1) I have often read in these pages and elsewhere that most multi-format disc players convert SACD to PCM. In all the designs I have studied (through service manuals and chip data sheets) this is just not the case. Demodulation of a DSD stream is trivial (a pair of stereo headphones connected to a DSD stream is all that is required to average the pulses), so why would anyone go to the trouble of DSD-->PCM conversion? ASAIK, all SACD players use a laser and DSD decoder board made by Sony. The output of this decoder board is DSD, and generally connects over to the output section of the player, where it *bypasses* the PCM DACs, but benefits from passing through the post-DAC lowpass filtering to remove ultrasonic components that will otherwise cause IM distortion in the power amplifier. Even though some designs appear to send the DSD stream into the PCM DACs, this is only an appearance, and is just for convenience. In every case I've studied, the DSD port on the DAC simply allows the DSD stream to pass through to the post DAC LPFs. There is no DSD-to-PCM conversion involved. In any case, conversion to DSD to PCM would be very problematic to do at the price required for a consumer player, as the ultrasonic components would be aliased back into the audio band as noise and distortion. If aliasing is an unknown concept to you, please go and read a good book on digital signal processing. Conversion of any analog signal (including quasi-analog signals such as DSD) can be done well only if there are *sufficiently small* out-of-band signal components. (It can be done perfectly only if there are *no* out-of-band components. Also a few other things must be done perfectly too.) If you apply the requisite lowpass filter to a DSD stream to remove the offending ultrasonic components, then what you have is the desired analog signal already! The suggestion (that I’ve read in these pages) that we then do sampling, followed by quantization, digital filtering, and D/A conversion to recover the SACD analog signal is lunacy!!! 2) As a corrollary to (1) above, the proof that SACD players (of any kind) do not convert the DSD stream to PCM is that there are still ultrasonic output components. 3) Even if it were true that 95% of all SACD recordings started out as PCM (can anyone provide some evidence for this?), it does not worry me unduly, for if the mastering studio uses a much higher quality D/A converter than I could ever afford to have in my disc player at home, then I am probably better off sonically. One big advantage of SACD/DSD over the DVD-A/PCM format is that the equipment required to decode DSD is trivial. As I said above all that is needed is the lowpass filtering action of a pair of headphones. Or we could use a high-speed op-amp-based multi-pole lowpass filter. The order of the filter is crucial, as most if not all power amplifiers react adversely (intermodulation distortion) to ultrasonics. By contrast, the quality of DVD-A playback is crucially dependent mainly on the quality of the D/A converter used. That is why we have multi-format players such as the APEX where the SACD performance is exquisite, but the cheap 25c DAC DVD-A is cr*p. Until DACs improve a lot, perhaps it is better to use DSD as the storage medium as it avoids one or two stages of D/A conversion. -Alastair Roxburgh
I think the thing with the stages is not that simple. There is a scientific work by two germans (doctor promotion) where they state that SACD (DSD)has no real advantage over DVD-A (PCM).
Hey Alastair, how's life these days? Hope you're doing well With respect to 1): Bass Management and Time Alignment tools are typically applied only to PCM data. So if you want to use these features in the vast majority of players you'll have to transcode from DSD to PCM. Sony claims that their ASIC can provide Bass Management in the DSD domain. However, they haven't been particularly forthcoming on their ASICs implementation details. I suspect that Sony's ASIC works the same way as their mixing station does, by converting it into 8-bit "Wide DSD" which some have referred to as "Narrow PCM". Once you go multi-bit you aren't DSD, IMO. All of the tools for practical, real world implementation are based on PCM. In most every case (Sony possibly excepted, see above) the Time Alignment and Bass Management circuitry is PCM-based. So when you engage time alignment and bass management the end user is forced into PCM in most players. The "for sures" are Denon and Pioneer. Many of the universals are based on Pioneer OEM transports, so they are "lumped into" the DSD to PCM convertors. What will we do with DSD data when Room Correction/Digital EQ are prevalent? Transcode it to PCM if we want to apply this processing. My solution? I'm getting a Pioneer player modified to have a triple S/PDIF output and feed the PCM output to my 861v4 What can you do with DSD with these tools? Diddly/squat. I would think the ASICs to do an on the fly conversion from DSD to PCM are relatively cheap in OEM quantities. The requisite LPF to make sure Nyquist is satisfied can be done in the same ASIC. With respect to 2): I don't agree with you that this constitutes proof. Even below 20K the noise components from DSD are a factor, since every SA-CD player I've seen measured shows lower dynamic range than Redbook CD @ 20kHz. Have you checked the frequency where ultrasonic noise stops? With an LPF of about 40kHz (assuming an 88.2 or 96K transcode) you'd still have an octave of DSDs "noise soup" left over in the signal even after transcoding to PCM. I've spoken with several manufacturers that build Universal players (both engineering and marketing) and they have all said that once Bass Management and time alignment are engaged, it's "Game Over" for DSD. Yes, the conversion from DSD to analog is relatively simple. That's fine, but you get a bunch of noise that's coming along for the ride too. The big gripes for me with DSD are: 1) You can't do any really interesting processing on the datastream. 2) The noise. 3) The sheer arrogance of some members of the DSD community who feel that it's ok to pass along incorrect information in the name of furthering DSD. The vast majority of the world is PCM based, and will continue to be PCM based. I don't see any big changes in the future, but I've been known to be wrong. Cheers,
Can I add the following points? 1/. The only energy above 22KHz in DSD is noise. Lots and lots of it. 2/. Noise shaped signals cause huge difficulty in multichannel mixes. 3/. Binary coded PCM is the most efficient recording format possible. DSD is the least. 4/. DSD bitstreams cannot be processed, EQ'd etc, as when they are they are no longer single bit data. Also, by design, the upper frequency limit of DSD is 22KHz. To borrow a quote from Mr John Watkinson, "all the subsequent gonads about phenomenal bandwidth came from the same people who brought you the emperors new clothes". In PCM mastering, the sound is captured at any suitable sampling rate and resolution, and all production is performed at this same resolution throughout. Only at the final step is noise shaping & bit reduction carried out, which is the correct way to use dither & noise shaping. When using DSD, the noise shaping is there from the start in the original material, and this just makes life difficult and will inevitably result in loss of resolution. SACD? Sad Alternative to Compact Disc.
