Quick reply Here : I just got my new SMD rework station. I am willing to test the idea of swapping the South Bridge.
If anyone has an IO card using the same chipset, send it to me, I'll post results.

I/O cards do not contain the southbridge - they only contain the Super I/O, which in my case did not correct the problem.

feipoa wrote:

I/O cards do not contain the southbridge - they only contain the Super I/O, which in my case did not correct the problem.

Ah !
Ok, now I understand.

Weird. How's that south bridge connected to the rest ? How can it still function for other things except COM and floppy ? That is one weird thing.

I am open to suggestions...

Maybe one or more broken traces on the board itself?

yawetaG wrote:

Maybe one or more broken traces on the board itself?

Sure, need to identify which.

Nobody has the schematics.

Basic thing would be to check which pin of the south bridge is supposed to go to the IO chip, then check for continuity. But witouth schematics or datasheet, tha't going to be pretty hard.

Synoptic wrote:

Sure, need to identify which. […]
Show full quote

yawetaG wrote:

Maybe one or more broken traces on the board itself?

Sure, need to identify which.

Nobody has the schematics.

Basic thing would be to check which pin of the south bridge is supposed to go to the IO chip, then check for continuity. But witouth schematics or datasheet, tha't going to be pretty hard.

All I would be able to do is verify if any pin on the Super I/O goes to any pin on the Southbridge, assuming it goes through a series resistance of less than ~50 ohms, or is a direct connect.

What I can do is verify which Super I/O pins go to the Southbridge on a known working board, then cross reference this knowledge to the defective board.

It sounds like you have an ISA card with the same I/O chip? In this case it should be rather easy to find out what each pin does. Many will go to the ISA bus, some are for the jumpers. The rest is for the I/O function. It would be interesting what the voltage levels on the pins for the jumpers do (do they match the card?). And.. is the replacement chip the same variant as the original (UM8663BF vs i.e. UM8663A)? Since this fine chip manufacturer doesn't release datasheets we can't know if there are important differences.

By comparing the connection between the Super I/O on a known good motherboard to that of the fault board, I determined that 32 pins on the Super I/O lead to the Southbridge. I matched up which pin # on the Super I/O went to which pin # on the Southbridge and determined that two pins weren't being connected. I have no doubt that the solder pads on the Super I/O and Southbridge are conductivity connected to the pads, so the reason for this was puzzling. I rewired these two traces using some thin wire, yet the floppy drive still does not work.

The fact that two traces were not connecting between these two ICs implies to me that there may be more somewhere else which will be extremely difficult to determine. I am inching more and more towards replacing the SouthBridge. I have 4 NOS spares.

To answer your question - yes, I replaced the Super I/O with the same sub-revision, that is, UM8663BF with UM8663BF (not *AF). The Southbridge chips I have are also of the same sub-revision - UM8886BF, not *AF. However, the datecode on the faulty board is 9640-FXA, while the replacement chips are 9530-DYT. I recall one or two accounts online in the past that the alphabetical suffix may have something to do with sub-revisions of the IC, with the further down the alpahbet of the first letter being newer revisions. So, I'd be replacing the *F with a *D revision. I'm not really sure what truth there is to this.

To answer another question - I do have an ISA card with a similar Super I/O chip and serial port IC on it, although the ISA card has UM8663AF rather than BF. The serial controller IC is the same - UM8667. I am not sure how much I can gain from this I/O card except that it is obvious that each pin on the Super I/O must go to either the onboard crystal osc, a jumper, a header, a resistor pack, a cap, diode, resistor, or the ISA bus pins. The most glaring piece of information I can gain from this piece of hardware is that for the SuperI/O to communicate with the Southbridge, the traces route through the ISA slot, which will make soldering easier.

In the mist of cutting off the Southbridge, I have been thinking about this a little further and ask myself: If using an ISA floppy disk controller card (UM8663AF) works with floppy access, does that imply that the Southbridge is working fine? Does plugging in the ISA card, in effect, fix a broken trace on the motherboard's PCB? I mean, I am not disabling the FDC controller in the motherboard's BIOS.

The curious aspect of using the ISA card is that the card's PCB contains its own crystal oscillator. Could the MB trace to the Super I/O's (UM8663BF) clock signal be broken? Using the ISA card as a guide, I believe I have determined which pin is the OSC input pin (pin 20). This pin on the motherboard is connected to a pin on the MB's PLL IC on, both, the working and defective motherboards. I have already tried swapping the PLL chip, so that isn't the issue.

I traced each of the MB's Super I/O chip's pins to the ISA bus and compared it with the defective board. There result was that all of these traces are intact.

I am now at a standstill and am not more than 50% sure that the Southbridge is the culprit. Where to go from here?

Basically you say that since the ISA IO card uses the southbridge from the motherboard, why the ISA io card fixes the floppy issue. Don't forget that in my case, the COM ports are not registering either. I think we should find a cause that explains both failure. Do COM ports need an OSC too ?

The Southbridge is connected to the RTC. If you look at a board where the RTC is not present, you see the traces going there : http://i.imgur.com/fwqTVSX.jpg

Maybe putting in an ISA card, bypass the RTC failed connection ?

Errius wrote:

This touches a nerve. I recently killed the floppy controller on a nice old mobo after pulling out and reseating the memory modules. There was no need to do this, I just wanted to see if a particular module worked (it did). But just doing this killed the floppy controller, probably due to flexing of the board. (The original modules probably hadn't been removed for 20 years and took some pulling to get out.) A DOS machine without a floppy drive is useless so I threw out the board and got a replacement. Fortunately it's a common model so this wasn't too hard/expensive. Lesson learned though: if an old computer is running OK leave it alone.

Hear Hear. Took a apart a tandy sensation to get at the bios battery. Now it just freezes once booting into the tendy dos gui. Was fine before that.

synoptic - I too do not have serial/COM access. There is something more going on here. But the serial port chip, UM8667, connects to the Super I/O. I plan on tracing out all the pins of the Super I/O just to ensure that they all are connected properly. Once I rule that out, and I will be stuck again.

> Does plugging in the ISA card, in effect, fix a broken trace on the motherboard's PCB? I mean, I am not disabling the FDC controller in the motherboard's BIOS.
The onboard floppy controller is completely absent in the address space and thus an external one can take its place.

> I am not sure how much I can gain from this I/O card
You can check i.e. which pin on the chip is responsible for the floppy enable jumper. Then check which voltage level it has when enabled/disabled. Check where this pin goes to on the mainboard. Check the voltage it has when the mainboard is on. Finally, you can try to disconnect it and override with the 'on' level.
Either these "jumper pins" go to GPIO ports on the chipset or are not connected. In the latter case the I/O chip is software configured.

Errius wrote:

This touches a nerve. I recently killed the floppy controller on a nice old mobo after pulling out and reseating the memory modules. There was no need to do this, I just wanted to see if a particular module worked (it did). But just doing this killed the floppy controller, probably due to flexing of the board. (The original modules probably hadn't been removed for 20 years and took some pulling to get out.) A DOS machine without a floppy drive is useless so I threw out the board and got a replacement. Fortunately it's a common model so this wasn't too hard/expensive. Lesson learned though: if an old computer is running OK leave it alone.

Some boards have very...tough memory slots that seem to be too tightly made no matter what modules you insert or extract from them, but I think it's a good habit to use one hand to keep the board from flexing when inserting or removing memory modules and also other pluggable things like expansion cards and data/power cables.

Flexing such old boards can't be a good thing and if it's difficult to insert or remove something, I take an extra breath of air and take aaaal the time I need, do things gently and in ehh...straight angles.

Just a couple of thoughts...

I would leave the southbridge intact. The IO chip sits on the ISA bus so if the ISA works, the southbridge is not to blame.

Can you make sure that you have all the voltages necessary to power the IO chip in place (this should be a voltage, 5V) on IO chip pins? Also check passives around (resistors/caps) for short/no circuit.

It might also be a good idea to check all output circuits that leave the IO chip for shorts/no connections, especially the COM ports, as there should be buffer chips there, and if they are dead, they might kill the IO chip (COM voltage span IIRC is 24V, from -12 to +12, this is more than enough to fry the IO chip dead). But any output signal that is short to ground or +5V can cause the IO chip to malfunction.

The only components that tend to fail spontaneously that I know of are tantalum caps. Usually very stable, but after 20 or more years one of them may blow up. However, if this happened you would probably notice, although I've never seen SMD tantalum caps fail, so I have no idea if they fail as spectacularly as the non-SMD variants.

There could also be some kind of minuscule metallic whisker causing havoc somewhere, and if that's the case it will be very hard to track down without a full service manual.

Yeahhh, I'd verify the connectivity and traces on that chip. It looks disgusting.

Thanks to everyone for their response.

h-a-l-9000: I do not know which pin on the Super I/O is the enable disable pin. And does the Southbridge control this? The datasheets for, both, the southbridge and superio are not available.

CelGen: The superio chip is connected properly. It has been check and rechecked. I have ensured that all the pins go to the same location on a known working board of the same model.

devius: I had checked several of the SMD caps, but I suppose it would not hurt to check them all for a short circuit. But if they fail, would they normally go open circuit? And if so, would there be reliability issues rather than complete failure?

devius: I have had issues with tiny metallic whiskers, but it is very rare. And that only occured because I was dragging my multimeter probe around pins too much, which caused some flaking.

GPA: I am going to leave the southbridge in tack for now. When I have the motivation to work on this again, which I will, I will try to map out the whole Super I/O's connectivity on this board.

GPA: It will be a lot more time consuming to check all the SMD resistors, resistor packs, and SMD caps for OC and SC. I am not really convinced the issue is there, but it may come to that at some point. I have ensured that the same 5V connetions on a known working board go to the same SuperI/O pins on the damaged board. I do not know which pins are I/O on the Super I/O. There is no datasheet. To probe every single pin on the Super I/O with a scope would be too time consuming. It would also require I solder a wire to each pin.

GPA: Could the buffers be built into the Super I/O? I am pretty sure that none of the pins go to the TTL DIP package buffers on the board, but it is something I will check again when I complete the wiring diagram for the Super I/O.

GPA: Also, if I suspected input circuits entering the superio for shorts and open circuits, how do you know that some pins aren't intentionally set to 5V, some set to GND, some are N/C? Without the datasheet, I am continually guessing. But as noted, the Super I/O as soldered, is connected to the correct locations.

I did have one more idea which seems like a long shot, but perhaps the serial chip, which connects directly to the superio, is malfunctioning and this is causing the superio to malfunction. I do have some spare serial chips of the same model I can solder in place.

feipoa wrote:

Thanks to everyone for their response.
GPA: Also, if I suspected input circuits entering the superio for shorts and open circuits, how do you know that some pins aren't intentionally set to 5V, some set to GND, some are N/C? Without the datasheet, I am continually guessing. But as noted, the Super I/O as soldered, is connected to the correct locations.

I think the only way to do it is to match the connection on the mother board versus the connections on ISA IO card. I don't see any other method to be honest.

if someone has any schematics for any UM8881/6 board that would definitely help, every board based on that chipset that I've seen uses same UM8663/7 for IO. I've tried googling HOT-433 schematics but couldn't find anything useful.

Judging by the layout though it looks like it is the UM8667 that implements physical COM port interface, so it must be this chip that buffers COM ports too. This is definitely a candidate to blame in that case.

There are not so many passives around these 2 chips though, you can check them, I see less than a dozen in UM8663/7 area. Of course you won't be able to test caps for their capacity while they are soldered on board, but a check for short circuit is still possible. Resistors are even easier, although don't trust the nominal you'll see - if they are in parallel connection with any other resistors, your multimeter would not show correct values. It still won't stop you from general inspection for short/open circuit, but SMD resistors rarely fail if they are not used as fuses (fuse resistors would usually be low Ohm resistors, and not in parallel, so easy to check).

There are also 3 diodes that I can see on the board, but I am not sure they are related to COM ports, although they may serve as wrong polarity protection. should be easy to check as well.

Having said that I do realize that this is not going to be a very fun thing to do, but, thinking logics:

1. ISA works -> South Bridge provides all the necessary connectivity.
2. You have checked that connection between IO and SB works. No issues here.
3. COM ports don't work -> IO chip doesn't work
4. Why can the IO chip not work? 3 reasons: (a) no power, (b) short circuit inside the chip, (c) chip is no initialized.

you can easily check (a) since you have a working ISA card, just buy manually checking every pin of IO chip on ISA card for continuity versus ISA's +5V and GROUND pins on the card.

AT PSU lets you easily check (b). Since AT power supply provides all voltages while it is on, a chip that is internally in short circuit will heat up intensively. Just let it run for a minute or two and touch the chip. It will be extremely hot if shorted. But since the board has been working like that for quite a while, if there would be a short circuit inside a chip, you would probably see that, visually. And could smell it.
(c) can be caused by the fact that chip shuts itself down or provides corrupt/no output due to some damaged circuit on its output. You can try desoldering the UM8667 and try running that DEBUG commands described earlier in this thread. If it was 8667 causing the 8663 not to initialize, the problem could be solved. Or not, depending on how the 8663 works. But since 8667 is one of the candidates to blame, you will still have to replace it. So why not running a test?

Another good idea would be check SMD capacitors for short circuit around the LPT connector (if I recognize it correctly, the CN6 is an LPT, but in this case, I do not see a floppy connector!). This should not take long.

One more thing. If your ISA IO card uses same UM8667 chip for buffering, you can test its pins for continuity versus ISA +12V/-12V pins, I am pretty sure it should use them to drive high voltage span of COM port. In this case you can see if the UM8667 is powered correctly on your motherboard. Who knows, maybe it doesn't receive all the voltages it need and doesn't start.