I've been making some more comparisons between ay8910 noise from real MSX and MESS and came to the conclusion that although very comparable, the noise sounds generated by MESS seem to miss some of the dynamics / "crispiness" of the real MSX.
I've tried to generate a good example of this, which unfortunately isn't easy, but hopefully the next one brings it across well enough.
It's a 4 second recording of the choppersound from the MSX game H.E.R.O. (a classic
), versus the same recording in MESS. In the audio sample the first 4 second part is from the real MSX and the second part is from MESS.
Please download soundsample here: http://www.filedropper.com/heronoisecomparison
If you listen to it a few times, you'll (hopefully) notice how the real MSX sound is more well defined, and the MESS emulation is sounding more "mushy" (for lack of a better word).
To me this is exemplary for most of the MSX noise emulation.
I've added the audiogram comparison in the picture below. It shows how the real MSX has more dynamics, versus the MESS part which is looking like a block diagram, cut-off at the bottom.
I'm hoping that a netlist implementation would improve this by some margin. I remember you had (or were working on) a preliminary / test implementation for an msx netlist based on the spectravideo schematics (referring to this post
). Would it be a possibility that we could test that implementation, just to see how it would impact the current example?
Hopefully this can provide a good objective testcase to get the MSX / ay8910 sound emulation another step ahead.