VOGONS

DTK PKM-0038S persistent keyboard error

Topic actions

First post, by majestyk

Posted on 
User metadata
Rank Oldbie
Rank
Oldbie

I got this nice SiS 85C471 based Socket 3 mainboard with some Varta leakage damages recently and decided it was worth being repaired. It was sold as "DTK PKM-0038S" and Gemlight "GMB-486SG" (mine is revision 2.2).

pkm_0038s_tot.JPG

So I removed all the components (KB controller socket, KB connector, headers, resistors and inductors) in this area, cleaned ecerything carefully, restored the interrupted traces, then checked all traces and vias very carefully for continuity and soldered all components back into place.

The system is starting now but hangs during POST with "CMOS checksum error - Defaults loaded" and you could "DEL" to enter setup or "F1" to continue, but there´s no reaction to keyboard entries at all. The POST-card displays "4E 4C" or "43 4D" in this situation.
When there´s no KB connected there´s a "no keyboard present" message. At startup all 3 LEDs of the KB flash twice.

I know there have been several topics about similar errors here on Vogons due to battery leakage and usually the reason is that one more interrupted trace or via had not been found yet.
But in this case I´m quite sure I have repaired everything.

pkm_0038s_tr.JPG

So far I tried with different BIOS images and different keyboard controller chips - to no avail.
I then checked the oscillator signal at the 8042 chip (pins 2 and 3) and found it was unstable. When the signal was missing there appeared a second message about keyboard error. The chip has it´s own 8 MHz crystal and because I don´t have one at hand at the moment I hooked pin 3 to the AT-clock signal and grounded pin 2.

8042_ho.JPG

I have started to take a closer look at the voltages / signals at the 8042. Vdd and Vss are present of course. The reset voltage rises to 5V when the system is turned on, then drops for a split second and is back to 5V right away.
I also checked all the connections between pins 1, 39, 37 and 38 to the inverters / buffers and then to the KB-connector. I even replaced the involved 7406 and 7404 chips (located between ISA slots 2 and 3).
The data-bus lines D0 - D7 are all connected to the BIOS chip and to the respective pins of the 85C407 and 85C471. I can see data signals with the oscilloscope on all of them.
CS# is connected to pin 20 (E#) of the BIOS chip and to the chipset (85C471) pin 87 KBROMCS#. It´s at high level.
KB_INH# is high as it should be.
RD# and WR# go to the chipset (95C471) pins 208 and 207 (IOW# and IOR#)
A0 goes to BIOS chip pin 8 (A2) and ISA slots pin A29 Address Line 2

Some of the remaining pins are unused or reserved for PS2-mode that is not being used here.

At the moment I´m not sure if the KBC fails at self-test or if any of the signals are wrong or not connected?

Reply 1 of 2, by majestyk

Posted on 
User metadata
Rank Oldbie
Rank
Oldbie

I was so fixated on finding the hidden damage the battery-leak had done, that I was blind to other possible damages.

When I tried to wipe away a piece of dandruff / dirt that seemed to be stuck under the BIOS DIL-socket, it wouldn´t go away. It had been well covered all the time in the shadow and could hardly be seen while a BIOS- chip was inserted.

pkm_0038s_sp.JPG

So I removed the 28-pin socket and found this:

pkm_0038s_sp1.JPG

I bet someone slipped under the socket with a screwdiver while removing the chip.

This was quickly repaired...

pkm_0038s_sp2.JPG

...aaaand we have a running system again.

IMG_1244_fin.JPG

I´m going to replace the 8 MHz oscillator for ther KBC now and I´ll check where the cut trace leads to.

Btw. - the original DTK BIOS has the "year 2094" bug.

Reply 2 of 2, by majestyk

Posted on 
User metadata
Rank Oldbie
Rank
Oldbie

There´s a 5x86 running now on this mainboard at 40 x 4 MHz. Therefore I flashed Jan´s 5x86 BIOS he uploaded for the "ABIT AB-AH4T" thread. The DTK/Gemlight BIOS is unable to support this CPU correctly although the Gemlight manual provides (most of ) the necessary jumper settings.
The Speedsys benchmarks look quite normal now:

PKM-0038S_sps.JPG

Write Back L1 is enabled, the connections for the "HITM", "CACHE" and "INV" lines are set correctly.
But then there´s this "Dirty TAG L2 N/A" issue.

PKM-0038S_ctcm.JPG

In the AWARD SiS 471 chipset BIOSes there´s no option to split the 8 TAG-lines into 7+1 (1 for dirty TAG). So I assume either BIOS should take care of the correct setting automatically or there´s still some jumper missing or incorrect. How could I find out what´s responsible here?
On the Abit AH4 and otheres this issue is not present with the same BIOS.

Edit: The CTCM output was from CTCM 1.5. For fun I tried CTCM 1.7:

PKM-0038S_ctcm7.JPG

So everything´s o.k. after all? (On the other SiS471 boards the results were consistent afair.)

Go to top of page Go to top of page
Back to General Old Hardware
Close window
Close window