WAV vs FLAC revisited

[amirm]

Mitchco, welcome to the forum. That is an exceptionally well done article indeed. Certainly goes beyond the call of duty there .

One question though. I saw how you used your Lynx card to capture the sine wave but now how you captured the FLAC and WAV files. Did you just compare them as they sat on disk, or did you play and then capture them as with the function generator signal.

Thanks and welcome again.

 

[Phelonious Ponk]

Ahhh, I felt better now: a good dose of Timism, or should that be Ponkism, goes a long, loooong way, shakes off the morning blues!!

I'm here for you, Frank.

Tim

 

[FrantzM]

Hi
I haven't read all the post so ...

THought experiment to clear up my mind ... I have an original File in WAV

I take it and convert it to FLAC ..
And back to WAV

Would the waveform be different? from the original WAV? If so we have a problem in FLAC ... Do we?

 

[amirm]

Hi
I haven't read all the post so ...

THought experiment to clear up my mind ... I have an original File in WAV

I take it and convert it to FLAC ..
And back to WAV

Would the waveform be different? from the original WAV? If so we have a problem in FLAC ... Do we?

No. That transformation is mathematically "bit exact." And will be the same no matter how many times you will go back and forth.

The scenario in question is if the *analog* output of the DAC changes if you play FLAC or WAV. Since the analog output of the DAC is dependent on both the digital samples and timing of the same, the above verification only proves half of this. To prove the latter, we need to capture the timing, or perform and analog capture of both and then compare.

 

[FrantzM]

Amirm

I haven't read the whole article, I will later but .. What i expect my DAC to do is to perform a mathematical transform of a series of numbers. If the numbers are the same, I would expect the resulting output to be the same. If the set of numbers are the same then I can assume and correctly that the timing data are the same since they are extracted from the same data/set of numbers.. What am I missing here?

 

[amirm]

I don't know that the article covers the scenario of interest and hence my question there.

As to the DAC, yes, you are missing something . The DAC gets the digital data to output alright. But it needs to know in the time continuum, when to raise or lower the output voltage to that level. Those steps are driven by a clock pulse. The design of our audio equipment puts the sender of the data in charge of when that has to happen, not the DAC. Variability in that clock directly modulates the analog output of DAC and is a form of FM Modulation which we call jitter.

The theory in this scenario is that if the PC is in charge of timing (or some other processing device), the act of decoding a format can subtly change the accuracy of the clock it sends to the other side, or noise on its power rails as to cause the DAC clock to vary and with it, its analog waveform. It is a very long shot and even longer shot that the changes are always in favor of .wav and not FLAC. But here we are .

 

[mauidan]

http://www.audiostream.com/content/further-adventures-flac

 

[Vincent Kars]

As far as I could judge, the article is about the null test.
Load the same song as on the HD in various formats in Audacity.
One of the many proofs that WAV and FLAC are bit identical at file level.

A nice one by Barry Diament, recording the SPDIF out.
Probably the ultimate test that Wav and FLAC are also bit identical in what is send to the DAC
http://www.computeraudiophile.com/content/first-test-and-some-food-audio-thought

 

[Bruce B]

As far as I could judge, the article is about the null test.
Load the same song as on the HD in various formats in Audacity.
One of the many proofs that WAV and FLAC are bit identical at file level.

I already did this and Amir said it wasn't valid.

 

[Phelonious Ponk]

The theory in this scenario is that if the PC is in charge of timing (or some other processing device), the act of decoding a format can subtly change the accuracy of the clock it sends to the other side, or noise on its power rails as to cause the DAC clock to vary and with it, its analog waveform. It is a very long shot and even longer shot that the changes are always in favor of .wav and not FLAC. But here we are

The same theory that hears hard drives and resource optimization, and computer minimization, and so many things digital. Many things are unresolved, one is certain: Give the audiophile a theory he can hang onto and he will hear things. I have one acquaintance who swears he hears dramatic differences between an optimized, dedicated music computer (same hardware, same OS, same player) and the computer electronically isolated and re-clocked outside.

Tim

 

[fas42]

Again, the point is not WAV vs. FLAC, but whether interference in some way is making its way into the analogue side of the situation. Whether you call it jitter, power supply glitches, clocking funnies or any of a million things, it doesn't really matter so long as NO interference gets through. And if you have a high resolution system it's extremely easy to hear the effects of this type of interference: it flattens, deadens the sound, makes it lose its sparkle. So it's vital to use by whatever means that work, to rid the setup of this phenomenon ...

Frank

 

[amirm]

The same theory that hears hard drives and resource optimization, and computer minimization, and so many things digital. Many things are unresolved, one is certain: Give the audiophile a theory he can hang onto and he will hear things. I have one acquaintance who swears he hears dramatic differences between an optimized, dedicated music computer (same hardware, same OS, same player) and the computer electronically isolated and re-clocked outside.

Tim

All the same Tim, I can't put forward the assertion that there is no factor here without measurement. What someone else might do from the other direction, doesn't eliminate our responsibility for being data driven, as opposed to assumption driven .

 

[rbbert]

All the same Tim, I can't put forward the assertion that there is no factor here without measurement. What someone else might do from the other direction, doesn't eliminate our responsibility for being data driven, as opposed to assumption driven .

But it seems that many of the negative or skeptical comments are indeed "assumption driven", in that there is no data on the audibility of most of these potentially interfering factors; rather there is the assumption that since many of these factors may be present, and they may be audible, then they are detrimental to the sound.

 

[fas42]

But it seems that many of the negative or skeptical comments are indeed "assumption driven", in that there is no data on the audibility of most of these potentially interfering factors; rather there is the assumption that since many of these factors may be present, and they may be audible, then they are detrimental to the sound.

Data is hard to come by because it's a dynamic phenomenon, the distortion doesn't just sit there nicely at some constant level for 30 secs or so, waiting for someone to come and measure it. This stuff is a will o' the wisp type of thing, as soon as you try and nail it, the very act of measuring may change the situation, and you're actually measuring the effect of attempting to measure! Ask someone who plays around with RF circuitry about such things ...

IF there is any interference, and the sound has changed, then by definition distortion has occurred and definitely it is then detrimental ...

Frank

 

[amirm]

But it seems that many of the negative or skeptical comments are indeed "assumption driven", in that there is no data on the audibility of most of these potentially interfering factors; rather there is the assumption that since many of these factors may be present, and they may be audible, then they are detrimental to the sound.

Well, we know as a matter of science these factors exist. We also have all of these anecdotal observations. Seems to me the natural next step is to quantify them. Saying they don't matter will do no nothing to the people chiming in otherwise as I am sure they have heard that before . I also think it will be a fun experiment .

 

[Phelonious Ponk]

All the same Tim, I can't put forward the assertion that there is no factor here without measurement. What someone else might do from the other direction, doesn't eliminate our responsibility for being data driven, as opposed to assumption driven .

I don't doubt there is a factor here, though I do suspect it is vanishingly small in properly functioning systems. Still, vanishingly small factor is reason enough for good galvanic isolation and clocking outside of the PC. I'm all for that becoming standard. Hobbling a perfectly good computer and its operating system? I won't be going there, thanks. I'll leave that to those with a desperate need to tweak.

Tim

 

[amirm]

Tim, personally I have no belief there and do not worry at all about the difference between lossless files and wave. But as I said, given all the talk around this, someone should do some testing and get us data to discuss.

 

[Phelonious Ponk]

Tim, personally I have no belief there and do not worry at all about the difference between lossless files and wave. But as I said, given all the talk around this, someone should do some testing and get us data to discuss.

I'm with you, Amir. Data. It's a wonderful thing to discuss!

Tim

 

[rbbert]

If I read the last issue of The Absolute Sound correctly, part II of their computer audio article will give some listening impressions pertinent to this issue. Not exactly objective testing...

 

[Mitchco]

Mitchco, welcome to the forum. That is an exceptionally well done article indeed. Certainly goes beyond the call of duty there .

One question though. I saw how you used your Lynx card to capture the sine wave but now how you captured the FLAC and WAV files. Did you just compare them as they sat on disk, or did you play and then capture them as with the function generator signal.

Thanks and welcome again.

Hey Amir, thanks. Yes, this is a basic null test using the files on disk...

I see there is further discussion that maybe during or after the the digital to analog conversion and analog line output stage, there is some thing causing a waveform difference between FLAC and WAV. The manufacturer of my card has these specs while installed in the computer:

ANALOG IN PERFORMANCE
(24-BIT MODE WITH CARD INSTALLED IN COMPUTER)
Frequency 20 Hz - 20 kHz, +0/-0.1 dB
Response at 44.1 kHz sample rate
Dynamic Range 117 dB, A-weighted
Signal-to-Noise 116 dB, A-weighted
Channel Crosstalk -120 dB maximum, 1 kHz signal, -1 dBFS
THD + N -108 dB (0.0004%) @ -1 DBFS
-104 dB (0.0006%) @ -8 DBFS
1 kHz signal, 22 Hz - 22 kHz BW
ANALOG OUT PERFORMANCE
(24-BIT MODE WITH CARD INSTALLED IN COMPUTER)
Frequency 20 Hz - 20 kHz, +0/-0.1 dB
Response at 44.1 kHz sample rate
Dynamic Range 117 dB, A-weighted
Signal-to-Noise 117 dB, A-weighted
Channel Crosstalk -120 dB max., 1 kHz signal, -1 dBFS
THD + N -97 dB (0.0014%) @ -1 DBFS
-104 dB (0.0006%) @ -8 DBFS
1 kHz signal, 22 Hz - 22 kHz BW

Plus since we know what ADC and DAC are being used, we can look up the jitter, etc.

But, given the specs, I am wondering what else could be measured to somehow quantify the differences some people hear? Do you have a proposal?

Cheers,

Mitch