed OPamp rolling on both its I/V and LPF functions. I tested several combinations, running 44 measurements per channel each time (THD, THD+N, IMD, bandwidth, phase, jitter, SNR, linearity, crosstalk, etc…). A Motu Ultralite Mk5 was used as the capturing device to perform measurements with REW.
A friend of mine did the same exercice using an old Yamaha CDX-390 CD player and using a Focusrite for measurements.

Here below are some results (with the Aiyima only), restricted to few measurements showing tangible differences (bandwidth and phase were always the same for instance). Note the relatively low resolution of the device anyways and fact I used 16bits/44.1kHz dithered test files as we initially wanted to test CD players only.

I’m currently repeating the same kind of tests with a DAC of much higher resolution than the Aiyima. Conclusions shared below are the same, that said.

A * next the the OPamp indicates the original one in its original slot.

View attachment 384959
View attachment 384960
View attachment 384961
View attachment 384962
View attachment 384963

In addition, we also organized ABX tests, including random people as well as audiophiles. We used two tracks of our choice (including the famous Tracy Chapman - Fast Track) and accepted any track from our audiophiles friends. We recorded the output of the DAC with high resolution audio interface, and gave the anonymized files to our testers.

From the above, you can see that the LPF slot is a little more demanding to our OPamps.

Only one of our audiophiles was able to spot the worst case scenario (here 0.4% THD) using a recording of his choice: it was a 24bits/192kHz file of the "Symphony No. 3 Op.55 Eroica IV. Finale“ and therefore slightly deviating from our initial idea to limit our tests to 16bits/44.1kHz.
I failed replicating his performance though (I tried hard). We then continued to include his file in the ongoing tests.

Conclusions have already been shared in this thread, nothing new under the sun. Don’t mess with the original design, and avoid randomly swapping OPamps, it’s useless.

We wanted to put all of that in a paper, but it’s still work in progress as we’re not done enjoying ourselves with higher end DACs and some CD players. It takes time.

Paging @amirm - worthy of pinning?

 

[antcollinet]

Paging @amirm - worthy of pinning?

Seconded - if it needs it.

 

[Down South]

Name one.

I think you go wrong when you make the error that is tragically common among audio fans of using "subjective" to mean "without controls." "Subjective" only means "evaluated with human senses." "Subjective" can be valid and objectively verified (basic ears only controls) or absolutely worthless (e.g., peeking, no rigorous level-matching).

With a proper 'hearing test' conducted in a small anechoic chamber you are presented with a detailed result for each ear - no two people ever have the same result ergo each person will hear sound/music differently - this is not an 'opinion' but a scientific fact.

 

[Killingbeans]

Yes, but just because your personal hearing has a unique frequency response, it can't in any way give you the ability to hear artifacts below -120dB.

No matter how much you want "unique" to mean "superhuman", it's just not going to happen.

 

[SIY]

With a proper 'hearing test' conducted in a small anechoic chamber you are presented with a detailed result for each ear - no two people ever have the same result ergo each person will hear sound/music differently - this is not an 'opinion' but a scientific fact.

Ok, so you can’t name one.

 

[NTTY]

Hello everyone,

As I see some interests in what I published so far, here are complementary information about what we’ve done and how, when it comes to measurements:

- REW was used as an analyser with a high-definition audio interface (Focusrite 4i4 or Motu Ultralite mk5).
- All measurements were taken from REW's "Real Time Analyzer" (see table below for specific RTA calibration).
- Gain was adjusted at the interface input to ensure no clipping.
- We followed the measurement recommendations of the "AES Standard method for digital audio engineering - Measurement of digital audio equipment". Exception for the three DC level check and Dynamic Range where we used undithered signal.
- We used a burnt CD whose test files have been created with the REW Generator (update: new test CD now includes files created with Multitone (@pkane) and others suggested by @AnalogSteph).
- The CD was burnt using Windows Media Player, without "Apply volume leveling across tracks" else it adds distortion. CD burnt quality was check.
- Measurements were taken on one output channel of the DUT, then the other, and always on the same interface input to reduce crosstalk within the ADC of the interface.

I already mentioned that we used 44 test signals for the measurements. Here below is the table with the track number, file name (I think explicit enough), and purpose of that single test file (there could be more than one outcomes).

EDIT (10/08/2024): I updated my test CD, so it is no longer 44 (or 46) but now 45 measurements. Some have gone and some are new. There are now test files to verify the oversampling filter headroom of CD players, which make sense with 44.1kHz. I also added files with noise shaping to lower the noise up to 6kHz and therefore allow for higher precision THD and SNR capabilities of newer players. I also updated the JTest file to use a better source (Thanks to @AnalogSteph). I also grouped the test files per specific requirements at the capture, to reduce errors and time to measure too. The new list is below:

Track	Title	Measurements	How to
1​	A_256kMeasSweep_1_to_22050_-12_dBFS_44k_PCM16_LR_refL.wav	Bandwidth, Phase, THD vs Frequency.	Auto setup by test file (44.1kHz, 256k)
2​	A_256kMeasSweep_1_to_22050_-12_dBFS_44k_PCM16_LR_refR.wav	Bandwidth, Phase, THD vs Frequency.	Auto setup by test file (44.1kHz, 256k)
3​	B_AES17_18k_20k_1_1_-5_dBFS_44k_PCM16_LR_Dither.wav	IMD AES	Input 24bits/192kHz, FFT length 512k, BH7, 32 Averages
4​	B_AES17_41_7993_4_1_-5_dBFS_44k_PCM16_LR_Dither.wav	IMD AES	Input 24bits/192kHz, FFT length 512k, BH7, 32 Averages
5​	B_CCIF_19k_20k_1_1_-5_dBFS_44k_PCM16_LR_Dither.wav	IMD CCIF	Input 24bits/192kHz, FFT length 512k, BH7, 32 Averages
6​	B_DIN_250_8k_4_1_-3_dBFS_44k_PCM16_LR_Dither.wav	IMD DIN	Input 24bits/192kHz, FFT length 512k, BH7, 32 Averages
7​	B_ISO-1 [email protected] 0.69dB	Intersample over test	Input 24bits/192kHz, FFT length 512k, BH7, 32 Averages
8​	B_ISO-2 7350@60 1.25dB	Intersample over test	Input 24bits/192kHz, FFT length 512k, BH7, 32 Averages
9​	B_ISO-2 11025@45 3dB	Intersample over test	Input 24bits/192kHz, FFT length 512k, BH7, 32 Averages
10​	B_SMPTE_ 60_7k_4_1_-3_dBFS_44k_PCM16_LR_Dither.wav	IMD SMTPE	Input 24bits/192kHz, FFT length 512k, BH7, 32 Averages
11​	B_TDFD_41_89_1_1_-10_dBFS_44k_PCM16_Dither_LR.wav	IMD TDFD	Input 24bits/192kHz, FFT length 512k, BH7, 32 Averages
12​	B_Triple_9k_10k_20k_-10_dBFS_44k_PCM16_Dither_LR.wav	Tripple Tone IMD	Input 24bits/192kHz, FFT length 512k, BH7, 32 Averages
13​	B_Triple_9k_19k_20k_-10_dBFS_44k_PCM16_Dither_LR.wav	Tripple Tone IMD	Input 24bits/192kHz, FFT length 512k, BH7, 32 Averages
14​	B_Triple_10k5_19k_20k_-10_dBFS_44k_PCM16_Dither_LR.wav	OverSampling Filter attenuation analyzis	Input 24bits/192kHz, FFT length 512k, BH7, 32 Averages
15​	C_Sine_100_0_dBFS_44k_PCM16_Dither_L.wav	Crosstalk @100Hz	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
16​	C_Sine_100_0_dBFS_44k_PCM16_Dither_R.wav	Crosstalk @100Hz	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
17​	C_Sine_997_0_dBFS_44k_PCM16_LR.wav	THD, THD+N with prime test tone	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
18​	C_Sine_997_0_dBFS_44k_PCM16_ShapedDither_LR.wav	High precision THD check	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
19​	C_Sine_997_-16_dBFS_44k_PCM16_ShapedDither_LR.wav	THD+N (to map with SACD Denon Test)	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
20​	C_Sine_997_-20_dBFS_44k_PCM16_ShapedDither_LR.wav	High precision THD check	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
21​	C_Sine_997_-60_dBFS_44k_PCM16_ShapedDither_LR.wav	Check dB gain in SNR with Noise Shaping	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
22​	C_Sine_997_-90_dBFS_44k_PCM16_ShapedDither_LR.wav	High precision THD and linearity checks	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
23​	C_Sine_997_-110_dBFS_44k_PCM16_ShapedDither_LR.wav	High precision THD and linearity checks	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
24​	C_Sine_997_-120_dBFS_44k_PCM16_ShapedDither_LR.wav	High precision THD and linearity checks	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
25​	C_Sine_1000_0_dBFS_44k_PCM16_Dither_L.wav	THD, THD+N, Crosstalk @1kHz	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
26​	C_Sine_1000_0_dBFS_44k_PCM16_Dither_LR.wav	THD, THD+N	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
27​	C_Sine_1000_0_dBFS_44k_PCM16_Dither_R.wav	THD, THD+N, Crosstalk @1kHz	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
28​	C_Sine_1000_-3_dBFS_44k_PCM16_Dither_LR.wav	THD, THD+N	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
29​	C_Sine_1000_-6_dBFS_44k_PCM16_Dither_LR.wav	THD, THD+N	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
30​	C_Sine_1000_-9_dBFS_44k_PCM16_Dither_LR.wav	THD, THD+N	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
31​	C_Sine_1000_-12_dBFS_44k_PCM16_Dither_LR.wav	THD, THD+N	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
32​	C_Sine_1000_-20_dBFS_44k_PCM16_Dither_LR.wav	THD, THD+N	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
33​	C_Sine_1000_-30_dBFS_44k_PCM16_Dither_LR.wav	THD, THD+N	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
34​	C_Sine_1000_-40_dBFS_44k_PCM16_Dither_LR.wav	THD, THD+N	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
35​	C_Sine_1000_-60_dBFS_44k_PCM16_Dither_LR.wav	THD, THD+N	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
36​	C_Sine_1000_-60_dBFS_44k_PCM16_LR.wav	DR (undithered)	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
37​	C_Sine_1000_-80_dBFS_44k_PCM16_Dihter_LR.wav	Linearity Test (Dithered)	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
38​	C_Sine_1000_-90_dBFS_44k_PCM16_Dither_LR.wav	Linearity Test (Dithered)	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
39​	C_Sine_1000_-100_dBFS_44k_PCM16_Dihter_LR.wav	Linearity Test (Dithered)	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
40​	C_Sine_10000_0_dBFS_44k_PCM16_Dither_L.wav	Crosstalk @10kHz	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
41​	C_Sine_10000_0_dBFS_44k_PCM16_Dither_R.wav	Crosstalk @10kHz	Input 24bits/96kHz, FFT length 256k, BH7, 32 Averages
42​	D_JITTER34416.wav	Jitter Test (https://hydrogenaud.io/index.php/topic,108927.0.html)	Input 24bits/44.1kHz, FFT length 128k, BH7, 32 Averages
43​	D_Sine_1000_-90.31_dBFS_44k_PCM16_LR.wav	3 DC levels representation	Input 24bits/44.1kHz, FFT length 128k, BH7, 32 Averages
44​	E_MT_Dec_20_20000_-10_dBFS_44k_PCM16_LR.wav	Multitone (1/10 decade)	Input 24bits/44.1kHz, FFT length 256k, BH7, 32 Averages
45​	F_White_PN_128k_Full_-10_dBFS_44.1k_PCM16_Dither_LR.wav	OverSampling Filter response (up to 96kHz)	Input 24bits/192kHz, FFT length 512k, BH7, 300+ Averages

————
Flo

 

[Sokel]

Hello everyone,

As I see some interests in what I published so far, here are complementary information about what we’ve done and how, when it comes to measurements:

- REW was used as an analyser with a high-definition audio interface (Focusrite 4i4 or Motu Ultralite mk5).
- All measurements were taken from REW's "Real Time Analyzer", at 24bits/96kHz with an FFT length of 256k (Window Blackman-Harris 7, overlap 93.75%, 32 averages, No Smoothing) at the output of the DUT (Device Under Test).
- We first calibrated the interface input so that a 0dBFS signal at the DUT output is as close as possible to 0dBFS at the interface input (by adjusting the input gain).
- We followed the measurement recommendations of the "AES Standard method for digital audio engineering - Measurement of digital audio equipment". Exception for the three DC level check where we used undithered signal (-90.31dBFS), of course.
- We used a burnt CD whose test files have been created with the REW Generator.
- The CD was burnt using Windows Media Player, without "Apply volume leveling across tracks" else it adds distortion.
- Measurements were taken on one output channel of the DUT, then the other, and always on the same interface input to reduce crosstalk within the ADC of the interface.

I already mentioned that we used 44 test signals for the measurements, I’ll try to put that in a table here. I’ll edit later.

Better on a new thread with all the data gathered in one place,in this thread it's lost.

It sure deserves a front-page thread of it's own!

 

[MCH]

Hello everyone,

As I see some interests in what I published so far, here are complementary information about what we’ve done and how, when it comes to measurements:

- REW was used as an analyser with a high-definition audio interface (Focusrite 4i4 or Motu Ultralite mk5).
- All measurements were taken from REW's "Real Time Analyzer", at 24bits/96kHz with an FFT length of 256k (Window Blackman-Harris 7, overlap 93.75%, 32 averages, No Smoothing) at the output of the DUT (Device Under Test).
- We firheyst calibrated the interface input so that a 0dBFS signal at the DUT output is as close as possible to 0dBFS at the interface input (by adjusting the input gain).
- We followed the measurement recommendations of the "AES Standard method for digital audio engineering - Measurement of digital audio equipment". Exception for the three DC level check where we used undithered signal (-90.31dBFS), of course.
- We used a burnt CD whose test files have been created with the REW Generator.
- The CD was burnt using Windows Media Player, without "Apply volume leveling across tracks" else it adds distortion.
- Measurements were taken on one output channel of the DUT, then the other, and always on the same interface input to reduce crosstalk within the ADC of the interface.

I already mentioned that we used 44 test signals for the measurements, I’ll try to put that in a table here. I’ll edit later.

Hey ntty, thanks for sharing all this work. What model is the little Aiyima DUT?
BTW, there is a similar comparison in a white paper from ess now retrieved, will try to find it.
There it goes:

 

[BDWoody]

With a proper 'hearing test' conducted in a small anechoic chamber you are presented with a detailed result for each ear - no two people ever have the same result ergo each person will hear sound/music differently - this is not an 'opinion' but a scientific fact.

It is also totally irrelevant to the question at hand.

What has that got to do with expecting someone to be able to back up their unlikely claims?

Everyone is different, so it could be true... That's what you are going with when someone claims green markers make CD's sound better, or any other unlikely claim?

I can jump 30' high... We're all different, so prove I can't.

 

[NTTY]

Hey ntty, thanks for sharing all this work. What model is the little Aiyima DUT?

It is the AIYIMA Audio DAC A5 Pro.
At full output, it was showing too much a distorsion, so I had to reduce its output to 0dBu (instead of the standard 8dBu) where it behaved good enough to allow CD player equivalent resolution. Since it was our initial intention to test from CD audio, that was good enough.

————
Flo

 

[solderdude]

With a proper 'hearing test' conducted in a small anechoic chamber you are presented with a detailed result for each ear - no two people ever have the same result ergo each person will hear sound/music differently - this is not an 'opinion' but a scientific fact.

Conveniently forgetting that the brain calibrates itself and such an audiogram is only valid at the bottom detection range over a narrow bandwidth (speech related) and can be different on another day or circumstances and even later in the evening.
What audiograms can tell you if there is hearing loss (opposite a standard) and about how much at certain frequencies.
So while no 2 people have the same stimulus of the auditory system they all hear music in a similar way.

When you are standing or sitting next to someone during a life performance the same sound-waves will hit the ears and it will sound real to both.
I have never heard 2 audiophiles during a life performance bickering how one feels the bass is tighter or something like the sound is more analog or other audiophile wording.
Well... maybe both will hear similar issues when the sound is reinforced but one won't find the treble 'smoother' than the other for instance.
They both hear the sound as real and listen to the same music. Regardless if one is 20 y.o. or 70 y.o. Chances are the young person finds the sound a lot louder though but both hear cymbals as cymbals and bass as bass.

 

[NTTY]

Better on a new thread with all the data gathered in one place,in this thread it's lost.

It sure deserves a front-page thread of it's own!

Thanks for the interest and suggestion.
I’ll do that when I’m back from vacation and in front on my computer.
In the meantime, I may continue to use this thread as a preview of a dedicated thread indeed, and also to collect your potential feedback.
As I said, tests are still ongoing with higher resolution devices, even if the story is already written.

Actually, the reason why we went for these tests, using a CD player in mind, is because people thinking one OPamp can dramatically change sonic performance/perception do not tell you it has to be done on super high end devices. So we went for middle class CD player of the past (Yamaha CDX-390), as it’s easy to find including its service guide to understand how the single stereo OPamp is used (LPF). I later added the Aiyima in the picture when I discovered it allowed OPamp rolling in its "pre-amp" and "LPF" functions.

We also wanted to satisfy those who can only believe their ears and/or do not understand measurements and how they relate to listening experience. So we recorded few test tracks allowing us to share WAV files for subsequent blind tests. We recommended the utilization of Foobar ABX plugin which automatically provides the p-value as an output of the test. Anyways, shared test files were anonymized, and since they were created through our interfaces, it was not possible to cheat to identify what was what.

Here are the results with the Yamaha, less tests since there was only one OPamp, but my friend tried two mono OPamp using an adapter in couple of cases. This is when he got the best measured crosstalk, without surprise:

It is to be noted that, in the Yamaha, the OPamp is powered this way: Vcc+ = 4,94V and Vcc- = -4,92V, while in the Aiyima both are powered Vcc+ = 9,08V et Vcc- = -8,76V. This alone can make a difference even when using the same OPamp in the same circuit and just changing the power voltage. This is were the specs of respective OPamp do not tell you everything, sometimes. In that respect, some vendors have much better data sheets than others.

————
Flo

 

[MaxwellsEq]

Thanks for the interest and suggestion.
I’ll do that when I’m back from vacation and in front on my computer.
In the meantime, I may continue to use this thread as a preview of a dedicated thread indeed, and also to collect your potential feedback.
As I said, tests are still ongoing with higher resolution devices, even if the story is already written.

Actually, the reason why we went for these tests, using a CD player in mind, is because people thinking one OPamp can dramatically change sonic performance/perception do not tell you it has to be done on super high end devices. So we went for middle class CD player of the past (Yamaha CDX-390), as it’s easy to find including its service guide to understand how the single stereo OPamp is used (LPF). I later added the Aiyima in the picture when I discovered it allowed OPamp rolling in its "pre-amp" and "LPF" functions.

We also wanted to satisfy those who can only believe their ears and/or do not understand measurements and how they relate to listening experience. So we recorded few test tracks allowing us to share WAV files for subsequent blind tests. We recommended the utilization of Foobar ABX plugin which automatically provides the p-value as an output of the test. Anyways, shared test files were anonymized, and since they were created through our interfaces, it was not possible to cheat to identify what was what.

Here are the results with the Yamaha, less tests since there was only one OPamp, but my friend tried two mono OPamp using an adapter in couple of cases. This is when he got the best measured crosstalk, without surprise:

View attachment 385113

It is to be noted that, in the Yamaha, the OPamp is powered this way: Vcc+ = 4,94V and Vcc- = -4,92V, while in the Aiyima both are powered Vcc+ = 9,08V et Vcc- = -8,76V. This alone can make a difference even when using the same OPamp in the same circuit and just changing the power voltage. This is were the specs of respective OPamp do not tell you everything, sometimes. In that respect, some vendors have much better data sheets than others.

————
Flo

Fascinating, your work just keeps on giving!

 

[NTTY]

To continue on the topic, and using the few measurements I stored on my phone, one element of surprise to me, was my inability to recognize the worst case scenario, using my ears and through all recordings we used.

That said, I sometimes enjoy listening to my test tones (I know, I know…). And if I failed through music, I could spot intermodulation distortion, as you know it’s easier to hear than harmonic distortion, and of course especially if there are few simple tones only.

Here you go below with an FFT of triple tones played with one of the best case combo and the worst in the Aiyima (graph in linear frequency scale) :

Besides the increased noise floor, note the really odd frequencies appearing at 1.04kHz, 9.94kHz, 11kHz, 18kHz, 19.04kHz and 20.1kHz with the two RC4560.

I don’t have an explanation for that and I did not investigate since it’s not the primary purpose of the activity, but rather demonstrating that random OPamp swap is not such a great idea, especially when we don’t fully understand what’s the role of the OPAmp. Do they make a sonic difference is the cherry on the cake.

All that said again, despite the obvious mess above, I reminded “ear-blind” to the two RC4560 flaws when I was listening to music. This proved, at least to me, that the masking effect of music is quite strong. If I can indeed hear defaults through simple tones, it becomes much more difficult with musical content. I was expecting this result, but not to that extend.

Note that even if only one of our audiophiles could hear an issue, it means that it can be heard, when it reaches that level (we’re talking 0.4% here, if I recall correctly). On our perspective, there’s no reason to disregard his result, even if not overall statistically relevant. It can be heard.

Last note about the RC4560: nothing against this AOP, it simply did not enjoy the LPF role here. It does not imply anything else, and is a good demonstration that random OPamp rolling is not a great idea in some cases.

————
Flo

 

[bgravato]

For my part, I think that this is just your own interpretation of science.
Science isn't as impervious as you seem to be, thankfully.
You can't pretend (under the guise of 'your' science) that if you can't (for now) prove something, it doesn't exist...

You clearly seem to have a different definition of science from the rest of the world...
Usually, in science, people consider things to be proven true and that they exist (for sure) if they can prove it by an empirical method, that is reliable, consistent and repeatable.
Some science is just theoretic, but it still needs to be coherent and consistent, based on logic and making sense...

What you are suggesting is (or may be) neither of that... which I find curious, crazy, hilarious and absolute non-sense, all at the same time...

I say this because for most human beings that I know, and there are many, all agree that they hear sound differences when replacing operational amplifiers.
For example, a friend who frequently comes to my house and with whom we listen to music (he is a musician but not at all a technician) asked me what I had changed as a 'setting' on my system because he found it improved.

Claiming you hear or see something is one thing... showing evidence of that is another...
I can throw a coin in the air and if it comes out heads up, I can claim "this coin always falls heads up". In reality 100% of the times I flipped that coin it came out heads up.
But that's not science. A sample of 1 is far from enough to claim anything, isn't it?

I only believe claims, if the person claiming it can pick the right one at least 8 or 9 times out of 10 in a true double blind test...
IF they really can hear a (clear) difference, then they should be able to tell consistently which is which even if they're blindfolded, right?
Though I'm yet to see any such test with such results on any of those "less scientific" and more "subjective" "claims".

Also what does "improving" mean? Yes opamp rolling can change the sound... but what did exactly change? Did it just get more distortion/noise? And if that's what happened, is that an improvement or not?

For some people (most people on this forum), improving means less distortion/noise, so an increase in noise/distortion is automatically worse/bad.

For others, like you, seems like distortion/noise can be seen as an improvement...

Edit: Tubes/valves add distortion, that's a widely known fact. Despite of that many people like the "tube sound", because they claim that kind of distortion is actually pleasant to their ears. Nothing wrong with that... electric guitar sound is often based in adding distortion to the signal. Also many people seem to like a V-shaped kind of sound, where treble and bass are boosted... Adding that boost to treble and bass may sound better (and some people will call that an improvement), but in terms of signal processing that's actually adding distortion to the original signal, so is that truly an improvement? Since this is a science based forum, the premise here is that improvement refers to accuracy reproducing the original audio signal, rather than "coloring" or "embellishing" some how the sound that may sound more appealing to someone's ears.

The ultimate goal should be enjoying music as much as possible and if adding distortion/noise to the audio signal makes you enjoy it more, there's nothing wrong with that... and I believe, for some people, replacing opamps can have that effect. All good, but from the audio signal processing point of view, that is not technically an improvement.

Don't take my intervention as a personal attack but I just wanted to say that I find quite curious this way that some people have to deny what is obvious though ?

Don't take this as a personal attack either, but to me it's pretty obvious that you are clearly delusional in most of the things you say

And science is not designed precisely on research...

Like I said you have a very curious and different (and very wrong in my opinion) definition of what science is...

 

[JSmith]

deny what is obvious though

You've provided nothing of substance to substantiate this apparent obviousness.


JSmith

 

[NTTY]

You've provided nothing of substance to substantiate this apparent obviousness.


JSmith

Hi,

Don’t waste time with ICETDIYEUR. He complained about a lot of you guys on a Swiss forum that I’m member of. And that’s the reason why I came here to follow this thread.

Fact is that he does not speak very good English and so he’s obviously using an online translator. Problem is that not only the translation is partially wrong from the beginning, but he will also interpret and assume many things you’ve never wrote (he does that is French too).

So not only he’s lost in translation both ways, but he has that tendency to make you say what you never thought about. He did that to me on the Swiss forum no later than yesterday where he copied a small part of what I published here, but the translation was wrong and misleading (that’s how I understood he’s using online translation). On top of that, his interpretation of what I wrote was completely out of the blue and out of scope, of course.

On this Swiss forum, as well as here, he’s been asked to stop quoting partially people, not naming who he is quoting, and to stop assuming things. He does not care. He’s not here (or there) to share or discuss, just to be disruptive in a wrong way.

 

[NTTY]

To come back to the topic, and if you looked at the results I published, it seems the LM4562 is at ease when being used in either the Aiyima DAC or the Yamaha CD player. Its resilience is to be noted… except when it followed the NE5532 (in Pre-Amp slot of the Aiyima). If the LME49720 is in “Pre-amp” slot, then the LM4562 behaved correctly. Could it be that the output impedance of the NE5532, being 30x that of the LME49720, is the cause? I do not know and I will let much more qualified people than me have a look and comment.

One more time, this means nothing is granted and there are risks associated to random selection and replacement of OPAmps.

Last but not least, I also wanted to come back on the listening experience and tentative to explain why those unhappy OPamps remained undetectable, at least to me.

As you’ve seen, in our tests, we checked the THD+N at different levels. And we also identified in several cases that those heavily distorted OPamps behaved normally at lower levels, generally starting at -20dBFS, sometimes earlier.

Here below three examples from my buddy in his Yamaha CD Player, at 0dBFS, -12dBFS and -20dBFS. Remember there’s only one stereo OPAmp in the Yamaha (whose function is LPF as per Yamaha’s service guide).

The two THD+N measures are side by side, left with the NE5532 and right with the OPA2604. At 0dBFS:

Then at -12dBFS, ENOB just above 9bits (ouch):

And finally at -20dBFS, we’re almost back to normal (ENOB=15bits):

So, with not so much compressed music and lots of headroom in the recording, as it recommended, it will render the detection of this OPA2604 a little more challenging. In addition, the increased distortion at higher levels could very well create a sonic difference on sudden high levels and which could find its way into our ears as a preference. It was not the case here, but it could be with another recording, maybe one using distortion intentionally.

Food for thoughts.

————
Flo