Information on Akai S1000's sinc interpolation (and code to generate its kernels)

[8bitbubsy]

Some time ago I found out that the classic Akai S1000 sampler has its windowed-sinc kernels stored in two EPROMs (lo/hi). After analyzing the data, I discovered that it has 4 kernels (varying sinc cutoff) with 6 (not 8!) points/taps at 2048 phases. The format is signed 16-bit integer with a scale of 15 bits.

I have created some code that generates almost perfect approximations of these kernels: https://pastebin.com/7nd5Xtqx
EDIT: Argh! Not enough post count to post links. Pastebin (dawt) com forward-slash 7nd5Xtqx

It also contains code for dumping the kernels from a IC11/IC12 ROM dump.

Anyway. This makes me believe that the S1000 uses 6-point sinc interpolation internally, and not 8-point which several people claim it uses. Of course, it would be possible for it to copy the kernels to RAM on start-up (which it probably does), then create 8 points out of them through additional interpolation, but I have my doubts. S1000's BIOS ROM is compatible with Intel 286 (or 8086?) assembly instructions, so it should "somehow" be possible to find out.

Sure, this is probably not all that interesting, but maybe it's useful for someone. Please keep in mind that the original ROMs are of Akai's property, so for that reason I will not share them, but for those that keep looking you will "maybe" find them somewhere.

[wikter]

Thanks!

[Alloyed]

hi! pastebin took down the link here, unfortunately. So if you have another way to share the code (or you can send it to me and i can mirror it) that'd be handy!

[8bitbubsy]

hi! pastebin took down the link here, unfortunately. So if you have another way to share the code (or you can send it to me and i can mirror it) that'd be handy!

https://16-bits.org/etc/akai_s1000_sinc_lutgen.c