VOGONS

Common searches


First post, by jal

User metadata
Rank Oldbie
Rank
Oldbie

I recently purchased an SAAYM sound card from TexElec, which sports two SAA1099s and should be Gameblaster / CM/S compatible. However, I noticed after writing a quick test program that in DOSBox the notes play a lot higher (8 halfnotes to be precise). Is this a known inaccuracy? If not, is there any way to properly test whether the problem lies with DOSBox or with the SAAYM? (I wouldn't expect the SAAYM is too blame, it seems to use the right clock frequency.)

JAL

Reply 1 of 7, by davidmorom

User metadata
Rank Newbie
Rank
Newbie

I recently built myself a custom CMS clone, and noticed the same problem. I also compared with the CMS emulation of PCem and with Youtube videos of real hardware, and all of them sound like my card, so DOSBox CMS emulation is definitely out of pitch.

Reply 2 of 7, by jal

User metadata
Rank Oldbie
Rank
Oldbie

Ah, thanks for the confirmation! I'll see if I can find where the problem lies in the DOSBox source code. It's quite annoying that my development (which I mainly do in DOSBox) sounds different than the real thing (my NuXT with SAAYM).

Reply 5 of 7, by davidmorom

User metadata
Rank Newbie
Rank
Newbie

I've just tested with DOSBox r4481, DOSBox ECE r4481 and DOSBox-X 2022.09.0 and all of the presents the same incorrect behaviour. As and example, I attach a record of the Monkey Island 1 intro, taken directly from my CMS clone with two real SAA1099 chips, which sounds exactly as 86box and PCem emulation: https://www.youtube.com/watch?v=Ur-zXDIUB3U

And a recording from DOSBox, that allows to crearly apreciate the issue:

Filename
MI_intro.mp3
File size
2.4 MiB
Downloads
50 downloads
File license
Fair use/fair dealing exception

Maybe that commit didn't fixed the problem correctly, or it was reintroduced by a later regression, but there is still an issue.

EDIT: tested also with DOSBox Staging 0.80.1 and it sounds correctly. Something weird is going on.

Reply 6 of 7, by ripsaw8080

User metadata
Rank DOSBox Author
Rank
DOSBox Author

In my tests, pitch changes occur when the saa1099 sample generation rate does not match the mixing sample rate. After the clock rate changes committed in r4010, subsequent updating to the then-latest MAME sound cores introduced a difference in rates. The attached patch (experimental for now) appears to resolve the issue in current source; however, the latest saa1099 core may (I'm not sure) have a method for changing the sample generation rate that is not solely based on the clock rate.

Edit: revised the patch with a better/safer default sample rate.

Attachments