Quote:

I guess it's a matter of making a very tiny read out loop. ... I guess there's a small preloader.

Yes , in both cases.

The casstte input reading loop looks like this:

loop  INC L
     IN F,(C)
     JP P , loop
 

Which takes only 26 tstates. IN MSX it takes 3 aditional tsates. With Spectrum it could take 0 or up to 6 aditional states. That's why MSX has more accuracy than Spectrm. The OTLA routine won't work at 19600 bps for Spectrum. But there are other faster reading loops that work with spectrum.

Quote:

Maybe we can combine MicroWAVer's compressing and RAM disk emulation abilities and OTLA's speed.

That would be great. Otla uses an external tool for converting . ROM to .BIN (only bin blocks can converted to wav with otla). You can change that msxr2b.exe tool for your personalize one.

Decicoder: at this speed, the idea of a virtual disk drive using mic-ear connected to a pc could work.
I guess that otla does not work in compressed form, but on uncompressed right?

A PC can listen to the mic audio input sampling at 96000hz or so, detecting the msx disk sector requests and sending the response throught the pc audio board into the msx.

If it's possible to work at 19600 baud (2.2 KB/sec) i expect with a light compression to reach 4KB/s.
For a virtual disk drive, virtualized via a mic/ear with no additional hw i think is good.

There's a project called Semi-Virtual_Diskette that emulates a disk drive. Schematics and source are available so it can be redesign to support larger drives.

http://www.thesvd.com/

http://en.wikipedia.org/wiki/Semi-virtual_diskette

There is a new version for OTLA.

http://code.google.com/p/otla

The mean feature for MSX users is that .cas files can be converted to high speed. (only bin files inside a .cas file)
Besides a phillips Vg8020 I could make hi-speed loadings on a Sony HB-10P. It works but not so fine as on the first one.