interface drives with ATA (IDE) drives. It's getting hard to find Maxtor 2190s at a decent price, and I'm not sure how many more years it will be possible to keep them running at all. The basic concept would be to simply sample the MFM (or RLL) channel code at somewhere around 50 MHz as it is being written, and store the data on the ATA drive. The interface would keep one track buffered in RAM. Whenever the host requests a head change or seek, the buffer would be written to the ATA drive (if it is dirty), and reloaded with the data for the new track. This would require a much larger ATA drive than the original drive it is emulating, 102K bytes per track. To emulate a Maxtor 2190, this requires a

ATA drive. You could even emulate drives larger than that by emulating

cylinders. The tricky part is that it might be necessary for the simulated drive to mimic the effect that write precompensation in the host's controller is intended to circumvent.