|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
192.251.226.206
From the HDTracks FAQ. I'm sure some have seen it before. It was new to me.
Follow Ups:
That's pretty much the only way it can be done.
The so-called "DSD" (a marketing term, not an engineering term, meaning anything that Sony wants it to mean) signal has a sharp rise in noise level starting at 20 kHz (not the 25 kHz stated in the HDTracks paper).
There is almost no instance where the energy of the music signal above 20 kHz exceeds the noise floor of the "DSD" process. This was just one of the duplicitous things about SACD. On the one hand they claimed frequency response "up to 100 kHz" and at the same time they said that the out of band noise was inaudible.
You can't have it both ways! If the noise above 20 kHz was inaudible, then why not just use 44/24 PCM? Or if the high-frequency content of the music was important, why bury it under massive amounts of noise? The real answer is simply that they wanted a system that they could license and make royalties from, and that was just about all.
The fact that SACD ended up sounding fairly good in most instances is a happy accident of the fact that there was virtually no filtering on the recording side. Then on the playback side, the Scarlet Book specified:
~~~~~~~~~~
To protect analog amplifiers and loudspeakers, it is recommended that a Super Audio CD player contain at its output an analog low pass filter with a cut-off frequency of maximum 50 kHz and a slope of minimum 30 dB/Oct. For use with wide-band audio equipment, filters with a cut-off frequency of over 50 kHz can be used.
~~~~~~~~~~
Thirty dB/octave is a steep enough filter to have a negative sonic impact. This could be dispensed with for "wide-band" audio equipment, whatever that means...
Some players had a switch on the rear panel (really only a handful) to select between the filter and no filter. But when converting to PCM for download, HDTracks (and anybody else) needs to filter out at least some of the noise to prevent problems with systems that are sensitive to that high-frequency noise.
An 88.2 kHz transfer will need a brickwall filter by 40 or 44 kHz, so most of the noise is filtered out anyway. A 176.4 transfer has more flexibility. They have to decide how much noise they want to let through, but they have the advantage of not requiring such a sharp filter. This results in better sound quality.
The fellow has good ears!
.............
nt
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Ditto, great post.
.
There is an interesting post about this subject by Bruce Brown
The Well Tempered Computer
In this discussion it is revealed that the low-pass anti-aliasing filter for the Weiss Saracon is exactly the same when converting DSD to either 88.2 or 176.4 kHz. Mr. Weiss makes some comments on that board defending his choice.
I disagree with that choice, as does the developer of the Pyramix digital audio workstation also, apparently.
Conversion to 88.2 kHz forces the use of a low-pass filter with a stop-band of 44 kHz to avoid aliasing of the DSD signal. This can either be a brickwall starting around 40 or 42 kHz or it can be a gentle filter starting around 20 or 25 kHz.
On the other hand, conversion to 176.4 kHz offers greater options with fewer compromises. The Scarlet Book doesn't specify the need for filtering until 50 kHz and even then, the 30 dB/octave requirement is quite gentle compared to a digital brickwall filter. So one possibility would be a filter that was flat to 40 kHz, and then with a *very* slow rolloff (in terms of digital filters) until the stopband at 88.2 kHz, over an octave above that.
This filter would give the benefit of the doubt to any musical signals that might be above the noise, all the way out to 40 kHz while still having an extremely gentle rolloff (for a digital filter) that would minimize the ringing. In this case there would be a definite benefit to a 176 kHz transfer over an 88 kHz transfer. This would be especially true if the source file were a double-rate "DSD" file sampled at 5.6 MHz, where the noise doesn't start rising until 40 kHz.
Given that all digital audio filters affect the sound and offer different tradeoffs with different settings, it makes no sense for a software product to do a conversion without providing a choice of filters. This may be inappropriate with a hardware product which is used in real time, but one of the advantages of doing sample rate conversion after the fact is the ability to evaluate several conversions for recording and select the best one. As you pointed out, because of the noise problem with DSD, the choice of filters is critical. The noise in DSD varies with the music, so if it is aliased into the audible band it will be particularly bothersome.
DSD has worse problems on the encoding side. If one really wants to be disgusted, take a look at the Saracon manual for the DSD encoder. Here there is a special "lockup" detector to cover the case where the overloaded sigma delta modulator dies. This is not Weiss's fault, BTW. This comes directly from the poor design of DSD. The Scarlet Book could have specified a firm ceiling on the input level (as is the case with PCM) but this would have meant giving away advertised S/N ratio which was already highly marginal with DSD64. No way is DSD64 a transparent system.
I know how to build a sample rate converter that goes from 44/16 to 88/24 and back to 44/16 and preserves the original bits. (None of the commercial units I've tried will pass this test.) However, I've been trying for quite some time to do the same thing with DSD (44/16 to DSD to 44/16) and haven't even been able to do this when the only signal to be passed through is DC! This absurd system created a huge number of engineering problems, which is just great if one wants to then make lots of "inventions" to "solve" them. What BS.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Great post, Charles.
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: