Refurb weekend: Commodore SFD-1001

The Commodore SFD-1001 is an oddball and a rarity even among Commodore IEEE-488 5.25" floppy drives, which nowadays aren't particularly common either (though my preferred IEEE-488 device is the MSD Super Disk Drive SD-2, which also conveniently has IEC serial). The SFD "Super Floppy Drive"-1001 is a low-profile single drive version of the CBM 8250 dual drive and stores a whopping 1MB per disk, which when the series was introduced in 1980 was really quite something. Unfortunately it requires 96tpi double density "quad density" floppies to do it — not the 48tpi double density disks you'd feed a more typical 1541 or 1571, nor the 96tpi high density PC floppy drives use — so there wasn't a whole lot of megabytes to store into even when these drives were newer. I also have a CBM 8050 which is the single-sided (but still dual drive) version of the 8250; it can store roughly a cool meg too but you have to flip it over for the second half.

This SFD came to me from a seller who said while it powered on it also made a weird noise and he had trouble inserting disks. Nevertheless, it's still an odd enough duck it was worth buying to see if it was repairable. We have an extra hour this weekend from the daylight savings change, so let's crack this sucker open.

We remove four screws from the bottom and take off the top.

The main circuit board is on top (a revision A from 1983). Commodore disk drives are intelligent peripherals and have their own CPU (a 6502, naturally) — Bob Russell once commented that the 1541 disk drive was the best computer Commodore ever made — but you're not seeing double here. There are indeed two 6502s in this drive. The reason is that Commodore drives were originally designed to have separate CPUs for the Interface Processor (what handles commands over the bus) and the floppy drive controller, and they operate in sync. As a vestige of this, although later Commodore drives starting with the 2031 (as well as the 1540, 1541, 1571, etc.) have only a single 6502 CPU, it alternates between IP and FDC modes as if there were still two. While the drive is in IP mode, an interrupt fires about every 10ms to put it in FDC mode and check the job queue the IP side left for it. In FDC mode it sets the interrupt flag, completes the tasks (if any), and then returns to IP mode.

The other notable chips are a handful of ROMs (16K), RAM (4K) and glue ICs, two 6532 RIOT (RAM-IO-Timer) combo chips used for GCR decoding, a 6522 VIA, and a little daughterboard between the 6502 at the top and the 6522. This replaces a 6530 RRIOT (ROM-RAM-IO-Timer), which has mask ROM and therefore is only interchangeable with exactly the same chip version. The daughterboard facilitates upgrades: the 901885-04 chip is likely this unit's original 6530-047, but its ROM is overridden by the 74LS14 and uses the contents of the EPROM (labeled 251257-02A, or Commodore DOS 2.7).

The problem was obvious: a floppy disk label had gotten wedged into the drive spindle.

I took the screws out of the logic board so it could give me a little more room to work with a pair of tweezers carefully pulling out the label. Initially I had to press down the spindle a bit to get the label free from the top and then pull the rest of it from the bottom. The label, fortunately, was not sticky (probably why it got stuck in the unit — it wasn't adhering to its floppy disk anymore), but it was brittle and required a lot of careful teasing. I got a couple chunks out of it unsuccessfully before being able to grab the main body.

The label came out in a couple fragments but looks basically intact, so I don't think there are any pieces left. Ironically it was apparently an ordinary double density disk which wouldn't have been readable in this unit anyway. The drive powers on fine and I can insert and remove disks, so now it's time to actually find it some media.