LaserDisc Database • Search

dewdude · **Posted:** 27 Oct 2015, 21:24

Ok...so I know I've been inactive for a while....and I came here to mostly check up to see what's happened during my long absence (plus I have a mildly related LD post to write in a little while). But I was excited to see a forum for SACD; because as an audio guy I've spent a lot of time studying the format and whatnot. I've actually authored an SACD disc a time or two using the so-called "SACD-R" format. Most of you might know this, some of you might not....so I'm going to share what I know.

Ok, so as you probably know, SACD uses the DSD format to store it's audio; and you've probably read where it's a 2.8 mhz sampling rate. "Wow!" you might think; "that's a lot more than 44.1khz i'm using on my lousy CDs"...well calm down there Fido. There's one little thing people tend to really overlook when discussing DSD.

DSD is a 1-bit format.

So what this mean? This means that rather than storing audio using a PCM method; in which discreet samples store an integer value that correspond to a voltage; 1-bit formats store the "change" in your waveform. It's functionally similar to delta-sigma modulation; and 1-bit is also how 99% of DAC's actually convert the digital audio to analog audio. This was done becuase, back in the 80's; building a DAC that worked directly from PCM to analog was difficult; plus the technology of the era was very limited and dynamic range was crap. The term for this is "oversampling DAC". The DAC converts your PCM signal in to a 1-bit delta-sigma stream; which can be converted to analog using something as simple as a low-pass filter to remove the ultra sonics.

The way it works is you have a sawtooth wave generator that is capable of running at an insanely fast rate; and every time you trigger the generator, it switches "directions". The idea is that you basically trigger this sawtooth wave generator (which is outputting a voltage) so that the sawtooth stays "right around" the area of your analog wave form. The output would be, from an analog stand point; a representation of your audio with a super-sonic waveform superimposed on it. IT's also worth mentioning that filters can be built easier for higher rates than lower rates; so dealing with a 2.8mhz output rate means you get to use a relatively simple filter on the output. Remove the ultrasonic sawtooth from the signal; and you're left with a relatively clean analog waveform.

DSD is basically storing this 1-bit audio format on the disc rather than enumerating it to PCM like we do for CD. So what advantage is this?

Well..the main advantage is if you're storing your audio in a 1-bit format already; you don't have to convert it to a 1-bit stream; you can just send it directly to your DAC. As I mentioned; 99% of DACs on the market (basically every dac capable of more than 16-bit) use an oversampling method; storing your audio in 1-bit actually eliminates a conversion step. Part of the perceived enhanced quality is because of that; converting PCM to DSD and vise-versa is not an easy or accurate task. While you do have this advantage; there are a number of disadvantages.

For starters, even with the insanely high sampling rate; the 1-bit suffers from horrible dynamic range in general. SACD overcomes this by using noise-shaping; a chunk of inaudible ultra-sonic spectrum contains nothing but noise...basically dithering. This allows you to increase the apparent dynamic range of the recording. It's used in GIF's to make the color downsampling look better; it's also used in better PCM Editors to convert 24-bit audio to 16-bit audio without encountering quantization noise...you put enough noise somewhere where it won't be heard so that you maintain an amplitude over the quantization level.

This is why it has been said "SACD is only as efficient as 20-bit 88.2khz". Basically, just about everything above 40khz in the audio spectrum is noise. It's not that the format itself cannot respond to something that high; in fact I've seen sweeps well over 100khz in a DSD file; it's just you have way more noise than signal up at that end. Most SACD players, in reality, have a 30 to 40khz lowpass filter.

The other issue is that a lot of people seem to want to play back DSD files on their PC without having a DSD capable DAC. So you're going though a DSD->PCM step..just to go from PCM->1-bit in your DAC. It's a pretty wasteful process; and it's one of the reasons people who attempt this start to say "DSD is stupid". It is not a horrible process as long as you realize what you are doing and use good filtering; I myself have a couple of DSD recorded tracks that have been converted to high-rate PCM and enjoy them just fine.

Some players, especially multi-format players; have a lousy habit of converting the DSD to 24/88.2PCM before decoding them; adding to the mess.

Now...another small truth about SACD...things hardly ever say true DSD. One of the problems with 1-bit files is that you cannot edit them; you can do some simple stuff to them; but the usual amount of post-production done to a CD during the mastering phase cannot be done in the 1-bit domain; likewise, you cannot record sessions in 1-bit, mix in 1-bit, and export in 1-bit. There are two ways they've worked around this; one of them is using a DSD format with a higher bit-depth. However, those stations are very expensive; and there is still the issue of we live in a largely PCM world. So...they came up with something else; DXD.

So what is DXD? DXD is basically high-rate PCM; specifically, it's 24(maybe 32) bit PCM audio at a sampling rate of 352khz! Whenever you are dealing with stuff recorded in DSD; 99% of times it gets converted to DXD PCM for production. In fact, I was told by one guy that just about *every* SACD on the market; at some point in the chain; was PCM.

Whether any of this is enough to turn one off of the format depends on your personal preferences. I happen to like DSD as a format; it, in a way..is future proof. You can take one DSD source and convert it to a higher rate PCM while..in theory..enjoying some of the benefits of being a higher rate. There is also a slight fact that most ADC's actually capture audio to a 1-bit pulse stream before that is converted down to a PCM rate. I have seen *many* 24/192khz files that show dithering noise in the upper end of the spectrum; a tell-tale sign of conversion.

Basically, it all boils down to how many times do you want to push the audio through format conversions; and whether or not you think what the professionals are using, or even what you using; do an adequate job of it. Of course, there are higher rate 1-bit formats out there; like 5.6mhz DSD128 and I've even seen some 11.2mhz DSD256 devices floating around out there.

Now...funky facts about SACD.

SACD medium is, physically, a DVD. The main difference is that the discs incorporate pit modulation as a form of copy protection; and as a result...cannot be read by anything except a SACD player. That's why he only way to extract DSD from an SACD is using an older hacked PS3; it not only contains the drive capable of reading SACD...but the software hacks let you get the DSD audio out. But they are 4.7GB/layer, just like a DVD.

The format is actually 2.8mhz *per channel*; this means you need 5.6mbps transfer rate for a 2-channel stereo content. If you do the math, this means you need close to 15mbps to support a 5.1 stream; much higher than a DVD itself can physically transfer per second. To overcome this; there is a special lossless codec for DSD called DST (digital stream transfer); it manages to get really good compression (2.5:1 to 3:1) on the content. It's this DST stream that is decompressed in the player to obtain a 5.1 DSD stream. It is mandatory on 5.1 audio; it is optional on 2.0 mixes. Some hybrid discs use DST on stereo content, just for space constraints.

Of course, by that math; you only need 5.6mbps to store this audio; where as 24/192khz requires 9.2mbps...then again it's not really all that better than 24/96 when you take in to account the noise-shaping; which only requires 4.6mbps.

Unlike the audio CD, which is a set length; the amount of audio you can put on a SACD is only limited by how much space you have. When using DST compression; you can actually cram about 4 or 5 hours worth of music on a 4.7GB disc. As an example; Hotel California in 2.0 DSD consumes about 1.3 GB of data; after conversion to DST, it's under 600MB. But don't try to burn that to a CD-R and expect it to work; the SACD players know the difference in the mediums (or at least my Oppo does). Even when you burn your own SACD (after obtaining the software to convert to DSD, merging your DSD files in to one large one; and figuring out the authoring software), it still shows up as an invalid disc on your PC.

I think I've said just about everything I can on the subject.