Yep, its a digital rip from a v1.07 rom, pcb 01 unit.
This is the HIGHEST POSSIBLE QUALITY you can extract from a v1.x mt-32 (15-bit digital)! Its digital all the way from the chip to the computer, no analog whatsoever!
(A v2.x mt-32 and a cm-32L/LAPC-1 do 16-bit digital, but have some differences in the la32 and the control code.)

To get 'true bit accurate' content, you have to make the following changes to KingGuppy's instructions:
Edit: 32000hz is correct, leave it at that.
Set a byte offset of 8 to the file, this fixes the fact that 4 extra samples were captured at the beginning.
Set channels to the following pannings:
1: right
2: right
3: mute
4: mute
5: left
6: left
7: right
8: left

The more informed among you will notice that the channels follow the time-slot order on the mt-32 schematic.
The channels are:
1: Right reverb send/out
2: Direct Right out
3: nothing
4: nothing
5: Left reverb send/out
6: Direct Left out
7: Right reverb recieve
8: Left reverb recieve

You can also set the +/- slider to boost channels 2 and 6 by 3db each, to counteract the 50% mixing of reverb wet and dry. (this is optional and may not be completely correct)

Lord Nightmare

Very interesting. I hear that some of the noise unfortunately is in that recording as well, so I guess that's just what the LA32 chip outputs, not a fault of the DAC or deglitcher as I hoped.

I would ask you how exactly you did the capturing, but if bit-perfect emulation is now within reach, why bother? 😀

These rips are exactly what's going to help make the emulation 'bit perfect' in future. That's how I read it anyway...

Here's a terrible camera photo of my setup for that rip:

http://www.kingguppy.com/files/MT-32-DAQ.jpg

I'm using the µChameleon to capture the digital data:

http://www.starting-point-systems.com/

There are some limitations:

• The firmware for doing this sort of streaming is still in beta.
• The streaming mode is very sensitive to the USB interface used. I only got it working reliably on one PC, and then only with a buffering USB hub attached (which I bought for the purpose on SPS's recommendation). They're working on this, but it's not certain that the problem will be solved.
• I can only capture 16 bits of data at a time, triggered by a separate line (rising edge, falling edge or - in this case - both). There are 2 more lines going to the demuxer that would be nice to capture since they indicate which time slot is being output at any time. However, in most cases it's easy to determine this from the data going to the DAC.

Positives:

• I didn't find anything cheaper to get the job done without requiring more electronics skills or time than I had.
• Starting Point Systems' email support is truly excellent.

I've now switched to capturing from my CM-32L instead of the MT-32, giving a full 16-bit resolution.

Thanks to Lord Nightmare for advice and keeping me motivated.

Actually, the early MT-32 has 16 bits as well; it's just that the 14th bit is not connected to the DAC, and the lower bits are shifted to the left to fill up the hole. The original 16 bits can be found on those pads to the left of the LA32 chip, right below that "M|1" print.
Not that this will improve the sound quality or anything, just something to keep in mind when using this to find out what the LA32 is doing.

By the way, could you post just one more example? I'm referring specifically to the music in Fate of Atlantis that plays either (a) when Sophia tells your fortune, (b) during the Séance, (c) when riding the crab raft. That tune sounds so noisy/grainy on all modules, I would like to hear how it sounds with the digital capture method. Only if it's not too work though. 😀

I have attached the mixdown for everyone's convenience. Hope I mixed it right 😉

NewRisingSun: I know, but I couldn't actually access the 15th bit (bit 14) easily. I'm using an IC socket wedged on top of the DAC to get at the data lines.

If you give me a MIDI file, and let me know which bit you'd like me to capture, I'd be happy to (preferably no more than ~16s).

Well, 18 seconds. 😉 (.zipped because this board won't accept .mid extensions.)

It has to be that long because I'd like to see if there's any zero-settling noise. If it's not possible at all, just snap off the final two seconds.

Thank you for your trouble.

Sorry about the delay - I'm having transient problems with missing the occasional timer edge transition. This should be OK, though: http://www.kingguppy.com/files/CM32L_NewRisingSun_TEST.zip

The channels are in the following order:

1. Right reverb dry
2. Right non-reverb
3. Nothing
4. Nothing
5. Left reverb dry
6. Left non-reverb
7. Right reverb wet
8. Left reverb wet

The rest is as previously described.

I'm not actually sure whether the sample rate is 256kHz (as I originally stated) or 262144Hz (as Lord Nightmare wrote). From my measurements - which are rather inaccurate due to buffering and other issues - it looked like the former.

You are right one, 32000 Hz is right, 32768 kHz sounds too high.

Thank you very much for your trouble.

I apologize for that mistake. Sorry.

Any news on development kingguppy?

Here I am enjoying Monkey Island 2 on my Nintendo Wii console with the munt mt-32 emulator enabled.

And I am thinking to myself, why this actually sounds good!

It's a little bit off, and there is notably a bit of tweaking to be done.

And yet still so close to sounding like the real thing.

CanadaCow & KingGuppy. What happened?

What happened?

Life, work, other projects, etc. 😢

If you have an iPhone you can grab my latest creation here: www.zinger-soft.com (I'll even hand out free coupons to forum users!)

Just the same I revisit the MT-32 emulator and the real deal occasionally. You're right. They're sooooo close, but at times, so far. That's kind of the brick wall that at least I ran into. Tweaking the sound to make one instrument work broke another instrument. Eventually my wife wanted to kill me because of all the tuning and note sounds. 'bout drove her (and even myself insane.) I had hoped by now someone else would have picked up the mantle and put new life into my project. Oh well. It'll get completely done one of these days.

canadacow wrote:

If you have an iPhone you can grab my latest creation here: www.zinger-soft.com (I'll even hand out free coupons to forum users!)

Nice one! I just purchased iSSH from the App Store 😁

I wish Munt Windows development somehow resumes.

canadacow, nobody can replace you. 😉

Too bad you stopped working on it.
Can you post the info you find helpfull to continue this project, so someone who for example would participate at Summer of Code, or anyone else for that matter, could easily pick up where you left?

Quick update on this: The project is very much alive.

Mok (who posted some interesting tidbits on this forum a while ago) has provided a powerful shot in the arm to the project. He's been providing us with an endless stream of insights into the MT-32's inner workings - I'm having a lot of trouble keeping up with him. The results should be available in the near future.

Canadacow made an announcement here: http://queststudios.com/smf/index.php/topic,2873.0.html

^now that's good news! 😀

Quick update on this: The project is very much alive.

Mok (who posted some interesting tidbits on this forum a while ago) has provided a powerful shot in the arm to the project. He's been providing us with an endless stream of insights into the MT-32's inner workings - I'm having a lot of trouble keeping up with him. The results should be available in the near future.

Exciting! Will linux users be able to build this and the alsa driver module with something other than gcc3.3? (I hope, I hope ) 😁