Sample transfer to/from Mirage using Alchemy 3.0 + others

[ioguie]

Hey all,

I've had some adventures transferring samples to/from my Ensoniq Mirage using various vintage software.

I can confirm that the Digidesign software (Sound Designer Universal, TurboSynth 2.0) works best with "basic" MIDI interfaces (MidiMan Macman, Opcode MIDI Translator II) connected to a serial port, on a 68k native machine.  For various reasons, these programs won't communicate with the Mirage in other configurations: 

• as noted in the OMS or Studio 4/Studio 5 manuals, some early Digidesign software for samplers used non-standard communication (baud?) rates when communicating with the standard MIDI interfaces, which the Studio 4 and Studio 5 can't handle, even in compatibility mode
• Even the native serial port on the G3 Wallstreet doesn't work;  it seems related to PowerBook serial ports not always being "powered on," and the Digidesign software not using the serial port correctly (which OMS would normally take care of);  it's possible that using OMS Preferred Device and leaving an OMS application running in the background would take care of this, except:
• it seems like there is some timing code in the Digidesign apps that is very tied to the speed of early Macintoshes, and perhaps runs too fast on PowerPC...?

Now, Alchemy is a different story.  Alchemy has support for OMS.  Unfortunately, I've found that Alchemy on the G3 Wallstreet still produces anomalies when transferring samples to/from the Mirage, whether using the native serial port with "MIDI Driver," or using OMS, or using OMS with non-serial drivers, using Studio 4 or Studio 5, etc.  Some pages (packets) get dropped without warning.  It is very easily reproducible:  I boot up with MASOS, then use Alchemy to grab sample #1.  Usually no problem here,  Sample #1 ends up in the Mac just fine.  But then I reverse the sample, and send it back to the Mirage.  I then transfer the sample *back* into Alchemy and compare it with the original.  Without fail, page 0 (or page 1), the first 256 samples, will still contain the contents of the original, non-reversed sample.

This extends to other samples as well:  I can grab sample #10, reverse it, send it back, and then grab it again, and random pages of it will still contain the old contents.

I believe this is a timing issue, but of course, where the timing issue exists is a mystery.

The interesting thing for me is that Alchemy does not miss pages on 68k native machines, like my SE, IIfx, or Centris 630.  Tested with MIDI Translator II, MacMan Midiman, and Opcode Studio 4 (running in compatibility mode with OMS Preferred Device and "MIDI Driver" mode in Alchemy, *or* in OMS mode in Alchemy -- both work fine).

Anybody else have experience with this ancient support for the Mirage?  From a historical standpoint, it is interesting to me that Blank Software and Digidesign both seem to have gotten their start with computer software that talked to the Mirage.  Not only for the Mac, but also for the Amiga and Atari.