[Danz03]

 

Interesting thread, I've always thought that 24bit/96kHz is only ever used in a studio, I think it's the same reason why digital artists would normally use 30-32 bit colors when we can really only see 24 bit true colors.

But the thing about 24bit audio is that one can manupulate it and then downsample it to 16 bit with no loss of quality. Let say, someone were to use a digital pitch shifter on a 16bit/44.1kHz audio file and lower the pitch by 10%, then the highest frequncy that's audiable would not be 20kHz anymore but 18kHz, which can be heard by most people and the resulting 16bit curves will not be as smooth either. I'm not saying a lot of people are going to pitch-shift a CD but a lot of people would use EQs in their digital amps etc. That will degrate the sound of 16bit/44kHz audio. But if the source was 24bit/96kHz, altering it with a digital EQ would not degrate the sound at all as the altering process was done in the frequency and resolution range beyond what we could hear, and then down sampled back to our audio range. So I guess if one were just listening to music without messing around with it, 16bit/44.1khz is fine, but if someone were to manupulate the music in anyway, 24bit/96kHz is a must.

 

[grokit]

Danz03, I find your posts interesting but somewhat hard to read. The text is smaller, the kerning between characters and spaces is less, and the line spacing is compressed. Plus you tend to write longish paragraphs, which would have better readability with the default formatting. I wonder if you are changing the text style from the default on purpose, or if you are pasting it in from another program. If it is the former, please stop it as it is just too small and scrunched together compared to the surrounding text, requiring me to either zoom in to read it or ignore your post. If it is the latter, please change to plain text before copying it to your clipboard and into the edit box, thanks.

 

[Danz03]

Thanks, better? For some reason, that was the default format on my computer, sorry. 

I managed to change the size but don't know how to change the spacing and kerning here. I tried cut and paste before but somehow it didn't seem to work. It looked fine when I type but after I submitted, it all got changed somehow! 

 

[grokit]

Hmm... now your text is bigger than normal, I wonder why it's different for you than everyone else! Do you type your text right into the edit box?

 

[Ham Sandwich]

Quote:

 

Interesting thread, I've always thought that 24bit/96kHz is only ever used in a studio, I think it's the same reason why digital artists would normally use 30-32 bit colors when we can really only see 24 bit true colors.

But the thing about 24bit audio is that one can manupulate it and then downsample it to 16 bit with no loss of quality. Let say, someone were to use a digital pitch shifter on a 16bit/44.1kHz audio file and lower the pitch by 10%, then the highest frequncy that's audiable would not be 20kHz anymore but 18kHz, which can be heard by most people and the resulting 16bit curves will not be as smooth either. I'm not saying a lot of people are going to pitch-shift a CD but a lot of people would use EQs in their digital amps etc. That will degrate the sound of 16bit/44kHz audio. But if the source was 24bit/96kHz, altering it with a digital EQ would not degrate the sound at all as the altering process was done in the frequency and resolution range beyond what we could hear, and then down sampled back to our audio range. So I guess if one were just listening to music without messing around with it, 16bit/44.1khz is fine, but if someone were to manupulate the music in anyway, 24bit/96kHz is a must.

16 bit vs. 24 bit and 44.1 kHz vs 96 kHz are separate issues.  You can have 16/96 or 24/44.1 if you want.
 
The pitch shift example you give is a concern with sample rate.  You want a higher sample rate if you're going to do things like pitch shifting.  The bit depth doesn't matter to a pitch shift operation.
 
What about if you upsample the 44.1 to 88.2 before doing the pitch shift?  I've been experimenting with doing a software based resample up to 88.2 before doing my digital EQ.  I'm not sure if it is making an audible difference.  I need to experiment some more and try some different EQs as well. 

 

[corsario]

Hi everybody,
 
I am replying here to a question arising from the following post in another topic :
http://www.head-fi.org/forum/thread/504567/hm-602-portable-music-preorder/120#post_6869534

 
Quote:

Sorry being of topic, but I find quite noticable differences between some commercial 16/44 and 24/96 tracks. Anyway your statement may be true and the difference may come from 24/96 mastering. (You can try Rebecca Pidgeon "Spanish Harlem" on Chesky for instance)

 
It sure must come from the mastering. Did you ABX this "noticable" difference by the way ?
 
The right way to go is to perform yourself the downsampling of the 24/96 track using R8brain (use the free version which appears to be better than the PRO one for this purpose. It is complicated to explain but let's say that the default dithering performed by the free version of R8brain is the good one, just set the quality setting to "Very high"). This way you will have the 16/44 version of the 24/96 track : no mastering difference, etc...
 
And there, go to foobar, use the ABX module and make the test*
 
If you can ABX it, then it will be time to consider if R8brain did not deteriorate the 24/96 track or did not performed an inappropriate dithering. But, so far, no ABX was sucessful, so the question is not on the table yet

 
 
*If your soundcard can not switch automatically between 16/44 and 24/96 (this is the case for EMU1212m for instance) then you will need to use R8brain again and upsample your 16/44 sample (no problem it is juste zeros added to obtain 24/96 res, nothing is changed to the file - you can not recreate what was lost in the downsampling process). If your sound card can deal indifferently with any input format (e.g. Lynx2B) then do not bother to upsample and just make the 24/96 vs. 16/44 comparison. If you can ever ABX on one sample, please be kind to indicate the 24/96 sample you used, and I am sure a lot of people, including myself, will be happy to try to ABX it as well.
 

 

[gatz]

 
Interesting topic !

I was always been convinced that 24/96 is superior to 16/44.
Well, until today.
I will try to make the story short.

It has started maybe 10 years ago when I was trying a SACD reissue of an old analog jazz recording. The SACD was in every aspect superior to the CD issue. Since then, I always thought that 24/94, DSD or any other “HD” audio formats are obviously superior. My last listening test was with my Hifiman HM801. It wasn’t to reconsider my position regarding HD/SQ but more to see if this 24/96 capability was really an asset for such a portable device. Then I purchased 24/96 reissue of some all time audiophile favorites I already own from Chesky label. Again, the 24/96 track was better than the (original) CD rip.
 

I made further testing today (thank’s to corsario remarks).


Test 1 (my original test) :

Track 1 : Spanish Harlem (Rebecca Pidgeon) 24/96 from “retrospective”
Track 2 : Spanish Harlem (Rebecca Pidgeon) 16/44 CD rip from “The Raven”
Result : Quite noticeable (well, I should say, obvious !) difference. No need to ABX.
Anyway I have installed foobar ABX and I tried the blind test : I always find the right track, in few seconds.


Test 2 :

Track 1 : Spanish Harlem (Rebecca Pidgeon) 24/96 from “retrospective”
Track 2 : same track converted to 16/44 with R8brain as corsario suggested
Track 3 : same track downsampled and dithered using soundforge
Result : No listening difference between the 3 using the HM801 + EM3pro.
Then I move to my DAW (RME fireface + dynaudio BM5a) : No obvious difference too !


But why my original 16/44 track is so different ?
OK, maybe my CD rip wasn’t that accurate?
I purchased (for the 3rd time...) “Spanish Harlem” from HD track, the 16/44 version, to compare with my CD rip.
They are identical.
 

The mastering of the original 16/44 track and the 24/96 one are obviously different.


Last test, back to soundforge, testing the tracks dynamics :
16/44 original track : peak level = -0,5dB, RMS level = -21,4dB
24/96 original track : peak level = -0,1dB, RMS level = -19,5dB

The 24/96 is slightly louder than the 16/44.
Having a different mastering for the 24/96, I would have expected the opposite!
This alone explains the more noticeable analog noise floor and soundstage I can ear on the 24/96 track.
 

The 24/96 track was part of a "best of", so this could explain the different mastering beyond the format. My next test would be to purchase (for the 4th time !) the 24/96 track from the original album : but it doesn’t exist. The current album version is advertised as “Bob Katz 15th Anniversary Remaster” (strangely in 24/88), so probably different from all versions I already have.

 
 
So, 24/96 (for the end listener) is just another marketing buzz ?
Why not reissuing directly 16/44 audiophile remaster ?

 

[xnor]

Quote:

Why not reissuing directly 16/44 audiophile remaster ?

Making (big) money doesn't work that way.

 

[grokit]

 
Quote:

I was always been convinced that 24/96 is superior to 16/44.
Well, until today.
I will try to make the story short.
 
-snip-
 
So, 24/96 (for the end listener) is just another marketing buzz ?
Why not reissuing directly 16/44 audiophile remaster ?

When all else fails just play a record

 

[leeperry]

 

 
So, 24/96 (for the end listener) is just another marketing buzz ?
Why not reissuing directly 16/44 audiophile remaster ?

 
Because it's far more complicated...most of the 16/44 crowd will listen to crappy mp3 on their ipod or on the radio or whatnot, and the higher the dynamics compression the better it will compress/decode. OTOH, the 24/96 crowd wants as little compression as possible as they have proper gear to boot. If I reuse the example I gave above...if they had released the untouched mastertapes dumps for the Beatles reeeditions, most ppl would have said that the sound was dull and lifeless. Audiophiles are prolly less than 0.1% of the population and the volume caps on the ipod will only make things worse.
 
Also, this doesn't help and will only make music clip even further: http://www.gearslutz.com/board/tips-techniques/334385-intersample-peaks.html
 
Ppl want to listen to clipped autotuned music these days...most of those singers are unable to sing w/o pitch correcting machines anyway

 

[Danz03]

 

[size=medium]I personally don't see much advantage with using 24bit with 44.8kHz (or 44.1kHz) sampling rate. It would be fine if the wave curve just happen to fall onto the grid conveniently, but more often then not, it doesn't, and what if the curve falls onto a point right in the middle of two bit points? Dither will kick in and randomly shift the point up or down to fit it to a bit point, but then when the 24 bit map is being down-sampled back to 16 bit, the same thing can happen, a point that has been previously shifted up by dither can be shifted up further again in the same way and creating more errors, and more errors in this case = more noise and distortions. However, if one were to use 24bt/96kHz, the mistakes would be halved in time, comparing to a 24bt/48kHz, as the time resolution is doubled.[/size]

 

[size=medium]Yes, if one were to just pitch shift a wave up or down by exactly one octave (double or half), then up-sampling it before-hand from 48kHz to 96kHz may be fine, as the bit positions will stay where they are, but what if one were to pitch shift the wave from an A4 (440Hz) to a C4 (261.63Hz)? Of course the bit positions would have to change, in that case, 24 bit would always be better than 16.[/size]

 

[size=medium]I don't think up-sampling would improve the sound quality of a digital EQ, since up-sampling would only affect the frequency and not the amplitude side of the EQ, ie: the Q factors or the selected frequency range would be a few milli-hertz more accurate. Probably a linear predictive coding would improve the sound quality more so than over-sampling.[/size]

 
Quote:

16 bit vs. 24 bit and 44.1 kHz vs 96 kHz are separate issues.  You can have 16/96 or 24/44.1 if you want.
 
The pitch shift example you give is a concern with sample rate.  You want a higher sample rate if you're going to do things like pitch shifting.  The bit depth doesn't matter to a pitch shift operation.
 
What about if you upsample the 44.1 to 88.2 before doing the pitch shift?  I've been experimenting with doing a software based resample up to 88.2 before doing my digital EQ.  I'm not sure if it is making an audible difference.  I need to experiment some more and try some different EQs as well. 

 
There could be a difference between the two, but one would need to have an extremely good monitoring system and hearing above 20kHz to notice the difference. Or, if one were to use a digital EQ, in that case, the 24bit/96 kHz should sound better in theory.

[size=small][size=medium]Like I said before, the main advantage of 24bt/96kHz is that one could manipulate the original sound signal and then convert it down to 16bit/44.1kHz with little or no degradation, so it is very useful for doing mastering.[/size][/size]

 
Quote:

Originally Posted by gatz /img/forum/go_quote.gif
I was always been convinced that 24/96 is superior to 16/44.
Well, until today.
 
So, 24/96 (for the end listener) is just another marketing buzz ?
Why not reissuing directly 16/44 audiophile remaster ?

 

[leeperry]

 

[size=medium]I don't think up-sampling would improve the sound quality of a digital EQ, since up-sampling would only affect the frequency and not the amplitude side of the EQ, ie: the Q factors or the selected frequency range would be a few milli-hertz more accurate. Probably a linear predictive coding would improve the sound quality more so than over-sampling.[/size]

Well, some professional EQ plugins think otherwise: http://www.gearslutz.com/board/5101666-post296.html
 
"The oversampling is necessary to avoid aliasing with the saturation, but also helps to get a more accurate response in the upper end. Just as you guys said, we use a steep IIR LP filter as oversampling filter."
 
Of course this EQ sounds out of this world, coz I've heard some upsampling EQ's that still sounded horrid ^^

 

[Danz03]

But you are talking about a piece of professional studio gear. Like I said before, most studios would use 24bit/96kHz at least for recording, so most professional EQs would be operating in 24 bit at least anyway, which is already better than what most people can hear, and the over-sampling is to make sure that it's accurate even in 24 bit. I was addressing @Ham Sandwich's comment on oversampling from 44.1 to 88.2kHz before going through a digital EQ, to me, that wouldn't make a lot of difference. I couldn't find any specs on how the A-Range EQ works but then I think most DAWs would up-sample the signal before going through any plug-ins anyway. It's quite different talking about studio gears and audiophile products, studio gears have to be accurate while audiophile products just have to sound good.
 
Quote:

Well, some professional EQ plugins think otherwise: http://www.gearslutz.com/board/5101666-post296.html
 
"The oversampling is necessary to avoid aliasing with the saturation, but also helps to get a more accurate response in the upper end. Just as you guys said, we use a steep IIR LP filter as oversampling filter."
 
Of course this EQ sounds out of this world, coz I've heard some upsampling EQ's that still sounded horrid ^^

 

[Ham Sandwich]

Quote:

 

[size=medium]I personally don't see much advantage with using 24bit with 44.8kHz (or 44.1kHz) sampling rate. It would be fine if the wave curve just happen to fall onto the grid conveniently, but more often then not, it doesn't, and what if the curve falls onto a point right in the middle of two bit points? Dither will kick in and randomly shift the point up or down to fit it to a bit point, but then when the 24 bit map is being down-sampled back to 16 bit, the same thing can happen, a point that has been previously shifted up by dither can be shifted up further again in the same way and creating more errors, and more errors in this case = more noise and distortions. However, if one were to use 24bt/96kHz, the mistakes would be halved in time, comparing to a 24bt/48kHz, as the time resolution is doubled.[/size]

 

[size=medium]Yes, if one were to just pitch shift a wave up or down by exactly one octave (double or half), then up-sampling it before-hand from 48kHz to 96kHz may be fine, as the bit positions will stay where they are, but what if one were to pitch shift the wave from an A4 (440Hz) to a C4 (261.63Hz)? Of course the bit positions would have to change, in that case, 24 bit would always be better than 16.[/size]

You're making arguments for 24 bit that are counter to gregorio's posts that began this thread.  The errors or difference caused by doing things at 16 bit shouldn't be audible or relevant to consumer processing and listening (listening to the music as an end product with a some digital post-processing being done, and not as tracks that will be further mixed or processed later).  In a studio you might be going through dozens of digital processes and if each of them bumped a 16 bit sample up to 24 bit to do its processing and then back down to 16 bit before sending the sample on to the next step then I could see some cumulative errors cropping in.  But for consumer style processing, even in cases of things like pitch shifting, I would need some convincing to see how 16-bit vs. 24-bit is an issue as long as the processing doesn't cause clipping at 16-bits.
 
The reason I'm playing around with upsampling before my digital EQ process is to see if maybe the higher sample rate moves the digital filtering that the EQ is doing further out of the audible range and maybe, just maybe, that will have an effect.  I have some EQs that I think may be dulling the upper high frequencies so that's why I'm curious.  I don't know what those EQs are doing internally, whether they're upsampling or not.  I'm going on the assumption that they are not.  I'm not changing the bit depth during the resample.  The resampler I'm using is the native resampler in J River Media Center.  It is claimed to be an "audiophile quality" resampler.

 

[leeperry]

 

But you are talking about a piece of professional studio gear.

I'm talking about a VST plugin that I'm using in my everyday media player.