VOGONS


Munt Reloaded - Development

Topic actions

Reply 60 of 965, by KingGuppy

User metadata
Rank Member
Rank
Member

Hi h-a-l-9000,

Coincidentally, through some driver tweaking I finally got my digital capture device working reliably yesterday. It's a µChameleon from http://www.starting-point-systems.com/ with a custom test firmware from the manufacturer.

The input to the DAC is 16(ish)-bit samples at 256kHz (8 multiplexed streams at 32kHz each, though two of the streams are always silent). See also discussion and examples of digital capture in this thread: Munt development ?

Reply 61 of 965, by h-a-l-9000

User metadata
Rank DOSBox Author
Rank
DOSBox Author

Ah, nice you got yours working. I just thought after programming a CPLD for digitally capturing OPL3 data I could extend that for parallel 16-bit data too, but with these 6 channels its a bit too much (I think).

1+1=10

Reply 62 of 965, by circle_theory

User metadata
Rank Newbie
Rank
Newbie

Awesome! I just downloaded Ykhwong's DosBox version and finally ... YES FINALLY ... Space Quest 3's opening theme song sounds PERFECT!! This is truly a glorious day for me since my Roland MT-32 has long been broken and the previous MT-32 emulation wasn't quite right ... but this implementation of the MUNT Reloaded patch is SPOT ON! Well done guys!! You have my UTMOST RESPECT FOREVER!!

Reply 63 of 965, by ponytear

User metadata
Rank Newbie
Rank
Newbie

I just tried Ykhwong's DosBox too. Great version but which version of the new MT-32 emulation is it? Legend of Kyrandia is sounding wrong again (and Budokan too), when it sounds great with the version I downloaded from the Munt Reloaded site two weeks ago.
Anyway, great work!!!

Reply 64 of 965, by sergm

User metadata
Rank Oldbie
Rank
Oldbie
ponytear wrote:

I just tried Ykhwong's DosBox too. Great version but which version of the new MT-32 emulation is it? Legend of Kyrandia is sounding wrong again (and Budokan too), when it sounds great with the version I downloaded from the Munt Reloaded site two weeks ago.
Anyway, great work!!!

Gentlemen, please be patient!

The work is in the progress. Some tunes are sound really good but just *close* to the true MT-32’s sounding. There are dozens of questions about how exactly it is supposed to do. While such parameters as pitch and modulation are corresponding almost exactly, others like filter and partial amplitudes are still not worked out. Unfortunately, there is no actual technical information in official docs.

While MUNT behavior tuned according to my guess for playing properly one instrument (or tune), then others instruments start playing wrong. And this way is certainly misleading.

I hope KingGuppy provide us with finest digital samples very soon, and most of our questions will be answered…

Reply 65 of 965, by ponytear

User metadata
Rank Newbie
Rank
Newbie

> Gentlemen, please be patient!

Sorry if I sounded impatient because I am really not! It's working great so far and I am more than satisfied with the current results! You and others have been working a lot on the project and I/we appreciate all your efforts! 😀

I was just curious about which version of the emu is implemented in Ykhwong's DosBox, if it was a newer version than the one available two weeks ago or not, since it sounds so different (for Kyrandia at least).

Reply 66 of 965, by sergm

User metadata
Rank Oldbie
Rank
Oldbie

Ykhwong's used in the DOSBox patch the latest sources published at GitHub. There are changes in sawtooth wave generator compared to that binary you have tried early. But as you can hear, these still don't make the MUNT sounding correct.

Reply 67 of 965, by sergm

User metadata
Rank Oldbie
Rank
Oldbie

It seems that the pulsewidth parameter handled by MT-32 in a way not corresponding to D-50 user guide. The latter claims that the pulsewidth parameter should change widths of upper and lower segments along the way but seeing BLIT WG produced squarewave it looks like there is no change in the shape of a waveform for pulsewidth <= 50. Listening instruments justifies this behaviour. So, pulsewidth handling changed in accordance.

EDT:
It's confirmed that MUNT really ignores pulsewidth values < 50 while calculating pwFactor.

Attachments

  • pulsewidth hanlding.gif
    Filename
    pulsewidth hanlding.gif
    File size
    3.37 KiB
    Views
    1316 views
    File comment
    Pulsewidth hanlding
    File license
    Fair use/fair dealing exception
Last edited by sergm on 2011-03-22, 15:16. Edited 1 time in total.

Reply 68 of 965, by sergm

User metadata
Rank Oldbie
Rank
Oldbie

If anyone is interesting in testing and have troubles with building sources, I enclose that dotnet win32 app precompiled. I think this version of "Linear" wave generator can pretend to be a "release candidate".

P.S. Corrected an exception condition caused by overrun of the filter coefficients array

Attachments

  • Filename
    mt32emu_win32app_CLR.zip
    File size
    63.37 KiB
    Downloads
    62 downloads
    File comment
    Precompiled dotnet win32 app for testing "Linear" wave generator
    File license
    Fair use/fair dealing exception

Reply 71 of 965, by sergm

User metadata
Rank Oldbie
Rank
Oldbie
Miki Maus wrote:

The parts of music still sound overdriven in Prince 2 even if I turn down volume to 30.

Yes, I've found Santour's overdrive of Prince 2 too. I think the reverb tuning is a cause. Ykhwong's DOSBox build even don't use freeverb at all.

But there is just one note: volume in that dotnet app is applied to the final mixed sound in the app. It isn't sent to the synth via SysEx as it does display application when you turn the volume knob. So, you cannot get rid of overdrives by lowering volume in the synth app. But this allows you to set volume higher up to 10 times gain (with overdrives of course). It's just useful for testing when you listen up to sound details or just one single note.

Reply 72 of 965, by Miki Maus

User metadata
Rank Member
Rank
Member

Oh, I see. I thought it lowered overall volume before mixing.

Maybe you can add more knobs for dynamic control of different parameters for development and fine tuning/testing purposes?

Reply 74 of 965, by sergm

User metadata
Rank Oldbie
Rank
Oldbie
ykhwong wrote:

I decided to use delayReverb rather than freeverb in order to resolve noises in some games. If that leads to the inaccurate tuning, then I will put it back.

I think the reverb is not a thing to worry about for now 😀 Although, freeverb sounds much more realistic, IMHO...

Reply 75 of 965, by sergm

User metadata
Rank Oldbie
Rank
Oldbie
Miki Maus wrote:

Oh, I see. I thought it lowered overall volume before mixing.

Maybe you can add more knobs for dynamic control of different parameters for development and fine tuning/testing purposes?

There are too many of them, IMHO 😉 But there are plans (not for now) to redesign the interface and to made it cross-platform.

Reply 76 of 965, by sergm

User metadata
Rank Oldbie
Rank
Oldbie

At last, I've got many-many samples from KingGuppy....

And I've made some changes to the munt_reloaded branch. Now, pulsewidth calculated in correct range (from sample analysis) and WG produces almost perfect sawtooth waves. Sounds beautiful... It's the good news...

But there is bad news... I think this is the very last munt_reloaded update 😉 The sample analysis performed clearly shows us the way it must be done properly. I’m beginning with a completely new WG code which (after thorough tuning) is supposed to bring us brilliant results. So, let’s see...

Attachments

  • Filename
    mt32emu_ext_int_CLR.zip
    File size
    63.52 KiB
    Downloads
    56 downloads
    File license
    Fair use/fair dealing exception

Reply 77 of 965, by Dominus

User metadata
Rank DOSBox Moderator
Rank
DOSBox Moderator

@yhkwong: with your last patch the birdy sound got much better but is still off. Possibly too much reverb or so. And then the bird sounds are much louder compared to the flute that starts playing at some point. Comparison at http://dl.dropbox.com/u/7737184/u7mt32.zip. U7introMT32-EMU.ogg vs U7introMT32.ogg, the emu one captured with Dosbox' wave capture and converted to ogg, the other one recorded with a real mt32.

Windows 3.1x guide for DOSBox
60 seconds guide to DOSBox
DOSBox SVN snapshot for macOS (10.4-11.x ppc/intel 32/64bit) notarized for gatekeeper

Reply 78 of 965, by ykhwong

User metadata
Rank Oldbie
Rank
Oldbie
Dominus wrote:

@yhkwong: with your last patch the birdy sound got much better but is still off. Possibly too much reverb or so. And then the bird sounds are much louder compared to the flute that starts playing at some point. Comparison at http://dl.dropbox.com/u/7737184/u7mt32.zip. U7introMT32-EMU.ogg vs U7introMT32.ogg, the emu one captured with Dosbox' wave capture and converted to ogg, the other one recorded with a real mt32.

The reverb comes from delayReverb. I reverted it back to freeverb again. Also, recent change of munt_reloaded fixed noise issue with freeverb. Give it a try to copy this to my build directory.

http://bigmail-file19.mail.daum.net/Mail-bin/ … k1PVN-Dj7O2pAV1
EDIT: If the above link does not work, then you can get it from my website.

Last edited by ykhwong on 2011-03-28, 05:33. Edited 1 time in total.