9

I obtained a quantity of these IDE 44 pin SSDs for use in my retro computers. At first I simply used an adapter to connect to the 40 pin IDE port, but it caused all sorts of weird problems e.g. the floppy disk change line didn't work and the POST info screen shows serial ports that don't exist.

I focused on determining why it's detecting 4 serial ports that don't exist since that seemed like the easiest. I built an IDE breakout and found disconnecting neither /CS0 nor /CS1 made the above issue go away, only disconnecting /IOR stopped the problems. I then further found that each time /IOR was asserted, it pulled down at least one data line even if neither IDE chip select was asserted. Looking at this on a scope, it's clear one or more data lines are pulled down partway, clearly not normal.

This SSD works great on a couple computers I have that have proper IDE interfaces, i.e. ones with bus transceivers. The two that are having issues have a stripped down multi I/O card where the data lines go straight to the ISA bus. Nonetheless, a regular HDD works perfectly with this card. If I connect this SSD up to power, pull both chip selects high and put pull-ups on the data lines, some data lines pull down hard for 14 or 15ms when I manually assert /IOR. I thought a drive's IDE interface is supposed to stay in high-Z unless one of the chip selects are asserted? Anyone know what's going on here?

6
  • Are they Compact Flash cards, or something else?
    – RonJohn
    Apr 16, 2023 at 3:30
  • 3
    The adapters are modern and either faulty by design or from a faulty batch of manufacture. They might be designed or tested with the assumption that they are not used directly on an ISA bus. Similar to how CF cards cause the floppy change to not work.
    – Justme
    Apr 16, 2023 at 8:07
  • Is the ssd or the adapter cabled directly to the power supply, or just sourcing power from the bus?
    – Sneftel
    Apr 16, 2023 at 8:35
  • They are not CF cards, they are disk-on-module boards with a 44 pin IDE connector. And the adapter is getting power straight from a 4 pin Molex, not the ISA bus.
    – eesz34
    Apr 16, 2023 at 10:54
  • Well please add make and model of what you bought. Maybe the answer why it does not work is in the manual. Also to which types of retro computers you are attaching them, like which era and do they support DMA or UDMA for IDE?
    – Justme
    Apr 16, 2023 at 11:53

3 Answers 3

14

If you are talking about a 44-pin 2,5" IDE to 40-pin 3,5" IDE adapter, that adapter is purely passive and can't be buggy. You are perfectly correct to expect that an conforming IDE device does not respond to /IOR or /IOW unless it is selected. There is a third way to select an IDE device other than /CS0 or /CS1 though, which is /DACK: So it looks like either your SSD is non-conforming, or /DACK is open circuit and read as low. In that case, "just" install a pull-up on /DACK.

5
  • 7
    Shoot, I already had a pull-up on pin 29 to +5. The datasheet for the SSD does indicate that pin is DMACK-. Looks like I'm back to trying to get my transceiver circuit working.
    – eesz34
    Apr 16, 2023 at 11:55
  • 1
    Passive adapters can be buggy. I've seen a lot of CF card adapters pretending to be compatible with ATA (at least working ok with socket 7 boards onboard controllers), but instead of listening for CSEL (pin 28) they just ground it directly on its PATA connector. Needless to say grounding the pin which carries ALE signal on older systems breaks everything. I had to cut the connection on that PATA pin to fix the behavior.
    – Vlad
    Apr 18, 2023 at 6:13
  • But grounding ALE would cause the system to not POST, I think? I wish the solution to this was to ground or pull-up some signal but it seems to be beyond that.
    – eesz34
    Apr 18, 2023 at 13:12
  • 1
    Grounding ALE can cause very strange behaviour, for example it can cause 16-bit VGA cards to misbehave in an otherwise working system. Checking for ALE trouble is a good idea if a card/adapter combination disturbs old computers, but you already identified that your card/adaper combination pulls some data lines down, which will cause issues even with ALE disconnected. Apr 18, 2023 at 20:40
  • Grounding ALE does not kill the system immediately. The length of the IDE cable allows some nanoseconds for signal and its reflection to propagate back and forth, it might be just enough for some cards to work, but not for the other.
    – Vlad
    Apr 19, 2023 at 10:32
7

The interfaces between really old IDE host interface and much more modern ATA can be difficult in terms of compatibility.

The 286 may have an IDE interface from the era before even ATA-1 standard existed. IDE drives appeared in 1986 and ATA-1 standard is from 1994.

The Flash Module implements UDMA4 mode, which was first introduced in ATA-5 standard in 2000.

However, the Flash Module also implements PIO mode 6, which is not an ATA standard, but introduced in CompactFlash 2.0 standard released in 2003.

So it is quite safe to say that the modules are actually CompactFlash modules in ATA/IDE mode with ATA/IDE connector directly on them.

And the early IDE interface really was different from ATA. IDE had no DMA pins. They were supposed to be disconnected, but there is no guarantee how your PCs have connects the unused pins, but they may be grounded or floating.

Also the ALE pin of IDE was reused as SPSYNC/CSEL pin in ATA, but that is unused on the Flash Module.

So safe to say, they should be just as problematic as CF cards with IDE adapter.

Also different IDE and ATA cables connect the pins differently (CSEL for example) the used cable may matter. Obviously the used 44 to 40 pin adapter as well.

5

I thought a drive's IDE interface is supposed to stay in high-Z unless one of the chip selects are asserted? Anyone know what's going on here?

This is correct, even the latest parallel ATA standard T13/1698D ATA8 still requires that. The specified maximum currents when chip select is not active is 200µA:

Screenshot from ATA standard DC characteristics

And the addressing:

Screenshot from ATA register addressing table

That said, I'm not surprised that there exists drives with such a mistake in implementation. IDE interfaces have been using separate drivers for so long that it's unlikely any manufacturer tests for compatibility for such old hardware.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .