Converting 16bit 44.1khz audio to 8bit 44.1khz, Correct method?

[voodooless]

Do you have access to Adobe Audition? It can do quite a good 8-bit decimation with noise-shaped and content-dependent dither.

 

[Ropeburn]

8 bits gives you a factor of 2.
You know FLAC will give almost the same with no loss?

Which is besides the point, because the audio is going to be used in a strictly 8bit playback system: an old game console. Certainly can't play FLAC.

Generally, for that use case it really doesn't matter to find the "technically best option", because the whole target system and purpose is as far removed from audible perfection and "audiophile" quality and goals as anything could be, and any tiny quality gains will complety vanish in the system grain.

Above suggestion to just compress the audio to keep it loud, then simply convert to 8bit using basic dither is perfectly sound. No need to overthink it.

It just so happens I posted an example of this just earlier in the electronic music thread. That track is 8bit 22kHz samples running on the old 16bit Protracker engine on Amiga 500/1000. To get the bass as loud and punchy as that on this antiquated audio engine, you need to do exactly the above: take care with sample (=instrument) preparation and compress the hell out of it. That is what's going to make all the difference in perceived sound quality from an 8/16bit system, not fiddling with sophisticated upsampling methods that will get you imperceivable results.

 

[Hayabusa]

Which is besides the point, because the audio is going to be used in a strictly 8bit playback system: an old game console. Certainly can't play FLAC.

Generally, for that use case it really doesn't matter to find the "technically best option", because the whole target system and purpose is as far removed from audible perfection and "audiophile" quality and goals as anything could be, and any tiny quality gains will complety vanish in the system grain.

Above suggestion to just compress the audio to keep it loud, then simply convert to 8bit using basic dither is perfectly sound. No need to overthink it.

It just so happens I posted an example of this just earlier in the electronic music thread. That track is 8bit 22kHz samples running on the old 16bit Protracker engine on Amiga 500/1000. To get the bass as loud and punchy as that on this antiquated audio engine, you need to do exactly the above: take care with sample (=instrument) preparation and compress the hell out of it. That is what's going to make all the difference in perceived sound quality from an 8/16bit system, not fiddling with sophisticated upsampling methods that will get you imperceivable results.

ok, get your file in audacity.
generate a 1.0 level white noise track.
filter that track with a hp filter band pass 18-22.25Khz stop band 15-17Khz
downmix the tracks
export audio to 8 bits and voila: a 8bit dithered file

 

[Blumlein 88]

Okay, so using shaped dither in Audacity (Sox), I compared going to 8 bit 384 khz and then to 44.1 8 bit. There was slightly more noise above 7 khz than going from 16 bit 44.1 to 8 bit 44.1. Also a bit more noise across the whole band. I think it is because when you dither the second time you are dither some of the noise again basically giving up another bit or so to dither.

First the two step process.

Versus going straight from 16 to 8 bit. Shaped dither in each case.

I then tried going to 384 khz at 16 bit and downsampling that to 8 bit 44 khz. The result was the same as going straight to 8 bit at 44 khz.

 

[Ropeburn]

Okay, so using shaped dither in Audacity (Sox), I compared going to 8 bit 384 khz and then to 44.1 8 bit. There was slightly more noise above 7 khz than going from 16 bit 44.1 to 8 bit 44.1. Also a bit more noise across the whole band. I think it is because when you dither the second time you are dither some of the noise again basically giving up another bit or so to dither.

First the two step process.
View attachment 394350


Versus going straight from 16 to 8 bit. Shaped dither in each case.
View attachment 394351

Good you tried, because now you confirmed it's going to be zero difference, especially when the whole thing gets played back on an 8bit game console engine or equivalent emulator.

As said above, no need to overthink it. The whole application is so coarse, you could probably omit dithering altogether.

 

[AnalogSteph]

The whole application is so coarse, you could probably omit dithering altogether.

Nah, that would be a wee bit crunchy.

Nothing wrong with loads of compression + fancy dither. Just make sure to leave a bit of headroom for dither to fit into (it may be just 10 or 11 values either side but if you only have 255 to begin with that's like -0.8 dB...).

Something to consider would be the frequency response of the target system. Small tinny speakers may be more revealing of a 10 kHz noise peak, which would tip the scales more in favor of the Foobar dither.

 

[Blumlein 88]

Don't know what kind of sound will be in the files, compression combined with gating might help too.

 

[Tell]

Yeah this seem overly complicated, just decimate it with some noise-shaped dither and it'll probably sound almost as good as a normal 16bit file without noise-shaping at normal listening levels.

 

[popej]

• You then reduce the sampling rate (e.g., to 44.1 kHz). The high-frequency noise introduced by the dithering process is now above the Nyquist frequency of the new, lower sampling rate.
• During the downsampling process, this high-frequency noise is effectively filtered out, leaving you with a cleaner, lower-resolution audio file that retains much of the original quality.

Except that filtering and downsampling computation don't produce 8-bit round numbers. You need to truncate them again, and this again require dithering. Then the result is similar to direct conversion.

 

[dasdoing]

Dynamic Range, Dithering and Noise Shaping

Dithering, explained and demonstrated.

www.audiocheck.net