CompuServe Thread

Well, it might turn out that we're both right or both wrong… 8) Allow me to counter your statements. You are correct in stating that S/N ratio is "a function of the WHOLE system… including the analog circuitry." This is also true of dynamic range. But since we are discussing A/D/A systems and bits per word, and since at eight bits the digital system is the major contributor of noise, I see no reason for bringing the analog front end/back end into it beyond acknowledging that it is a factor. For this reason I did not mention it earlier, and don't feel we need to discuss it further now. You also say "s/n is not a function of the number of bits in the system…". This is completely and wholely incorrect. S/N ratio can be directly linked to the word width of an A/D/A system; the well known and well documented formula is, generically 6n db, where n is the number of bits. You have further stated that "… the Amiga is… capable of much better S/N ratio than any cassette recorder… since the quiescent output of the d/a converters is very low…". While the second part of your statement is correct, it is misapplied in the measurement of S/N ratio in a digital system. Yes, if you output zeros continuously, no digital noise will be generated. But how often do you listen to music consisting of all zeros? Your statement is tantamount to saying that a cassette deck has a 150 db S/N ratio because it gets really quiet if you push the STOP button… Remember that in a digital system, if you generate _ANY_ output at all, you will generate noise; in the case of the Amiga running 8 bit words, it will never be more than about 48 db below the signal, and most of the time will be WORSE than this. You said "A clean digitized signal, with a very quiet background, will continue to have a very quiet background when played through the Amiga…" NOT IN REFERENCE TO THE MUSIC, which is what matters. Don't believe me? Dump a signal from a CD directly into an Amiga digitizer. Pick a selection with some hot peaks and a quiet passage; a soft flute will do. Digitize a slice of it and then listen to the result over a good stereo or headphones. Yes, when there is NO OUTPUT there will be no noise of significance, but the flute will be couched in noise and distortion. You also said "… although the ratio of the softest… sound to the loudest is only 48 db." Again, incorrect. The Amiga has a six bit gain control on each A/D converter. By varying input gain during digitizing to use as many of the eight data bits as possible, and then applying the proper envelope using the gain controls on playback, it is possible to reproduce audio with approximately 80 db of dynamic range; yet AT ANY GIVEN INSTANT the S/N ratio will be 48 db OR LESS. Again, don't believe it? Get ahold of a Mimetics sampler, which does gain riding. Record a single, fading piano or guitar note with the AGC turned off, and then play it back. Listen to how bad the release envelope sounds. Now record the same thing again with AGC enabled, apply the proper envelope, and listen to the result. It will sound noticeably better, because the combination of compression on digitizing and expansion on reconstruction PUSHES THE NOISE FLOOR DOWN, keeping it approximately the same distance — 48 db — away from the signal at all times. Confusing? You bet! Professional quality? Never! I've never said otherwise, and generally take the devil's advocate position of explaining the situation to those folks who believe "the Amiga can sound as good as a CD player". No way, no how. Steve, it's entirely possible that we both agree on this issue, and just have our terminology in juxtaposition. But using the terminology I know and use, you are in error (and obviously you feel the same way about my comments). For a third party, might I suggest Chamberlin's "Musical Applications of Microprocessors" as an excellent reference for study? (And if you have some generally available references which support your statements, I'd be happy to take a look at them.)