I have been chasing a problem on one of Greg's ST536 test setups, and it has turned up something I would like more information on from anyone who has used CF or SD to IDE adapters.
I used AI (Claude) to do the analysis for this, going through the TOS 2.06 and EmuTOS sources, the EmuTOS commit history and some old forum threads, and each point below has been checked back against those sources.
With one particular CF to IDE adapter connected, TOS 2.06 would boot to the desktop, but as soon as the floppy was used the machine crashed and rebooted. With that adapter unplugged, TOS 2.06 booted and ran fine from both the floppy and the UltraSatan. With the same adapter connected under EmuTOS, everything worked. Greg then tried two SD to IDE adapters, both FC1307A based, and all was perfect on two different ST536 boards. The same CF adapter does work fine on Greg's older ST536 in the same machines, so it is something about how the newer boards or firmware react to it, which we are still narrowing down.
The interesting part is how TOS and EmuTOS handle the IDE drive.
TOS 2.06 enables the drive's interrupt before every IDE command, then waits for it by polling bit 5 of the MFP GPIP register, the same bit the floppy and ACSI use for "command done". Once it sees the interrupt it reads the IDE status register, which clears it. On the newer ST536 firmware the IDE interrupt is passed through to GPIP bit 5 by the CPLD, so TOS 2.06 can autoboot from IDE.
EmuTOS does it the other way round. Its IDE driver sets the "no interrupts" bit (nIEN) in the drive's control register and polls the drive's own status register instead, so the interrupt line is never used. EmuTOS has done this since its IDE code was reworked in 2013, but I cannot find a reason written down anywhere, either in the source comments or in the commit history.
The Falcon has something similar in hardware. In the "Falcon bizarre floppy problem" thread here, SteveBagley pointed out that the IDE and floppy share the same interrupt to the MFP through a NOT gate and an AND gate, so if something holds the IDE interrupt line high, the floppy and SCSI interrupts never reach the MFP. That thread was about a museum Falcon with an FC1307 based SD adapter and floppy problems, and it was never resolved. There is also a thread on AtariAge where a Falcon had floppy errors whenever a card was in an SD or CF to IDE adapter, and it went away after changing to HDDRIVER:
https://forums.atariage.com/topic/34200 ... oppy-issue
The common chips on the cheap SD adapters are the FC1306 and its successor, the FC1307 or FC1307A. The FC1307A is often hidden under a "QC Passed" sticker, some boards have it relabelled as STCA1, and it reports itself as "SINTECHI HighSpeed SD to CF Adapter". Greg's FC1307A adapters work fine, but the museum Falcon above had problems with an FC1307 one, so the chip alone does not seem to be the answer. Years ago I had two SD adapters in my Falcon where one worked and one did not, and I think they used these two chips, but I need to open the Falcon to check. A plain CF to IDE adapter has no chip of its own, so there the CF card itself is the IDE device.
My theory at the moment, which is not proven, is that the CF card in the problem adapter holds the IDE interrupt line active when it should not. Under TOS 2.06 with the newer firmware, that would make the floppy and ACSI think every command has finished straight away, while EmuTOS turns the drive's interrupt off, so it would never show up there. The next steps are to measure the interrupt line, pin 31 on the IDE connector, and to try the older firmware that does not pass the interrupt through.
So a few questions for anyone who knows:
Has anyone else had floppy or ACSI trouble that came and went with a CF or SD to IDE adapter, on an ST, Falcon or anything else? If so, which adapter, which card or chip, and which driver?
Does anyone know why EmuTOS chose to poll the IDE drive rather than use its interrupt? Was it to avoid this sort of problem, or just simpler?
Does anyone know whether HDDRIVER uses the IDE interrupt or polls the drive?
Does anyone have a full FC1306 or FC1307A datasheet that covers the interrupt pin and the timings? The FC1306 sheet I have shows the interrupt on a pin that is shared with the master/slave select and a card ready output, but there are no timings in it.
SD to IDE adapters, IDE interrupt and floppy problems
Re: SD to IDE adapters, IDE interrupt and floppy problems
As this was done long before I joined the EmuTOS project, I, too, can only speculate. Considering that EmuTOS supports IDE also on Amiga and ColdFire targets, I would assume polling the IDE status register removes platform dependencies. The IDE registers exist everywhere (when there is an IDE device), the MFP does not.exxos wrote: Sun Nov 15, 2026 12:56 pm Does anyone know why EmuTOS chose to poll the IDE drive rather than use its interrupt? Was it to avoid this sort of problem, or just simpler?
PS: Anecdotally, I know of more than one case where the IDE interrupt line caused problems. E.g., in one case a (presumably bad) CF card crashed sometimes and kept the interrupt asserted, taking out not only itself but also the floppy access.
Re: SD to IDE adapters, IDE interrupt and floppy problems
That makes sense, the IDE registers are there on every platform but the MFP is not, so polling the status register is the one way that works everywhere.
Your CF card example is interesting, as it is pretty much exactly what I suspect is happening here. If a card holds the interrupt line asserted, anything that shares that interrupt with the floppy, like the Falcon's AND gate arrangement, would see every floppy command as finished straight away, while EmuTOS would never notice as it does not use the line.
Your CF card example is interesting, as it is pretty much exactly what I suspect is happening here. If a card holds the interrupt line asserted, anything that shares that interrupt with the floppy, like the Falcon's AND gate arrangement, would see every floppy command as finished straight away, while EmuTOS would never notice as it does not use the line.
