MFM & ESDI Adapters

@DDFF.ADF IBM ESDI Fixed Disk Controller
   (ADP CDDFF.ADF not required / rename to @DDFF.ADF)
@DDFF.ADF IBM ESDI Fixed Disk Controller
CDDFF.ADF Init file for @DDFF.ADF
@8EFF.ADF IBM PS/2 SCSI Adapter w/Cache (DEAD)
   (modified, needs no ADP, ROM selectable / German comments)
   Use this for IBM ESDI and SCSI Controllers in the same system.
@DFFD.ADF IBM ST506 Fixed Disk Adapter

ESDI Fixed Disk Drive Adapter/A - Technical Reference

Early MFM Adapter 72X8540
Later MFM Adapter 90X8643
Early ESDI Controller
Later ESDI Controller
   ESDI BIOS and Microcode
     BIOS Extensions ROMs
     Microcode ROM
     Updating ESDI Microcode
   Install Second ESDI Drive
   Using Non-IBM ESDI Drives
   8560/8580 HD Cabling Schematic (Signal and Data)
   MFM/RLL Cable Source
Is the IBM Integrated HD Adapter ESDI?


Early MFM Controller 72X8540

U3 Intel P8051AH
U4 1700824 (metal can)
U5 1700874 (TI CF60025FN)
U7 6301209 (metal can)
U30 ?
U31 6127783
Y1 20AKSS6M xtal

U30 is covered with a grey-black resilient compound that has a heatsink? pushed into it??

The card-edge connectors at the top are not labeled.

Card Identity

From Peter:
   This "Mystery Card" is a MFM controller, IBM P/N 72X8540. It's an early downlevel card, which has been withdrawn with ECA 002, service code 33, available from 87-06-17 / IBM Boca Raton. This adapter has been used in early PS/2 Models 8560-041 with serial No. range 8001342 - 8009651 (US production only).

   The ceramic shield has been obsolete on the new redesigned MFM adapter for Models 60 & 80. These adapters were already "factory reworked" cards - the very first series had the U30 module without the shield and experienced "sudden death" due to some sensibility against electrostatic discharge. Therefore the shield. The P/N for the various cards stayed unchanged.

Source: IBM Engineering Changes Group 819 - PC-Family / PS2 Family Service Information Manual, IBM Doc.No. SR28-0280-2 / 3rd Edition Nov. 1987


MFM Controller 90X8643

J1 Data Cable Connector
J2 Data Cable Connector
J3 Control Cable Connector
U1 6127893
U2 T5627
U3 Motorola MC3486P
U4 AMD AM26LS31PC Line Driver
U5 SSM 8736 24D 220/330 ohm resistors
U6 5960904
U7 6127784
U15 Hitatchi? HM6116LFP-3
U18 Intel P8051AH
U27 83X3202
Y1 20.0T7J
Y2 12.0MC TDK

Charles Lasitter asks:
   I've had an inquiry about how the 72X8540 ST-506 / MFM adapter works in Model 8560 computers, and specifically I'm wondering about the usual stuff: Is this some Unique IBM flavor of ST-506, such that "Don't Bother!" is the word of the day when it comes to considering non-IBM MFM drives?

   Any hope of substituting larger drives for this machine, and if so, what's the point at which it will freak out over translation issues and require exquisitely unique device drivers to step in front of the operating system and hide the messiness?

From Peter:
   From the principle the stuff IBM used there is the usual Western Digital stuff... with the major difference of a fixed BIOS-resident lookup table with fixed values and no "User Type". Back in the glorious old days we used to solve this problem with a software called "SpeedStore". You enter the BIOS type of a drive which comes closest to the one you want to install and then override the CMOS settings with the software and a boot-sector resident driver.

   Another significant difference: IBM has castrated the 4-device ST-506 interface down to 2 devices with altering the device addressing a bit. All drives have to be set to "second drive" (DS1 when counting "0"-based from DS0 to DS3). The two possible drives are addressed with the motor-on and drive select lines - and a twisted cable for the first drive, which "corrects" the false addressing logic. The IBM PS/2 BIOS also and consequently supports only two MFM drives (and two ESDI as well... they repeated the mistake there again).

   This part is still missing in the PS/2 Reference PDF section. I *think* I have the MFM controller HITRM or TRM anywhere... but I might be wrong. I PDFed the ESDI and SCSI controllers - which seemed the more important to me.


Early ESDI Controller P/N 90X8063?, PCB P/N 90X6858 [P] [P]

Single-side load (components on one side only). All major parts are the same as on the later version (including the IDs).

Later ESDI Controller P/N 15F6586 or 15F6805, PCB P/N 72X8588 [P] [P]

J1 Control Cable Connector
J2 Data Cable Connector
J3 Data Cable Connector
U4 Hitachi HM6264ALFP-15T 8K SRAM
U6 Adaptec AIC-010FL
U12 72X7408
U14 Intel D27128A 128K ROM BIOS Low
U16 Intel D27128A 128K ROM Microcode
U18 Adaptec AIC-300FL
U22 72X8305
U25 Intel D27128A 128K ROM BIOS High
U33 Intel N8031AH
U37,38 Hitachi HM6264ALFP-10T 8K SRAM
Y1 10.00 MHz xtal 68X6852

Y1 is a unique flat, square clear plastic cased crystal.

The later version has a double-side load PCB with a slightly modified layout. All resistors and bypass caps were moved to the solder side of the PCB.


ESDI BIOS and Microcode

BIOS Extensions ROMs (ROM images from David Beem)

U14,25 2x 27128 EPROM (128KB, 16x8)

FRU P/N (low & high)Internal P/NDate
90X8064 & 90X806590X8066 & 90X806702/11/87
90X8969 & 90X897090X8971 & 90X897206/05/87

Microcode ROM (ROM images from David Beem)

U16 1x 27128 EPROM (128KB, 16x8)

FRU P/NInternal P/NDateVersion
90X739990X68533 February 19870002
90X8635UnknownUnknownlikely 0001 or 0003
15F658715F65887 October 19870004
15F680715F680913 January 19880005
91F7430UnknownUnknownlikely 0006
04G375904G37615 April 19910007

15F6587 caused a diagnostic formatting problem and an intermittent hardfile delay during system operation (the hardfile light would remains "on" for approx. 13 seconds). also, in rare instances, a write fault could result in a data shift problem during error recovery, which would be detected during read operations and during diagnostics as a "10473" error (ECC error; read error).

15F6807 caused a highly intermittent problem of undetected write faults on the last 1/3 of the last sector written (detected during system read operations and by diagnostics as error code 10473, ECC read errors).

91F7430 experienced a highly intermittent system "HANG" only on 115MB ESDI fixed disks.

David Beem says:
   Version 0001 may have not been released since it would have been prior to the initial PS/2 models that used the ESDI controller coming out in April 1987.
   One of my 04G3759 EPROMs (I should look below the other two I have) had a confusing label on the underside: "COMPATIBILITY SOFTWARE (C) 1985 PHOENIX SOFTWARE ASSOCIATES LTD ALL RIGHTS RESERVED".

Updating ESDI Microcode

   If U16 is 04G3759, then this ECA has already been applied. Modules with any other P/N should be replaced by using this ECA.

Downlevel ROMs: P/N 90X7399, 90X8635, 15F6587, 15F6807, and 91F7430.

Note: Some older versions of direct driver software, which bypass BIOS (basic input/output system) may experience failures accessing the Fixed Disk after the installation of this ECA. This may occur because changing this module may alter how the Fixed Disk subsystem "appears" to the software. Software which uses BIOS is not affected and will function normally. DOS and OS/2 use BIOS.

   If the user software fails after this module is changed, the original module should be re - installed, and the appropriate software support function should be contacted for any possible software patches or updates.

   After replacement of the module, FRU P/N 92F0062 (P/N 04G3759) advanced diagnostics ESDI fixed disk(s) routine should be run to insure proper hardfile operation.


How Many Drives are Supported?

   Two are supported. ESDI natively supported 7 to 8 drives - but IBM (and others) cut that down to 2 or 4... the original IBM / WD controller has two ports for drives.

Installing a Second ESDI Drive

From Joe Kovacs
   You will need another data-cable for the new drive. The wide control cable has a second plug already. To make it a D: drive, you take out the resistor (Or some models use a DIP switch).

   Run automatic configuration, low level format it (CTRL-A on the main menu), fdisk it, DOS high level format it, and you're away.

Using Non-IBM PS/2 ESDI Drives

>Will the HD run in my 8580 even if it is not the original IBM-HD?

   As I understand it, the ESDI drives for the 80-class machines had identity data stored on the drive itself. If it's not an original equipment ESDI drive, or if it *IS* an IBM drive but has since been low-levelled in another (non-IBM) machine, it can't be put back in an 80 unless the Reference Diskette is "cooked". For Peter Wendt's recipe, look HERE.

8560/8580 Harddisk Wiring Schematic (from Peter)

                    to Power Supply
                      |        |
  +----------------+  |        |    +----------------+
  |                |H-+        |   H|                |
  |                |        +------H|                |
  |                |H       |  |    |                |
  |     HD #1      |H-------------\H|      HD #2     |
  |    (Rear)      |H       |  |  /H|     (Front)    |
  |                |      +------/ H|                |
  |                |H     | |  |    |                |
  |                |H---+ | |  +---H|                |
  +----------------+    | | |       +----------------+
                        | | |
                        | | |
                        | | +---------+
                        +-|------+    |
                          |      |    |
         Rear             |      |    |        Front
         L_U-----------HHHHHHH--HHH--HHH-------+---+
          |            J1       J2   J3        |   |
          |                                    |   |
          |                                    |   |
          |    IBM HD-Adapter (MFM or ESDI)    |   |
          |                                    |   |
          +-+HHHHHHHHHHHHHHH+------------------+---+
            +HHHHHHHHHHHHHHH+

Cable from J1 to HDs #2 and #1 is twisted for 5 lines 6 to 10 between HD#2 and #1
The segment between J1 and HD#2 is wired 1:1
Cables from J2 to HD#1 and J3 to HD#2 are both wired 1:1 with no twists


ESDI Terminator (from Peter)

How to build your own ESDI terminator:

 +----o----o-- ... --o----o----+
 |    |    |         |    |    |
 |    |    |         |    |    | +++  +++  +++       +++  +++   |
| |  | |  | |       | |  | |   |
|R|  |R|  |R|       |R|  |R|   |
|0|  |0|  |0|       |1|  |1|   |
|1|  |2|  |3|       |0|  |1|   |
| |  | |  | |       | |  | |   | +++  +++  +++       +++  +++   |
 |    |    |         |    |    |
 |    |    |         |    |    |

12   11   10   ...   3    2    1 = Pin No.

Pins 12 - 2 are 150 Ohms against Pin 1
Pin 1 is the common contact
All resistors are 150 Ohms / 0.25 Watts

MFM/RLL Cables from RadioShack (Dated, for reference only)

Dual MFM/RLL Drive Kit 950-0325
   28" dual data cable and a 28" dual control cable (?)
MFM/RLL Replacement Cable 950-0326
   18" 20-pin IDC to edgecard socket
MFM/RLL Replacement Cable 950-0327
   28" 34-pin IDC to edgecard socket


Maxtor 8760E ESDI drive problems on IBM ESDI controller

   What could be causing so many 10480s (seek errors)- The drive light flickers on the disk, but is constant on the top HD light, and only gives 10480, even though it looks like it works. The drive was pulled from a 486, what could be wrong with the drive/controller in the model 80? I've read a post about setting a 380MB and other nearly alike ESDI drives similar to mine, but none of the tips work/apply so far. I've even tried custom cables, and different types of 34pin cables. What do I need to do to either get IBM's cable for this card (number please?).

From Peter:
   The IBM ESDI controller is a 10MHz controller that has a limit on the speed (10Mb/s disk-to-interface) and the sectors (36). So most likely the XT-8760E will not work with that controller. It is a 52-sectors drive and seems to be an ESDI 15Mhz device as well.


ESDI in a 9577 Bermuda?

From Werner Förtsch:
   I have a 9577 with an onboard SCSI with one hd drive which was up to now my boot disk. I found from an old PS/2-80 an ESDI controller and two ESDI drives which I installed in the 9577. After long I got the system up running. My problem now is that my 9577 now boots from the first ESDI drive. Is there any possibility to boot from the SCSI harddrive in changing something in the firmware?

Peter
   1. The ESDI controller has *not* been announced for use in the later models after Mod. 80 - so it is no good idea to use it in a 77 of any flavour.
   2. If any ESDI drive is recognized during setup the machine BIOS handles it directly on the BIOS-Int Level as system hardware extension (INT 80h device) just like an MFM-drive. The SCSI BIOS is in this case "one step behind" and the MFM (if any), IDE (on "Lacunas") and ESDI-drives like in your case will called first and attached to the Int80h device-call.
   3. It *might* be possible to use the "Selectable Startup Sequence" in the machine setup ("Features" in the main menu) - but I truly doubt that the startup will "know" the ESDI-drive *because* the adapter is not supported in that machine. However worth trying and looking at anyway.
   4. The 16-bit MCA Stage 1 ESDI-Adapter will most likely have some influence on the systems performance. I would recommend to remove it - in case you really plan to do something with the machine and not only do that for curiosity only. The investment in a new faster and larger SCSI hardisk (like the IBM DCAS-32160, 2.16GB Ultra-SCSI) is not wasted money. The system acts a lot more lively with that.

>Thank you anyway for your helpful information.
   Nothing to thank for. I even forgot to mention another nasty effect of this combination: you cannot run Win95 or WinNT with it. Both adapters, the IBM SCSI and the IBM ESDI are hardwired to use IRQ 0Eh (14) and are tied up at the same time. This interrupt-sharing is a technical feature of the MCA - and causes no problem under DOS / Win 3.x or OS/2 ... but Win95 / 98 or NT cannot handle that, because it runs against their "one device / one resource" strategy. So much for the "guys in Redmont" and their understanding of modern technologies.

   So if you just tried it for curiosity - you better leave it. I tried something similar back in 1989 with the Mod. 80-311 to add an SCSI adapter for larger drives and wanted to boot from the SCSI ... did not work. The ESDI always started first. This misbehaviour is (as far as I know) buried in the different handling of ESDI and SCSI from the BIOS.

   If anyone else finds a way - okay - I am interested. But as far as I know - and from my own experimenting - it does not work. (Also: Mod. 70 with IBM SCSI and SCSI-HD: also starts from the DBA-2 ESDI drive first)

(Ed. Peter points out the 16 bit compatibility mode the SCSI and ESDI controllers create. So you can run W95 with this setup, but...)

>You are right it will be much better to invest some money for a new SCSI drive.
   Please keep in mind that the 9577 with the onboard SCSI is limited to a drive size of 3.94 GB (corresponding to IBM) for the "first drive to boot from and which holds the system partition". This point was topic on an older (or: several older) threads in this group. Therefore I recommended the 2.16GB IBM and not the 4.2GB ... ! But any modern 2GB - 3.5GB drive will do fine. Quantum makes (made ?) a Fireball with 3.5GB capacity. This would mark the maximum installable in the Model 77. The "over 4GB" appear to be installable, are even recognized with the exact capacity - but the IML-partition will not be installable. Now: will install - but will not work. And then you ran in a nasty IML-error of the I999 00nn category. That for completeness.


Is the IBM Integrated HD Adapter an ESDI Controller?

From Peter
   First off: The "IBM integrated harddisk adapter" (Card-ID DF9F) as it can be found in 50Z, 55SX, 70 and P70 is not a real ESDI drive. It is more or less technically an MFM RLL 2.7 drive - but combined with a MCA harddisk adapter in one physical unit. The "ESDI or not misunderstandment" is caused by the PS/2 BIOS.

   They (IBM) treated the drive as ESDI, because back in those days the MFM harddisk standard was limited to 17 sectors per track (and still is for pure Non-RLL MFM drives) and while the "modern drives" used to be smaller and use lesser platters and -therefore- lesser heads it was easier to translate the physical geometry with e.g. 929 cylinders, 56 sectors and 4 heads into a scheme with 64 heads, 32 sectors and "downscale" the number of cylinders accordingly.

   The above example (929 x 56 x 4) would result in 208.096 data blocks á 512 bytes = 106.545.152 bytes. The translation into the 64/32 ESDI scheme would result in the more handy 101 cylinders ... by cutting down the total capacity to 105.906.176 bytes total. However the values 101 cylinders, 64 heads and 32 sectors give a better match into the old XT/AT controller scheme - particularly the cylinder register was -according to the basic WD1007 controller- the problem. It could not hold values over 1.024 ... the ESDI translation in the BIOS opened a more handy way to handle bigger drives up to 1GB IIRC.

   So after all the "ESDI" in the desktops using the integrated harddisk thingy is only imaginary. The towers (60 and 80) used "Real" ESDI controllers and harddisks.

Secondly the DF9F HD / controller combo was primarily designed as "single device". The later @DF9F.ADF allowed to set one as "primary" and one as "secondary". But as far as I know this has been included to match an early draft of the PS/2 Mod. 90 hardware .... which *had* two integrated harddisk controller ports at the front end of the sysboard. These however had been made non-functional in the later platform BIOSes and don't work. I have played around with them in the early 90s but found no clue to get them working with any Type 1 - 3 platform. As well as Alfred Arnold tried recently - don't know if he gave up yet.

I hadn't been that desperate to try installing the 386DX-20 (Type-0) platform in my 8590 and see if I get the front drivebays going with that.

Due to the lack of appropriate connectors none of the PS/2 machines support two integrated harddisk adapters. These *are* MCA connectors. The 72-pin layout of these drives is basically a slightly stripped-down 16-bit MCA connector. And the planar ADF for e.g. a Mod. 55SX says "4 slots" where the Slots 1 - 3 are for expansion cards and Slot 4 is at the end of the riser card - extended with a flat-ribbon cable over and down to the harddisk.

And - No - you cannot just crimp another 72-pin connector in that cable. There are signals that select the slot number - and that for this "cable port" is fixed set to Slot 4... so any other connector on that cable would signal "Slot 4" to the sysboard. It is -as said- "stripped down"... means: apart from some DC- and GND-wires also "other unimportant signals" are not passed over to the HD-connector.

Content created and/or collected by:
Louis Ohland, Peter Wendt, William Walsh, Kevin Bowling, Jim Shorney, Tim Clarke, David Beem, Tatsuo Sunagawa, Tomáš Slavotínek, and many others.

Ardent Tool of Capitalism - MAD Edition! is maintained by Tomáš Slavotínek.
Last update: 12 Jun 2021 - Changes & Credits | Legal Info & Contact