ADF Sections for Integrated Fixed Disk and Controller

- the revised 55SX board FRU 27F4667. Components slightly rearranged. Riser cards have more capacitors and / or the 74LS-something that fixes a problem with 8514/A-style cards.

It is for a 2K CMOS Ram extension chip (6116 or something) that adds to the one in the Dallas RTC - needed for some purpose on totally medialess 55SXs (no FDD, no HD, only NIC).
I think it is intended to hold some system status / shutdown infos that are usually written to disk.

Early Riser P/N 27F4625

This riser is bare of anything to the right of the MCA slots but the battery/speaker connector.

Newer Riser  P/N 27F4625

This riser has an addition of a resistor, a 74F08N, and a bit of wiring on the back.

Late Riser P/N 27F4666

  This bad boy moves the battery/speaker header down and adds a few resistors and capacitors.

After a power cycle and reboot from the reference diskette the silly game repeates.

"I have exactly the same problem with a 55SX.  Only in my case I was trying to install OS/2 on it.  The install progressed to the first reboot and then couldn't boot off the drive.  What's weird is that it
manages to copy all the files to the drive during the installation, so it does see the drive, it just won't boot from it.

Aron suggested the following, but I haven't had a chance to try it yet.  I have to look for a DOS floppy to boot from.  If you try this and it works please post it here.

"I did some searching on the 55sx problem after replacing the Dallas RTC chip, and found one suggestion that says you need to clear the CMOS before the configuration will take.

The method is to boot from a floopy, run DEBUG, and type

  o 70 13
  o 71 00
  q

Peter replies:
   I have seen this twice in the past that a Mod. 55SX "reacts funny" on memory upgrade. In both cases it was a *board* failure: #1 had a micro-crack in one copper-line at the underside of the board #2 had the pins of the riser-card connector too long and these touch the grounded chassis after re-insertion of the riser card. Both errors showed up only with 2 memory modules installed. Pretty strange.

I would suggest to try out a "bread board" installation. Remove the board from the chassis, remove the power-supply also and place it on a non-conductive surface (Warning: underside pins may scratch furniture). Install the memory modules, the riser board and lay HD and FDD in a place where it can be connected but do not touch board or power-supply. When the system runs fine without the chassis inspect the underside of the planar if there are cracks or too long component pins, especially on the riser-card connector and around all places, where the board fixing screws go.

And: there are some combinations of planars / riser-cards known as non working. If the planar FRU is 27F4667 the appropriate riser-card must be 27F4666 (P/N 64F0809 on a white sticker, contains a 74F08-chip and some condensors / resistors. The almost totally empty riser-card (P/N 274630 on white sticker / 6 condensors between slots / 2 larger tinned areas at the rear) belongs to an early down-level 55SX-planar and is known to cause problems. The 2nd series risercard .

Peter wrote:
   But .... some early Mod. 55SX suffer on a sort of systemboard bug that changes the LPT timing a bit and they choke on faster transmissions in bi-directional mode - like with using EPP devices (CD-ROM, Iomega Zip drive) or Interlink. Those EPP devices run but rather slow and if you connect that 55SX to another computer with LapLink you will get a hell of time-out errors when copying larger files (until the system hangs).

   It is pretty hard to tell which boards are affected. I found out during testings, that mainly those old boards tend to fail that have the "dead cockroach" chip at the mid/front. It is sitting between the two bigger square chips and is accompanied by a silver capped chip glued to the board. Both are wired to the system with patch-wires. These machines have the old Level-1 riser card 27F4630 (printed on a decal), which is almost bare apart from six small condensors between the MCA connectors. The Level-1 riser board has no chips and no "component printing" in white. The board P/N is 33F5064 (sticker over 33F5060 P/N - which was the buggy, unfixed original board P/N).

   There was a revised systemboard without the patch-on chip, which came with the Level-2 riser (several resistors, some el-co's and the small condensors - came with an ECA from IBM due to problems running 8514/A style cards), which has been replaced by the Level-3 riser 64F0809 with a 74F08 chip (ZM1). The early 55SX planar cannot be used with the Level-3 risers ... the later systemboard should not be used with the bare Level-1 riser. 
   The later sysboard is FRU P/N 27F4667, P/N 85F0419 (55SX) or 57F3003 (55LS / LEO/LTO - which have the additional 2K CMOS NV-RAM), the riser is FRU P/N 27F4666, P/N 64F0809 (55SX) or 64F3732 (55LS). Sadly the FRU alone does not tell, which P/N is behind ... IBM always delivered the latest P/N at a FRU request.

I installed about 350 - 400 Mod. 55SX in the early '90s for the german Automobile Club ..... so I guess I knew them quite well ... :-)

The harddisk is in fact a camouflaged MCA-adapter with a harddisk mechanism atop.
The entire MCA-bus is fed over the cable to the drive. This drive is the fourth MCA slot, which shows up in the configuration. In case the contacts between board and riser, riser and cable, cable and harddisk are oxydated -or- in case the cable is slightly damaged there will be a communication problem between board and harddisk - which in the end leads to 1048x HD-error or "hang condition" during Power On Self Test (boot-up).

The maximum system board memory capacity is 8MB (4MB memory module kits installed in memory-module connector 1 and 2).  A memory-module kit must be installed in memory
module connector 1.  The system can support a total of 16MB of memory.

My experience is that the 55sx is fine with one SIMM, but it must go in the slot closer to the power supply. The 55sx takes PS/2 SIMMs, 1mb, 2mb, or 4mb, either 85ns or 100ns.  I'm pretty sure it can also take 80ns SIMMs (definitely can in the 4mb size), but I'm not certain.
------- 
 -- Aron Eisenpress, 

486 CPU Upgrade: IBM PS/2 Model 56, 57 
>Any reason it won't work on the 8555?

Older 55SX had a slightly buggy mainboard / CPU where the "CPU disable" pin did not work as supposed. You could use "clip-over" upgrades only on the 55SX anyway since it has no "upgrade" socket of any sort (the 387SX socket does not feed all required lines through ...).

If your 55SX planar has the "dead cockroach" fix (silver capped chip with patchwires around) suspect it as one of the earlier ones that might or might not work. The early models with the "bare" riser card without any TTL-logic chip and without lots of capacitors had multiple troubles - not only with the CPU.

>But could you ground out a pin directly on the cpu?

Ahem ... as far as I recall IBM reported having got a quantity of 386SX where the disable pin does not work *at all* - and therefore any clip-over upgrades won't work.

The "dealer confidential" paper I have in mind mentioned the 55SX / 65SX only - none of the other SX-machines (like L40, N33 etc.) where the case forbids to use any updates anyways. The L40 however had been on the "CPU upgradeable" list at Hantz & Partner (www.upgrade.de) years ago, which was a "send-in upgrade" with soldering on board level. So it seems as if only very early 16MHz 386SX were affected by this general fault.

The corresponding pin is -FLOAT ("FLT#", active low - pin 28), which "floats all Intel 386SX bidirectional and output signals, including HDLA. Asserting FLD# isolates the Intel 386SX from the surrounding circuitry." (Intel Datasheet 24018708.PDF, Page 60) Originally this pin was added to allow in-circuit emulation without the need to unsolder / remove the chip from the board.

Simple test: ground pin 28 and the system may not POST.  If it does the chip is one of those faulty few.

386SX/16 CPUs in general do not have a (working) disable pin, except they bear a 'C STEP' writing (most of the ones I've seen don't...).  All 386SX CPUs with 20 MHz or more use at least the C stepping mask, so that's not a issue for them...

Best regards, Alfred   

Remove Password
To remove the password:
- open the cover (hardest part of the operation...)
- locate the speaker cable from the front to the vertical bus riser card
- unplug the cable at the bus riser card
- turn it around by 180 degrees
- plug it back to the bus riser card
- close the cover.

Password is gone.