| Is There A Future For DSD And What Is It? [message #334950] |
Sat, 19 April 2008 12:58  |
 |
 Barry Hufker
Messages: 6013
Registered: October 2004
Location: between heaven and hell |
Platinum Member |
|
|
I've become fascinated with DSD because I finally got to do a comparison on my own between DSD, a very high quality PCM at 96/24 and analog tape on a Studer A-807 (15 ips). The PCM didn't come close to either of the other two. Tape was nice but "mushy" in comparison to DSD.
I have a wonderful Sound Devices 722 recorder that is quite robust with great features. It'll do 192k but I wonder if that is "needed". I'm thinking about the more expensive Korg unit (which I never thought I'd say). But I don't want to sell or discard the 722.
So I need some knowledge and advice. Help please?
Thanks.
http://www.hufkerrecording.com
Torture “is basically subject to perception,” CIA counterterrorism lawyer Jonathan Fredman told a group of military and intelligence officials gathered at the U.S.-run detention camp in Cuba on Oct. 2, 2002, according to minutes of the meeting. “If the detainee dies, you’re doing it wrong.”
|
|
|
| Re: Is There A Future For DSD And What Is It? [message #334963 is a reply to message #334950 ] |
Sat, 19 April 2008 13:34  |
Bruno Putzeys
Messages: 669
Registered: November 2006
Location: Rotselaar (Belgium) |
Gold Member |
|
|
Most was said in this thread: http://recforums.prosoundweb.com/index.php/t/18762/0/
My experience is that DSD and PCM do not differ that much sonically (if at all) if both are equally carefully implemented. After all, a DSD converter followed by a decimator is also known as a PCM converter. So a DSD to PCM comparison is best done using a DSD converter and a separate decimator. If you then convert the resulting PCM back to DSD you have two DSD files to compare. One the original, the other more or less representative of PCM. That way you can compare only the formats without changing converters as well.
I'm not at liberty of posting my own DSD->PCM->DSD software, but if you're curious you could send me a DSDIFF (.dff) file, which I can then convert to, say 88.2/24 and back to DSD using the most transparent filters I could come up with. You can then compare the two DSD files on the same replay system, and hear the PCM file on another converter for comparison.
Your question about a future for DSD is a difficult one. DSD has been pushed onto the market as a production format, and has in the past largely been promoted using one specific brand of converters.
To try doing audio processing on 1-bit signals is difficult at best and irrational at worst. The result is that the only people actually doing DSD production do very little processing (say, classical music). A tour of mastering studios with the Grimm converters was also very informative. The most common reaction was "I didn't know DSD could sound right". Apparently the DSD plug crew had been equating DSD with one specific make of converter. Users believed that and attributed the sonic quirks of those converters to the format. This has not done DSD much good either.
The simple fact is that 1-bit sigmadelta is a fine way of making converters (at least if you make them discrete), but it is not such a good idea to couple a data format and a conversion method. If the quality edge in your comparison holds up after everything is converted to the same format and replayed through the same DAC, the difference is in the converter, not in the format.
Have more faith, Bambi. The only right way of dealing with headlights is staring them down.
|
|
|
|
| Re: Is There A Future For DSD And What Is It? [message #335131 is a reply to message #334963 ] |
Sun, 20 April 2008 11:18 |
John Roberts {JR}
Messages: 112
Registered: April 2004
Location: MS |
Active Member |
|
|
Well said. I find it true for so many things in audio, execution trumps many other variants latched upon by merchandisers to differentiate finished products, or audiophools to understand often subtle differences.
I have a question. In your experiments converting back and forth between 1 bit and multi-bit PCM formats, have you tried a simple null test between otherwise identical outputs derived from the same input file? It seems this would reveal all differences. Of course this would also reveal innocent delay offset errors too, but you're a clever guy and could parse out what the differences are.
JR
Resotune at http://circularscience.com/
|
|
|
| Re: Is There A Future For DSD And What Is It? [message #335148 is a reply to message #335131 ] |
Sun, 20 April 2008 13:37 |
Bruno Putzeys
Messages: 669
Registered: November 2006
Location: Rotselaar (Belgium) |
Gold Member |
|
|
The nulled output from the input and output of a DSD->PCM->DSD chain is the sum of:
*everything from the original DSD file outside the passband of the decimation filter (ie a lot of outband noise plus some signal)
*All of the noise from the digital modulator in the PCM to DSD process.
*A bit of LF that doesn't null because of finite in-band ripple in the decimation/upsampling filters
*A tiny bit of distortion from the second modulator (one-bit modulators are not entirely distortion-free, there's some goings on below -130dB)
The most interesting item is whether the noise floor gets modulated by the signal or not. Suffice to say I tweaked the modulator until that was no longer the case.
As you point out, unless very special care is taken, linear errors will completely swamp the interesting part. This is why I rarely do actual null tests. Usually "normal" measurements using test signals will tell you everything you need to know.
Have more faith, Bambi. The only right way of dealing with headlights is staring them down.
|
|
|
|
|
|
| Re: Is There A Future For DSD And What Is It? [message #335504 is a reply to message #335421 ] |
Tue, 22 April 2008 03:27 |
Bruno Putzeys
Messages: 669
Registered: November 2006
Location: Rotselaar (Belgium) |
Gold Member |
|
|
Not sure what conclusions to draw from that. I've got the tascam as well and although I felt the onboard converters sounded commendably good, I wasn't exactly blown away.
Have more faith, Bambi. The only right way of dealing with headlights is staring them down.
|
|
|
|
| Re: Is There A Future For DSD And What Is It? [message #337073 is a reply to message #335692 ] |
Sun, 27 April 2008 16:54 |
 |
Schallfeldnebel
Messages: 591
Registered: October 2004
Location: Europe |
Gold Member |
|
|
I own also a Tascam DVRA1000, and I was surprised about the sound quality of both AD and DA converters, in PCM and DSD.
The rest of the machine I find a disaster, both the menu software and the recording reliability. I have never seen such stupid menu structure as in this machine, like programmed by a bunch of trainees with no experience in pro-audio. I cannot trust on the machine during recording. In the middle of live recitals it just stops (error 61 etc) recording. I have used different brands of DVD-plus, but I get failures anytime.
I was surprised this machine got some kind of an award somewhere. I use it as my main listening and back-up machine, and Genex 8500 for recording. The best things on the Genex are the + 10 and -10 seconds button and the fast back-up copying from HD to MOD and vice versa. I use the AD converter from the Tascam to record DSD in the Genex.
Sound of the Month, Steinway, Het Concertgebouw Amsterdam:
http://webzoom.freewebs.com/eriksikkema/Steinway.mp3
|
|
|
|
| Re: Is There A Future For DSD And What Is It? [message #337217 is a reply to message #337074 ] |
Mon, 28 April 2008 03:03 |
|
Schallfeldnebel
Messages: 591
Registered: October 2004
Location: Europe |
Gold Member |
|
|
For those who want to learn about DSD, the Tascam DVRA1000 is the right machine, because it costs peanuts. ( Don't yell when you have to hit the menu and cancel buttons ten times to get your DVD disc out, because you happended to have it on EXT clocking, when there is no clock available temporarely, what a stupid software !!!)
But it sounds marvallous, and the unreliability issues I have got, can simply be solved by taking the ugly blue painted version with the harddrive on board. ( Who has choosen that colour!)
Soundwise it is about the best machine I have had.
Sound of the Month, Steinway, Het Concertgebouw Amsterdam:
http://webzoom.freewebs.com/eriksikkema/Steinway.mp3
|
|
|
|
|
| Re: Is There A Future For DSD And What Is It? [message #337301 is a reply to message #337295 ] |
Mon, 28 April 2008 11:43 |
Bruno Putzeys
Messages: 669
Registered: November 2006
Location: Rotselaar (Belgium) |
Gold Member |
|
|
| zmix wrote on Mon, 28 April 2008 18:21 |
I had a converstion at the AES with a designer of digital audio interfaces and if I am remembering correctly, it was his opinion that introducing DSD to a production environment is a huge mistake, as it is imposssible to perform any processing as simple as changing the level without damaging the signal integrity...
|
It's actually quite obvious. If I want to add 1dB of gain I have to multiply every sample by about 1.1 times. The number "1.1" can no longer be represented using one bit so we then have to word-length reduce. In this context that means running the signal through a sigmadelta coder anew.
Sony and Philips have taken two different roads to get round the issue. Philips simply went back to PCM for the entire production process with the sampling rate and data format chosen as to be uncontroversially sufficient. The DAW operates at 352.8kHz 32bit float (now known as DXD and commercially implemented as Pyramix) with the signal passing only a single re-modulation stage when the SACD master is created. Sony, on the other hand, opted for 8 bit, 2.8224MHz which still requires noise shaping after every processing stage. Apparently the use of 2.8224MHz and noise shapers was needed to retain "that DSD essence". This method is seen, for instance, in the Sonoma and Sadie DSD workstations.
I like to think this perfectly illustrates different cultures' take on engineering issues, but commenting further would be Politically Incorrect.
| Greg Reierson wrote on Mon, 28 April 2008 18:26 |
That's always been the rub, and I have heard from a few people that the magic goes away when moved to PCM for processing even at high sampling rates.
|
Once engineering is submitted to thinking in "essences" such beliefs are unavoidable. Yet, how come that no ME would dream of trading their Pacific Microsonics for "Sony's preferred brand of DSD converter". Apparently, magic is not enough. I'm extremely pleased with the sonic results I'm getting with my own discrete 1-bit design, but I don't have a sense of loss when I convert to PCM.
Have more faith, Bambi. The only right way of dealing with headlights is staring them down.
|
|
|
Members
Search
FAQ
Register
Login
Home
