Atari 1050 Firmware Revisions and Drive Mechanisms

Year
1983
Type
Storage
Category
Technical reference — 1050 ROM revisions E to L, Tandon and WST mechanisms, WD2793/2797
Manufacturer
Atari, Inc. / Atari Corp. (built by Atari-PCI, Singapore, and Atari Hong Kong)
An Atari 1050 disk drive with its cover removed, showing the drive mechanism, stepper motor and power-supply capacitors

Overview

From the outside every Atari 1050 looks the same, but three production runs in two countries, two different drive mechanisms, two floppy-controller chips and at least five firmware versions lie behind the badge. The differences rarely matter for everyday use — the users who catalogued the ROM versions found no practical differences between them — but they matter a great deal when repairing a drive, swapping ROMs, fitting an upgrade or chasing a timing problem. (For the drive in general see the main 1050 entry.)

Where and how it was built

Production run Maker Mechanism Firmware
May 1983 – Dec 1984 (most common) Atari-PCI Enterprises, Singapore Tandon Rev. J (1983), Rev. K (1984)
Nov 1984 – Feb 1985 (fairly rare) Atari, Hong Kong World Storage Technology (WST) An alternate Rev. K
Oct 1985 – Jan 1986 (fairly rare) Atari-PCI, Singapore Tandon Rev. L

Inside, every 1050 has a 6507 processor at 1 MHz, a 4 KiB custom 2732 ROM (part FC100541), a 6810 RAM, a 6532 RIOT and a Western Digital floppy controller — normally the WD2793. Late units from autumn 1985 use the WD2797 instead, and these work only with ROM Rev. L. The drive's mechanism is an 80-track type stepped in half-tracks to read 40-track disks.

Firmware revisions

Rev. Notes
E Earliest released firmware. Reading or writing "sector 0" makes it try to seek to track 56; format fills sectors with $E5.
H Many reliability fixes: a timeout while waiting for SIO data (so the drive can no longer hang), sector 0 rejected, formats fill with $00, cached FDC status.
J Fine-tunes the serial receive loop from 51 to 52 cycles per bit, improving the timing margin with PAL computers (whose bytes arrive slightly more slowly).
K One more cycle (53) for PAL margin; fixes the ROM self-checksum; returns an error for "no disk" without first spinning the motor (H and J spin up first).
L Changes the controller command bytes so the drive works with either the WD2793 or the WD2797.

All revisions automatically detect whether a disk is single or enhanced density: when the latch is closed the drive seeks to track 1 and issues a Read Address command to see whether the disk is FM or MFM. The drive senses disk insertion through the latch, which is wired to the controller's ready input.

Hidden diagnostics

Atari's service software used undocumented test commands. Command $23 sends a 128-byte buffer whose first byte selects a test — RPM measurement, motor start time, a head step/settle test across all 40 tracks, half-track steps, a lengthy double-density burn-in, or a seek to a given track — and command $24 runs the test and returns results. As a side-effect $24 also returns a ROM signature in its last three bytes: the revision letter in ATASCII followed by the model number 1050 in binary. Revs. H and J have an addressing bug that shifts the signature by one byte (returning $AA $48 $04 or $AA $4A $04); K and L return the intended $4B/$4C, $04, $1A. Few third-party firmware replacements implement these commands.

Quirks worth knowing

  • Tone: the 1050 sends data at about 19,608 baud using a 51-cycle bit loop, so the "beeps" heard through the TV while loading are noticeably lower than with an 810.
  • Seeks: it steps 10 ms per half-track, settles for 20 ms, and always ends a seek with an outward step, so inward seeks are slightly slower.
  • Write-protect errors make it recalibrate to track 0 and seek back, even though that cannot help.
  • Enhanced-density layout: 13:1 interleave with odd and even sectors separated, and about a two-sector skew between tracks.
  • Deleted sectors: because of its 279X controller it reports deleted-record status differently from the 810's 1771 ($DF instead of $9F), something copy-protection checks can trip over.

Practical advice

  • Swapping ROMs: any revision runs on a WD2793 drive, but a WD2797 drive must keep Rev. L.
  • PAL users: Rev. H's serial receive timing is marginal with PAL computers (by the stop bit the error approaches 40% of a bit); Revs. J and K widen the margin.
  • Upgrades: many 1050 enhancements (Happy, US Doubler, Speedy and others) bring their own firmware, so the original revision only matters when you put the stock ROM back.
  • Emulation: Altirra's firmware-level 1050 emulation can run the original ROM revisions, reproducing these behaviours exactly.

Credits