'I wasn't willing to pay money for a copy of IEC standard 61119, so I went looking for the info elsewhere. I believe RFC 3190 contains a description of the 12 bit non-linear encoding:
https://tools.ietf.org/html/rfc3190'
Cool!
'Sending encoded 12 bit values to a linear DAC will surely give horribly distorted audio.'
Aha. If that is the case, then can we presume that when output via S/PDIF, a 12/32 DAT has already had its bits mapped to 16-bit linear values? That would be good news for my purposes, I think...

May 2 2017 4:53 PM, "John Chester" <[log in to unmask]> wrote:
> On 5/2/17 4:32 PM, Eli Bildirici wrote:
>
>> Fascinating. Thanks for explaining this. I wasn't 100% sure what non-linear meant in this context
>> and didn't want to derail, but hey, might as well. So the output after having been remapped using
>> decode_lp_sample or similar is 16-bit. How did e.g. Mr Taylor in that function arrive at those
>> values for this look-up table though?
>
> I wasn't willing to pay money for a copy of IEC standard 61119, so I went looking for the info
> elsewhere. I believe RFC 3190 contains a description of the 12 bit non-linear encoding:
> https://tools.ietf.org/html/rfc3190
> (see page 2).
>
> decode_lp_sample is a lookup table which converts a 12 bit encoded value to a 16 bit linear value.
> It appears to agree with the encoding shown in RFC 3190. I haven't followed the code in read_dat.c
> far enough to understand how the 12 bit values are packed on the tape.
>
> Sending encoded 12 bit values to a linear DAC will surely give horribly distorted audio.
>
> -- John Chester
Eli Bildirici
(347) 837-8337