Listening right now-

1. Instruments sound wrong. Ignoring all else, only the drum kit sounds right, and it's ok. 😀 The actual instruments the song uses sound incorrect. Maybe 1 or 2 are right, but they sound not SQ3's SysEx instruments. What are you using?

2. Pitch is totally screwy.

Interesting. Nothing special at the moment, but interesting. Good luck with this, this sample's pretty 'interesting' at the moment but keep trying. I'm listening.

- Alistair,
http://www.sierramusic.tk

Spikey wrote:

Maybe 1 or 2 are right, but they sound not SQ3's SysEx instruments. What are you using?

My MT-32 emulator is doing its best to interpret the MT-32's original SysEx parameters to make the identical sounds. Specific documentation on what those parameters mean is hard to come by. I mean, heck, I have 6 manuals that say stuff like "When the TVA Velosity Sensitivity is high the amplitude is greater for higher velocities. When the TVA Velos Sens is low, the amplitude is lower" Of course, nothing more specific than that. No actual formulas or degrees to how much things are affected. What this means in the long run is that slowly, ever so slowly, I am inching my way closer to the correct sounds as I tediously compare the output of my emulator with the output of my MT-32 and try to figure out what the MT-32 is doing differently.

2. Pitch is totally screwy.

No. The pitch is dead on. Getting the pitch right is one of the easiest things since all MIDI devices perform the same way in this department. What's making you think that the pitch is screwy, however, is the lowpass filter that's not interpreting the cutoff frequency and TVF parameters correctly. If I were to disable the lowpass filter (which I have done occasionally to verify other things) you would agree that everything is playing at the right pitch, only then all synthesized sounds would sound much brighter than they should.

Interesting. Nothing special at the moment, but interesting. Good luck with this, this sample's pretty 'interesting' at the moment but keep trying. I'm listening.

Thanks. I'm glad I still have people following my progress. For a little bit of humour, my real sounding board isn't this group, but my wife rather. She couldn't care less about this project and forces me to wear headphones when working on it. Occasionally, however, I'll force her to compare the real MT-32 Monkey Island theme with my version. Its very reassuring to hear even her say I'm getting closer but ultimately its still "too flat" and not quite right. When she says its correct I'll know I'm done. 😜

Could you please upload a sample of, let's say MI1 with the low pass filter disabled? I'd like to hear how it sounds, musically, then...

Personally, I'd rather hear 'overly bright' instruments than instruments played with/through a crappy low-pass filter.
Put me down for a SQ3 version, too. 😀

- Alistair

Spikey wrote:

Personally, I'd rather hear 'overly bright' instruments than instruments played with/through a crappy low-pass filter.

I certainly wouldn't call it a "crappy low-pass filter". I'm rather limited in my options with regard to my choice of low-pass filter algorithms. Since the filter needs to be applied to every sample of every partial of every instrument, what I implement needs to be fast since I could be filtering up to 32 waveforms to generate a single sample of sound output. I therefore (Unless you have a dual Athlon-64's running at 3Ghz) can't use a trigonometic four pole/four zero moog filter as I ideally would. Of course, I can't just forgo the filter either. That's not how the MT-32's subtractive synthesis works. FM synthesis, like the Adlib are known as additive synthesizers. They take clean waveforms and add them together to make sounds. Old school synthesis, like the MT-32 and ealier synthesizers with analog components worked in the opposite direction. They take harmonic rich waveforms (a square or sawtooth wave) and then filter out the unneeded harmonics.

As for posting what my emulator sounds like without the filter... I'll post what I have along with a filtered example (using SQ3 as an example). I'm getting pretty close to posting a new binary, though not quite yet. I'm still not completely happy with the LFO.

I know, I know. I'm saying crappy as compared to the MT-32's. Why can't you use the MT-32's anyway? </layman>

- Alistair

Examples posted at their usual place:

http://www.artworxinn.com/alex

Spikey wrote:

I know, I know. I'm saying crappy as compared to the MT-32's. Why can't you use the MT-32's anyway? </layman>

- Alistair

I'm not exactly sure how the MT-32 implements its filter. There's no specific documentation on that. I assume its a traditional Moog lowpass filter. There are several different variations and implementions of such a filter, each with its advantages and disadvantages. The main advantage I need to look out for, however, is speed. There's just so much that has to be done for each generated sample that time is of the essense. One of the reasons that its taken 17 years to see the production of an MT-32 emulator is because until now, general purpose processors (like the Intel x86 series) just weren't fast enough.

I'm gonna close this thread...canadacow please make a new one for every new release with dated subject so that runoff between the versions is minimized.