Future Format Frenzy: Digital Audio and You

23 Comments

View All Comments

• sprockkets - Saturday, June 12, 2004 - link

  Good points. I was trying to think of any reason why increasing the sample rate would help, there seems to be a reason, cause it does look better, but then I think that the only reason you need to sample faster is cause the frequency changes faster than can be measured.
  
  Think about it, you say that if you don't sample enough you miss the peak of the sine wave. Sounds like good reasoning. But then again, the only reason why you would miss the peak is cause the frequency is changing too fast for the turrent sampling rate to graph correctly. Either way sounds reasonable.
  
  Well, also, just about all I've said about SACD was on Sony's website when SACD first debuted. They along with Philips explained why they when this route instead of simply increasing the resolution. Their first player also employed some really neat tricks as well to counter noise problems when switching from 0 to 1, and also using current instead of voltage in most stages since current is more stable.
  
  What convinced me is the fact that just about all cd players use the delta/sigma converter to reproduce sound using the 1 bit strategy. So Sony and Philips said let's just store it that way to begin with, since according to them music nowadays is recorded that way, though that sounds unlikely.
  
  Good point about the picture. Most of the time it's easier to represent all colors with just blue, green and red. Monitors, tv's, and LCD and DLP projectors do this. What DVD audio or Intel says though is it's better to make a monitor with 4.3 billion different color tubes to each represent 32 bit color instead of using on and off timing between the tubes, LCD pixels, or the color wheel with on and off of a DLP's mirrors. Which implementation sounds better/easier/more realistic to work to you?
  

• stephenbrooks - Saturday, June 12, 2004 - link

  That explanation of SACD was great, and very interesting - a lot like an audio version of dithering! One thing I'm concerned with slightly is how does it cope with large dynamic ranges (for instance orchestral music)? A very quiet signal (thousands or tens of thousands of times below the max) would have to rely on detecting that one part in 10000 in an otherwise 50-50 mix of ups and downs. You'd probably need 10000 1bit samples to do that. In comparison to 24 bit and 32 bit formats it gets (theoretically) worse, as these can represent within one of their samples down to millionths or billionths of the maximum volume. However I _think_ those correspond to -120dB and -180dB attenuations respectively, so unless you're listening to a rocket launch interspersed with quiet string instruments... :)
  
  --[There is no point to increasing the sampling rate though in normal PCM formats. You can say that the sine wave looks smoother since you are graphing it more precise, but it makes no difference]--
  
  I don't think there'd be much point going above Azalia's 192kHz, but sampling a sine-wave with only about 5 points per cycle does have some effect because although you've got enough to define the _frequency_, the samples sometimes miss the peak of the sine by a little bit, changing its reconstructed amplitude. They do this periodically as the sine frequency beats with the nearest subharmonic of the sampling frequency. This beat frequency is (annoyingly!) quite often back down in the audible range, so I think that explains why some hi-fi nuts can hear it even when the sampling is 44 or 88kHz.

• Odeen - Saturday, June 12, 2004 - link

  Sprockkets, thank you for that wonderful treatise on SACD. It seems to boil down to dramatically increasing temporal resolution to make up for shallow bit depth.
  
  Those of us who are into imaging can see the tradeoff: a 320x240x32bit image is still low resolution, even though every pixel can take on one of four billion values.
  
  On the other hand, if we increase the resolution hundredfold in each direction (to 32000x24000 pixels), each pixel can be a dot of a solid color but they're so small that a group of them can represent any shade you'd like.
  
  The only problem with using digital output for SACD or DVD-A is that.. you can't. SPDIF is defined for 24/96 stereo max, and DVD-A's are 24/192 stereo, or 24/96 in 4-6 channel (4.0, 4.1, 5.0 or 5.1). SACD is "weird" enough in and of itself that I don't think ANY receiver can decode it.
  The possible exception is the Pioneer Elite series which have a firewire input from compatible DVD players for audio. Now I'm no audiophile, but it strikes me as, well, efficient to keep the signal in digital format until the last possible stage..
  
  

• KristopherKubicki - Saturday, June 12, 2004 - link

  Sprockets; all good points.
  
  "Since you can't hear over this what is the point?" - I think as someone mentioned earlier; although you cant audibly detect it, it does affect other frequencies you can* hear.
  
  As far as playing DVD or SACD on PC; i think to some extent things do converge this way a little - look at how many people use their laptops this way. Furthermore, most decent motherboards and certainly every board with 8 channel output uses optical outputs with the intention you output to a receiver or something fairly capable.
  
  But yes, i basically agree with all your points.
  
  Kristopher

• sprockkets - Saturday, June 12, 2004 - link

  Direct Stream digital addresses the issues of quantization errors and decimation due to first off recording in a 16 bit format to decoding. The best way to represent music is the simplest way. Instead of graphing music in 16 or 32 bit resolution, DSD samples the music at an extreme rate then gives a 1 if the value at that time went up or 0 if it went down.
  
  A way to explain it is you can either try in vain to have 65,536 (now 16,777,216 with 24 bit and even 4,294,967,296 with 32 bit) different voltages reproduced in the analog output or you can turn the voltage on and off fast enough to make the different voltage levels necessary. It's like turning on and off the pixels in LCD displays to generate other colors, or turning on and off the power switch to a light to get the desired light level.
  
  This problem was discovered early on with cd players. They had 16 bit d/a converters which were poor quality. It was much easier and with much higher quality to change the 16 bit format into a 1 bit format. Nearly all cd players do this except for the so called bur brown 16 or 32 bit decoders which claim to be better than the usual 1 bit converters.
  
  This is the advantage of SACD: It gets rid of the resampling and conversion of 16 bit to 1 bit. It simply is recorded that way and stored that way.
  
  There is no point to increasing the sampling rate though in normal PCM formats. You can say that the sine wave looks smoother since you are graphing it more precise, but it makes no difference, cause you sample the signal fast enough to define the full 20-20k Hz that people can hear, though most can't hear over 18k.
  
  For instance, if the frequency of the wave changes from + to - only 30 times a second, the all you need to make sure you get it right is 30 times 2, plus a little more for safety. If you sample it any more than necessary you are simply wasting data because it will also report the same 30 Hz. But if it changes 100 times a second, then you will not see this cause you are not sampling enough. You can notice this cause when you record music at only 22k sampling rate, it sound dull because you are missing the higher frequencies that make it sound bright and clean.
  
  I also get a kick of how to properly use these new formats you have to have systems that can reproduce sound over 20k Hz. Since you can't hear over this what is the point? Moreover, it seems that while we have higer end formats coming out, people are claiming that the cd and such are going away in favor of music downloading. Great, the quality of that music is disgusting.
  
  There really is no point also to playing DVD or SACD audio on a PC. A computer with the many frequencies being generated internally makes any analog output sound bad compared to it being outputed digitally then decoded in a real audio receiver. But what is nice about SACD is that you can still rip the cd part of the disc if it is a hybrid and then listen to the SACD part in a proper player.

• stephenbrooks - Saturday, June 12, 2004 - link

  Actually I guess that's an encoding issue anyway. Seeing as we can send signals at 10Mbps through wires with no trouble, 192kHz isn't at all wasteful for the output. On the DVD itself you might want to use a form of high-frequency MP3ing if you wanted to get more on there.

• stephenbrooks - Saturday, June 12, 2004 - link

  Sorry I wasn't explaining my line of thought. I know that increasing the sampling rate beyond 44kHz improves the quality of the lower frequencies because they can then be represented more accurately as sine-waves rather than blocky things. I was just wondering if there could be a completely different way to improve the accuracy of the lower frequencies (a direct representation as a curve rather than a discrete set of samples, for instance), which is how MP3 works, at a lower level of fidelity.

• dgrady76 - Saturday, June 12, 2004 - link

  I've posted this before, but something to remember about those frequencies you can't hear- they do affect the frequencies you CAN hear.
  
  If you've every heard music on a more accurate, flat response Hi-Fi system that can actually produce those frequencies (and not on MP3s blared through Klipsch speakers) you would hear depth, clarity and a three dimensional soundstage. The first time I listened to music like that, from an analog source as well, I was completley floored, it was not unlike the first time I saw HD television.
  
  I just hope that these 8 channel devices come with some kind of self-calibration because most folks don't even know how to set up a 2.1 system to be accurate.

• stephenbrooks - Saturday, June 12, 2004 - link

  Surely after a certain point, your ears don't get any better? I mean 192kHz includes a load of frequencies only your cat can hear. While I don't argue that it wouldn't be cool to have a button on your PC that could deter your cat on demand, I'm wondering if the next stage from this shouldn't be an MP3-like format that concentrates on encoding the frequencies we _can_ hear better. Perhaps it could even be decoded to many different raw sampling rates dependant on the hardware... it'd take up less room too.

• Oxonium - Saturday, June 12, 2004 - link

  I don't think either DVD-A or SACD will catch on until there are mobile audio systems that support those formats. I listen to far more music in my car than I do at home or work. The Acura TL offers DVD-A support so hopefully it will soon spread to other manufacturers and the aftermarket. I have a few DVD-A's and they sound excellent through my JVC DVD player. R.E.M.'s "Automatic for the People" is just wonderful in DVD-A.