Joined: 17 Aug 2007
|Posted: Sun Jun 29, 2008 6:20 pm Post subject: Expanded DXD definition
|We've received a suggestion to enhance the description of DXD in our SACD FAQ.
We're not sure to what extent we can subscribe to some of the statements (like "in essence, DSD is PCM anyway with noise shaping built-in") but there's a lot to be said for the other facts.
Although the FAQ paragraph is currently rather short, the following may be a little too extensive for it. That's why (for now) we'll publish it here.
In the discussion (carried out off-board), Dobyblue wrote:
DXD is closely related to DSD. The frequency response is markedly similar to DSD and the information for Mono is three times the clock speed of DSD64, which is why Philips recognizes DXD as an acceptable editing format for DSD.
The idea behind DXD is that you can deliver a DSD or PCM recording with no loss of signal and very little change in the noise ceiling. You also aren't as processor intensive as the DSD-Wide idea that Sony came up with, which is 8-bit/DSD.
In essence, DSD is PCM anyway with noise shaping built-in, so the high sampling rate and bit depth of DXD is perfectly acceptable for mastering to get to DSD or PCM where SACD or Blu-ray is the destination medium. Where multi-channel is concerned I still think you're getting a more accurate representation of a DXD source that you do with DVD-Audio, but Blu-ray with its 24/192 capabilities may be the most desirable delivery of DXD to date.
However, not everyone agrees that DXD is the most desirable tool for editing DSD, some prefer Sadie, from Sony.
Here's some good reading for other people regarding DSD, DSD editing, DXD, etc., in these forums and some differing opinions:
|PCM stands for "Pulse Code Modulation," a way of encoding digital information. We modulate a signal in pulses using a code that is binary. The result is a numerical stream of data that contains "words" of information that represent, within a determined proximity, the amplitude of the waveform intended.
In a traditional 24 bit 44.1kS/s system the digital data is 24 bits long, encoded in the binary format, meant to represent the amplitude of the waveform at even intervals of 44,100 times per second.
In a DSD system the digital data is 1 bit long, encoded in the binary format, meant to represent the amplitude of a given waveform at even intervals of 2,822,400 times per second.
By any definition you can throw at this they are both, unequivocally, "PCM" systems. One of the aspects of DSD is that it can be looked at in many different ways because it is one bit data. It can be looked at, for example, as 1 bit "PAM" data - Pulse Amplitude Modulated. True enough, it still fulfills all of the requirements of Nyquist, represents the same data and all if we call that data "PAM" as it would if we look at it, read it as, and call it "PCM."
It can be looked at, for example, as 1 bit "PNM" data - Pulse Number Modulated. True enough, it represents the same data and all if we call that data "PNM" as it would if we look at it, read it as, and call it "PCM." It still fulfills the rules of Nyquist - has a constant sample rate and intervals, has numerically encoded data that describes the amplitude in given moments in time, and the data represents an encoded set of numbers that represent a waveform that has no frequencies higher than half the sample rate.
We can also look at the data as "PWM," but when we do so we find that it is no longer fulfilling the requirements of the Nyquist theorem. As PWM data it no longer uses even interval sampling periods. Having said this, it still works as digital data if we call it "PWM," but describing why it actually works gets ugly-complicated - unnecessarily so. There is really no reason to rewrite digital architecture and formulas just so we can choose to look at the data as PWM data. It is rather elegantly PCM and so long as we choose to look at it that way it still makes sense according to all digital theorems, still fulfills the rules, and we can treat it as PCM data.
It's kind of like looking at a car and choosing to call it an elephant. Sure, we can do that if we want, but in doing so we're going to have to rewrite the definition of "elephant" and along with it a lot of traffic laws. It is far more elegant to call the car a car and call elephants elephants.
The same applies here. The data CAN be many different things. One of those things that it IS is PCM data. That is how it is treated in all A/D converters in the PCM world, in all DSD editing systems, etc. It is PCM data. Just because it sounds different than some OTHER set of PCM data does not mean it doesn't qualify, just like because my car is spray-painted gray and looks different than other cars on the road doesn't mean it's not a car. It simply is NOT an elephant!
Many people are confused about what DSD data really is - confused partially because of the marketing press of Sony. DSD is just 1-bit sampling data, recorded at a very high sampling rate, and with a "noise-shaping" algorithm built in - much like POW-r. Think of it as a 1-bit PCM A/D converter running at a very high rate and with POW-r built into it. It has 6dB of dynamic range, just like a 1-bit system should according to the work of Claude Shannon and others. It has a Nyquist frequency of 1/2 the sampling frequency. It is, in all ways, a 1-bit PCM system.
1 bit PCM data streams sampled at 2.8224MS/s and with 9th order noise shaping built in may sound different than 24 bit PCM data streams sampled at 2.8224MS/s and then downsampled to 44.1kS/s with linear phase filters - in certain testing situations. They are both, however, PCM.
|There is a HUGE difference between how SADiE and Pyramix handle DSD editing. Same for Sonoma vs Pyramix.
All start with a 1-bit DSD source. All can deliver a final 1-bit DSD "master." How they get there is a different story, though.
The Pyramix, the instant you put the system in edit/mix mode, converts the DSD source to 24bit,325.8kHz pcm for all editing and then re-converts back to DSD for the output to DACs or mastering. All DSD audio, from beginning to end of what you're working on, has gone through the conversion/re-conversion process. This is NOT just simple math being performed. The audio goes through internal pcm and sigma/delta converters for the process.
SADiE (or Sonoma) keeps the DSD source as DSD (no conversion) except at the actual edit crossfade, where an insert occurs the length of the crossfade of 8bit,2.8Mhz pcm and back to DSD. This conversion occurs ONLY at the crossfade. All DSD source on either side of the edit is untouched as long as the "bypass mixer" button is pushed (how I almost ALWAYS run the SADiE DSD). I defy anyone to hear the conversion effect for the duration of an edit crossfade! However, even if the mixer is inserted, the audio integrity is more transparent than that of the pyramix to MY ears. There is no low-pass filtering used or needed. The "annex" DSD metering verifies there is no increase in UHF noise.
(Here's where I pull on the flame-proof clothing ) In MY experience, there is a big difference in how these system sound as well. The Pyramix, as soon as the DSD source hits the 325 mHz pcm mixer (that's every mode other than source recording), changes the sound in an unacceptable manner, IMO. Philips and Pyramix claim otherwise, but I have had 6 very experienced DSD editors and skilled listeners here pick out the Pyramix vs SADiE or Sonoma audio every time. The SADiE and Sonoma keep my DSD source intact.
As with any tool in any format, taking care in how intrusive one is on the source audio will make a difference in the final outcome. The vast majority of my DSD projects can stay in native DSD except for the edits. For any processing or mixing, I come out to analog via EMM Labs dacs and record the process or mix back to DSD via EMM Labs adc. All properly-designed analog gear has no problem whatsoever with the DSD source - no low-pass filtering needed here either! This conversion to analog and back in MY experience is actually less intrusive than the Pyramix 325 mHz conversion process (I'm pulling on the second layer of flame-proof clothing...). Others may disagree (they may own a Pyramix), but this is the way I hear it!