DSD comparison to PCM.
- Thread starter U47
- Start date
Hi Rene! Good to see you here! And thanks for the link to the Bruno Putzey correspondence ... such a refreshing change from the usual PR guff from reviewers and vendors.
From a production point of view, it seems like DSD is the SECAM of audio: a useful distribution format, but almost impossible to use in the studio. (Cross-fading SECAM gives bizarre color distortions, which is why PAL was used in the studio, and then converted at the transmitter to SECAM.) The intriguing little trick of converting a tiny section to DXD, editing that, and then converting back to DSD for the edit interval is fascinating. Not a bad method of concealing a momentary change in quality.
I especially like the point of BP mentioning that A -> D -> A comparisons that share a common clock for encoding and decoding are not valid, since DSD has very low latency (transmission delay) while PCM is many times larger, which decorrelates the jitter on each end. Better to play it back and listen to the track recorded on two different formats, and decide from there.
The first time I heard pro-quality DSD at the studio in Boulder I noticed the reverb on the singer sounded quite different than the acoustic instruments - it actually stood out from the sound of the vocalist. I asked, and it turned out that there's no way to do DSD reverb, so it's done in PCM format, and mixed in. EMT plates still exist, I understand, but are getting pretty rare these days.
Nearly all modern recordings originate as high-res PCM, and have been made that way for nearly twenty years. Given that reality, I wonder what quality losses happen when 192/24 PCM is transcoded into 64fs DSD ... compared to leaving 192/24 in its native format. Hard to believe that PCM -> DSD conversion would actually *improve* the quality ... unless the PCM converter at the consumer end is much worse than the equivalent DSD converter.
From a production point of view, it seems like DSD is the SECAM of audio: a useful distribution format, but almost impossible to use in the studio. (Cross-fading SECAM gives bizarre color distortions, which is why PAL was used in the studio, and then converted at the transmitter to SECAM.) The intriguing little trick of converting a tiny section to DXD, editing that, and then converting back to DSD for the edit interval is fascinating. Not a bad method of concealing a momentary change in quality.
I especially like the point of BP mentioning that A -> D -> A comparisons that share a common clock for encoding and decoding are not valid, since DSD has very low latency (transmission delay) while PCM is many times larger, which decorrelates the jitter on each end. Better to play it back and listen to the track recorded on two different formats, and decide from there.
The first time I heard pro-quality DSD at the studio in Boulder I noticed the reverb on the singer sounded quite different than the acoustic instruments - it actually stood out from the sound of the vocalist. I asked, and it turned out that there's no way to do DSD reverb, so it's done in PCM format, and mixed in. EMT plates still exist, I understand, but are getting pretty rare these days.
Nearly all modern recordings originate as high-res PCM, and have been made that way for nearly twenty years. Given that reality, I wonder what quality losses happen when 192/24 PCM is transcoded into 64fs DSD ... compared to leaving 192/24 in its native format. Hard to believe that PCM -> DSD conversion would actually *improve* the quality ... unless the PCM converter at the consumer end is much worse than the equivalent DSD converter.
Last edited:
garylkoh
WBF Technical Expert (Speakers & Audio Equipment)
With the collected wisdom here - any opinions about the PCM1795 vs PCM1704? I have a M2Tech Young using the PCM1795 that I like very much but I don't really know why.
Yes, DSD won the war - but not the battle.
Once Sony saw that DVDA was dead, they withdrew from the marketplace. How gracious of them to take their toys and go home.
A Sony engineer at one of the hifi shows back in the 90's told me that DSD was developed so that payers could be made for less $, needing only 1-bit DACs.
I wish I'd gotten his business card. I wonder if he's still out of work.
The PCM1795 is a 32-Bit/192kHz Stereo DAC, but way more cheap than the Mono 24-Bit/96kHz PCM1704K DAC.
* My pre/pro (Integra DHC-80.3) has six PCM-1795 DACs, but to me they are inferior sounding to the PCM-1704K (top grade) DAC.
Architecture; you bet! ...Lynn, Rene, Richard?
* My pre/pro (Integra DHC-80.3) has six PCM-1795 DACs, but to me they are inferior sounding to the PCM-1704K (top grade) DAC.
Architecture; you bet! ...Lynn, Rene, Richard?
Yes, the latest price in small quantities for the PCM 1704 is around US $75 each - and those are mono converters, not stereo, so you must buy two (or more, depending on whether you want to parallel or operate differentially). My understanding is the ESS Sabre 9018 is around US $40, and it has 8 channels, which are ganged together to make it a 2-channel converter. Other delta-sigma converters are less - some, much less, with the cheaper ones having voltage-mode output.
The rumors I've heard about the PCM 1704 is that TI has tried several times to discontinue it (thus, the scary note on the spec page), but military and aerospace customers demand a ladder converter with well-defined input/output characteristics. This goes all the way back to the original form of digital, telemetry from missiles.
The rumors I've heard about the PCM 1704 is that TI has tried several times to discontinue it (thus, the scary note on the spec page), but military and aerospace customers demand a ladder converter with well-defined input/output characteristics. This goes all the way back to the original form of digital, telemetry from missiles.
-- ...And in comparison, the PCM-1795, which is a Stereo (2-channel) DAC, is roughly only three dollars each (when purchased in fair quantity).
Well, Sony is famous for a long line of battery-powered consumer gadgets, going right back to the first mass-marketed transistor radio, followed by the (cassette) Walkman, and then the Discman ... which is probably why Sony developed the 1-bit converter in the first place, making the best of the limited power supply. Sony's acquisition of a movie studio at the same time might have had something to do with the mania for copy-protection that characterized the SACD disk format.
Still curious how ESS implements DSD playback in the 9018 converter; in a way, it's the converter's best feature, and certainly sounds different than other delta-sigma converters when fed a DSD source.
-- I forgot to mention: The PCM-1795 Stereo DAC will do both PCM and DSD.
- The PCM-1704 Mono DAC, only PCM.
_______________________
* An easy question for you people: Who built (designed) the ESS DACs?
...And from which country, and town?
- The PCM-1704 Mono DAC, only PCM.
_______________________
* An easy question for you people: Who built (designed) the ESS DACs?
...And from which country, and town?
Its a jelly bean digital part, whereas the PCM1704 needs time on the laser trimming machine fixing up all those internal resistors to get the performance specs. Hence $$$.
The DAC chips I'm using cost $0.08 each, recycled out of junked Creative Soundblaster PC cards
I dunno, what's the sound of 8 cents? I do need more than one, currently the DAC BOM subtotal is 64 cents
-- I dunno either Richard; as long that your ears like what they sound like, that is all it truly matters at the end of the audio musical expedition. :b
Analog vs Digital ... PCM vs DSD ./././ and all that jazz.
Analog vs Digital ... PCM vs DSD ./././ and all that jazz.
It cost a bit more than than 8 cents, but is was also considered the Rolls of the DAC chips for some time - the Ultra-Analog 20400a. It cascaded two BB PCM63 K in a single die and it was hand trimmed for lower distortion. I have owned several DACs using them and all (Krell KPS25, Audio Synthesis DAX and a PS Audio Ultra Analog) all of them shared a full bodied sound. Later versions of these units had 24 bit DACs that sounded more detailed, but less powerful.
Do you have experience with it? Do you think it can be considered NOS?
EDIT - the the 20400A used a BurrBrown 12 bit multiplying DAC (the DAC7541A) coupled to a discrete 8 bit DAC, not the PCM63K as I incorrectly wrote.
Do you have experience with it? Do you think it can be considered NOS?
EDIT - the the 20400A used a BurrBrown 12 bit multiplying DAC (the DAC7541A) coupled to a discrete 8 bit DAC, not the PCM63K as I incorrectly wrote.
Last edited:
Was it ever verified that 20400a was 2 PCM63? I thought that was a rumor from diyaudio.
Dan Lavry was the creator of that DAC. From his old website
You are correct - but I also found the rumor spread all over the net! Just found the truth in a old review - the 20400A used a BurrBrown 12 bit multiplying DAC (the DAC7541A) coupled to a discrete 8 bit DAC. Thanks for noting it! Surely it must be considered a true NOS!
Lynn have you considered a DHT on the input as well, such as 26, or not really worth it?
Just found the truth in a old review - the 20400A used a BurrBrown 12 bit multiplying DAC (the DAC7541A) coupled to a discrete 8 bit DAC. Thanks for noting it! Surely it must be considered a true NOS!
Its worthy of note here that the clever sauce of this module was in the calibration algorithms to select/trim the resistor values in that discrete DAC. And also that it had an output deglitcher (sample/hold) as CMOS ladder DACs like the 7541 are notoriously glitchy.
You are correct about the internals of the UltraAnalog 20400. The standard part did use an internal sample and hold for deglitching. The factory had an algorithm for calibration. As an employee of Pacific Microsonics, I spent a few Saturdays at the plant doing the cal. Keith Johnson insisted on true 20 bit parts, so we had to be especially careful with selection and trim, taking somewhat more time than the standard production units. PM developed a way to get around using the sample and hold, so our parts were custom. I believe this trick also omitted the internal NE5534 opamp.
Similar threads
