Finding fault in early 90's ISA PC Card

[rx8pilot]

This post is semi-related to the thread I started for an old pick and place machine I got recently. https://www.eevblog.com/forum/reviews/finally-have-a-pick-and-place-on-the-way-quad-ivc/

Getting it up and running has revealed at least one bad PCB that is part of the vision system critical to the operation. In general, I would not bother reviving old-tech except for the fun of it. In this case, finding a fault in the board is quite important for me to be able to assemble PCB's without having to spend the money on contract manufacturing. This machine is full of PCBs that are hard to find so sharpening my PCB repair skills may be worth it.

The board is an ICOS MVS922 which is a controller that goes into an ISA slot of an industrial PC running wIn XP PRO. The original configuration was DOS based, the XP is an update that happened around 2009. The exact function of the board is unknown and I have not found any documentation at all. The company was purchased by KLA-Tencor and I have reached out to them to see if an archive exists. This is a commercial/industrial product so there is at least a small chance that some helpful docs exist. I have done some troubleshooting before, but my day to day is designing circuits. Without a schematic or block diagram it is hard to know where to start. The PC does not flag it as a problem, but the software cannot get a response from it. That is all I know. It worked for some time with the exact same PC configuration.

I pulled some of the datasheet on the components which are largely common discreet 5v logic, RS-485 differential driver/receivers, and some memory. I did a thorough visual inspection under a microscope. Everything looks good, no evidence of previous repairs. No thermal damage or cracks.

Checked the +5v to GND at around 170 Ohms which seemed a little low but not alarming. When powered up, it pulls 400ma (2 watts) which seems a little high with nothing connected to the I/O but not alarming. Nothing gets hot enough to find with my fingers and I do not have a thermal camera.

Powered it up outside of the PC and checked to verify the input to output relationships on the logic chips as best I could based on the data sheets. Nothing seems abnormal. Its hard to check flip-flops and shift registers in a simple way.

The chips that have the HEX value labels are unkown. At this time, I have not removed the stickers because these have a trade-in value if I were to replace it. A company called PPM keeps these and repairs them for $500 plus your trade in. Without a trade, it costs a lot more. If anyone has a guess what those might be doing without pulling the stickers, I am all ears.

Any suggestions where to go from here?

[wraper]

400 mA does not seem to be very high current at all, considering ICs used. 74F series consume a lot of current. 74LS consume significant current too. IDT7202 has 20 mA  max standby current, typ unknown. SN75179B has 57 mA typ. supply current, so those two will eat 114 mA alone.  But considering that there is oscillator, this board might be not in standby at all.

[PA0PBZ]

Just some suggestions:

 - Are you sure the jumpers are in the right position?

 - Does the card need an input signal to be initialised properly?

 - If you are stuck start with replacing the F374's and the ACT244's on the bus

[SeanB]

Labelled devices are Lattice Semiconductor logic gate arrays, used to fit what would otherwise be another board full of TTL into a single small package.

http://www.latticesemi.com/

At these are programmed and then have a protection fuse blown, it will be near impossible to see if they are working or not other than if they work properly. My bet is that they have the correct programming for these to replace faulty ones on the board, and simply change all of them on faulty boards, or do some simple tests to see which one is faulty, then reprogram it ( likely it has just had eeprom failing so a rewrite will make it work again).

If it was working then it is likely that Windows Plug-n-Pray might have reassigned an IRQ and IO address, and then it conflicted with this board because the board is not even aware of P-n-P, having jumpered settings. Check the registry that there are blocks fixed for the board, I get hazy, not having done much with hard wired non pnp boards and XP. 98 did need some massaging.

[wraper]

BTW why are you sure that this boards is faulty? Does the program give some error code? What I see that there is some serial connection going from the single connector on this board, so obviously it need to communicate something on the other end. There could be some jumper put in the wrong place too.

[rx8pilot]

Just some suggestions:
 - Are you sure the jumpers are in the right position?
 - Does the card need an input signal to be initialised properly?
 - If you are stuck start with replacing the F374's and the ACT244's on the bus

The jumpers have been confirmed by the company that pseudo supports these. The initialization fail is reported by the software with all signals attached. Since the error reporting is cryptic, it's hard to guess what is happening. Simply "Initialization: FAILED". It was previously working in this state from what I have been able to learn from the previous owner. Information is not 100% reliable though.

Are the 374 and 244 prone to failure?

BTW why are you sure that this boards is faulty? Does the program give some error code? What I see that there is some serial connection going from the single connector on this board, so obviously it need to communicate something on the other end. There could be some jumper put in the wrong place too.

It is still a guess based on the troubleshooting I have done on the PC side and what the technician at PPM has seen before. I have verified voltages on the ISA BUS and the driver is loading without reported errors. The serial connection is dual RS485 based on the drivers next to the connector.

Labelled devices are Lattice Semiconductor logic gate arrays, used to fit what would otherwise be another board full of TTL into a single small package.

http://www.latticesemi.com/

At these are programmed and then have a protection fuse blown, it will be near impossible to see if they are working or not other than if they work properly. My bet is that they have the correct programming for these to replace faulty ones on the board, and simply change all of them on faulty boards, or do some simple tests to see which one is faulty, then reprogram it ( likely it has just had eeprom failing so a rewrite will make it work again).

If it was working then it is likely that Windows Plug-n-Pray might have reassigned an IRQ and IO address, and then it conflicted with this board because the board is not even aware of P-n-P, having jumpered settings. Check the registry that there are blocks fixed for the board, I get hazy, not having done much with hard wired non pnp boards and XP. 98 did need some massaging.

I don't think this is a PNP card, the BIOS is forced to adssign IRQ10 and 11 to this "LEGACY ISA" card. I confirmed that it is physically connected to IRQ10 and 11 although I dont think it is actually using 11. The software is a custom re-write done long after this card was in service to make it work with XP.

I will take a closer look at the logic gate arrays, glad to know what those are. Thank you.

Also, I don't see any type of clock signal on the board with a quick pass with a scope. May mean nothing, but the crystal should be driving something but maybe it wont start without a PC connection.

continuing to search........ thank you for the tips!

[SeanB]

Crystal module could be faulty, check it has 5V and ground on 2 of the pins, then pop the scope on the other 2 pins, one will be a 20MHz something like square wave ( it will not look like one, just has to be close enough for a TTL gate to like it) and if it is not there you have your fault.

[PA0PBZ]

Are the 374 and 244 prone to failure?

Since it looks like there is a faillure to communicate with the card it could be that the part(s) sitting between the card and the pc are failing. If there is something wrong in the processing that the card should do I'd expect it to initialize properly but then being unable to do what it's supposed to do, resulting in the card outputting wrong data. I have seen and repaired ISA cards that had the bus drivers failing. Also, this is easy to do, much easier than trying to replace the gate arrays.

[SeanB]

Plug and pray always has an issue with legacy ISA cards, as it often will, for some reason, ignore the assignments and decide to allocate those used locations and interrupts, so the card and some other thing ( like a serial port or a second parallel port) stops working correctly.

[rx8pilot]

Crystal module could be faulty, check it has 5V and ground on 2 of the pins, then pop the scope on the other 2 pins, one will be a 20MHz something like square wave ( it will not look like one, just has to be close enough for a TTL gate to like it) and if it is not there you have your fault.

....checking. Will report.

Since it looks like there is a faillure to communicate with the card it could be that the part(s) sitting between the card and the pc are failing. If there is something wrong in the processing that the card should do I'd expect it to initialize properly but then being unable to do what it's supposed to do, resulting in the card outputting wrong data. I have seen and repaired ISA cards that had the bus drivers failing. Also, this is easy to do, much easier than trying to replace the gate arrays.

Definitely easy to replace....I will try crystal first followed by a DigiKey order for all the common logic and possibly the RS485 drivers.

[6502nop]

It looks to me (at 200% mag) that your oscillator is in backwards. Rotate it 180 degrees and see if it works...

{These tin can modules have one "sharp" corner, the others are rounded. The sharp is pin 1. If you look at the silk screen, you'll see that the sharp corner is the lower left, but the module shows all three rounded corners (I can't make out if the upper right of the can has the sharp edge from that angle). This would also explain no clock on the board, as the fourth pin is often NC.}

If it still doesn't work, it may be that the module was damaged from reversed polarity. No biggie, as they're only a dollar or two.

nop

[German_EE]

You should be able to get a clue about the fault before ordering parts.

1) Go to the XP Device Manager and make sure that the card can be seen. Does it show the correct IRQ?

2) Check for clock pulses using a scope.

3) The 74F374 and 74ACT244 chip pinouts are easy to find so follow the signal flow through, if you get an input then there should be an output.

4) Now it gets more difficult. Data on the rest of the 74 series chips can be found but you need experience in logic circuits. A 74F74 for example is a pair of flip flops so if you get inputs but no outputs it could be a faulty chip but it could also be a reset line that is stuck.

My guess is the 74F374 and 74ACT244 buffer chips.

Edit: Yes, it does look like the oscillator is the wrong way up.

[T3sl4co1l]

It's dead, Jim.

There's no way 400mA accounts for everything on there... at least, I don't see that it could.

Is this card really so precious?  It appears to be an old fashioned 8 bit RS-422/485 UART.  Evidence: SN75179? transceivers; the big IDT parts are FIFOs, and everything else is glue logic (bus latches, address decoding, etc.).  No clue what port(s) it shows up as, but I'd be willing to bet (knowing what the FIFOs are) that it wouldn't be hard to find what addresses make them blink.  Or just DEBUG the driver program and see what it IN/OUT's to!

Anyway, it doesn't seem like an unusual enough card not to have any generic substitutes.  You could probably save piles in repair cost if you can identify a sub.

All the CMOS parts should draw about bupkis current, at least until things start really cooking (to be fair, it does have an onboard crystal, so who knows what it does during startup/run?).  The 74F logic might be a bit more hungry, I forget, but there's only five of them.  The five PAL/GALs (almost certainly what the stickered chips are) could be pretty awful, now that I think of it, but they'd have to be doing, like... at least 50mA each, were they really that bad?

Tim

[T3sl4co1l]

You should be able to get a clue about the fault before ordering parts.

1) Go to the XP Device Manager and make sure that the card can be seen. Does it show the correct IRQ?

FWIW, unless the card supports Legacy PnP (unlikely in logic??), Windows literally knows only what the driver tells it.  If the driver were not installed, Windows probably won't automatically detect the card.  Likewise, the driver itself may not have any good way to check if its card is installed or not; hopefully it has a testing register or something that can be used to at least show that.

But in short, yeah; there's very little that can be said about ISA cards, generally speaking.  They're just kind of there, and if the CPU sends out data to certain ports, if something just happens to be listening... then it works, or if not, it's just a fart to the wind.

Frankly, I'm impressed that they support this thing so thoroughly that they actually wrote an NT kernel driver for it!

Tim

[rx8pilot]

1) Go to the XP Device Manager and make sure that the card can be seen. Does it show the correct IRQ?
2) Check for clock pulses using a scope.
3) The 74F374 and 74ACT244 chip pinouts are easy to find so follow the signal flow through, if you get an input then there should be an output.
4) Now it gets more difficult. Data on the rest of the 74 series chips can be found but you need experience in logic circuits. A 74F74 for example is a pair of flip flops so if you get inputs but no outputs it could be a faulty chip but it could also be a reset line that is stuck.
My guess is the 74F374 and 74ACT244 buffer chips.
Edit: Yes, it does look like the oscillator is the wrong way up.

It does show up in the device manager with correct IRQ. The BIOS 'forces' the OS to assign IRQ 10/11 to this card and I verified that the card is physically connected to IRQ 10 via the jumper. I also looked at the IRQ conficts and nothing else is attempting to use IRQ10.
I checked for clock pulses yesterday while the card was out on the bench and did not find anything. I had accidentally put the crystal in backward when I was fiddling with it earlier - hope that did not damage it. When I first pulled the card, it was correct. I will look a little closer to see if crystal is oscillating or not. If it is not doing anything, that would be a critical point in the investigation for sure.

It's dead, Jim.

There's no way 400mA accounts for everything on there... at least, I don't see that it could.

Is this card really so precious?  It appears to be an old fashioned 8 bit RS-422/485 UART.  Evidence: SN75179? transceivers; the big IDT parts are FIFOs, and everything else is glue logic (bus latches, address decoding, etc.).  No clue what port(s) it shows up as, but I'd be willing to bet (knowing what the FIFOs are) that it wouldn't be hard to find what addresses make them blink.  Or just DEBUG the driver program and see what it IN/OUT's to!

That was my gut feeling as well. This card is paired with a machine vision processor in a Quad IVC pick and place machine (made by ICOS). It is hard to tell what secret sauce is in there but it is definitely more than just an RS485 card. PPM had to jump through hoops to get this classic hardware to work in XP. They have since written support for Windows 7 as well. From what little I know, it is a dual channel RS485 with some sort of buffer to send patterns to the vision system. The vision system in turn does the analysis on the image and sends geometry back to the PC. I wish there was an easy way to substitute something, but all the secrets are so old they probably backed up on 5.25" floppies in someones closet.

It certainly appears to be a dumb card as far as the OS is concerned which is why the BIOS had been set to reserve the IRQ. In the end, I can get a refurb one with a guarantee for $500 so I can't spend too much time on this one. I purchased one in unknown condition that will arrive in a few days. If it works, I can do some comparison troubleshooting and maybe have a spare.

[rx8pilot]

Without too much surprise, the crystal is not doing any crystalling anymore. Not sure if that is the original problem or if it simply dies from reverse polarity.

I don't have any of the 4 pin crystals like that so I will have wait for some to be delivered. There is a surplus electronics shop down the street from me that may have something. So bummed that I bodged it like that. Silly mistake.

[pickle9000]

- Are you using the original motherboard?
- If a particular slot was used clean it and use it.
- Mark down all settings for the bios, I assume they are correct. Then re-enter them. Check the battery voltage on the mb.
- check your power supply and clocks (mb and all cards). Start a log you may need these in the future.

Do the above first

- RAM is always an issue, 7202LA50. I wouldn't even test it just install a couple sockets and install new. Cut the old ones out remove pin at a time and test the thru holes with a meter before installing a socket. The board is more valuable than the chips so treat it with care.

Even if you have to change all the chips (all the ones possible) it's only a couple hours work and will make servicing the board in the future easier (sockets). Even though I do agree with the learning approach, this is a service operation no need to get fancy. More important to get the equipment up and running so it can earn it's keep.

[T3sl4co1l]

Oh, and try another electrical test and see if the rest of the board seems okayish or not.  Hopefully no [other] chips were harmed?

Tim

[rx8pilot]

- Are you using the original motherboard?
- If a particular slot was used clean it and use it.
- Mark down all settings for the bios, I assume they are correct. Then re-enter them. Check the battery voltage on the mb.
- check your power supply and clocks (mb and all cards). Start a log you may need these in the future.

The card came out of a pre-packaged controller PC that was assembled by the company that converted the machine from DOS to XP. It does not appear to have been changed at all. I tried three ISA slots although it should not matter with that bus. I checked the PSU but the MB is a single board PC form factor so nothing is really accessible without a major effort. It is plugged into a PCI/ISA backplane.

Even if you have to change all the chips (all the ones possible) it's only a couple hours work and will make servicing the board in the future easier (sockets). Even though I do agree with the learning approach, this is a service operation no need to get fancy. More important to get the equipment up and running so it can earn it's keep.

Every single chip on the PCB except the gate arrays is <$30 including sockets for most of them,  so I added them to my DigiKey order. If I am lucky, the gate arrays are just fine and the glue logic is the culprit. I have a fantastic setup for replacing the chips with little chance of damage to PCB. My guess is that it will be less than an hour. At that point, it could only be the gate arrays or the PCB itself.

Still hoping the replacement that is on the way will work and I can put this off til I have more time. Taking multiple paths should give me a better chance of having a pick and place machine working.

Oh, and try another electrical test and see if the rest of the board seems okayish or not.  Hopefully no [other] chips were harmed?

No changes....

[pickle9000]

You can pick up an ISA POST card reader off for under 5 bucks. Pretty handy to mount a test points on. They are getting harder to find so it's a good thing to have kicking around.

The reason I mentioned the slot is procedure. Some cards have a favorite slot be that mechanical, electrical or whatever. As a general rule put things back where they where so you don't have a compounded issue.

If the betting window is open I'll take the RAM.

One handy tip for ya. You can get machine contact ic sockets without a carrier. The bare pins are just on a holder to help with install. Super handy if you lift a pad on the upper side of the board.

Good luck to ya.   

[Kjelt]

Is it a crystal or is it an active oscillator? I guess the latter.
Can't you use a signal generator then to replace the oscillator till you have a replacement?

[rx8pilot]

Is it a crystal or is it an active oscillator? I guess the latter.
Can't you use a signal generator then to replace the oscillator till you have a replacement?

It is indeed active, not just a crystal. The good news is there is an electronics shop about a mile from me so I was able to replace it this morning. I verified the crystal first, installed it, powered the card, checked for clock pulse in a few places I figured they should be. Put the card back in the PC and still cannot initialize.

This result is not surprising since it was behaving this way before I damaged the original oscillator. I will wait until Digi-Key and the other 'new' card show up. At this point, I need to be careful with the time spent. Up til now, I was getting some real benefit from the troubleshooting giving me a picture of the whole system so future issues will be easier to understand.

[rx8pilot]

Ok, an update.

replaced 20Mhz Crystal - No change
replaced the RAM LA50P chips - no change
replaced 74ACT244 - no change
replaced 74F374 chips - no change

The mail man should be delivering a 'new' card tomorrow. That will provide some reference. I am also continuing to check the PC configs to make sure I am not chasing my tail. The only thing I can report is that the PC behaves exactly the same with OR without the card installed.

Hopefully the other card will help....

[Bryan]

Is there some tantalums I see. Are they ok?

[abyrvalg]

I can try looking into the software with disassembler to check the cause of that error message if you wouldn't mind to share the sw.

[rx8pilot]

Is there some tantalums I see. Are they ok?

They are not shorted and not burned. If they are weak, I don't think they would cause a full card failure - but who knows. They appear to be just global filters not critical to any timing or bypassing.

I can try looking into the software with disassembler to check the cause of that error message if you wouldn't mind to share the sw.

The software is hardware keyed which will make that offer impossible. I was wonder though, if there is a utility of some kind that can monitor to the ISA bus so see what data is passing through. I cannot tell yet if it is doing anything or nothing. Since the computer behavior is the same regardless of whether the card is installed or not is weird to me. The drivers appear to be made with WinRT Toolkit which is a shortcut to make a driver if you don't know how to make drivers. The relative simplicity of this card would make it a good candidate. Anyway, the OS reports everything as OK and working properly even when the card is absent. There are no errors, flags, additional warnings that the card is missing, etc.

No IRQ conflicts that are reported by the OS. I will try to replace more of the chips on the ISA bus side of the card to see if I get lucky.

[abyrvalg]

I don't need to run that sw to analyze it (doing static analysis in IDA Pro instead). No data files required also, just the exe and driver sys.

You've told that entire setup was converted from DOS to XP. Was the main sw replaced with some never Windows version or it is the same old DOS app running under XP now?
For "DOS under XP" case the driver should be pretty dumb: just make card's IO ports range accessible to DOS app which does all IO. No easy ways to sniff IO in this case.
But there are some ways for native XP app: first you can try DebugView https://technet.microsoft.com/en-us/library/bb896647.aspx to see if the driver outputs something (some drivers do). Next is a heavier tool that shows all activity between apps and drivers - http://www.osronline.com/article.cfm?article=199

[rx8pilot]

I don't need to run that sw to analyze it (doing static analysis in IDA Pro instead). No data files required also, just the exe and driver sys.

You've told that entire setup was converted from DOS to XP. Was the main sw replaced with some never Windows version or it is the same old DOS app running under XP now?
For "DOS under XP" case the driver should be pretty dumb: just make card's IO ports range accessible to DOS app which does all IO. No easy ways to sniff IO in this case.
But there are some ways for native XP app: first you can try DebugView https://technet.microsoft.com/en-us/library/bb896647.aspx to see if the driver outputs something (some drivers do). Next is a heavier tool that shows all activity between apps and drivers - http://www.osronline.com/article.cfm?article=199

I see what you are saying now. The original DOS software is no longer used at all. The software is a completely new application written for XP and I believe they used WinRT Toolkit to make the driver for the old ISA card. There are a number of other I/O's that have been consolidated into a single USB that is broken out at a custom interface card that has 4 FTDI serial interfaces (which all work well). There is also one dedicated RS-232 being used as well.

The entire system was pre-built including all software and configurations by the company that wrote the software/drivers. At least I know that the entire hardware system has, at one time, worked well together. I got the docs on the SBC (single board computer) which is a Portwell ROBO-8712EVG2A with a Pentium 4 CPU and 1GB RAM. It is attached to a Portwell PBP-14A7 backplane with PCI and ISA slots. The SBC has a built in PCI-ISA bridge, but I am not sure if that bridge needs a specific driver,service, etc to be active. I do not see any obvious devices or services that seem like they have anything to do with an ISA bridge. Any idea if that is a hardware element that runs on its own or is software needed to get it connected to the IO of the PC?

I now have another card (function unverified) that behaves the same way. In fact, the PC responds the same way whether the card is present of not. The System Information says everything is OK even if there is no card attached. Device Manager does not show any errors when the card is absent either.

I am going to look at the links you provided, hopefully they will provide some clues about the status the ISA bus in general.

[rx8pilot]

DebugView output:

Line 40 shows the "HS3L Task"
HS3L refers to the ISA Card (Model MVS922). Not sure what the significance of the mapped memory address 0x258 (guessing that is an address)

[abyrvalg]

The PCI-ISA bridge should be brought up by BIOS and remain active, nothing special there.

For non-PnP card it is normal to show itself in device manager when it is removed actually. There is no generic way for OS to check the presense of such card or detect it's IO/IRQ params (they are specified by driver installer instead). It is driver's responsibility to check for card presense (in some card-specific way, read some "magic" value from some register usually), but it is optional (or the hardware is too simple for that).

That repeated "re-sending ABORT" can be the cause - something like no reply to some ABORT message from the remote side of RS485, but it is hard to say where is the problem - transmitter/receiver/cable/remote side disconnected/powered down/broken/etc. Observing RS485 lines with a scope/LA/or even a multimeter (is there any voltage change during this activity?) can show more.

I wonder why they didn't replaced this board with some USB-Serial converter (just another FTDI chip + ADM485)?

[rx8pilot]

I wonder why they didn't replaced this board with some USB-Serial converter (just another FTDI chip + ADM485)?

I think this is because they don't know what it is doing. It is paired with an industrial machine vision processor from the early 90's and it seems to be doing slightly more than just a simple RS-485 driver. It is buffering and/or manipulating the data in some way (or so it seems).

More troubleshooting notes that could be relevant. There was originally a PCI USB card in the chassis, but the system will not boot with it in there. Take it out, all is well, put it in - no boot. Also some strange (onboard) NIC behavior. The NIC's seem to load ok but when I plug in an ethernet cable and it tries to get an IP - the whole system freezes. Not a blue screen crash - it just freezes. No caps lock, mouse movement, etc requiring a reboot. If I don't plug in the Ethernet, it works fine. It could be related to a bus problem or not. On board USB works, COM1/2 work ok, and PS/2 mouse and KB work ok.

[Towger]

Might be time to track down another 'computer'.  Is the NIC card PRO or PCI?
Would you get away with a standard P4 board with PCI and ISA slots?

Edit.. Another thought.
Is it possible the card is working, but it not getting a response in turn, from whatever it is connected to.

[abyrvalg]

The NIC behavior can be a sign of resources conflict. Try two tests:

1. Remove as many removable cards as possible (including MV922 of course) then power up and connect the network cable - will it freeze the system as usual?

2. With all cards back disable the NIC in BIOS (is there such setting?), then run the pick&place app - no change?

[rx8pilot]

Might be time to track down another 'computer'.  Is the NIC card PRO or PCI?
Would you get away with a standard P4 board with PCI and ISA slots?

Not sure, the hardware configuration is pretty specific to the software. Hard to say....

Is it possible the card is working, but it not getting a response in turn, from whatever it is connected to.

It is possible. Based on the error messages I am truly guessing. I could hook up a scope to the RS485 RX and TX lines and see if any activity is there. If there is some Tx and no Rx, could be a clue.

The NIC behavior can be a sign of resources conflict. Try two tests:

1. Remove as many removable cards as possible (including MV922 of course) then power up and connect the network cable - will it freeze the system as usual?

2. With all cards back disable the NIC in BIOS (is there such setting?), then run the pick&place app - no change?

The MVS922 is the only card in the machine. All other functions are integrated in the single board PC. Graphics, Dual LAN, IDE controller, USB, Serial, Parallel........
I will dig through the BIOS to see if the NIC can be disabled. For the moment, I disabled them in the device manager but that does not seem to change anything yet.

[T3sl4co1l]

No idea what the later bus standards (PCI, etc.) do about it, but hard freezing is a traditional 'panic' mode in the IBM-PC when /NMI is pulled on the ISA bus.  This is usually reserved for "shit's broke" signals, and the usual BIOS ISR (interrupt service routine) is an infinite loop (JMP +00 or something like that).  So interrupts remain disabled, the processor doesn't read any memory but the BIOS that it's looping on, and peripherals are left alone.  Since the keyboard ISR isn't used, CTRL+ALT+DEL will not clear this fault, and a hard reset is required.

I would be surprised if this sort of signal and code path is still present, but it could be that something vaguely similar is happening.  The traditional "panic" of a 386+ user-kernel system (typical of *nix and Windows NT+) is an escalating processor fault (e.g., page fault, general protection fault (GPF)) that causes the kernel code itself to crash, triggering a reset (and on Windows, a BSoD, at least briefly before the reset vector is executed).

Could very well be that the driver is so poorly written that it doesn't play nicely with any of the systems integrated on the SBC.  I have no idea what driver or kernel programming is like in Windows, so... could very well be, they don't know either..

Tim

[poot36]

Try booting the computer off a linux or windows live cd to see if the network card works then.  If it does then there is a software issue that needs to be looked at.  Try running chkdsk /f /r and sfc /scannow (you will need a windows install cd for the sfc check) and see if that fixes the problem.  If that does not work you can try running tweaking.com's system repair tool but make sure to make a registry and system restore point (and maybe a image of the hard drive) before you run the repairs making sure that you click the select all button for the repairs.  If it still locks up the boot disk then there is either a bios configuration error or corruption or a hardware issue.  Record all of the bios settings and remove the bios battery and power to the computer and short out the battery holders contacts for 30 seconds then replace the battery and repower the system and enter in the bios settings again and see what happens.  Also check the voltage of the bios battery as well.  You can also try a bios update as some bioses hold the configuration data for the network card in them as well.  I have personal experience in a very low bios battery causing a industrial pc to not even power on and have also seen linux boot disks fix totally dead usb ports on a Dell xps system (the bios would not even see the usb keyboard, I measured the d+ and d- of all the ports on the system and they were hovering around 3.3V or so so they were not been initialized properly by the software in the bios but the linux boot disk was able to get it out of that weird situation).

[abyrvalg]

Old IO port manuals says that NE2000-style NICs can be configured to 240-25F port range - that overlaps 258 where your HS3L card lives (according to DebugView). I would recommend looking at NIC's settings carefully (is there any option in BIOS? check IO range in windows device manager also)

[Shock]

Disable any unused BIOS functions that may chew space on execution. e.g. PXE booting. That may cause failure to boot and a hang situations if some other BIOS function is compromised. Especially if there has been a defaults restore, backup battery failure, or board changed.

[Rasz]

try another P3 era computer

[rx8pilot]

I think this repair attempt has met the point of diminishing return. I replaced the majority of the logic, the RAM, and the RS485 drivers with no forward progress. The parts are cheap, but my time limit is probably up. Next stage is get one from the company that supports the machine at a high cost, but less than a useless pick and place machine. They will support the project only after I buy the card from them and they seem to have the secrets since they wrote the software and built the PC setup.

Arrggg......I was hoping to get lucky. Thank you all for the help.

[senso]

They have you hanged by the balls..
There is no support/warranty from the company that sold you the machine, or is it the same company that is now "selling" you the card for 500$?

Given the price and rarity of the machine and what makes it tick I would be very worried that you might have similar problems in a couple months and they will charge you thousands to apply duct-tape..

[Rasz]

I think this repair attempt has met the point of diminishing return. I replaced the majority of the logic, the RAM, and the RS485 drivers with no forward progress. The parts are cheap, but my time limit is probably up. Next stage is get one from the company that supports the machine at a high cost, but less than a useless pick and place machine. They will support the project only after I buy the card from them and they seem to have the secrets since they wrote the software and built the PC setup.

Arrggg......I was hoping to get lucky. Thank you all for the help.

look for a hackerspace in your vicinity, offer straight up cash/time on fixed machine HS donation for help, let people that do this sort of thing _for fun_ do their magic.

[coppice]

Have you checked the condition of the edge connectors? Its very common for the connectors on 90s era boards to be so brittle by now that they crack with barely a touch. The cracks aren't always obvious unless you look carefully, but can remove almost all pressure from the contacts.

[amyk]

ISA is slow enough to probe with a scope or LA... and since the card is mapped to a single port, capturing the communication to/from it should be relatively easy.

How many layers is the card? It shouldn't be that hard to RE and derive a schematic. The logic arrays (likely GAL/PAL) might be a bit more troublesome but it shouldn't be hard to figure out their possible function once the components connected to them are determined.

I have a feeling that the problem is NOT in the card itself.

It's not uncommon for companies like this to lead you on a wild goose chase...

[rx8pilot]

Have you checked the condition of the edge connectors? Its very common for the connectors on 90s era boards to be so brittle by now that they crack with barely a touch. The cracks aren't always obvious unless you look carefully, but can remove almost all pressure from the contacts.

I looked at them under a microscope early in the investigation and did not see anything unusual. The traces coming from the gold fingers are senselessly small though. A tiny crack could be beyond what my low power scope can see. I should check again with a higher power.

ISA is slow enough to probe with a scope or LA... and since the card is mapped to a single port, capturing the communication to/from it should be relatively easy.

How many layers is the card? It shouldn't be that hard to RE and derive a schematic. The logic arrays (likely GAL/PAL) might be a bit more troublesome but it shouldn't be hard to figure out their possible function once the components connected to them are determined.

I have a feeling that the problem is NOT in the card itself.

It's not uncommon for companies like this to lead you on a wild goose chase...

It may or may not be the card. The investigation of the card and the PC have not revealed a smoking gun yet. The software and drivers are as closed as possible and no information is publicly available. As much as it sucks, this is how companies stay in business - they have secret knowledge that someone else needs to be able to keep their business going. I would not fault PPM for operating this way, but I would prefer to stay away from closed systems as much as possible. In this case, I got a fixer upper Pick and Place machine for $5k. If I have to put a few $k into fixing it, so be it. The reality is that it will make it value make each and every month - and PPM knows that. I could have chosen a $5k low-performance TM-240a or spent north of $50k on various low-end NEW vision based systems out there. The Quad, if working, will be great and in all likelihood will be fully functioning for under $10k.

Not too discouraged......yet. ;-)

[krivx]

Have you checked the condition of the edge connectors? Its very common for the connectors on 90s era boards to be so brittle by now that they crack with barely a touch. The cracks aren't always obvious unless you look carefully, but can remove almost all pressure from the contacts.

I looked at them under a microscope early in the investigation and did not see anything unusual. The traces coming from the gold fingers are senselessly small though. A tiny crack could be beyond what my low power scope can see. I should check again with a higher power.

I wouldn't bother. Plug the card into the mainboard and check for continuity between vias/pins on both the card and the mainboard. Should take 10-20 minutes and will be much more reliable than a visual inspection.

[rx8pilot]

SOLVED!

It was indeed a connection issue, just not with the ISA card at all, not an OS issue, not a BIOS issue. I re-seated the SBC on its back plane along with the RAM - FIXED. I am so embarrassed I did not try that earlier. Nothing was ever really broken - it was just not connected.

I went down many paths looking for the root cause in the most complicated places. Software, bugs, replacing chips, etc. In the end, the solution was about 30 seconds and free. 

[PA0PBZ]

SOLVED!

 

[krivx]

SOLVED!

I went down many paths looking for the root cause in the most complicated places. Software, bugs, replacing chips, etc. In the end, the solution was about 30 seconds and free. 

Hey, you could have forked out money for replacement parts that wouldn't have helped, this sounds like a victory to me.

[Towger]

Good to hear. Might be worth while cleaning all the motherboard connectors etc. You now have two of these cards, are there many others required to build a second backup computer, using a 'standard' ISA/PCI motherboard.
BT any chance of more photos?

[rx8pilot]

Here are a few pics from the effort. I replaced a bunch of the chips hoping to get lucky. Fortunately, those are cheap and easy. As pickle9000 suggested, I put all of them in sockets except for the RS485 drivers. It reminded me of how much I hate through hole boards. Most of my own PCB's don't have a single hole expect for vias. Super easy to replace parts compared to these things.

The failure was not obvious to me because the symptoms were that some thing sort of worked. It seemed like maybe some sort of software related corruption.

Started with the RAM chips....being careful to get clean holes and not lift any pads.


Here you can see some of the new chips in sockets. The RS485 drivers are the 8pin DIPs next to the DB9 connector.


PC layout. Single board computer on a backplane with ISA slots on one side and PCI on the other.



The bad connection from the SBC to the backplane created a fault with the PCI USB card, the ISA bus, and the on-board network interface which is also on the PCI bus.

And yes, I have a verified working spare vision card. I also have an option to buy a spare SBC with processor and RAM rather cheap. Right now, the HDD is being cloned and imaged. Converting to SSD for main drive as well so I should have some options if things get crazy in the future.

[wagon]

Well done!  If I had a dollar for every complex fault that had a simple fix (or the cause was yelling out at me to be fixed) I'd be a rich (OK, less poor) man!