mkarcher wrote on 2024-08-09, 22:07:

mita wrote on 2024-08-09, 21:13:

I can not close out bad ram chips but the thermo camera showed Uniform Thermal image of the ram chips. I suspect a bad chip shoud have a bit different thermo signature but it is not 100%.

The mb bios should be disassembled to.get the answer of the.cause of the 8 beep.

8 beeps is obviously borrowed from AMI. The AMI BIOS expects a CGA compatible or an MDA compatible video card. This might either be an actual CGA or MDA card, or an EGA/VGA card intialized to test-mode 0-3 (CGA compatible) or mode 7 (MDA compatible). They typical check is to read/write the CRTC registers for the cursor position (on of the few 6845 registers that were not write-only) and read/write some bytes of video memory. If this fails for both the MDA compatible addresses (3B4 / B000) and CGA compatible addresses (3D4 / B800), the 8 beeps will be sounded.

Thank you the info! THis should point to the check of the ET3000 chip's interface to the ISA bus. Because no Et3000 documentation one way would be to check the ET4000 documentation. With a little luck the interface section would be similiar enough to be able to figure out how it is implemented in case of the ET3000.

Started to read the following document's section 4 to understand how the MB BIOS initializes and checks a subsystem like VGA:
http://bitsavers.trailing-edge.com/pdf/ibm/pc … rence_Apr87.pdf

Dorunkāku wrote on 2024-08-09, 22:08:

Mita have you seen this post?
https://dosreloaded.de/forum/thread/4791-matt … 4760#post154760
It mentions bad electrolytic capacitors on ET3000's.

Thank you the link, it is a really interesting blog for DIY guys. In my case the capacitors seems to be in good shape, but it is worth to be sure, I will replace them. I have already replaced the program BIOS's socket because of corrosion. I found the VCC, GND pins connected to the internal layer which is a huge heat sink. I was able to desolder these pins only using heat gun and the soldering tip together.

Recapped the board, produces the same error.
Started to draw the schematics of the ISA connector bus interface part of the board to understood the ET3000 a bit better and to be able to compare to the ET4000 schematics. I hoped they are similiar but it is not: the ET4000 bus interface is quite straight forward, everything is understandable and checkable. On the other hand the ET3000 uses a PAL which is a black horse for debugging.
The ISA buses are handled by 2xLS245 for 16 bit data bus and 2xLS244 for address bus. ET4000 is using a muliplexer scheme to process the address-data bus to the same pins using the LS245-LS244 drivers. The output enable and direction pins of the LS245 and LS244 should be controlled tightly in order to avoid bus collision. I hope the ET3000 is using the same bus handling mechanism. I found one of the Ls245 output enable pin is high all time leaving half of the data bus floating dead. This definitely explains the 8 beeps.
The OE pin is controlled by the PAL IC. That was the point when decided to draw the schematic to see what the PAL is controlling from what input signals.

Finally cracked this beast!

Couple of observations:
- do not shoot the sparow with cannons. Before sw debugging make sure the HW is ok, every connection is connected. Hardware corrosion in the vias or in a through hole could cause a trace break practically invisible. Only the continuity measurement can pinpoint this kind of problem.
- for old hw the primary suspect is the corrosion by the MB battery or/and caps on the card. Save yourself a lot of time and replace the caps. When I desoldered the first cap I got a clear message about the corrosion based on the bottom of the cap. It was awfull. Do not fooled getting a clean board from somebody. The truth is below the caps on the component side.
- if the caps are leaked than pay extra attention to the traces around the caps. Grind the solder mask off from the traces and check with the microscope. Sometimes, like in my case it was not enough. I suggest to grind the soldermask off from the traces at the ISA bus connector. Measure them for continuity, they tend to break when they changes 45degree of direction.
- If nothing is helping than do the last resort: decode the schematic. It has an advantage: the net`s continuity is checked. That was the way I found two internally corroded through hole break and one break at the ISA connector. It takes a lot of time and patience, prepare yourself!
- regarding the ET3000 schematic: it is completely different from the ET4000 based system. The bus multiplexing is different. I decoded the ISA-LS244-LS255 connections but the other sides of the drivers are a mess. Difficult to decode because of the vias under the chips. I was lucky to have breaks where I was able to trace.
- do not be fooled by the live signals on the eprom socket: the internal data bus are multiplexed and it can hide a broken trace floating.

Again, thank you all tried to help me in this process! Lessons learned.

As usual Murphy works!

After cleaning the flux residue, cover the bare traces with green UV soldermask I tried the card again. To my surprise the screen showed only garbage! Murphy is knocing in.

ET 3000/ET4000 has high and low side dram Write signals. One of them showed constant high. Checked the connection to the ET3000. I found 3 loose pins. At first inspetion I checked all the pins of the ET3k and all the pins were solid. The component soldering step of the manufacturing process was not qute right. 3 side of the chip have solid soldering but one side soldered with minimal solder. It seems like the pin is only glued to the pad, no visible sign of the solder material. Using toothbrush for cleaning was enough to break some of the pins from the pad.

After soldering the loose pins the card worked again. Lessons learned.

mita wrote on 2024-08-15, 11:20:

As usual Murphy works! […]
Show full quote

As usual Murphy works!

After cleaning the flux residue, cover the bare traces with green UV soldermask I tried the card again. To my surprise the screen showed only garbage! Murphy is knocing in.

ET 3000/ET4000 has high and low side dram Write signals. One of them showed constant high. Checked the connection to the ET3000. I found 3 loose pins. At first inspetion I checked all the pins of the ET3k and all the pins were solid. The component soldering step of the manufacturing process was not qute right. 3 side of the chip have solid soldering but one side soldered with minimal solder. It seems like the pin is only glued to the pad, no visible sign of the solder material. Using toothbrush for cleaning was enough to break some of the pins from the pad.

After soldering the loose pins the card worked again. Lessons learned.

Could you run tvdiag (mentioned at my ET3000 thread) on it? I'm really curious how it's 8 font mode is handled. Does it display the fonts in jumbled order? Or a good 01234567 on the screen? What happens to the background bits (bits 3-6 and perhaps bit 7)? Are they forced to 0? What about blinking?
The ET3000 thread is Re: Real ET3000 8 simultaneous fonts effects on attribute controller?
It has a link to the Excelgraph post where I have a link and images from my emulator rendering it's 8 simultaneous font mode (before disabling the attribute bits 3-6). Of course report the results on my thread, if you can get it running.
I can't find any documentation how the real card handles that mode, other than the character glyph mapping.

Unfortunately I can not test the card anymore, I repaired it for my friend who was happy to extend his video card collection. Anyway I will ask him to test it, we will see.

Meantime I go back the card because it had some issues again. As it turned out some of the pins of the ET3000 had cold soldered or cracked. After the repair I was able to run the tvdiag app, see the result attached.

mita wrote on 2024-10-08, 16:39:

Meantime I go back the card because it had some issues again. As it turned out some of the pins of the ET3000 had cold soldered or cracked. After the repair I was able to run the tvdiag app, see the result attached.

Interesting the dot below the "4" character. Could it be a ROM error?

mita wrote on 2024-10-08, 16:40:

mita wrote on 2024-10-08, 16:39:

Meantime I go back the card because it had some issues again. As it turned out some of the pins of the ET3000 had cold soldered or cracked. After the repair I was able to run the tvdiag app, see the result attached.

Interesting the dot below the "4" character. Could it be a ROM error?

Either that or a stuck 1-bit in plane 2 of video memory probably. Perhaps run CheckIt. It will test video RAM (just to be sure) for such errors (up to 256K(262144 bytes) of memory).

superfury wrote on 2024-10-08, 18:14:

Either that or a stuck 1-bit in plane 2 of video memory probably. Perhaps run CheckIt. It will test video RAM (just to be sure) for such errors (up to 256K(262144 bytes) of memory).

Checkit 3.0 only performs quick video memory test, which may fail to detect subtle failures.
Checkit Pro and X-VESA are much more compherensive.

superfury wrote on 2024-10-08, 18:14:

mita wrote on 2024-10-08, 16:40:

mita wrote on 2024-10-08, 16:39:

Meantime I go back the card because it had some issues again. As it turned out some of the pins of the ET3000 had cold soldered or cracked. After the repair I was able to run the tvdiag app, see the result attached.

Interesting the dot below the "4" character. Could it be a ROM error?

Either that or a stuck 1-bit in plane 2 of video memory probably. Perhaps run CheckIt. It will test video RAM (just to be sure) for such errors (up to 256K(262144 bytes) of memory).

Thank you your reply! Could it be a rom error? All the other test screen was OK, no extra or missing pixels. Anyway lets run the test and see the result.