|
|
|
To follow up on a comment of mine in another thread that greater (higher) sample frequency yields better results only because of filtering issues - the increased data is not actually giving us anything! - I thought I would post this link to an experiment done by respected mastering engineer Bob Katz.
Granted, one experiment does not make scientific certainty (pardon the oxymoron), but in terms of my discussion with Nick about perfecting the current state of the art before just adding more bits, I believe that this info is enlightening.
http://www.audiomedia.com/archive/features/uk-0400/uk-0400-listeningtest/uk-0400-listeningtest.htm
|
|
|
|
|
|
Dan Popp said:
To follow up on a comment of mine in another thread that greater (higher) sample frequency yields better results only because of filtering issues - the increased data is not actually giving us anything! - I thought I would post this link to an experiment done by respected mastering engineer Bob Katz.
Granted, one experiment does not make scientific certainty (pardon the oxymoron), but in terms of my discussion with Nick about perfecting the current state of the art before just adding more bits, I believe that this info is enlightening.
http://www.audiomedia.com/archive/features/uk-0400/uk-0400-listeningtest/uk-0400-listeningtest.htm
Wasn't "digititis" caused by the standard CD only being able to hold 1/4, or every fourth bit, of the original recorded data? The SACD can hold *all* the info, whatever the format, right? I've been surprised people are upset when they discover that their SACDs are 96Htz/24bit-derived; while not pure DSD, I think that they sound *so* much better in any case and I'm assuming that it is because the SACD can now hold *all* the data.
|
|
|
|
|
|
jdaniel@jps.net said:
Wasn't "digititis" caused by the standard CD only being able to hold 1/4, or every fourth bit, of the original recorded data?
Huh?
Red book standard (Compact Disc - Digital Audio) is 44,100 x 16-bit samples per second x 2 channels interleaved. Even despite rather robust error-correction, if you have a CD that has more or fewer samples than that, your player won't play it.
The SACD spec is 64 times that, but in either case the medium is "able to hold" all of the "original recorded data" of that format.
|
|
|
Post by Johnno May 13, 2004 (4 of 26)
|
|
I've both read and printed out the article for further examination. Thanks for making us aware of its existence.
I do recall an article written some years ago when domestic DAT recorders were making their appearance, in particular ones which had dual sampling frequencies of 48kHz and 96kHz and I remember Tony Faulkner was trying one out by making recordings with the London Philharmonic Orchestra at both frequencies. He had invited members of the orchestra who were sufficiently interested into the control room after the sessions to listen to excerpts from both series of recordings. I don't actually have the article to hand, but I remember that there was one female clarinettist who said she could distincly hear differences between the two as the ones sampled at the higher frequency brought out the characteristics of her playing more accurately than those at the lower frequency; that is, she could tell more easily when she was playing.
Now what factor or factors involved in employing a higher sampling frequency would bring about such comments? Phaseshift in the filtering, time smear when the lower frequency was used? I found that most interesting.
|
|
|
Post by mdt May 14, 2004 (5 of 26)
|
|
Johnno said:
I've both read and printed out the article for further examination. Thanks for making us aware of its existence.
I do recall an article written some years ago when domestic DAT recorders were making their appearance, in particular ones which had dual sampling frequencies of 48kHz and 96kHz and I remember Tony Faulkner was trying one out by making recordings with the London Philharmonic Orchestra at both frequencies. He had invited members of the orchestra who were sufficiently interested into the control room after the sessions to listen to excerpts from both series of recordings. I don't actually have the article to hand, but I remember that there was one female clarinettist who said she could distincly hear differences between the two as the ones sampled at the higher frequency brought out the characteristics of her playing more accurately than those at the lower frequency; that is, she could tell more easily when she was playing.
Now what factor or factors involved in employing a higher sampling frequency would bring about such comments? Phaseshift in the filtering, time smear when the lower frequency was used? I found that most interesting.
I believe that most of the effect of higher sampling rates is due to filters being moved further away from the audible range.Other tests have shown that in PCM hihgher bit-rates are more important than higher sampling rates. (a true 28bit/44.1 converter sounded better than e.g. a 24/96 converter.)Makes sense to me because lots of the information making a recording sound life-like is at very low levels (ambience).This can only be captured by increasing the dynamic range, meaning the wordlenght. The negative effects of the filters however can also be approached by bettering their quality as Dan says whilst leaving the Fs where it is.
But higher quality means higher costs which means inadequate filters are likely to be used especially in consumer products.
When looking at the facts they convince me even more of staying with SA-CD as consumer format, since DSD has the most simple D/A conversion possible.
On the recording side a coexistence of DSD and PCM seems sensible at the moment.Since apropriate editing tools are not yet avaylable in DSD i imagine PCM could be used in that area since higher quality filters and converters should be affordable in the pro-field.
DSD could be used as mix-down format or as recording format for recordings calling for only minimum processing.
We've only mentioned the effects of the low-pass filtering involved but not the quantization process necessary in PCM, something DSD can do without.
Also considering which frequencies could actually show up in an analog (musical) input signal and DSD's capability to capture up to 100k, i should say no low pass would be needed on the input at all.
So even if a combination of DSD and PCM may be a good solution for now, i still think the aim should be to develop more DSD-tools, making it possible to switch to pure DSD productions in the future.
|
|
|
Post by Johnno May 17, 2004 (6 of 26)
|
|
I found the first part of your contribution interesting and agree with it regarding the necessity for a high degree of quantization to retain ambience and low level detail in recordings. However I don't think Martin Colloms of "Hi-Fi News" would agree with you when you say that quantization is more important than sampling frequency. Some time ago, in collaboration with Tony Faulkner (how often does his name appear when discussing digital recording processes?), who supplied the recordings, Colloms carried out some experiments involving 24/96 and 16/176.4 PCM recordings -- and came out strongly in favour of the latter, as being more realistic and easier to listen to at high levels! It's obvious that, ideally, both quantization and sampling frequency should be as high as technically possible for a variety of reasons. Actually another article in "Hi-Fi News" on DSD and DVD-A attempted to show that if one translates the former (i.e. single bit sampled at 2.8224MHz) into an equivalent multi-bit PCM format, it doesn't measure up particularly well and that the sampling frequency should really be much higher still -- a technical impossibility!. Whether that is a fair assessment or not, I leave for others to decide.
|
|
|
Post by Johnno May 17, 2004 (7 of 26)
|
|
mdt said:
On the recording side a coexistence of DSD and PCM seems sensible at the moment.Since apropriate editing tools are not yet avaylable in DSD i imagine PCM could be used in that area since higher quality filters and converters should be affordable in the pro-field.
Do you know anything about SAdIE? Tony Faulkner told me about this DSD editing tool some time ago and thought it would revolutionise SACD production.
|
|
|
Post by mdt May 17, 2004 (8 of 26)
|
|
Johnno said:
Do you know anything about SAdIE? Tony Faulkner told me about this DSD editing tool some time ago and thought it would revolutionise SACD production.
I dont know about it personally. I read about it on the website of the german audiophile label "Stockfisch Records" (and the according studio Pauler Acoustics).The say it offers fare more editing tools in DSD than previous systems, allowing for productions entirely in DSD. They are presenting a pure! DSD production on that site.
I suggest visiting their site, they are offering for people interested in the subject to contact them.
They are also the people having made the comparisions between PCM converters that i mentioned. They are writing about that as well and are giving information as where to get the technical reports from these tests.
www.stockfisch-records.de
the site has a german and an english version an provides a link to Pauler Acoustics.
|
|
|
Post by mdt May 17, 2004 (9 of 26)
|
|
Johnno said:
I found the first part of your contribution interesting and agree with it regarding the necessity for a high degree of quantization to retain ambience and low level detail in recordings. However I don't think Martin Colloms of "Hi-Fi News" would agree with you when you say that quantization is more important than sampling frequency. Some time ago, in collaboration with Tony Faulkner (how often does his name appear when discussing digital recording processes?), who supplied the recordings, Colloms carried out some experiments involving 24/96 and 16/176.4 PCM recordings -- and came out strongly in favour of the latter, as being more realistic and easier to listen to at high levels! It's obvious that, ideally, both quantization and sampling frequency should be as high as technically possible for a variety of reasons. Actually another article in "Hi-Fi News" on DSD and DVD-A attempted to show that if one translates the former (i.e. single bit sampled at 2.8224MHz) into an equivalent multi-bit PCM format, it doesn't measure up particularly well and that the sampling frequency should really be much higher still -- a technical impossibility!. Whether that is a fair assessment or not, I leave for others to decide.
About the translating of DSD to PCM. This is a mathematical consideration of the issue looking at the amount of information. But in audio not only the amount of information is important but also the quality.So the absence of certain distortions may be more important than the technical values of dynamic range or frequency range.Certain parts of the audio signal are of special importance to reproduce it faithfully, a system superior in this area may sonically outperform another one with technical specs seeming superior at first glance.
I suggest reading Ed Meitner on this, he has done much work developing PCM and DSD converters and favours the latter for the kind of reasons mentioned above.
|
|
|
Post by randy May 17, 2004 (10 of 26)
|
|
mdt said:
(a true 28bit/44.1 converter sounded better than e.g. a 24/96 converter.)
Try as I might, I have not been able to find specifications for any ADC or DAC that credibly claims more than 20 bits of resolution. I would be appreciative of anyone who could provide a link to the spec sheet for such a beast, as I trust it would make for a fascinating read.
|
|
