Head-Fi.org › Forums › Equipment Forums › Computer Audio › FLAC 192/24 & Computer audio
New Posts  All Forums:Forum Nav:

FLAC 192/24 & Computer audio

post #1 of 21
Thread Starter 

Okay, I felt the need to start this thread because I am having problems finding a concrete solution to this specific issue.

I have converted my entire music collection to FLAC. - my living situation does not offer me much space, nor does it give me much leeway in terms of "noise leakage" (have to use headphones), and my income greatly limits my pursuit of aural perfection.  Anyway, my current setup is thus:

 

Headroom Ultra Desktop Amp + Headroom Desktop Power Supply

Denon AH-D7000 Headphones

HP DV-9000 Laptop

PS3 80gb model (SACD playback but optical output only 16/44)

Laptop music program: Foobar. WASAPI plugin, but soundcard only capable of 16/44 output.

 

What I am looking for is a way to play my FLAC files without going through the Windows mixer, and also to do so via 192/24.

Currently I am eying the M-Audio USB which "only" offers 92/24, and around 100 bucks.  But I really like the specs of the Pixel Magic Systems Audiophile MB200.  Normally not offering FLAC support, it seems like it does now.  According to Pixel Magic they offer a

24-bit/192KHz audio DAC from Analog Devices Inc, and low noise Op Amps from Burr-Brown®.

 

The M-Audio is tempting because it costs much less, and it will also let me continue to use Foobar as my music player, which I am very fond of....

post #2 of 21

Why do you need to play them at 192/24, if I may ask?

Playing at 96/24 may make sense if the reconstruction filters of your DAC are not to notch, but 192/24 is just useless.

And even if you have some 192/24 files, it's not like there's any information in this spectrum, downsample it to 96/24 and you're good.

post #3 of 21

I'm interested at upsampling my music too. While I have NEVER heard any of my music upsampled to 192/24, it is widely accepted that doing so increases the fideleity ( for lack of a better word ). Most all better DAC and CD manufacturers are supporting this feature already. But again, I don't know for sure and am curious about eaking every drop of fidelity from my music. If upsampling really helps, and can be achieved with a media player and DAC then I'd like to learn more about this. Has anyone here actually managed to properly upsample their music to 192/24? What did you use to upsample? Is it done in the PC's media player or at the DAC? Or both? How did you get the stream to the DAC? 

 

Jimmy

post #4 of 21
Thread Starter 

...and since I've never had the correct setup to hear the difference between these resolutions, I'm curious as to what others notice. How much difference is there really between 192/24 and 92/24?

According to HDTracks.com "the enhanced clarity is remarkable: improvements in dynamic subtlety and shading are the first things you notice, followed by the absolute perfection of the high-frequencies."

 

The DAC in my amp supposedly upsamples and pads all incoming digital data to 24-bit 196 kHz. But not only that, the Toslink optical and coaxial digital inputs natively accept signals at sample rates up to 24-bit,192 kHz.

post #5 of 21

The first thing is that HDTracks's quote is actually a selling pitch, not an excerpt from the conclusion of a scientific experiment.

 

Secondly, let 's get to the basics of digital audio. What does 44.1/16 means?

There are 2 parameters here, sampling frequency and bit depth.

 

The Shannon-Nyquist theorem (mathematically proven) states that

Quote:
If a function x(t) contains no frequencies higher than B hertz, it is completely determined by giving its ordinates at a series of points spaced 1/(2B) seconds apart.

translated, it means that if you signal contains no frequency above 20 kHz (which is the hearing limit of a very good ear anyway), the entirety of the signal can be reconstructed by using a sample rate of 40 kHz, which the CD format permits by using a sampling rate of 44.1kHz.

Of course, that's the theory, in real life, reconstruction filters are not perfect, that's with in the first place CDs were made at 44.1 kHz and not 40 kHz and why some CD players/DAC implement an upsampling function. This leads to the creation of 96 kHz sampling, in those conditions the signal, up to 48 kHz (which you can't hear anyway), can be reconstructed, and any imperfections in the reconstruction filter (that usually affect higher frequencies in the reconstructed signal) will be in the inaudible range.

 

Do I sound like as of I meant that for playback, even 96 kHz is useless? Probably, upsampling to 96 kHz should give the same result. But, maybe your ear is exceptional and you could hear it, possible, but quite unlikely, the only means to be sure is to make the experiment yourself, take a 96/24 track and downsample  (there's a plugin called Sox for foobar) it to 44.1/24. Use the foobar ABX plugin, see for yourself your you get statistically meaningful results.

 

The second parameter is bit depth, 16 bit means that you've got 96 dB dynamic range, ie the smallest signal variation is 1/65000th or 96 dB lower than the peak signal, would you ear be able to pick up the tienest variations? Maybe, but cerlainly not unless you play your peak signals above 96 dB, and tha's in a totally quiet environement (a quite room has already ~ 30/40 dB noise floor). Only one way to know if you can do it, logics says you can't, maybe you can, test it using ABX (with a file of which you reduced the bit depth yourself).

 

So playing at 96/24? Why not, for peace of mind and certainty that the reconstruction filters don't harm the music... but it's already supposed to be inaudible, but 192/24, totally excessive IMHO.

Do I hear a difference between those two? Yes, when I know I am playing the 96/24 file, there's more air, a sense of microdynamics, the treble is better etc. but once I do the test blind I can't tell which is which anymore, placebo effect is that strong.

 

PS: I insist on taking the same original 96/24 track and downsampling yourself because quite often the places where you buy your tracks don't offer the same master for the 96/24 and the 44.1/16, resulting in an audible difference that has nothing to do with samplign rates or bit depth.

 

Woo... long long post, hope things are clearer.

 

post #6 of 21

There are several interesting questions here. Firstly, the easiest one. Just use the inexpensive J. River Media Center (JRMC) software to bypass Windows Media Player (WMP). As you may know, WMP on Vista and Windows 7 always dithers the output, so you cannot get bit-perfect data to your DAC in this way.


Use the WASAPI Event Style mode of JRMC and guaranteed bit perfect out of your PC and you're done with this task.

 

Regarding 192/24, firstly, my recommendation is to start with files recorded natively in this format. 2L of Norway has many free sample tracks available. Reference Recordings has many releases in 176/24 also.

 

If your ears are sensitive enough, and even with relatively modest playback gear, you should notice a definite improvement in smoothness that becomes really apparent as one goes to higher and higher levels of gear.

 

The high resolution files are available as uncompressed .wav, Windows Media Lossless, or FLAC, as you prefer.

 

And yes, often high resolution files are or rather have to be mastered quite differently from the 44/16 type, as khaos pointed out. Watch out for junky 44/16 that have been magically transformed into 96/24 or higher. Although upsampling may help in various ways, better to start with real, native high rez recordings in the first place.

 

TOSLINK is a second-rate interconnect compared to S/PDIF on RCA  or AES3 on XLR. It's higher jitter, and will sound a little dull or blurred compared to the others.

post #7 of 21

Where did you read that Win 7 or Vista dither the audio stream? As long as it is set properly (ie. audio output at 24 bit/ xx Hz), all audio streams will be output at this setting, this may involve some resampling but not dither, and that's in the Direct Sound mode.

 

Using Wasapi (free in Foobar2k) will even solve any resampling issue by letting the player take exclusive control of the sound stack.

IMHO, any decent DAC reclocks the incoming datastream, making the importance of the level of incoming jitter irrelevant.

 

Finally, upsampling can be done both by the player (PPHS or SoX component in Foobar) or by the Dac.

post #8 of 21
Thread Starter 

khaos974, thanks for the info.  I'm going to have to mull it over for awhile since you have a lot there to think about and contemplate, and I'm not an audio specialist!

 

One reason I am interested in such high resolution playback is my experience with SACD. But what you're saying seems to imply that there is no discernible difference between the SACD layer and redbook layer. Yet, on hybrid SACDs, when I compare the hi-res vs. the lower res playback, I definitely notice a significant difference, even doing a blind test. Is there some different process going on here?  It seems unlikely that in every case there is a different master for each.

 

You mention taking a 96/24 track and downsampling to 44/16.  What is the best way to do this?

 

On a side note, here's a link to a forum discussing Toslink vs. SPDIF: http://www.hometheaterforum.com/forum/thread/238775/spdif-coax-vs-toslink

post #9 of 21

The fact that the SACD layer sometimes uses a different master is actually quite a well known fact, and even of the master is not different, it is just sometimes half a dB louder.

That half dB difference can be easily detected blind, but the brain doesn't interpret it as sounding louder but as sounding better.

 

It could also be that the DAC decoding the SACD is simply better than the one decoding the CD layer, or related to the fault of  single blind testing.

 

On a side note, did you know that SACD, which uses the DSD encoding is actually 1 bit PCM with a lot of noise shaping?

 

To convert 96/26 to 44/16, you could download Audacity, open your 96/24 file, use the resample function and export at 16 bit (don't forget to set the resample function to very "high quality" and the dither setting to "shaped"). Alternatively you sould resample using the SoX plugin in foobar and leave the 24 ->16 (+ dither) to Audacity. And don't forget to use replain gain to get the same volume out of both files when you ABX.

 

Side note: dithering is a mathematical "trick" that injects random noise to avoid the errors due to a quantization at a reduced bit depth, noise shaping involves "moving" that noise in the less audible rane to get better results.

post #10 of 21
Quote:
Originally Posted by khaos974 View Post

The first thing is that HDTracks's quote is actually a selling pitch, not an excerpt from the conclusion of a scientific experiment.

 

Secondly, let 's get to the basics of digital audio. What does 44.1/16 means?

There are 2 parameters here, sampling frequency and bit depth.

 

The Shannon-Nyquist theorem (mathematically proven) states that

Quote:
If a function x(t) contains no frequencies higher than B hertz, it is completely determined by giving its ordinates at a series of points spaced 1/(2B) seconds apart.

translated, it means that if you signal contains no frequency above 20 kHz (which is the hearing limit of a very good ear anyway), the entirety of the signal can be reconstructed by using a sample rate of 40 kHz, which the CD format permits by using a sampling rate of 44.1kHz.

Of course, that's the theory, in real life, reconstruction filters are not perfect, that's with in the first place CDs were made at 44.1 kHz and not 40 kHz and why some CD players/DAC implement an upsampling function. This leads to the creation of 96 kHz sampling, in those conditions the signal, up to 48 kHz (which you can't hear anyway), can be reconstructed, and any imperfections in the reconstruction filter (that usually affect higher frequencies in the reconstructed signal) will be in the inaudible range.

 

Do I sound like as of I meant that for playback, even 96 kHz is useless? Probably, upsampling to 96 kHz should give the same result. But, maybe your ear is exceptional and you could hear it, possible, but quite unlikely, the only means to be sure is to make the experiment yourself, take a 96/24 track and downsample  (there's a plugin called Sox for foobar) it to 44.1/24. Use the foobar ABX plugin, see for yourself your you get statistically meaningful results.

 

The second parameter is bit depth, 16 bit means that you've got 96 dB dynamic range, ie the smallest signal variation is 1/65000th or 96 dB lower than the peak signal, would you ear be able to pick up the tienest variations? Maybe, but cerlainly not unless you play your peak signals above 96 dB, and tha's in a totally quiet environement (a quite room has already ~ 30/40 dB noise floor). Only one way to know if you can do it, logics says you can't, maybe you can, test it using ABX (with a file of which you reduced the bit depth yourself).

 

So playing at 96/24? Why not, for peace of mind and certainty that the reconstruction filters don't harm the music... but it's already supposed to be inaudible, but 192/24, totally excessive IMHO.

Do I hear a difference between those two? Yes, when I know I am playing the 96/24 file, there's more air, a sense of microdynamics, the treble is better etc. but once I do the test blind I can't tell which is which anymore, placebo effect is that strong.

 

PS: I insist on taking the same original 96/24 track and downsampling yourself because quite often the places where you buy your tracks don't offer the same master for the 96/24 and the 44.1/16, resulting in an audible difference that has nothing to do with samplign rates or bit depth.

 

Woo... long long post, hope things are clearer.

 

 

Additionally I found sth. interesting regarding the sampling theory. Maybe this helps..

 

http://www.lavryengineering.com/documents/Sampling_Theory.pdf

post #11 of 21
Thread Starter 

Good stuff, very detailed.

 

I found this article very interesting as well:  http://www.users.qwest.net/~volt42/cadenzarecording/DitherExplained.pdf

post #12 of 21
Quote:
Originally Posted by khaos974 View Post

Where did you read that Win 7 or Vista dither the audio stream? As long as it is set properly (ie. audio output at 24 bit/ xx Hz), all audio streams will be output at this setting, this may involve some resampling but not dither, and that's in the Direct Sound mode.


 

There are a couple of statements floating around on the internet that Vista/Win7 converts everything to 32 float, do the DSP and applies dither when converting back to fixed. The dither is needed because of the mixing

Check this one: Woodinville (system architect of Vista audio) in post 45

 http://www.hydrogenaudio.org/forums/index.php?showtopic=55459&st=25

post #13 of 21
Quote:
Originally Posted by 1Jimmyneutron View Post

I'm interested at upsampling my music too. While I have NEVER heard any of my music upsampled to 192/24, it is widely accepted that doing so increases the fideleity ( for lack of a better word ).


No, upsampling your music doesn't make it sound better. That's like converting 128k MP3 to FLAC and saying it increases quality. It's simply impossible.

post #14 of 21

As said before, the aim of upsampling is not to re create missing data (which won't be head anyway since they are above 22.5 kHz, the aim of upsampling is to allow the use of better reconstruction filters (ie. not a brick wall filter).
 

Quote:
Originally Posted by Matt08642 View Post

No, upsampling your music doesn't make it sound better. That's like converting 128k MP3 to FLAC and saying it increases quality. It's simply impossible.
post #15 of 21

DACs are optimized to use their internal upsampling, so it is better to feed them native sample rate. I just downloaded a couple of free players: XMplay and AudioGate to use with my musiland interface. You can read here: http://hifiduino.wordpress.com/

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Computer Audio
Head-Fi.org › Forums › Equipment Forums › Computer Audio › FLAC 192/24 & Computer audio