Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Better 8-bit -> 16-bit conversion? (Read 3504 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Better 8-bit -> 16-bit conversion?

Converting an 8-bit PCM source to 16-bit is of course easy enough in CoolEdit or whatever, but that simply gives you a 16-bit file with 8-bit audio. Is there some utility that can try and do a (not sure of correct term here... dither? anti-alias?) to try and smooth out the big 8-bit steps to make it sound less "8-bit"?
 

Better 8-bit -> 16-bit conversion?

Reply #1 –
I believe the term would be "interpolate," and I have no idea what apps will do it, though I'm sure there is such a creature.

Better 8-bit -> 16-bit conversion?

Reply #2 –
I don't think it is possible, the frequency is kept the same. You could interpolate upto 2x the frequency but you would end up with exactly the same sound.

Once that quality is gone, it is gone.

Better 8-bit -> 16-bit conversion?

Reply #3 –
In Sound Forge, there's a "Smooth/Enhance" function that's the audio equivalent of the "soften/sharpen" available in graphics editing programs.  Something like that might help blunt the perceived harshness -- perhaps the judicious use of some kind of EQ might also deemphasize certain problematic frequencies creating a harsh, grainy quality. 

You could also try using the expander functions (opposite of compression) to gain a "false" dynamic range, while deephasizing the noisies below about -30 to -40dB.

Just some ideas, dunno if they work or not...

Better 8-bit -> 16-bit conversion?

Reply #4 –
Quote
Once that quality is gone, it is gone.

The original signal is obviously gone, but I'm hoping it's possible to "fake it", similar to how de-clipping filters work (which I've never actually used, so I maybe shouldn't comment too much), but they also face the task of re-creating data that is gone.

Better 8-bit -> 16-bit conversion?

Reply #5 –
well, clipping restoration is a little bit different, in that you at least have a beginning slope and an ending slope that you can use to guess how the rest of the missing curve might have looked like.  I think the 8bit problem is a little harder to tackle, as the DAC is probably already doing most of what you're trying to do anyways. You don't exactly get a stair-step output from a DAC (or it wouldn't be a DAC, kinda) but rather a smooth curve between sample points, more or less.

Better 8-bit -> 16-bit conversion?

Reply #6 –
It's a shame the original didn't have noise shaping dither, because that can make 8-bit sound almost as good as 12-bit at 44.1 kHz sampling rate.

If it was truncated rather than flat-dithered to 8-bit, you might have serious difficulty because the truncation would have introduced harmonic distortion and new frequencies, especially noticeable in fade-outs.

You could try one of the noise reduction routines in Cool Edit 96 or above, Exact Audio Copy, Soundforge or Audacity. You'd need to feed it a sample of noise at the 8-bit level. If you have a sample with no signal, just noise, and that noise is representative, that may do a fairly good job and make it more pleasant to listen to, though some quiet sounds and transient effects may well vanish along with the noise.

You might even try something odd like convert to 16-bit, amplify by -48.2 dB (i.e. use only 8 least significant bits), encode with, say LAME into an MP3, then apply mp3gain of +48.2 dB. The ATH assumption at that level might just happen to not encode the low-level noise, but you could also lose some signal.

Better than applying -48 dB gain in a WAV editor would be to use a commandline with a scale in it, such as:
lame --alt-preset standard --scale 0.00390625 file.wav file.mp3

Then apply about +48.2 dB of gain in mp3gain to make it as loud as it began.

It's not likely to work very well, and a Noise Reduction routine is likely to be better, but it might be worth trying.

Better 8-bit -> 16-bit conversion?

Reply #7 –
I did try the CoolEdit noise reduction function, but it didn't work particularly well in this case - before and after just had different types of noise and I'm not sure which I liked better.