This one started with an IBM PC 5150 motherboard that wouldn’t POST, wouldn’t beep, and gave me just enough signs of life to keep me from tossing it back on the pile. It was one of several early IBM systems I had pulled from e-waste, and I went with the 5150 first. The board was the early green 16K-64K revision, fully populated to 64K with 4116 RAM, and it still had the original ROM set installed.
I checked the board over, identified the original BIOS revision, and verified the power rails for shorts before applying power. The +5V line measured about 265 ohms to ground, so there wasn’t a dead short. I also went through the DIP switch settings so the board matched what was actually installed: no floppy drives, no coprocessor, and just the onboard memory. With a spare power supply, a speaker, and a dummy load on the supply, I powered it up and got nothing useful back.
One obvious fault turned up pretty quickly: the variable capacitor near the crystal was physically broken apart. I replaced it with a fixed 20 pF capacitor, but that still didn’t bring the board up. From there I pulled the socketed chips, cleaned the sockets with DeoxIT, reduced the RAM configuration to the first bank, and tried again. Still no POST beep, so I went to the oscilloscope.
Using the empty 8087 socket as a safer place to probe, I checked VCC, clock, reset, and some bus activity. The board had a proper clock, reset behaved normally, and I did see address and data activity at power-up, which told me the CPU was at least executing code for a short time. That pushed me toward memory or ROM trouble, so I swapped in RUUD’s diagnostic ROM and added a display card. That finally gave me output, and the diagnostic stopped in the first 2K of RAM with a bad bit indication.
I went ahead and desoldered the first bank of soldered RAM and installed sockets so I could test chips more easily. That turned into its own mess. The board’s solder mask was delicate, the ground plane made desoldering stubborn, and I had to use a microscope to inspect traces and pads closely while I worked. Once the sockets were in, the diagnostic suddenly reported that all the RAM in the bank was bad, which clearly wasn’t right.
That sent me down another rabbit hole. I checked address lines, data lines, RAS and CAS activity, and then even the Retro Chip Tester Pro started acting up and falsely failing known-good 4116s. I cleaned the tester, worked around a flaky ZIF socket connection, and eventually confirmed that only one of the original RAM chips was actually bad. The real board fault after the socket work turned out to be a tiny solder bridge between a trace and a nearby socket pin. Once I found and removed that bridge, the diagnostic passed the first bank and then the full memory test.
With the original BIOS back in place, the board finally gave the expected unhappy beep codes with no keyboard or drives attached. I then installed a later 5150 BIOS so it would support option ROMs, added an XT-IDE card, an EGA card, and a memory expansion card, and brought the system up to 448K. It booted DOS 5 from CompactFlash, then ran Alley Cat and Arkanoid successfully.
— Aaron
Links
Timestamps
00:00 – First Look
03:19 – Identifying Board And Chips
09:31 – Checking For Shorts
11:50 – Setting Switches
15:26 – First Test
18:17 – Variable Capacitor Problem
19:47 – Bad Socket Connections
21:05 – Probing Signals w/ Oscilloscope
29:12 – RUUD’s Diagnostic ROM
34:17 – Desoldering Chips
42:42 – Troubleshooting RAM Signals
46:18 – Retro Chip Tester Pro Problems
51:31 – Troubleshooting Bad Traces
56:34 – Final Tests And Games
