Just some info
Hi,
I thought I'd try to add some more info to anyone following this thread about digital audio, upsampling, conversion, and so on. Some of you obviously know all this; I just wanted to consolidate things which came up for anyone not versed in audio.
1. First, this is pretty basic and I think folks are clear on this now, but:
-- A/D means analog-to-digital (usually for recording)
-- D/A means digital-to-analog (usually for playback)
Pros will often refer to both as simply 'converters' and the context should make it clear which direction it's going.
2. Pro recording systems often have digital inputs. This is different that having A/D converters or D/A converters. Optical inputs on recording systems allow you to use whatever outboard converters you like. It isn't too hard to spend $25k on a good microphone, mic preamp, and A/D converter. In fact, with this gear, you can then plug it right into any crappy soundcard with digital input and have *recorded* sound quality as good as plugging the same gear into a pro system with digital inputs. Obviously if you do *playback* of youur perfect signal through your crappy soundcard, you won't be able to tell how nice your recording is, but that's a different point.
3. Oversampling, upsampling and filtering.
-- Oversampling means increasing the effective sampling rate of a file, and usually implies some interpolation is done. If you take a file sampled at N samples per second, you can easily convert it to 2N samples per second just by adding a second sample for each sample in the file and playing it back twice as fast. This obviously does nothing for you in terms of quality, good or bad. As discussed above, if you interpolate data points this may improve the perceived sound quality, though the biggest benefit is that you don't need to use a crappy filter to chop off frequencies above N/2 Hz.
However, if you go from N samples per second to 1.1N samples per second, it's not so clear how to convert the file without artifacts (since the sample boundaries don't line up nicely). For example, doubling every 10th sample will produce a file at the desired effective sampling rate, but that'll sound terrible. So you probably end up doing some something reasonably sophisticated to more closely approximate what you think the original (anaog) signal must have been, given the digital samples you've got, and thus introducing artifacts (which can be good or bad). In general, you oversample to an integer multiple of the original rate to avoid such issues.
-- "Upsampling" refers specifically to oversampling when you do it before sending the signal to the output D/A converter. This may improve sound quality, perhaps because you're sending the upsampled signal to a better D/A converter than you would be otherwise. The D/A converter can be considered a filter of sorts, and if you've got good filters, you can make audio more listenable. As an analogy, consider photoshop filters. If you edge enhance an image, you haven't actually added any new information to the image. But you've modified it to make different use of the bits available in the file format. You've used the original file's information, but presented it in a way which exaggerates things which the human visual system is sensitive to. For an extreme example, just consider normalizing an audio or image file. You haven't added any information; you've just made better use of your available bandwidth.
-- So anyway, point is that there's no theoretical magic to upsampling. It doesn't add anything to the signal. But in practice it may turn out to improve perceived audio quality on existing hardware.
4. mp3 and aac, while quite practical, introduce significant artifacts. Everybody knows this. Will mp3 files sound better through better hardware? Of course. Will they sound as good as vinyl or CD audio with good converters? Of course not. Can people actually tell the difference? Yes, it's not that hard, especially if you can A/B them.
5. It is CLEARLY impossible to have both digital and analog output from a single connector and anyone who suggests otherwise is a mouth-breathing cretin.
Just kidding about #5.
Anyway, I hope someone finds this useful.