Help me bring back retro 80386 test system to 2016!

[voidptr]

Everything has been said  ^^^

Except that I might have one in my closet museum 

I booted up my first Pc 8088 8MHz last summer,  It booted right away,  those were so expensive, but those green characters are still pretty   

You might find some 80386 in good condition around.

[IconicPCB]

Let me add to the everything.

Talk to your local industrial computer supplier.

Inform Yourself about PICMG 1.0 backplane boards and various plug in processor cards. You may find a board with the required number of ISA slots and You may find a processor card runing a 33 MHz  80386 CPU.

All the other caveats apply... ESTABLISH A PAPER TRAIL unless you want to be the bunny.

[MrMetthew]

Thanks for all the feedback guys! We are currently reviewing what we are going to do. Btw there are indeed replacement cards. The machine is also only used for legacy tests and for the production of still supported spare parts, but none the less still generates some revenue. First thing what we are going to do is to calculate how much revenue the thing actually generates. Possible solutions than range from stopping support for the specific parts the machine is responsible for, testing and calibrating a certain stock of parts while the machine is still functioning and then retiring it. Or rebuilding the tests that are still profitable on a new platform (my personal favorite is an fpga/ARM SOC + labview interface). Again thanks for the advice, we now really realized how unsupportable this platform actually is !

[CJay]

Sadly, the right decision.

As a technical challenge it is very interesting indeed and I'd have thoroughly enjoyed being given that sort of task but from a business/career point of view, it was a nightmare.

[pelule]

Thanks MosherIV, I am feeling not such crazy anymore.
My GPIB based lab-pc is an old ISA slot PC OS is Windows-XP (old HP system, thus not such old at the 386 AT's).
All my new PCs failed after some years. This PC is running very well 24h a day, 7 days a week without any issues since 1997.
I backup all data regulary, as I am not such crazy.

[mikeselectricstuff]

The correct way is to start again with modern hardware and to re-specify the tests and software and get it properly documented this time.

There is no such thing as "correct" or "inccorrect", only more or less appropriate for a given situation.

In this situation, if you can find/build  of drop-in replacements for all parts of the system as backups this would seem by far the cheapest & easiest soluiton, without the risks of breaking something in the process of re-doing everything from scratch.
It ain't broke - just get some backups for when it does need fixing.
 

[Ian.M]

In this situation, if you can find/build  of drop-in replacements for all parts of the system as backups this would seem by far the cheapest & easiest soluiton, without the risks of breaking something in the process of re-doing everything from scratch.
It ain't broke - just get some backups for when it does need fixing.

+1.   
Spares of the custom cards exist.  I suspect a 486 board clocked down to 33MHz with Cache disabled will have near enough compatible timings, if not, then try to source a 386 board.   Once its up and running off a CF card SSD, and all tests have been validated, run the new system for a month, obtain another identical motherboard and build up a spare the same way, then the old system can be retired.

It also seems that management realises they've been running on pure luck for the last decade so the risk of the project going toxic is much lower.   If the parts it tests still generate enough profit, or customer loyalty to justify keeping them in production, its worth doing.

[Fungus]

There are challenges and there's certain death.

Personally my approach would be to try and re-engineer the test procedure,

Yep. Trying to replace the PC is suicide. Replicate the functionality/results instead.

eg. All those ISA cards can probably be replaced with a few Arduinos.

[timb]

A 386 era PC is likely to have pre-ATA hard drives, maybe early IDE, (incompatible with modern controllers or USB adaptors), or SCSI-I, ESDI or ST-506.

First off, ATA *is* IDE, they are synonymous. Basically, IDE was renamed to ATA (AT/Attachment) when the ATA-1 standard was ratified in 1994.

"Early IDE" drives are perfectly compatible with modern controllers, why wouldn't they be? So long as you've got a PATA port (or SATA to PATA adapter) it should work...

I've actually got one of the *first* models of IDE drives to be produced; it came out of a Compaq Portable III (which was the first commercial machine to use the IDE interface). I can plug this 30 year old drive into a USB to PATA/SATA dongle and read the contents just fine...

Anyway, by the time the 386 was popular (let's say 1990), IDE had pretty much become the defacto standard. SCSI was the only other real competition, however it was expensive so relegated to servers and high end workstations. ESDI was very short lived, being displaced by IDE. ST-506 was pretty much only used for legacy applications by this point as well.

[Ian.M]

I have a couple of 8 bit IDE drives that pre-date the ATA standardisation that definitely aren't compatible with modern controllers.  One is a 40MB  WD drive on a hardcard in a legacy system I've owned since the late '80's and the other is IIRC a 30MB drive I keep as a spare for the legacy system. See https://en.wikipedia.org/wiki/Industry_Standard_Architecture#XT-IDE

[westfw]

A 386 era PC is likely to have pre-ATA hard drives, maybe early IDE

You probably mean the old MFM disks, complete with separate controller card.

rebuilding the tests that are still profitable on a new platform

That would be time better spent, especially if you document carefully as you go along.

(my personal favorite is an fpga/ARM SOC + labview interface).

Generic PC with a set of Arduinos to replace the custom logic.  (You haven't mentioned what the custom cards DO, or how complicated they are, but we've established that Speed Is Not An Issue, so FPGA is probably overkill.  And FPGAs "go stale" at about the same speed as PCs.)

[kfnight]

At the Tektronix service depot they had a test system that was an 8086 running MS-DOS 3.2. Its software was programmed in QuickBASIC and was still going strong in 2008.

[Ian.M]

A 386 era PC is likely to have pre-ATA hard drives, maybe early IDE

You probably mean the old MFM disks, complete with separate controller card.

No I don't. I mentioned the possibility of ST-506 MFM drives, then there's ST-412 RLL, which looks like ST-506, but uses a different controller and could increase the data capacity of a drive by 50% over MFM,  and I also mentioned ESDI.  That's three different incompatible systems over the same type of cabling.   Even if it has IDE, if its a generic PC, its almost certain to have a separate controller card - usually an IDE/FDC/IO combo card.   The leading brand systems (e.g. Dell, Compaq) tended to have a higher level of integration and tended to have controllers on the motherboard.
Then you get all the different drive geometries.  Its one big can of incompatibility worms, hence my recommendation to use Laplink or Interlnk to get the data off.

[timb]

I have a couple of 8 bit IDE drives that pre-date the ATA standardisation that definitely aren't compatible with modern controllers.  One is a 40MB  WD drive on a hardcard in a legacy system I've owned since the late '80's and the other is IIRC a 30MB drive I keep as a spare for the legacy system. See https://en.wikipedia.org/wiki/Industry_Standard_Architecture#XT-IDE

The hardcard doesn't count, because it's mounted on an ISA card, so it's not really IDE (well it is, but not in the traditional sense).

As far as I know, there were no 8-bit only IDE drives (aside from CF); there was an 8-bit transfer mode that was part of the IDE standard. Basically, you'd install an XT-IDE card into your XT machine. It contained a BIOS and IDE connector and would talk to a standard IDE drive in 8-bit transfer mode.

In fact, there's a modern project called XTide that has an open source BIOS available for XT and AT machines that provides large drive support, LBA and lots of other goodies. There are even several open hardware ISA cards, some with IDE connectors and some with CF connectors (plus ROM slot). I actually used the AT version earlier this year to get a 2GB CF drive working on an old 286. I just popped the BIOS into the option ROM slot of an EtherLink III card and it just worked. It's pretty slick.

A friend used it to get a circa 1999 4GB hard drive drive going on an original PC XT. That drive was ATA-2, but it supported 8-bit transfer mode, so it worked fine.

Edit: Despite what that Wiki article says, "XT-IDE" does *not* pre-date IDE. It came after, as a way to use IDE disks (that supported 8-bit transfers) with an XT class machine. The original IDE interface was first launched on a 286 (the Compaq Portable III) in 1986.

Furthermore, even if you had a drive that only supported 8-bit transfers, it should still work today, as that's a mode required by CF cards, which are IDE by nature.

[Ian.M]

See http://nerdlypleasures.blogspot.co.uk/2014/04/the-original-8-bit-ide-interface.html

Also the XT-IDE drive WD93044-X-11R I have in the legacy system is pre-1990, and the ATA-1 standard was ratified in 1994 (ANSI X3.221-1994, AT Attachment Interface for Disk Drives). 

I certainly had problems attempting to clone the drive on a  more modern PC circa 2000, (I needed an image for an emulator) and in the end had to resort to copying the filesystem over a proprietary 100Kbps LAN to the HW emulator.

[timb]

See http://nerdlypleasures.blogspot.co.uk/2014/04/the-original-8-bit-ide-interface.html

Also the XT-IDE drive WD93044-X-11R I have in the legacy system is pre-1990, and the ATA-1 standard was ratified in 1994 (ANSI X3.221-1994, AT Attachment Interface for Disk Drives). 

I certainly had problems attempting to clone the drive on a  more modern PC circa 2000, (I needed an image for an emulator) and in the end had to resort to copying the filesystem over a proprietary 100Kbps LAN to the HW emulator.

Yeah, and I've got an original 20MB IDE drive from 1986 that I just plugged into a USB to SATA/PATA adapter and was able to read the data off just fine.

Just because the ATA standard wasn't ratified until 1994 doesn't mean that IDE as a standard wasn't followed from 1986 up to that point. It was.

Anyway, XTA != XT-IDE.

XT-IDE was basically putting regular IDE (ATA) drives into 8-bit transfer mode so they could work on the 8-bit XT bus. Most ATA drives of the era supported that, as do pretty much all CF cards. This was done with an XT-IDE ISA card that had a regular IDE connector and a special BIOS.

XTA was a short lived standard that used different wiring, addressing and only used 8-bit transfers (except one drive that supported 8/16-bit).

That WD drive is XTA, not XT-IDE.

[ebclr]

Possible solution

http://www.ebay.com/itm/HM386SX-Intel-386SX-25MHz-8MB-SIMM-Video-512KB-ISA-I-O-CABLES-WORKING-/282261294074?hash=item41b8158bfa:g:IEIAAOSwV0RXssfA

[CJay]

As far as I know, there were no 8-bit only IDE drives (aside from CF); there was an 8-bit transfer mode that was part of the IDE standard. Basically, you'd install an XT-IDE card into your XT machine. It contained a BIOS and IDE connector and would talk to a standard IDE drive in 8-bit transfer mode.

I'm pretty sure there were some 8 bit only XT IDE drives, they were pretty rare but they were used in Commodore A590 external hard disk modules for one, horrible slow stepper motor driven heads as well.

I certainly remember searching high and low for the mythical hard disk drives which supported both modes but never managed to find any and eventually settled on buying SCSI drives instead after discovering a (very) cheap source of Maxtor XT4380S drives (old ICL mainframes and ICL/Fujitsu PC servers).

[Fungus]

Possible solution

http://www.ebay.com/itm/HM386SX-Intel-386SX-25MHz-8MB-SIMM-Video-512KB-ISA-I-O-CABLES-WORKING-/282261294074?hash=item41b8158bfa:g:IEIAAOSwV0RXssfA

I dunno. What state will the capacitors be in on a motherboard that old?


Edit: I take that back. Looking at the photos, there aren't any electrolytic capacitors on that motherboard. 

[Fungus]

PS: Embedded 386 machines are still a thing, you can actually buy new ones...

https://encrypted.google.com/search?q=embedded+386+system+with+isa

[ebclr]

For IDE drives

http://www.ebay.com/itm/Super-Talent-DuraDrive-ET3-64GB-2-5-inch-IDE-Solid-State-Drive-MLC-/321695435495

[timb]

For IDE drives

http://www.ebay.com/itm/Super-Talent-DuraDrive-ET3-64GB-2-5-inch-IDE-Solid-State-Drive-MLC-/321695435495

That's massive overkill, especially when a 512MB-2GB CF card would work just as well and you won't run into any LBA issues.

[technix]

If you can still find the original specs for the test signals, try rewrite the test suite moving it to a microcontroller. I think that a PIC18F4550 @ 48MHz or STM32F103RC @ 72MHz have enough horsepower to replace that aging system testing the DUT. Those USB microcontrollers also allow you to use a modern laptop to diagnose the target system.