VOGONS


First post, by mita

User metadata
Rank Newbie
Rank
Newbie

Hi,

Try to revive a GNB-486UNL V1.1 board. Powers are OK, clock is ok, reset ok but no sign of code execution on the post diag card. Checked the bios and found the CE - output enable pin short to ground. It seems this pin is connected on the internal layer there is no trace on top or bottom layer on this pin. I am unable to find where this pin is connected. Checked other 486 boards and none of them has shorted CE BIOS pins. I believe the BIOS is controlled by the chipset which is a Toshiba 4800 or the other controller a SIS 85c206. I was not able to find datasheets for these chips, if somebody has access for these please reply.

If somebody has suggestion what to check next please share with me.

Thank you.

Reply 1 of 5, by BitWrangler

User metadata
Rank l33t++
Rank
l33t++

You could flip the pin out, tie it high and see if everything seems good after, in which case if it works perfect then, it's a bit of a wild goose chase to find a problem possibly buried between layers.

Unicorn herding operations are proceeding, but all the totes of hens teeth and barrels of rocking horse poop give them plenty of hiding spots.

Reply 2 of 5, by mita

User metadata
Rank Newbie
Rank
Newbie

Thank you your reply! The problem is shorting low the OE results the 8 bit data bus of the BIOS drives the data bus lines all the time. It will conflict with the controller when the controller wants to drive the data lines. If the OE will be put high all the time than the BIOS data lines will be in high impedance state, no way to get code from it. But maybe I am wrong. I tried to find a 486 MB schematic to understand how the BIOS is controlled regardless of the chipset but I was not able to find any.

Reply 3 of 5, by mita

User metadata
Rank Newbie
Rank
Newbie

My bad.... The BIOS has two pins for enable: chip select and output enable. This particular circuit use only one control not two. FInally the board came to life.
Conclusion:
In case of battery damage board the precise and full cleaning is mandatory. Use dremmel with rubber tip or fiber glass pen to clear the coating from the traces if you see any small point of corrosion. You never know when will you find a broken trace under a small spot.
Clean all the corroded vias as well. Repair them if necessary. Check the continuity of the traces with via.
5-10cm2 area around the leaked battery should be clean. In case of heavily corroded boards more.
Check the outer edge of the board with long traces next to the board edge. Inspect them via microscope. In my case I found a very long trace on the component! side of the board with a small spot turned out a micro break. I figured it out by flexing the board a bit and do the reset. Bang! the board started to post.
One more note: do not jump into a deep hole overanalyzing the post codes. Post codes can point to a given direction suspecting one element, like a keyboar controller. If you do not have full documentation about the chipset - most of the case it is true - than do not spend too much time measuring signals because it is unlikely to find the culprit. I think most of the case there is a broken wire somewhere. If it is not blocked by a component than you can get result quicker to check the traces via microscope.

One more conclusion - correct me if I am wrong - :

It is interesting that it is no way to find a 386 or 486 pc schematic diagram regardles of the chipset. Chipset documentation is very rare as well. Because of these the repair is very difficult and time consuming.

Just one example, a simple problem:
Board is not starting at all at power on. But starts on reset. This part of the circuity should be quite common and almost the same for most of the board. But measuring the power good signal and the reset signal is a real pain. Add to the nonexistent chipset documentation and we are dead: do not know the power on sequence the conditions when the chipest will find these startup conditions eligible to startthe CPU.
If somebody found a solution for this problem please share with me - and the viewers -.

Thank you!

Reply 4 of 5, by BitWrangler

User metadata
Rank l33t++
Rank
l33t++
mita wrote on 2024-03-05, 17:00:

One more note: do not jump into a deep hole overanalyzing the post codes. Post codes can point to a given direction suspecting one element, like a keyboar controller. If you do not have full documentation about the chipset - most of the case it is true - than do not spend too much time measuring signals because it is unlikely to find the culprit. I think most of the case there is a broken wire somewhere. If it is not blocked by a component than you can get result quicker to check the traces via microscope.

Yes, POST codes can be quite precise, when you have the reference for the exact BIOS version for the exact board from the manufacturer, but that only happens about one time in twenty or less. Frequently you can't even find the code table for the specific revision of the BIOS, and many manufacturers modify them a lot, so you are comparing between a couple of adjacent code tables. At that point it's more a Magic 8 ball kind of indicator "signs point to maybe..." and you have to take a broader "something probably to do with X" kind of view. Of more use sometimes is knowing which parts it has passed, so you can eliminate things to look at. A lot of times it's "wiggle stuff, make number go higher" since bad connections on old hardware are very common.

Optical inspection is often neglected, really detailed inspection can find stuff a lot quicker than probing it, as you say. A particular trick is to make use of light reflections at different angles, you see a line of pins on the edge of a QFP chip, shine the light off them, they all shine bright and one is dull, because it's slightly tweaked, it has lifted off the board, or is twisted and fractured or something. Then also reading the shine off packages can be useful, the surface can sometimes tell you whether it has burned out, by a duller brown reflection when light hits at an angle, where as straight on, it looks fine, just black. Sometimes this is in color or texture, but oblique light shows it up. Bad solder joints often show dull also.

Temperature is also a useful tool, in the 90s I only had finger prodding, then IR thermometers got cheap, but have a large cone of detection so gotta point it near and accurately, some guys have got thermal cameras to help them out, but I haven't sprung for one yet. Anyway, if things are getting too hot too quick, they either have a problem, like wedged on DC, no clocked signal, or their support capacitors etc for oscillation have failed or gone out of spec. So you backtrack from there or suspect the device if it's all good upstream. Then also you can use freezer spray or inverted can of air for "would it work if it was colder" to see if semiconductors are still working at all, or heat gun for "would it work if it was hotter" mainly useful for marginal caps. Also "frosting" the board can let you see if some devices are completely inactive, making no heat, or some are making heat way too quickly, by where it melts off real quick or doesn't melt off at all.... you don't want to do this on high humidity days though or you will have too much water condensation.

Anyway, those are relatively fast ways to hunt out problems before you hook up test leads and start obsessing over exact voltages and ohms readings. That is, as if you say, any references are available. Sometimes service manuals are a mixed blessing... the procedures are designed to find faults in normal failure modes at EOL of the most likely to fail components. If however something else happened, End Of Life components already replaced, and you're getting into decade plus long term component wear that was never anticipated, or something as simple as mechanical (or chemical/corrosion) damage, that was a little too subtle to spot on an inspection, then those procedures will have you going round and round in circles... yes, they are from the manufacturer, but it's not the entire scope of everything that could ever go wrong.... just the low hanging fruit. So maybe go through them carefully once, then forget it, they don't apply, you are off script, do the things you do when you haven't got a manual.

Unicorn herding operations are proceeding, but all the totes of hens teeth and barrels of rocking horse poop give them plenty of hiding spots.

Reply 5 of 5, by mita

User metadata
Rank Newbie
Rank
Newbie

Thank you your post! Lot of things to consider.

An important update! I found a freaking schematic of a 486 PC DIY. I do not know wether this porject is finished or not but a lot of effort were made to kick it on. FInally have a canche to study the whole system imcluding the chipset's co-operation with the processor. There is a chipset documentation! PC AT reference guide! I am very happy camper now!
https://www.retrobrewcomputers.org/doku.php?i … linux-sbc:start

Happy digging.