The MT-32 --- even crappier than we thought.

[Rhizome]

This could be a dumb question but... has anyone tried to track down the engineer(s) who designed the MT-32?

Bloody hell, I would love to get some information from them if we were to ever find them.

[Cloudschatze]

This could be a dumb question but... has anyone tried to track down the engineer(s) who designed the MT-32?

I've mentioned this elsewhere, but "Masa" was the main engineer, and Adrian Scott was the sound designer.

I've spoken with a former Roland software engineer, Adrian Bruce, who implied that he still maintains some contacts.

Yes, I did ask for Masa's contact info. No, I wasn't given it.

However, I wonder if he might be willing to forward on a question or two. I'll find out...

[Laust]

If the LA32 chip is a Fujitsu DSP, the most likely candidate would be the MB8764, which was introduced in 1983. It was even available in the same package as the original LA32 PGA chip (you can find a datasheet if you search those online datasheet archives). The idea that it's a DSP instead of a totally custom IC is not that far fetched. It would certainly have reduced Roland's development time, I think.

However, the idea is not without problems either:

• the MB87/89 prefix is used not only for DSPs but a whole range of chips, video encoders/decoders, PLLs, SCSI controller chips, etc. (as said archives will also reveal). It could be a prefix that simply means integrated circuit or maybe large scale integrated circuit (LSI/VLSI). So the number tells us nothing.
  
• The pinout doesn't match (a few pins do though ), and more importantly, the various busses don't add up either.
  
• I don't see how an MB8764 could generate the SHx signals. The MB8764 definitely not an MCU, there are no timers, interrupt handling, general purpose I/O, (some of) the things you would need to generate those signals. The MB8764 is simply designed for reading data in, processing it, and spitting it out again. The LA32 chip does slightly more than that.

I couldn't find much info on the MB89232, but it seems to be a much larger chip (both capabilities and physically), with high speed serial busses (as opposed to parallel in the old one) and was released in 1987, so it's a bit unlikely Roland would have a product with it out the same year.

Of course, it is always possible the chip is a custom version of some DSP tailored to Roland's requirements. If they were willing to pay and buy in enough quantity, I'm sure Fujitsu would have obliged.

[Cloudschatze]

Of course, it is always possible the chip is a custom version of some DSP tailored to Roland's requirements. If they were willing to pay and buy in enough quantity, I'm sure Fujitsu would have obliged.

This seems the most likely scenario. Most of the service notes refer to it as some variation of "synthe custom IC."

[DragonBoy]

This is all heresay, but:

There was an interview with Eric Persing in Keyboard Magazine once where he said that at the time he worked with it, the D-50 was in breadboard stage and entirely represented in software.  The original implementation had five LFO's, but when they finalized the on-chip hardware implementation implementation it couldn't quite keep up with it so they had to back down to three, and the chip definitely wasn't powerful enough to do a multitimbral D-50.  He said the D-10, D-20, D-110, and other follow-on products were "low-end" market products that sacrificed sound quality for multitimbrality.  According to him, the next successor synth that rivaled the D-50 in sound quality was the JD-800., so I went out and bought one.

[Zemus]

Heh, good thing I saw this thread. Looks like there's no point in completing my SPDIF mod for the CM-64.

[Rhizome]

Heh, good thing I saw this thread. Looks like there's no point in completing my SPDIF mod for the CM-64.

Wouldn't that mean you'd need two SPDIF outs (one for the PCM and LA board)?

Unless you're only interested in the LA board

[Zemus]

I never use the CM-32P part so it'd be only for the LA board.

[NewRisingSUn]

Actually a SPDIF mod would still be a good idea, since you're at least getting rid of all analog noise. So if you have the electronics knowledge, go for it!

[Zemus]

I got a CS8404 for free, but since it only accepts serial data, I'd have to put together a shift register converter so that's what I've been fiddling with.

[NewRisingSUn]

You'll also need to demultiplex the 256 kHz signal, which is an interleaved 8x32 kHz signal, to the stereo 2x32 kHz. I take it you have already considered that.

Will you rerecord your soundtracks digitally? Shouldn't be too hard if you still have the MIDI files.

[Zemus]

Yeah, I'm not sure how to do the demultiplexing right now, so I'll have to find a solution to that one...

The actual recording of the soundtracks into OGGs is the most boring part of making them, so I'll really have to see about that.