Blue or red pill?

[adamdea]

1 minute ago, sandyk said:

 

 That's the $64 question !

Maybe. Maybe not. I asked PeterSt whether he thought it could survive and ADC/DAC chain and he seemed to say he thought it could. If so the files can be distributed and the experiment can easily be repeated using an established protocol .

[PeterSt]

14 minutes ago, adamdea said:

If the tascam is not capable of recording that effect, why does it matter what an FFT shows? Sorry if I have misunderstood.

 

Adam, I think it is you who is saying that it may matter what an FFT shows. I don't see Mani saying that (at least not lately). In what you quoted from him he explicitly says the opposite.

[adamdea]

Just now, PeterSt said:

 

Adam, I think it is you who is saying that it may matter what an FFT shows. I don't see Mani saying that (at least not lately). In what you quoted from him he explicitly says the opposite.

No I'm not saying that. I'm saying that the only thing that matters is whether the effect can be recorded. If it can -fine lets AbX it. If it can't then ex hypothesi no FFT of the recording will be relevant. 

Let's just stick to the point. 

[PeterSt]

In general, people always think that what is audible can also be captured by a random (or better) ADC (btw this includes Mani, usually ). But I wouldn't know why this is true. Look :

 

The changes applied (to the output of the DAC) are so minuscule that we may wonder whether they can be captured by anything, properly (yes, on analogue tape). So envision that the output includes noise and let's (definitely) assume that noise is part of the audible game here. What is the frequency of this noise ?

Shall we state it is higher than the best capable audio ADC hence higher than 384KHz ? Mind you, this noise is not a frequency as such, but it appears for its peaks and dips at a way high frequency and depending on the bandwidth downstream, you will notice some of it. Also it will influence your amplifier and in the end your tweeter too.

And this we now try to capture with a 176.4 / 192Khz ADC.

 

All what it will capture for real is differences. So the sampling is way too low but it will capture a spike here or there, accidentally. And especially when the higher frequencies are not emphasized (read : less present) in the source file it will be a too difficult task for the ADC and in the end your own sensory system.

 

I also think that because of what happens for real during the ADC process and what is captured for real, the differences are not really "correlated" as such. I mean, you won't be able to say that the one sounds crispier than the other or such. Different probably yes (because the recordings ARE different - this has been proven if all is right, by Mans).

 

 

[PeterSt]

9 minutes ago, adamdea said:

If it can -fine lets AbX it. If it can't then ex hypothesi no FFT of the recording will be relevant. 

 

Huh ?

 

If an FFT shows differences but you can't perceive that, then you are deaf or your system is not on par.

Aparenty you think the other way around, somehow.

 

If an FFT does  not show differences, the sound can still be vastly different.

(I'd say that everybody knows this and it is a sad thing it is so)

[adamdea]

4 minutes ago, PeterSt said:

Different probably yes (because the recordings ARE different - this has been proven if all is right, by Mans).

 

 

Not that I have seen. I have have seen no analysis showing that the variance in As is significantly less than the variance in ABs 

[adamdea]

2 minutes ago, PeterSt said:

 

Huh ?

 

If an FFT shows differences but you can't perceive that, then you are deaf or your system is not on par.

Aparenty you think the other way around, somehow.

 

If an FFT does  not show differences, the sound can still be vastly different.

(I'd say that everybody knows this and it is a sad thing it is so)

Peter I find that when posting on the internet there is always a risk of misunderstanding. But as between us the risk appears be around 99%. Please can you read a little more carefully. I'm sure there is some chance of meaningful exchange, but life is too short to try to deconstruct half of what you read into my posts.

It really is this simple:

I am only interested for now in whether anyone can ABx a difference in the recorded files. If not I see no point in analysing the files further because the only point in doing so would be to see whether the audible effect could be identified whether by FFT or any other technique.

 

 

[fas42]

Sorry, I've had to deal with those boring other things in life for the last day or so, have not read the recent posts  - just fired up Audacity, and almost immediately came across a distinct variation in a waveform peak, beween A and B - in the analogue captures of course - enough to be audible, was it a foible of the recorder? Is there a pattern, or was it a single glitch? Will have little time to investigate further for a day or so, but there is a first tiny glimmer of something ...

[PeterSt]

9 minutes ago, fas42 said:

just fired up Audacity, and almost immediately came across a distinct variation in a waveform peak, beween A and B

 

Frank, if it was only well after 50ms in the file(s) ...

[jabbr]

1 hour ago, PeterSt said:

If an FFT does  not show differences, the sound can still be vastly different.

(I'd say that everybody knows this and it is a sad thing it is so)

That’s only because the FFT is not done adequately. 

 

There is a mathematical equivalence between a signal and its Fourier transform. 

 

Assuming you accept that if two signals are identical that  they will sound identical — I hope everyone agrees!!! — then the proper acquisition and limits on the FFT will be also identical and conversely that if two signals are different and sound different, then the properly acquired and limited FFT will also be different.

 

I hope everyone is seeing that getting a proper FFT is more than pushing a button and believing what pops up on the screen.

[PeterSt]

6 minutes ago, jabbr said:

and conversely that if two signals are different and sound different, then the properly acquired and limited FFT will also be different.

 

There we go again. 

Take your DAC, take my software, grab your tool and show us.

 

So Yes you are right. But No you are wrong because it is theory only again (with existing practice you won't be able to show - nobody will). Only if you show it, it is practice.

 

Hint : I tried.

 

So here we are (and the ideas about it become stronger and stronger) :

 

Mans and Mani set up a hoax for us. It is a hoax because nobody can show any real measurement showing the difference.

Claim A : When we hear a difference an FFT can show it.

Exhibit B : The FFT shows no difference so we can't hear a difference.

 

 

Quote

then the properly acquired and limited FFT

 

Yes, I read that.

[opus101]

19 minutes ago, jabbr said:

That’s only because the FFT is not done adequately.

 

Consider the case where the signal's time-reversed. The magnitude (square root of sum of squares) of the FFT will be identical but obviously the sound will change fairly substantially.

[jabbr]

7 minutes ago, opus101 said:

 

Consider the case where the signal's time-reversed. The magnitude (square root of sum of squares) of the FFT will be identical but obviously the sound will change fairly substantially.

Phase opposite. FFT that includes only magnitude and not phase, or real but not complex, is thus clearly inadequate 

[opus101]

Sure, but who is plotting FFT phase in any graphics here? I've only seen magnitude plots.

[manisandher]

26 minutes ago, jabbr said:

That’s only because the FFT is not done adequately. 

 

It looks to me like @testikoff did the analysis adequately:

 

Wouldn't you agree?

 

If a higher resolution ADC, say 32/705.6, had been used, do you think the delta would have been any different in this plot?

 

Mani.

[PeterSt]

1 hour ago, adamdea said:

1 hour ago, PeterSt said:

Different probably yes (because the recordings ARE different - this has been proven if all is right, by Mans).

 

 

Not that I have seen. I have have seen no analysis showing that the variance in As is significantly less than the variance in ABs 

 

I assume that you understand what you are talking about. I read very carefully but all I see  is turning things upside down. But I said it : it becomes too difficult (my English as well ).

Mans showed the difference in the recordings of the both (SFS) situations of the test signal, remember ? the test signal was there for a reason, namely the sheer fact that that could show something while all else (for music recording) would probably not. This difference IS there. It bears no value (if you ask me) but the difference is there with proof.

So what do you want ? ... testify that this shown difference can't exist because people can't hear (or differentiate) the difference sufficiently ? Something which was all predicted (btw by Mani as well who now can't do it himself from the recorded files).

 

What could be difficult to understand or grasp is that the recording of the test signal is allowed to be representative for a same difference in the recorded music files. So, true, Mans did not show anything from those, but which is impossible.

Well, not quite, IF you followed my 10 posts about exactly that with plots and all (2, 3 weeks ago).

 

So it really is not easy and at least I don't blame anyone for that. And I hope that the example of your quote shows that it is easy to misread each other (both having assumptions of how our world looks like).

I will still try to read better.

[jabbr]

12 minutes ago, PeterSt said:

There we go again. 

Take your DAC, take my software, grab your tool and show us.

 

So Yes you are right. But No you are wrong because it is theory only again (with existing practice you won't be able to show - nobody will). Only if you show it, it is practice.

 

Hint : I tried.

Let’s see, suppose I want my measuring ADC to have 10x the resolution of the DAC I am measuring (this is a standard measurement rule of thumb) 

 

What specs? 7 Msps x ? 20 bits ... or perhaps 7 Gsps x 16 bits ...

 

Could be done ... or I might be able to coax by HP contraption while ? how many weeks/months work?

 

To prove to you that Fourier transforms work ?

 

Anyways how did you try? Do you have an ADC that’s “better” than your DAC ??

[jabbr]

19 minutes ago, opus101 said:

Sure, but who is plotting FFT phase in any graphics here? I've only seen magnitude plots.

Where to start? So ... throwing out 1/2 the information ... ??‍♂️

 

Telling me that you cant see the differences that you can hear ??‍♂️

 

Lets see: we are looking at a phenom that is likely phase related but throwing out the phase info? ??‍♂️

 

This isn’t just theory, it’s math

[PeterSt]

4 minutes ago, manisandher said:

If a higher resolution ADC, say 32/705.6, had been used, do you think the delta would have been any different in this plot?

 

Honest answer ? This is hard to tell because the granularity is way too low.

First zoom in on the Y-axis 20 times or so.

 

Remember what I told way earlier in the thread : the differences are not about "dB's" as such. It's that low that no dB number will express it (because it is an average (SPL) over time (and frequency) and it is merely at the noise level.

 

If anything, I'd say that the differences emerging beyond 14KHz or whatever it is, are not real. They are too large.

And they indicate exactly what Mans (I think it was him) mentioned yesterday : when not aligned properly this shows more the higher the frequency.

Of course we all like to see that the difference which is audible in the crispiness of the lower SFS exhibits as higher frequency, but trust me, you won't see it back like that. Btw :

 

15 hours ago, manisandher said:

So, the analogue captures are virtually identical up to 14kHz. My ears are good to 12kHz nowadays. And yet I heard clear differences in the A/B/X.

 

[jabbr]

25 minutes ago, PeterSt said:

Mans and Mani set up a hoax for us. It is a hoax because nobody can show any real measurement showing the difference.

Claim A : When we hear a difference an FFT can show it.

Exhibit B : The FFT shows no difference so we can't hear a difference.

 

No Peter, I’ve said that the analysis does not use the proper technique. I’m not saying that doing this analysis the right way would be a casual endevour. 

[PeterSt]

5 minutes ago, jabbr said:

What specs? 7 Msps x ? 20 bits ... or perhaps 7 Gsps x 16 bits ...

 

Could be done ...

 

Yes. But sadly the bit depth is not sufficient.

So *now* you suddenly talk about speed = possible jitter measurement. But I never said that could not be done. That is only lacking equipment (OK, $).

OK, you didn't say jitter for real, but we could make it that. Anyway that is not an FFT which requires the bit depth and btw memory depth. AND speed.

The combination does not exist.

 

9 minutes ago, jabbr said:

Anyways how did you try? Do you have an ADC that’s “better” than your DAC ??

 

No. Which is exactly my pose.

 

Side note, for fun and FYI : Yes I have. 32 bits, SNR of 160dB or so BUT 4K sampling rate only. I really have this on an existing board and with some FPGA trickery you will be able to envision that this will be able to do what we want here. Mind you, this is (thus) explicitly NOT jitter but is FFT stuff. Reads a 44.1 file  in ten times the real length (or more when I oversample more than it does itself) and then what ...

The board is made for something very different, but in the back of my mind is this application.

[STC]

5 hours ago, PeterSt said:

 

OK STC, you take that part on you, OK ? Let us know whether you hear a difference and what it is.

 

Ok. Just tried with my main system. There is a slight slight shift of the vocal between the two version. Even the piano had slight bite in one version. But this is with JRiver. When I play the file in the native format the difference is probably more. I use probably because it is hard to do DBT. Will use headphones with Foobar in pc and update later. 

 

Btw, can anyone check the phase difference between the left and right channel on both version ?

[manisandher]

5 minutes ago, STC said:

Just tried with my main system.

 

Just to be absolutely clear...

 

Are you comparing the two digital captures to each other, or the two analogue captures to each other?

 

Mani.

[STC]

5 minutes ago, manisandher said:

 

Just to be absolutely clear...

 

Are you comparing the two digital captures to each other, or the two analogue captures to each other?

 

Mani.

 

File 3 and 4. I preferred 3. Are the digital? Then placebo is works well on me.  

[adamdea]

2 hours ago, jabbr said:

 

 

There is a mathematical equivalence between a signal and its Fourier transform. 

 

 

This is obviously the case. It shouldn't need saying.

The thread is all over the place since there are least two different points which are getting mixed up. 

1) is there an audible difference in the files. So far I see no evidence in favour. At least 2 negative ABx

2) if so does that correspond with any measurable characteristic.

 

Almost unbelievably, so far as I can understand it, some participants are wishing to draw conclusions about 2) without having established 1), or even whilst possibly answering 1) in the negative. It is difficult to keep up with the convolutions (no pun intended).