Sega Master System / Mark III / Game Gear|
SG-1000 / SC-3000 / SF-7000 / OMV
Home - Forums - Games - Scans - Maps - Cheats - Credits
Music - Videos - Development - Hacks - Translations - Homebrew
Sample rendering on MS I and MS IIPosted: Wed Jul 19, 2017 1:59 pm
My friend vingazole is searching/testing a sample player for my Cimmerian dev.
Recently, we tried GP Rider
Surprisingly the rendering is completely different between Master System 1 and 2...
on MS1 (and Megadrive) the sample is really clear/nice.
on MS2 its very noisy (like what you hear on emulators, plus seems to be lower).
Maybe people like Maxim, Sverx have some supposition on the code side, and hardware mens like Charles McDonald, Eviltim or Tmee on the technical side.
||Posted: Wed Jul 19, 2017 4:29 pm|
That's unexpected. Can you try Heliophobe's PSG tester at high frequencies and see what it does? All my testing was on an SMS2 (my original one from the 90s).
GP Rider uses an 8 bit sample underneath and I'm just converting the format to WAV, so it sounds like the SMS1 is applying some filtering.
||Posted: Tue Sep 05, 2017 5:42 pm|
I've tested heliophobe's PSG soft but i do not know what i should do with it...
So, i've taken the LUT from Robocop vs Terminator and samples are far more clearer than before (emulator and real SMS 2 hardware, a little bit more clean on SMS 1)
I've tried some LUT :
- GP rider
- Robocop vs terminator
Only RvsT is really "clean".
Others produce nrarly all the same output with a lot of noise.
These LUTs are coded on 4 bytes, but can be on 3 to save 256 bytes as noise channel is not used.
My samples are at 11025hz
Maxim, is it possible to generate a cleaner LUT than the one used by RvsT ?
Maybe you know other games (GG ?) which are using LUT (3 or 4 bytes) for PCM
||Posted: Tue Sep 05, 2017 11:37 pm|
Use pcmenc, it beats any LUT. However, mastering for loudness is hard, and the biggest determinant of quality is still sample rate (and data size).
I think I graphed all the LUTs in my sample extraction thread. They naturally suffer from the intermediate state during updates, which pcmenc is designed to minimise. You could probably brute force an optimal LUT given assumptions about the perfection of the volume table (the SMS2 one seems to truncate the loudest ones, for example).