Possible Clean Conversion of 16 to 8 bit WAV without Dither Hiss or Optional Distortion?

[yellowmix]

I'm not a fan of dither either, but I remember doing experiments with SoundEdit 1.0 back in 1990 and you will have quantization noise in 8-bit. In all other higher bit audio, I feel that that the dither and quantization is usually invisible to our ears whether it's there or not. I quit using dither back in 2007 and nobody ever complained about it nor noticed. But I really do think you're maybe just hearing quantization noise and not dither. So you'll have to just cope with it and maybe do some manual edits if possible.

Yes, when FastTracker 2 introduced 16-bit samples, the noise floor lowered from -50 dB FS to -98 db FS. Quantization rounding error drops from 0.1958 dB to 0.001497 dB. That's two orders of magnitude. It was good enough for compact discs which billions of people listened to. The majority of streaming music is encoded from 16-bit PCM.

[Danno]

I remember 808/909 style kicks never sounded right because the 'clean' tails were always fuzzy. I'd have to raise the gain until the transient was clipping before the tail sounded good. Before FT2, Ultratracker was my first 16bit experience.

I kinda don't think there IS a way to make 8bit sound good. 12bit was kinda my threshold for 'acceptability'.

[gearwatcher]

SoX will let you do anything and is easy to batch using CLI commands.

https://madskjeldgaard.dk/posts/sox-tut ... processing

Off course, what it can't do is remove quantization noise inherent to 8bit audio.

[WackyZoundz]

Looks interesting. Does SoX also support high samplerates like 786 kHz?

[BertKoor]

Looks interesting. Does SoX also support high samplerates like 786 kHz?

Being a command line util and just dealing with data, I don't see why it won't convert to 1bit @ 6MHz.

[kidslow]

Wouldn't it need to be at least 2-bit?

[WackyZoundz]

No, you can have high quality sound with a single bit as long as the samplerate is high enough (like DSD with 2.8224 MHz): https://en.wikipedia.org/wiki/Pulse-density_modulation

[BertKoor]

Silence in 1bit can be '0101010101010101'
(with a low-pass filter at 1/4 of the sampling rate)

[gearwatcher]

No, you can have high quality sound with a single bit as long as the samplerate is high enough (like DSD with 2.8224 MHz): https://en.wikipedia.org/wiki/Pulse-density_modulation

Delta Sigma Delta still needs a Sigma (even if the integrator is in the analog domain for the DAC) to convert the 1-bit digital noise to the electronic sound signal.

It's one big "uLaw meets dithering" trick, it's not remotely the same thing as PCM.

8bit DSD wouldn't be the same as 8bit PCM and would still need to be converted to higher bits-per-sample PCM to be replayed by things expecting PCM.

[gearwatcher]

Looks interesting. Does SoX also support high samplerates like 786 kHz?

Being a command line util and just dealing with data, I don't see why it won't convert to 1bit @ 6MHz.

I am not sure that the DSD patches are in the standard build, as I have never used it.

But there's a DSD patch for SoX and you can find a version with it online for sure

[BertKoor]

I am not sure that the DSD patches are in the standard build, as I have never used it.

I was talking about 1bit PCM, actually...

[gearwatcher]

1bit PCM wouldn't be useful at all unless you're encoding 0dBFS square wave

[BertKoor]

I'm quite sure 1 bit PCM sounds totally fine at a sufficiently high sampling rate and proper low pass filter (which PDM also requires btw)

[EfreetiSultan]

Audacity, cubase, rx etc

[gearwatcher]

I'm quite sure 1 bit PCM sounds totally fine at a sufficiently high sampling rate and proper low pass filter (which PDM also requires btw)

Wrong. It would GREATLY depend on how it was encoded.

it would need to be DSD encoded for the reconstruction with the filter as integrator to work.