The 386s-16 and 386-33n have those same 7 switches and they are for the onboard WD video, not for the IDE controller.

Horun wrote on 2023-03-19, 16:54:

The 386s-16 and 386-33n have those same 7 switches and they are for the onboard WD video, not for the IDE controller.

Well poop. There goes that idea 🙁

Check your CF-to-IDE adapter, see if it has pin 28 connected to ground on the IDE connector.
On old systems that pin is the ALE signal directly from ISA bus, grounding that signal on the IDE cable breaks ISA logic in weird ways that is hard to explain.

vstrakh wrote on 2023-03-19, 18:39:

Check your CF-to-IDE adapter, see if it has pin 28 connected to ground on the IDE connector.
On old systems that pin is the ALE signal directly from ISA bus, grounding that signal on the IDE cable breaks ISA logic in weird ways that is hard to explain.

Thank you for pointing me in this direction. I hope this is what is going on with my system. Below is a quote from a website where this guy is building a multi I/O card https://www.alexandrugroza.ro/microelectronic … face/index.html. The key part is bolded and underlined but the reset is very interesting as well.

One thing that was corrected in VER. 1.4 REV. D is the fact that now the SPSYNC:CSEL signal of each individual IDE interface can […]
Show full quote

One thing that was corrected in VER. 1.4 REV. D is the fact that now the SPSYNC:CSEL signal of each individual IDE interface can be switched either to the GND or the ALE signal. When I drew the original schematics, I implemented the IDE interfaces as described by the earliest IDE standards. Pin 28 was originally called Drive Address Latch Enable (DALE signal) and was designed to connect to the ALE signal of the ISA bus. It was used to indicate a valid address from the host system. But then, in 1994 the ANSI X3.221-1994 standard appeared. It described the AT Attachment Interface for Disk Drives (ATA-1) and pin 28 was re-purposed as SPSYNC:CSEL signal. While the SPSYNC signal is of no interest to me whatsoever, the CSEL signal actually is. Let's find out why.

After finishing the assembly of PCB VER. 1.4 REV. C, I tested it with a bunch of mechanical hard disk drives ranging from 40 Mb stepper motor models to 4.3 Gb voice coil models. They all worked perfectly in all sort of master-slave combinations. Then I also tested four identical 512 Mb Compact Flash cards in all sort of combinations. And they all worked. The issue was when I tried to test Transcend Industrial 2 Gb and 4 Gb Compact Flash cards. These refused detection by all means. They would register and work only in a certain combination: Transcend Industrial card as master and 512 Mb STI Flash card as slave. All other combinations failed for some reason.

So I started reading the ATA-1 papers and the ALE signal is not required anymore for modern mechanical drives or CF cards. That means pin 28 of the physical IDE interfaces should be tied to something else. The paper describes the connection of the CSEL signal as follows.

if CSEL is tied to GND signal then the drive is configured as master
if CSEL is not connected then the drive is configured as slave

I immediately modified the CF adapters that I was using so that they reflect this hard-wired connection of pin 28 to either the GND signal or left unconnected and surprise: the Transcend Industrial cards came to life. Also, while reading the datasheet for the Transcend CF220I series of compact flash cards, I learned that card's CSEL signal is internally pulled up to allow configuration of the device as either master or slave in true IDE mode. This fulfills the X3.221 standard requirements of pulling this line up with a 10 kΩ resistor. Thus nothing to do on the PCB level in this case.

You're saying to not ground it but he is. Perhaps you were mixed up? Perhaps I am interpreting this wrong - after all I am only slightly above a layperson here.

I went ahead and connected pin 28 to pin 40 (which I read was ground) and I tried to install FreeDOS again but once again it failed with an error that the drive was not ready.

I did confirm my soldering job worked by checking continuity.

Well this is dismaying 🙁

It needs to be grounded on the card side, but you cant ground ISA ALE signal, that will break all kinds of things
TLDR: use multimeter to check if pin 28 on your CF converter is grounded, if it is isolate it from IDE controller

Dare I say not all systems can use a IDE-CF adapter and many are are not designed for pre-Pentium boards ? I know they are popular but they also have many issues on many older boards. A quick forum search proves that 😁

rasz_pl wrote on 2023-03-20, 03:31:

It needs to be grounded on the card side, but you cant ground ISA ALE signal, that will break all kinds of things
TLDR: use multimeter to check if pin 28 on your CF converter is grounded, if it is isolate it from IDE controller

I have several of the same CF adapter so on another one I did not modify I used the multimeter and touched one probe on pin 28 and then, one at a time, put the other probe on all the other pins on the back of the PCB. No beeps. That means it is not grounded, right? There's no connection between pin 28 and any other spot?

It could still be pulled low with a resistor, and there are CF cards that ground that pin internally. Imo safest bet is to isolate pin 28 between IDE controller and CF converter.

rasz_pl wrote on 2023-03-20, 05:23:

It could still be pulled low with a resistor, and there are CF cards that ground that pin internally. Imo safest bet is to isolate pin 28 between IDE controller and CF converter.

Please let me know if I'm understanding this correctly: So pin 28 was used in IDE for ALE but then in 1994 a new ANSI standard repurposed pin 28 to become cable select. Grounded meant master, open meant slave. Because this 386 system and the BIOS predates the 1994 ANSI standard, the onboard IDE host adapter is still using pin 28 for ALE though. According to a data sheet for a CF card I have, pin 39 on the CF card is for CSEL. Does that mean that the CF adapter will route pin 39 on the CF card to pin 28 on the IDE connector? The CF card was most definitely made after 1994 so If the CF card has pin 39 grounded internally then that is going to mess with the ALE signal right? If it DOES have pin 39 grounded inside the CF card, is there any way to correct this on the CF adapter? Is that what you meant by isolating pin 28 between the IDE controller and CF converter? How do you isolate it? Sorry, I don't know the terminology.

geiger9 wrote on 2023-03-20, 06:35:

Does that mean that the CF adapter will route pin 39 on the CF card to pin 28 on the IDE connector?

It should, and then you need a proper Cable Select cable, with gap in line 28 between two device connectors.
But typically it does not. The CF card does not use the Cable Select line in PC cards mode, but it should not be left floating. It must be pulled up or low, and the cheapest way is to ground it with some trace on the CF adapter PCB.

geiger9 wrote on 2023-03-20, 06:35:

If the CF card has pin 39 grounded internally then that is going to mess with the ALE signal right?

Problem is not with CF card pin 39. That pin is grounded on CF connector almost always, with few exceptions that actually has jumpers for CSEL signal.
The real problem is that many adapters also ground pin 28 on the IDE cable connector. They should not do it.
CF will work already ok anyway because its pin 39 is grounded. And CF adapter do not route IDE pin 28 to CF pin 39 (with some exception like here: Re: Troubleshooting Compact Flash to IDE on 486 System with American Megatrends BIOS , where pin 39 can be switched between ground or IDE pin 28).
The issue manifests when such CF adapter grounding both IDE and CF endpoints is plugged to IDE controller that is not Cable Select capable, because such IDE controller exposes ISA bus ALE signal, which then will be grounded by CF adapter, not even reaching CF card.
So the solution is to isolate IDE cable's pin 28 from the CF adapter. Either with jumper on the CF adapter if it's there, or cutting the traces that connect IDE pin 28 to ground on CF adapter PCB. CF will still work as before, because typically it has CF pin 39 grounded with traces directly under CF connector.

The 386s/20N does work with 2 of the most common adapters out there, so there is nothing special with this particular motherboard.
Suggestions to try:
-Do not use another machine to detect CHS values, but something like the IDEINFO utility on the same machine you intend to use the card on (if the values match that's not the case of course).
-Even if it sounds weird install some DDO software as a test, to make sure it's not the detection/translation of the particular card you are using
-Remove all cards while trying the above
-At this point I'd question the CF card too and try another one (different brand/model). I don't mean it's broken, just quirky.

vstrakh wrote on 2023-03-20, 07:09:

So the solution is to isolate IDE cable's pin 28 from the CF adapter. Either with jumper on the CF adapter if it's there, or cutting the traces that connect IDE pin 28 to ground on CF adapter PCB. CF will still work as before, because typically it has CF pin 39 grounded with traces directly under CF connector.

Okay so I'm very confident that pin 28 is isolated on the CF adapter. I checked for continuity between pin 28 and ANY other point on that PCB and there was none. There are no resistors on this PCB. This is the exact card I have https://a.co/d/3eltNk5

There's something about the IDE controller in this machine that doesn't like what the CF adapter is doing. I have ordered two different CF adapters than the usual looking chinese copies so we'll see if they work this week.

konc wrote on 2023-03-20, 08:26:

The 386s/20N does work with 2 of the most common adapters out there, so there is nothing special with this particular motherboar […]
Show full quote

The 386s/20N does work with 2 of the most common adapters out there, so there is nothing special with this particular motherboard.
Suggestions to try:
-Do not use another machine to detect CHS values, but something like the IDEINFO utility on the same machine you intend to use the card on (if the values match that's not the case of course).
-Even if it sounds weird install some DDO software as a test, to make sure it's not the detection/translation of the particular card you are using
-Remove all cards while trying the above
-At this point I'd question the CF card too and try another one (different brand/model). I don't mean it's broken, just quirky.

Thanks for letting me know about this. So strange that I am having such a hard time then. I posted a link just above with the CF adapter I am using.

I used whatide.com in the target machine to get the CHS values for my CF cards yesterday. They matched what the autodetect picked up on the other machine.

All expansion cards were removed.

It would be strange if the CF card was the issue because this same CF card performs fine in the early Pentium machine. I have also tried several of these CF cards in the XT-CF lite and they work fine there too.

I have yet to try the DDO software. I wonder how that would work though? I suppose I AM able to access a DOS prompt and copy stuff TO the CF card.

Here are some additional things I noticed today:

I can copy files to the CF card from a boot disk at a DOS prompt.

I can delete a file from the CF card here as well.

If I copy a file from the CF card to the floppy drive it gives me an error after about three minutes "not ready reading drive C"

If I try to format c: it will pause for about three minutes and then give me an error "not ready"

If I format the CF card in another machine c: /s in another machine and put it into this 386, it will pause during POST, for three minutes just like when I tried copying from the cf card to the floppy, then tell me "hard disk controller failure"

Update from the future:

I could not get the 386s/20n to work with a compact flash adapter even after using a BIOS from MR BIOS. I guess this particular chipset was not supported by MR BIOS.

I was able to get a Samsung 386s/16 to work with a compact flash adapter by using another BIOS by AMI. That BIOS is located https://theretroweb.com/motherboards/s/intel- … 5-v-2#downloads if anyone has the same issue as me in the future. I did try this BIOS with the 386s/20n because it does not work. Perhaps because the 386s/20n has a cache chip.