PCM VS FLAC ; Personal Experience
Jun 27, 2014 at 8:47 PM Thread Starter Post #1 of 35

JosephTheGreat

New Head-Fier
Joined
Jun 27, 2014
Posts
3
Likes
10
Hello audiophiles of the planet that float on emptiness.
 
I usually read your website, I always found interseting wars between FLAC aficionados & WAV aficionados.
 
Music is my life, It's the friend that never left me, forever faithful.

I am from Casablanca, I am not rich to affroad MBL gear, I am just poor question of money.

I always listened to Music trough walkman, always felt that walking & listening to music was kind of miracle !

I never heared top end gear, but I think the best music quality that my ears ever heard was trough Sony Minidisc, still have two. I tell you that to let you know that my ears are well trained & I can fell differences in dynamic of music.
 
I know secret of great music is SPEAKERS & QUITNESS, so listen to music in early morning or late at night will make great difference, if you don't have a studio in your home
 
I created this account, to let you know my personal opinion about WAV vs FLAC.

My gear is simple :
= PC windows 8 onboard audio chip, 44.1 24bits (24bits to avoid noise)
=cheap 2.1 speakers
=same VOLUME LEVEL, same options
 
Don't say anything about my gear, I am not a your level. 
Nothing Fancy, see !
 
Music that I listen, I always listen to, I find great pleasure listening to Julio, gipsy kings, george michael, foreigner, ten sharp.
 
I download or RIP a CD to FLAC album that is 100% from CD
 
I listen to it with FOOBAR, great music dynamics AWESOME, I am far from mp3 & aac.
then I decompress the FLAC to WAV, & I play it with WMP, DAMN that's heaven, DYNAMICS & REAL, it is BETTER you can't deny it no placebo !!!
 
Yes decompressed FLAC is REPLICA of the original WAV, it is scientifically real.
but Playback WAV sounds fuller, yes flac is good, better than MP3, but I want the BEST in music reproduction, I don't accept concessions, I don't care about HDD space !  & WAV can be tagged, I don't need cover or fancy tag stuff, all i care is music.
 
I believe that CD PCM 16bits 44.1khz is a PERFECT techncology, more than that you are wasting your money
 
If you adore MUSIC why stand for less, keep it WAV . Everything is converted to PCM before playback.
Flac is good for storage though
 
Now listening to YOU; ten sharp in WAV

You
You are always on my mind
You
You're the one I'm living for
You
You're my ever lasting fire
 
 
sorry for language errors, it typed this fast at 01:50 UTC+1 I am such a geek :)
 
Jun 28, 2014 at 1:48 PM Post #2 of 35
Flac and wav result in identical PCM streams being sent to the sound card. This is not debatable.

If (for whatever reason) your computer is so old and underpowered or poorly designed that the marginal cpu load imposed by decomressing flac vs wav results in increased electrical noise, then the problem isn't flac vs wav.

A $35 raspberry pi computer can play flac and wav equally perfectly. I suggest you buy one.

Cheers


Ps, if you can hear a difference between flac and wav, you should post abx results demonstrating so.
 
Jun 28, 2014 at 3:05 PM Post #3 of 35
 
I listen to it with FOOBAR, great music dynamics AWESOME, I am far from mp3 & aac.
then I decompress the FLAC to WAV, & I play it with WMP, DAMN that's heaven, DYNAMICS & REAL, it is BETTER you can't deny it no placebo !!!

The differences you hear might be explained by that FLAC requires more computing power to decompress the file, and some minor CPU lags can really make a difference with the sound (actually reading WAV file from disk could take more time than decoding FLAC but let's assume it doesn't matter). But that would mean there is something wrong with your PC, not the format itself.
 
If you are sure it's not placebo then why not try to prove it first by doing some ABX testing with Foobar ABX plugin? If you can pass it positively with confidence >99% then your observation might be very interesting.
 
About 44kHz being perfect, I'm not so sure -  according to Nyquist you can store up to 22kHz, and it's above human hearing abilities, but there is a roll of, if I remember good, at >18kHz, depending on which phase you are (unf)fortunate to sample (especially at 22kHz, if the sampling goes at sin(pi*k), which is zero) so in theory you are losing some audible information.
 
Jun 28, 2014 at 3:37 PM Post #4 of 35
The Nyquist-shannon sampling theorem is pretty clear: any bandlimited signal sampled at more than twice the highest frequency component is mathematically perfectly captured. That includes phase and amplitude.

To help prove it to yourself (within truncation error) i recommended playing around with some test signals in matlab and sampling them at different rates and sample rate converting them.

44.1kHz captures every frequency < 22.05 kHz to within whatever bit depth you use.

Cheers

Aside: Implementations of antialiasing filters require steep filters which are difficult to do. Dont confuse implantation challenges with limitations of the format.
 
Jun 28, 2014 at 3:55 PM Post #5 of 35
  The differences you hear might be explained by that FLAC requires more computing power to decompress the file, and some minor CPU lags can really make a difference with the sound (actually reading WAV file from disk could take more time than decoding FLAC but let's assume it doesn't matter). But that would mean there is something wrong with your PC, not the format itself.
 
If you are sure it's not placebo then why not try to prove it first by doing some ABX testing with Foobar ABX plugin? If you can pass it positively with confidence >99% then your observation might be very interesting.
 
About 44kHz being perfect, I'm not so sure -  according to Nyquist you can store up to 22kHz, and it's above human hearing abilities, but there is a roll of, if I remember good, at >18kHz, depending on which phase you are (unf)fortunate to sample (especially at 22kHz, if the sampling goes at sin(pi*k), which is zero) so in theory you are losing some audible information.

 
I've only ever had playback issues when using USB soundcards on really old crappy systems and even then the difference was dropouts rather than qualitative differences. I don't know if anyone has shown any good quality evidence that FLAC playback is in any way qualitatively different, however it is something you could test fairly easily by resampling the analog outputs bunging them through an audio program and looking for differences or by the DBT test you suggest.
 
As for 44K sampling to date only one AES paper (Pras and Gusatavino) has suggested that there might be audible differences between different sampling rates, most published papers  do not find any such difference. One paper (Oohashi et al) tentatively found a difference in the physiological responses when cutting off ultrasonics but this has never been replicated and Ashihara et al found that the effect may have been due to IMD between the different speaker units woofer/midrange and supertweeter. Way back in 1978 researchers from JVC (Sampling-Frequency Considerations in Digital Audio, TERUO MURAOKA, YOSHlHlKO YAMADA, AND MASAMI YAMAZAKI) experimented with different roll-offs to see if the sampling rates at the time were adequate. Back then there were a few different sampling rates used for digital recordings for different media. They experimented with roll-offs at 20K, 18K, 16K and 14K using a setup capable of good response to in excess of 40K and used musical samples with loads of high frequency energy (way above 22K) . Their listeners were engineers and musicians. They found with musical samples that roll-offs at 16K and 18K were not reliably detected (15/20) and with the 20K roll-off none of their 30 subject could detect the roll-off better than 13/20 and only 3/30 managed 13/20 which is well within the range for guesswork. To date the majority of data points suggest that the material above 20k is not actually missed when removed from musical content
 
Jun 28, 2014 at 4:24 PM Post #6 of 35
The Nyquist-shannon sampling theorem is pretty clear: any bandlimited signal sampled at more than twice the highest frequency component is mathematically perfectly captured. That includes phase and amplitude.

To help prove it to yourself (within truncation error) i recommended playing around with some test signals in matlab and sampling them at different rates and sample rate converting them.

44.1kHz captures every frequency < 22.05 kHz to within whatever bit depth you use.

Cheers

Aside: Implementations of antialiasing filters require steep filters which are difficult to do. Dont confuse implantation challenges with limitations of the format.

So for example, with 44.1kHz sampling rate, can we perfectly reproduce a 22.04kHz sine wave with the exactly same amplitude, no matter what phase are we sampling at (assuming that our DAC has infinite computing power)?
 
As for 44K sampling to date only one AES paper (Pras and Gusatavino) has suggested that there might be audible differences between different sampling rates, most published papers  do not find any such difference. One paper (Oohashi et al) tentatively found a difference in the physiological responses when cutting off ultrasonics but this has never been replicated and Ashihara et al found that the effect may have been due to IMD between the different speaker units woofer/midrange and supertweeter. Way back in 1978 researchers from JVC (Sampling-Frequency Considerations in Digital Audio, TERUO MURAOKA, YOSHlHlKO YAMADA, AND MASAMI YAMAZAKI) experimented with different roll-offs to see if the sampling rates at the time were adequate. Back then there were a few different sampling rates used for digital recordings for different media. They experimented with roll-offs at 20K, 18K, 16K and 14K using a setup capable of good response to in excess of 40K and used musical samples with loads of high frequency energy (way above 22K) . Their listeners were engineers and musicians. They found with musical samples that roll-offs at 16K and 18K were not reliably detected (15/20) and with the 20K roll-off none of their 30 subject could detect the roll-off better than 13/20 and only 3/30 managed 13/20 which is well within the range for guesswork. To date the majority of data points suggest that the material above 20k is not actually missed when removed from musical content

I know that extreme highs are hardly audible, but let's say we are audio purists, who don't want to lose any frequencies (even if they border with audibility), then increasing the rate might make sense. The fact that test groups couldn't tell a difference doesn't prove that some audiophile might not hear it. But if higher sampling rate adds audible IMD then I think it's better to not use it.
 
Jun 28, 2014 at 5:43 PM Post #7 of 35
  I know that extreme highs are hardly audible, but let's say we are audio purists, who don't want to lose any frequencies (even if they border with audibility), then increasing the rate might make sense. The fact that test groups couldn't tell a difference doesn't prove that some audiophile might not hear it. But if higher sampling rate adds audible IMD then I think it's better to not use it.

 
You pays yer money and you takes yer choice. Yes, even a test that every one to date fails does not mean that nobody can ever pass it, but even if one person in 10,000 can tell the difference how important would that be in the scheme of things. 
 
 
As for the "audiophile" tag many of the big tests such as Meyer and Moran's do actually use "Audiophiles" or those involved in musical production or performance , not that calling yourself an audiophile really means anything one way or another,
 
Jun 29, 2014 at 2:53 AM Post #8 of 35
  So for example, with 44.1kHz sampling rate, can we perfectly reproduce a 22.04kHz sine wave with the exactly same amplitude, no matter what phase are we sampling at (assuming that our DAC has infinite computing power)?
 

Yes.
 
Jun 29, 2014 at 10:05 AM Post #9 of 35
I listen to it with FOOBAR, great music dynamics AWESOME, I am far from mp3 & aac.
then I decompress the FLAC to WAV, & I play it with WMP, DAMN that's heaven, DYNAMICS & REAL, it is BETTER you can't deny it no placebo !!!

You have of course volume-matched to within 0.1dB, right ?
Because what you describe is EXACTLY what happens if the levels aren't matched ..
(NO, you can not do that by just fiddling with the volume-control, you have to measure it -
OK, actually you can do it by just fiddling with the volume-controls : When it sounds the same, you have matched volume to within 0.1dB)
 
So for example, with 44.1kHz sampling rate, can we perfectly reproduce a 22.04kHz sine wave with the exactly same amplitude, no matter what phase are we sampling at (assuming that our DAC has infinite computing power)?

Can you show A SINGLE piece of recorded music that contains musical frequencies above 20kHz ?
(NOT just some 'noise' )
 
The fact that test groups couldn't tell a difference doesn't prove that some audiophile might not hear it.

Well, then it shouldn't be to hard for 'The Golden Ears' to set up a DBT and prove their supra-natural hearing-abilities, should it ?
Of course, they won't, because they can't - And that's because 'DBT is flawed'  .... It MUST be, because when they make a DBT - THEY CAN'T HEAR ANY DIFFERENCE !
 
 
Also :
Even if you use a real old computer, it shouldn't matter the least, computer-audio isn't played back in 'real-time', it's 'loaded' to a buffer and played back from the buffer .
But one thing that can't be ruled out 100% is 'anti-virus' software - It may actually be scanning the flac-file when it's read from disk, scan it again as it is being de-compressed and maybe even scan it once again as it's loaded to the buffer ..
 
Jun 29, 2014 at 10:19 AM Post #10 of 35

Ok, thank you. For some reason I thought there is a roll off at some point when frequencies get close to sampling rate / 2. I'll try to experiment a bit with scilab when I'll have some time.
 
 
Can you show A SINGLE piece of recorded music that contains frequencies above 20kHz ?

It's usually a noise, but I have noticed that for example in rock music cymbal crashes can extend up to the end of scale (22kHz). You can find one example in my older post:
http://www.head-fi.org/t/717472/is-flac-better-than-320-mp3#post_10523366
 
Jun 29, 2014 at 10:40 AM Post #12 of 35
Yes, crash-cymbals CAN extend beyond 20kHz but unless everything else, including the fundamental, is dead-silence you won't be able to 'hear' it -
You might be able to 'feel' it - As annoying pain .
 
@  ieee754 :
That image doesn't 'prove' anything .
What  equipment was used to generate those pretty colours ?
 
Jun 29, 2014 at 10:48 AM Post #13 of 35
It's just audio file spectrum, made with Spek. This image proves that there is a musical content above 20kHz. Also I have seen some people claiming to hear a difference between 320kbps vs. lossless (and proving it with ABX test result). So I won't say that these frequencies are not audible for some people, but personally I can't find any with ABX testing.
 
Jun 29, 2014 at 11:46 AM Post #14 of 35
  Ok, thank you. For some reason I thought there is a roll off at some point when frequencies get close to sampling rate / 2. I'll try to experiment a bit with scilab when I'll have some time.

In all real-world applications, it will roll off, yes. In the absolutely ideal case, no, it won't.
 
Jun 29, 2014 at 11:54 AM Post #15 of 35
Nyquist criterion  is the reason why the the sampling frequency double audible range.  We need to faithfully reproduce those sounds in the audible range.  I don't see why you wouldn't have wave output up to 22.05Khz as there are waves beyond the audible range, and with 44.1Khz sampling rate why not?  Looked up Cymbals, a source says harmonics goes up to 15KHz.  Anyway, just looking up how far the instruments extend, I don't see anything beyond 16KHz.  What sound would be beyond that?  
 
I would say harmonics could be out there.
 

Users who are viewing this thread

Back
Top