Here's *ANOTHER* idea for a new format

So anyhow end result...
Old crummy CDs: 44,100 samples per second x 16 bits per sample x 2 channels = 1,411,200 bits per second or 176,000 bytes per second
Custos 5.1 UberAudio DVDs:88,200 samples per second x 24 bits per sample x 5 channels = 10,584,000 bits per second!

As for the half-speed mastering, I think you're misinterpreting the process: the music is laid on at half the speed (and twice the time) but it's still played back at full speed. The difference in amount of time you can lay on the media would be nil, because both source and media would be running at half-speed.

Aargh! Please, how is representing a digital signal in base10 an improvement on binary? The fact that it's just a bunch of 'on-offs' is its great strength - much less margin for error. It's either above a certain threshold or below it. Implementing a laser system this is required to *measure* a value on the surface of the disk rather than merely verifying that it's a 'pit' or a 'land', introduces more potential for inaccuracy. How would you do error correction in base10? Is it even mathematically possible?

And why is CD 44.1kHz 'if you're lucky', 'often less' than 16-bit and has a dynamic range of 'half of' 96dB? Where do you get this stuff from?

DVD-A is 24-bit, 192kHz and (though possibly not at the same time) 5- channel, btw.

24-bit mastering (most CD's have 16-bit or less.); 96dB range (Most cd's have approx half that.)

16 bit => 20 * log(2^16) = 96.33 dB

24 bit => 20 * log(2^24) = 144.49 dB

Where did you get your numbers from?

Create a multi-layered medium in which each layer requires more energy to "burn". This way the waveform can be represented in base10 instead of base2 (binary). Sure, its still not analog, but its closer than a bunch of on- offs.

Integers are integers, whatever way you represent the data. Presumably your understanding of how CDs work is 1 bit/sample?

Ah, you're probably right. I'm just struggling to think of a way to force base10 numbers into modulo-2 Galois field calculations.

In an entertainment media sense, what would go on there? Not, presumably, channel data rates which outstrip the human ear/any transducing devices for musical purposes - but what to do with all that storage? You could finally do Varese and Xenakis justice (if you had a geodesic dome).

Okay. Enough self-defensive backpedalling. Back to you...

But it can already discern all 65536 whole numbers in that range, and a 24 bit system can percieve (theoretically) 16777216 voltage levels.

(The on-offs do not directly correspond to the output voltage, 16 of them do. Those 16 bits together encode a single voltage level, hence allowing 65536 possible voltage levels)

And yes, I accept your point that a higher quality factory pressed disc is one way put people off now substantially lower quality MP3s.

I guess one question that is being begged: how do you plan to represent those fractions of a whole on the CD/DVD media itself? In other words, how will the laser know the difference between 12456.4 and 12456.5? Right now the player reads a bunch of pits representing 1 or 0 depending on the reflectivity, and possibly at the depth in the media for multi-layer media like DVD. If not a simple on/off switch like a pit, what?

But, as this was mostly addressed to me...

Q: Please, how is representing a digital signal in base10 an improvement on binary? A: Closer to analog...

It *really* isn't, y'know. Representing your sequence of sampled values as 7-5-3-9-7-6 on the disc rather than 1-0-1-1-0-0-1-0 does not a continuous medium make. They're still *samples*, they still have to be reconstituted as their original quantised-level values, and having an optical scanner that's sensitive to more than two states of amplitude doesn't make this any easier (it does the opposite). As Graham has pointed out, it's the *length* of the digital word that determines resolution, not the range of possible values inherent in a single element of that word. (Yes, I know '32768' takes up less than a third of the 'characters' in base10 than in binary). The encoding to digital should be the hard bit, with the greatest attention to accuracy, the reading of the disc should be a doddle (and is, with just two states to deal with). Too far down the 'multivariate amplitude sensing' route and you've got yourself a system akin to the Laser Turntable (www.elpj.com) reading your shiny discs.

If you want to get yr teeth into something which might more reasonably be called 'closer to analog', check out Sony's Direct Stream Digital (and its playback counterpart SACD). It's high-speed single-bit delta-sigma encoding *with no decimation filter*, so rather than a string of absolute integer values at 20- or 24-bit precision, it's a string of 1-bit relative values, each one an increment or decrement of the previous one. There's some very clever noise-shaping going on to get it to work at all. *But*, you can bet it's encoded as binary on the disc.

Again, I'm counting on better DAC/ADC tech. [for the base10 system]

This really wouldn't be anything to do with the DAC/ADC technology, but would involve a much more difficult-to-implement and less robust optical scanning mechanism. Probably not worth persuing.

Now when we're looking at the digital media we currently have, we're looking at a surface that's either reflective or non-reflective, which represents on or off, 1 or 0. Part of this is just sheer simplicity: the patches that are marked as such don't have to be particularly precise, and if you look at the media under an electron microscope you'll see that the dots are more or less huge craggy canyons that sort of resemble a circle. You won't get more precise than that, for example having the laser inscribe a little 9.9 on the surface...just won't happen, at least not with any accuracy, and even if it could be done, optical recognition would certainly have to improve manyfold. Perhaps a better way to do this would be with colour--to achieve base ten you'd only have to be able to distinguish between ten discrete colours, and maybe a couple more for error correction or tracking or some such. Just bounce the laser off it and then interpolate the colour based on how the signal returns, after taking into account the colour of the laser beam. Probably there are serious issues of physics here that I'm not taking into account, but I'm just putting that out there as a way of getting past the ol' 0/1 dilemma.

I wonder how a DSD A/D convertor works - can it do anything but convert from [perhaps v. high frequency] PCM data?

No, you still don't get it. The binary code is merely a representation of an *enormously high number of discrete levels* (over 16 million in the amplitude domain for 24-bit). Putting it on the disc in binary means [i] the optical reading system doesn't have to be terribly refined to read the right numbers (CD readers can be made cheaply which run at enormous speed with no errors) and [ii] the data can be interleaved inside all manner of 'redundant' data which ensures that, even if there are scratches or fingermarks on the surface of the medium, virtually perfect data retrieval is possible.

What you're edging towards is encoding the continuously-varying analog waveform *directly onto* some form of optical media and measuring its changes in intensity with a laser. Which is what the Laser Turntable does with vinyl LPs. Which means leaving the digital domain entirely - encoding samples in base10 is not any less 'crude' than binary in your terminology.

Binary is just a very, very robust means of storing the very precisely measured digital samples.

* to all the electronics experts: Yes, I know there is no such thing as a Bazillion Watt stereo system. At least not yet. But as soon as the gubmint gives me my research grant....

As for a gazillion-watt sound system, there's apparently this new gadget that you can attach by suction cup to any surface in your house that resonates (desktop, window, etc), and it will convert it into a speaker of sorts. I am suddenly imagining a suction cup the size of a city block turning the entire planet into a loudspeaker, with the resulting audio waves rippling through the crust of the planet cleaving the entire Earth into small chunks.

The current state-of-the-art PCM system is 24/192k - essentially transparent to the mic feed and capable of greater resolution than any analogue recording system ever devised. Now, wouldn't it (in a crazy audiophile way) be worth developing systems where the solid- state electronics was actually up to passing this level of detail undiminished (24-bit is not actually achievable in current analogue circuitry), where a greater sense of the soundfield generated by a live musical event was captured by having *lots* of discrete channels (not just 5.1). Keep in the digital domain up to the transducers? Wireless 360-degree flat-panel systems which sense room acoustics and tailor response accordingly? Wouldn't that be a step more in the direction of 'realism' than wasting extra data capacity on describing a single audio channel with ever more numbers (when what we have now seems to outstrip the ear)?

Applications of vastly increased media capacities on *pop music*, though? What would you do with all that data?

Q: 'If we can "encode" sound onto wax with a needle, why can't we encode the same exact waveform into a CD?'

There are few inherent sound quality advantages from optical (CD) vs physical (wax, vinyl). The advantages come from robust error correction and your apparent arch-enemy BINARY.

it was big in japan.

http://www.elpj.com/

And the beauty of cross-interleaved error correction in binary is that scratched disks can still play back PERFECTLY. I think you're arguing very well *against* yr format now.

Those extra 8 bits are the transparency settings, to allow merging of images. A similar thing happens in digital mixers/DSP chips - altering 24-bit audio data will increase the wordlength, so the results of interim processing are held at 56- or 64-bit precision (the blending of digital images = the mixing of digital sound). There's absolutely no need for the final result to be at any more than 24-bit/channel though.

Now...imagine if your audio was 32 bits wide with the audio equivalent of "alpha channel bits"...wouldn't you be able to notice whole new vistas of sound in that John Coltrane solo, or hear new vibes from that ambient trance CD?

If you wanted to remix the thing from the original multi-track you'd need your extra capacity. If you're just playing the thing back, no you wouldn't. A 32-bit transfer of a John Coltrane master tape would devote immense amounts of storage space to the faithful transcription of *tape hiss* - the 'window' of the target medium being so much wider than the musical information on the source medium.

Utterly wrong. The laser is doing no such thing. The laser is READING (not translating) the binary code - the error correction information ensures that, in most cases, EXACTLY what went on the disc is read back. That is ALL that is happening in the READ process. Do you FINALLY understand this? Whether or not that binary code, when unpacked into 16- or 24-bit words is an accurate rendition of the origin analogue waveform is entirely dependent on the quality of the ADC in the studio and the DAC in your playback device. The beauty of binary data storage and the associated Reed-Solomon error correction is that a (virtually) perfect read is possible, so that part of the process is something we don't have to worry about.

Morse code is fine for simple messages, but terrible for mimicing a Jimi Hendrix guitar solo.

This statement says so much about your lack of appreciation for even the fundamentals of this subject that I really wonder whether it's worth continuing with this thread.

very faint sounds (that aren't tape hiss or power grid hum) can be lost because theres no way to express them in binary.
This is true only if the resolution of the digital signal is very coarse, and there are only a certain number of steps for volume. Improving the resolution of the audio signal can get around that pretty easily by adding just a couple extra bits (keep in mind that when you're talking binary, adding just ONE extra bit doubles the effective number of steps, so adding a few extra bits increases it exponentially). Yes, there will always be a cutoff point where there's sound one bit and then the audio door is slammed shut the next, but it can be argued that we can already make that sound so quiet at this point using binary that it makes very little difference: most analogue consumer devices have so much noise in the circuitry that you wouldn't hear something that subtle anyhow.

That's not to say that we shouldn't strive for something better. I think this discussion is serving a purpose, but I'm not sure that getting rid of binary is necessarily the answer: it's quick, absolute, and doesn't require as much decoding on the playback side (if you're talking a purely digital device, there's probably going to be a lot of binary on the playback end anyhow, so making the signal binary to begin with may make the whole process a bit easier). A new system would have to blast wayyyy past the efficiencies of the on/off switch to make it worth considering, especially because any improvements in technology that would make say, Base Ten, worth considering in the future will almost certainly increase the processing power of the binary signal as well. (Incidentally, I'm not convinced that Base Ten is necessarily any better at this stuff than binary...it is, after all, a rather arbitrary numerical base that's only really easy for us to get our heads around because we have ten fingers/toes and it makes it very easy for us to count on 'em. Hexadecimal is every bit as legitimate for such a process, and if we're going to go that far, hey, why not make it Base 100 to add even more to the subtlety? Know I'm say'n?)

Please, for the sake of my sanity, go and read the first two or three chapters of any book on digital audio. Go and learn what it actually means for a continuously-varying analogue signal to be sampled 192000 (or 44100) times per second, with 16.7 million (or 65000) discrete levels used to describe the amplitude domain. Understand what limits this places on the system, and compare these to the limits of human audibility.

(I was really hoping Mark's wookie comment would end this).

My cat's breath smells of cat food.