HDTracks Offers 352/8/24 Downloads

[Boris75]

That's a subject I do wonder about. Is it exactly the same to our ears if the system plays back a trumpet with its 30+kHz harmonics, as with only the <22.05kHz effects of those harmonics?

 

The other reason to have the capability of reproducing higher harmonics is to "build" transients with rise time faster than that of a 22.05kHz wave. There is academic research indicating that different neurons are responsible for processing transients than are used in "hearing" tone (frequency); and additional research indicating we are able to detect transients with rise time approximately twice that of a 22.05kHz wave.

 

Thanks for a very interesting, information-rich post. Since we are capable of detecting transients with rise time matching that of a 44kHz wave, 96kHz sampling should be close to "perfect sound forever" though without much leeway for proper digital filtering. At any rate, 192kHz should be more than enough (perhaps ~120kHz would be optimal to have a bit but not too much headroom); 352kHz sounds like wasting a lot of bits.

[tailspn]

352kHz sounds like wasting a lot of bits.

 

Well then, why is it pretty easy to detect the degradation difference, especially spatial cues, from 64fs DSD to 352,8KHz DXD?

[tailspn]

That's a subject I do wonder about. Is it exactly the same to our ears if the system plays back a trumpet with its 30+kHz harmonics, as with only the <22.05kHz effects of those harmonics?

 

To add some additional numbers to Jud's post, the highest note on a piano is C8, which has a fundamental frequency of 4186 Hertz. Applying Jud's 7th harmonic criteria, that would extend the necessary bandwidth to about 29KHz for accurate reproduction. In practice, that piano note, or trumpet could be band limited to way less than 8KHz, and it would be still very recognizable. I believe the brain part of our ear-brain recognition system fills in the blanks. Another reason I laugh every time I hear someone say "trust your ears"

[Boris75]

Well then, why is it pretty easy to detect the degradation difference, especially spatial cues, from 64fs DSD to 352,8KHz DXD?

 

Perhaps because of PCM vs DSD performance differences in the way DAC circuitry operates.

[tailspn]

Perhaps because of PCM vs DSD performance differences in the way DAC circuitry operates.

 

Well, that certainly is part of it. But that's an affect. The primary contributing reason is the conversion(s) from the original Pulse Density Modulation bit stream of all A/D converter front end's to PCM, back to the PDM of the delta-sigma modulator used in the vast majority of DAC conversion stages. It's a necessary evil if any post processing (outside of editing is performed), but can be managed for minimal degradation by using just one conversion step, and converting down to the highest PCM sampling rate the delivery media allows.

[Teresa]

Thanks, I just downloaded the 1GB Mozart snippet. Looks like 96kHz sampling would have been ideal.

[ATTACH=CONFIG]12301[/ATTACH]

[ATTACH=CONFIG]12302[/ATTACH]

 

That looks a lot like my 24/352.8kHz download of 2L's Blagutten by the Hoff Ensemble I've compared the 24/96, 24/352.8 and DSD versions and I prefer the sound of the 24/352.8 version despite the fact that it's -90dB down at 25kHz and there is raising ultrasonic noise beginning at 70kHz.

 

I believe that natural rolled-off recordings, as opposed to brickwalled filtered ones benefit from high sampling rates, not because of extended frequency response but because there are MORE samples per second, thus more accurate sound able to respond to really quick dynamic contrasts, in other words faster impulse response. Human beings can hear changes as quick a two millionths of a second, and one of the reasons I believe live music sounds live as it is quicker than recorded music.

[Teresa]

...I believe the brain part of our ear-brain recognition system fills in the blanks. Another reason I laugh every time I hear someone say "trust your ears"

 

I believe music sounds more relaxed when one doesn't have to fill in the blanks and that is why I prefer authentic high resolution PCM, DSD and well-recorded analog, less work for my brain than crippled 16/44.1kHz PCM.

 

I laugh when anyone ignores their ears. I agree with Harry Pearson and Wilma Cozart-Fine Trust your ears!!

[wgscott]

Well then, why is it pretty easy to detect the degradation difference, especially spatial cues, from 64fs DSD to 352,8KHz DXD?

 

Can you do it double-blind?

[Paul R]

Can you do it double-blind?

 

It would probably be a good idea to setup a DBT or at least an ABX test to get that issue out of the way. I am pretty sure I could, but if we get it tested and the results do not agree with your preconceptions - or mine - will we still have to go around with this subject time and time again?

 

Or can we finally put it to bed? Preferably about six feet under some convenient lava bed in a subduction zone?

[Jud]

Can you do it double-blind?

 

Not speaking for Tom, but for myself: That's a fraught subject. I was completely unable to accurately tell two specific versions of Audirvana Plus apart in a double-blind test where I felt sure I heard differences non-blinded. On the other hand, Superdad and I, in non-blinded listening over a period of months, consistently came up *independently* with identical sonic impressions of a whole series of Audirvana Plus versions, down to specifics like pianos sounding natural through a particular new build. So to the question "Can you do it double-blind?", no matter what "it" is, I personally have to answer "Heck if I know."

[Jud]

Thanks for a very interesting, information-rich post. Since we are capable of detecting transients with rise time matching that of a 44kHz wave, 96kHz sampling should be close to "perfect sound forever" though without much leeway for proper digital filtering. At any rate, 192kHz should be more than enough (perhaps ~120kHz would be optimal to have a bit but not too much headroom); 352kHz sounds like wasting a lot of bits.

 

Along the same lines as I asked Bill: Why is it that way back in the early days of digital audio, before this stuff about humans detecting transients with double the rise time (edit: sorry, half the rise time, double the rise *rate*) of a 22.05kHz wave was in the academic literature, 352.8/384kHz ("8x oversampling") became *the* standard for DAC chips? The answer is in something I've mentioned a couple of times in my responses to arw: The problem is not reproducing frequencies or transients (following the waveform), the problem is avoiding audible artifacts from filtering. It was felt that a 352.8/384kHz sample rate allowed the necessary filtering to take place without creating artifacts that deleteriously affected the resulting sound quality. Interesting to wonder, if those rates were felt to be adequate to do filtering on a maximum frequency of 22.05kHz, whether 705.6/768kHz rates could be necessary when the modern research re transients is taken into account.

[manisandher]

Hey Jud, I just wanted to say that everyone should take time to read your posts #20, #46 and the last, #61, in this thread. Really nice explanations of things that are pretty counter-intuitive.

 

Thanks for taking the time and effort.

 

Cheers, Mani.

[Melvin]

Hey Jud, I just wanted to say that everyone should take time to read your posts #20, #46 and the last, #61, in this thread. Really nice explanations of things that are pretty counter-intuitive.

 

Thanks for taking the time and effort.

 

Cheers, Mani.

 

Jud .. I quite agree with Mani. Thanks.

[tailspn]

Can you do it double-blind?

 

I believe so. It's very evident if, first of all, you do it allot, secondly, know the cues to listen for, and have material that contains the cues. It's also much easier to differentiate in multi-channel, than stereo.

 

In Pyramix, you monitor an edit (crossfades) of a DSD file realtime in DXD. Then, when you're satisfied with all the edits, you render the file in DSD without the file itself having been converted to DXD. Bottom line, I'm switching all the time between hearing the DSD file in both DSD and DXD. The cues are in the very low level detail that gives spaciousness a realistic feel, massed strings the velvety feel of individual strings playing as a group, as opposed to one monolithic big instrument etc.

 

All of this requires an unsweetened, unprocessed acoustic recording of instrument(s) in a natural space that contains the necessary low level detail in the first place. It has to be something that you have had experience hearing allot live, so that subconsciously it just feels correct when you hear it recorded. As I pointed out earlier, not only does the ear/brain fill in the recognition gaps, it also has no detail memory (or at least mine doesn't). We memorize experiential aural cues (think hearing as a defense mechanism-snake in the grass noises cues), not literal sound images as the eye/brain does.

 

Sorry for the long winded post, but it just didn't feel right saying "sure I can"

[wgscott]

Sorry for the long winded post, but it just didn't feel right saying "sure I can"

 

Thanks. I've never heard a DSD file, except converted and down-sampled to 96kHz, so I haven't any way of knowing.

[joey192]

Thanks, I just downloaded the 1GB Mozart snippet. Looks like 96kHz sampling would have been ideal.

[ATTACH=CONFIG]12301[/ATTACH]

[ATTACH=CONFIG]12302[/ATTACH]

 

Next time you pust a screenshot, please change the defenition of the spectrogram to say, 32 768 ? It's an HD music file, so don't show us a low defenition representation of it [oh and don't tell me it doesn't matter, come ooonn ;-) ]. Also change the range to 144dB if not done already, but it looks right to me.

[wgscott]

Next time you pust a screenshot, please change the defenition of the spectrogram to say, 32 768 ? It's an HD music file, so don't show us a low defenition representation of it [oh and don't tell me it doesn't matter, come ooonn ;-) ]. Also change the range to 144dB if not done already, but it looks right to me.

 

You seem to be making a lot of demands on behalf of "us" for someone who is making his very first post here. The second of the two figures shows the entire bandwidth. If it looks right to you, you really need to take a closer, and perhaps more informed look before you start tossing around this kind of bullshit.

 

 

Anyway, welcome to ComputerAudiophile.com.

[wgscott]

I agree. This gives a much better appreciation for the quality of the (same) signal vs. the ultra-sonic noise:

 

[wgscott]

This changes everything, too:

 

[Boris75]

@wgscott

Many thanks for the new images. Both clearly show that all the 352kHz file does is to add a great deal of noise to the signal. It seems to me that this 352kHz file can sound better than its equivalent downsampled to 96kHz only in a DAC that would add even more ultrasonic noise than present in the 352kHz when fed with a 96kHz stream. In other words, this 352kHz file looks like two steps down, not up, from the equivalent 96kHz file. I'd like to have the time to test this with my ears on my system.

[Boris75]

This changes everything, too:

 

[ATTACH=CONFIG]12586[/ATTACH]

 

I would be very interested in the opinion of an audio engineer regarding this recording. It seems to me that ultrasonic noise is very insufficiently filtered on this file.

[Jud]

@wgscott

Many thanks for the new images. Both clearly show that all the 352kHz file does is to add a great deal of noise to the signal. It seems to me that this 352kHz file can sound better than its equivalent downsampled to 96kHz only in a DAC that would add even more ultrasonic noise than present in the 352kHz when fed with a 96kHz stream. In other words, this 352kHz file looks like two steps down, not up, from the equivalent 96kHz file. I'd like to have the time to test this with my ears on my system.

 

Seems a very reductionist way of looking at things. Why concentrate on the inaudible shown by the graph, when there is so much audible that it does not show (e.g., impulse/transient behavior)? Also, I think it is perhaps not an accurate way of thinking about what is going on to say that this ultrasonic noise is being "added." The energy is coming from somewhere (noise originally in the audible range, I would guess, but am not sure), and is being filtered as high as possible into the ultrasonic band so it will be inaudible.

[Boris75]

Seems a very reductionist way of looking at things. Why concentrate on the inaudible shown by the graph, when there is so much audible that it does not show (e.g., impulse/transient behavior)?

 

Well, any information about transients is surely drowned in this sea of ultrasonic noise.

[Jud]

Seems a very reductionist way of looking at things. Why concentrate on the inaudible shown by the graph, when there is so much audible that it does not show (e.g., impulse/transient behavior)?

 

Well, any information about transients is surely drowned in this sea of ultrasonic noise.

 

I do hope you get to hear this and a 24/96 version so you can let us know how listening with your ears compares to doing so with your eyes.

 

Edit: BTW, I mean that sincerely - after all the discussion I'd be quite interested to know how the thing actually sounds.

[One and a half]

Most good preamps have a frequency response well into the ultrasonics, some of them to 200Khz. The amp I use maxes out at 150kHz. If the ultrasonic noise was present to this level as shown in the Audacity graph, then the VU meter would never be zero...should it not?

 

There's a filter that removes the ultrasonics. I don't understand the enormity of the discussions when the DAC filters out this gunge with a very gentle filter. What's the issue again???