Hi all,

I've purchased 8 of these internal CF to IDE adapters from eBay

Right off the bat, I'm going to report their shoddy build quality - One didn't work outright, another bent pins after only the 2nd or 3rd insertion, causing a short (erasing my data on the CF card as well), and one was so poorly soldered onto the board that several pins from the leads to the PCB (CF card side) broke off when I was inserting it into an empty bay backing. Really dodgy stuff. On one, I had to re-solder the LED as well. Luckily, I got a refund for the three with little to no questions asked, without the need to return them to mainland China: Just goes to show how shitty the product is. There are additional issues with these adapters on computers with more than one disk I/O, read on.

I also got two of these (mine are x266) to serve as media:

I also have a USB to CF adapter, which sees both disks under Ubuntu's "disks" - I can format them, partition them, etc. I can read the contents of the 540 Mb partition after DOS 5.00 created it and put some files to it, as well as the full 2gb partition that can be seen by DOS 6.22, Windows 95, etc.

I have these adapters up on four computers, with all BIOS parameters being Cyl 3949, Hd 16, Sectors 63 - 1944 Mb with varying degrees of success. In every case, the role of the CF card is a secondary, rather temporary drive that can be swapped between the computers semi-easily for file swapping, supplanting my ZIP100 system. I will be doing a separate project using a permanent CF (or an SCSI2SD, if I can get my wife to order me one for Christmas!) on my 486DLC after that one is thoroughly refurbished.

Here they are:

1989 - 386SX - CF integration level of difficulty/time spent 4/10.
CF position - Pri Slave
CMOS info - ENSX-1131-101589-K8 (NEAT SX) AMI

The first computer is a 386SX 16 with DOS 5.00 - BIOS sees the drive (manual input), formatting it takes seconds, but when it asks you for a label (hit enter for none) it does nothing for a long time. Eventually it formats it as a 540Mb (no surprise there). Copying files starts off promising, then hangs on a file for a while. Currently I did a CHKDSK on it, and it's been hung on Volume Serial Number is xxxxx - It finally spat out Proccesing cannot continue = File Allocation Table bad. After creating a nice little slotted location for the CF card behind an RS232 25 pin blank and soldering on a small floppy power connector (the power supply came with none at all, just MOLEX) I was able to access the 2gb partition simply by booting with a DOS 6 or higher. I eventually reformatted the drive and put DR-DOS 7.02 on it. Status - Working well.

1994 - 486 DX2/66 CF integration level of difficulty/time spent 1/10.
CF position - Pri Slave
CMOS info - 40-PD07-001169-00101111-072594-SIS471G-H

The 486DX2/66's BIOS autodetects the card and in DOS 6.22 there are no issues at all - runs files off it A-OK. Copies them over at break neck speeds - FDISK nukes partitions and creates new ones no problem. Standard card mounting in the rear - blank removed, this one added. Status - Working fine.

1995 (later after upgrade, 1996) FX and VX chipset, respectively. CF integration level of difficulty/time spent 5/10 and 8/10 post-upgrade.
CF position - Pri Slave
CMOS info - i430VX-10031996C-00

The P166 I have behaves strangely indeed. First, whatever IDE channel it is hooked up to (Pri/Sec) completely disables the other channel. I've tried putting known good IDE equipment on the other channels, both CF cards, several CF card readers, and no matter what combination I use, if the CF reader is present with a card installed on either channel, the other one is disabled. I got around that by putting my primary HDD and the CF on primary IDE, and the ZIP100 and CDROM on secondary - I don't care about their loss while momentarily copying files. Works fine when booted into DOS - Windows does not see the CF drive.

Eventually, I upgraded the board to one with a VX chipset (same specs otherwise) after fixing this:

Surprise of surprises, a DALLAS RTC with a PLASTIC CAP? Never seen that before. I still managed to break a pin off the chip, necessitating drilling and some fine soldering skills. I put the CF card adapter in place, and it seemingly works after booting into Windows 98SE - You can copy files straight from Windows with ease. Interestingly, it kills off the CDROM (Zip drive does show, I didn't test it for operation). Booting from DOS with an XCOPY/DOSSHELL enabled diskette seems to work better and faster.
Status - Working well-ish.

And finally, the Dell Pentium Pro 200 from 1995. CF integration level of difficulty/time spent 12/10.
CF position - Depends.
CMOS info - Dell XPS BIOS v A03

Oh, this one was a goddamn pain. First off, use of the CF adapter on any channel generally disabled the other channel EXCEPT when configured like this:

Believe me, I've tried all combinations, using other hard disks, putting the drives on the secondary channel, no, this is the ONLY way it could work. HOWEVER, Dell needs the Pri Master to be the boot disk! That's right, it will NOT boot from Pri slave (or any other position). What a pain in the ass this is: To use the CF card, open the computer, set the jumper to Master (or leave it in Master and plug the IDE cable in - it will interfere in Slave - I do not recommend that as these CF adapters are shite and break easily) and the hard disk to Slave. Then boot from floppy. Every. Time.

Fortunately, I did this:

However I later found that this SPDT switch doesn't force the hard disk into Slave - Now, luckily the drive will POST as Master without any jumper in it (but if the CF jumper is also set to Slave, the hard disk doesn't like it) so really all this has to do is set the CF reader to Master and the hard disk to Slave. I probably COULD have used the same switch contacts to do both (master for CF, slave for HDD) but I am unsure of how signalling works - Even if I put diodes inline to isolate the circuits (as I am quite sure there's no oscillation or signalling, just a simple ground or +5v) I decided against risking grounding out (or sending +5V, or whatever the high bit voltage is on the IDE controller) and simply revamped my idea with a DPDT switch I had lying around, with a HDD LED header pulled from spares connected to the WDC2500's SLAVE jumper location. The result:

Drive side:

Yes, now all I have to do is put the CF card in, flip the switch, and boot up. I still have to boot from floppy (I've since made a NWDOS boot diskette with XCOPY, self loading mouse driver, and DOSSHELL from MSDOS 5.0) but it is suitable to its engineered task.
Status - Gimped, but working.

That is my CF story for now -in the future I will do two more systems - My 486DLC, and my 286, once I get the Keyboard Error sorted.

I hope someone can use some of this information for their own projects or for a good chuckle.