A simple ohm-meter proves instrumental in solving a frustrating case involving an early IBM 360 computer balking at booting
By Drew Berding, Contributing Writer
There was a time when IBM sold add-on core memory for the 360 model 65 computer for 11 cents per bit. That’s right—almost one dollar per byte! As a comparison, today we can buy a 4 Giga-byte memory stick for $12. Using IBM’s pricing, today’s memory stick should cost almost $4 billion!
Obviously IBM’s customers were desperate for lower cost alternatives.
A company I helped form made the first add-on semiconductor memories for the 360 model 65 at a very attractive price. Like IBM’s, it was in a separate cabinet the size of two refrigerators that cabled to the computer. Unfortunately, when we attached our memory to the 360 computer, the whole system would not boot correctly. Part way through the boot process, it would crash. Embarrassingly, when our memory was disconnected, the system booted properly.
A team of engineers spent two weeks trying to determine the cause of the problem. They discovered that at some random time during the boot process, a Reset pulse was occurring which of course aborted the process. However, they could not discover the cause of the Reset pulse and they were getting very frustrated. I was pulled off another project to give them a fresh look at the problem.
Since the errant pulse was occurring at random times and random widths, I decided it probably wasn’t a logic problem but a noise problem. I tried the obvious experiment of looking at the power supplies with a high-speed oscilloscope but they were clean—almost no noise.
Then I asked about ground loops. They answered that there was a single-point ground between DC ground and frame ground at the CPU. This of course is as it should be. I asked if they would mind if I checked it. No problem.
So I disconnected the cabling between the memory cabinet and the computer to isolate the two boxes. I checked with an ohm-meter and discovered that there was an unwanted solid short between our memory DC ground and frame ground. Eureka!
Then I started disconnecting DC power supply connections and was able to isolate the short to one gate and then to one row of cards. By systematically unplugging cards in that row, I found a control card that was causing the short. Examining the card carefully, I saw that the card had been fabricated improperly so that the internal ground plane on the card was exposed at the edge. When the card was inserted, the DC ground plane at the card edge shorted to the metal card guides. Therefore when the card was unplugged, the short went away.
As a temporary fix, I put tape on the edge of the card, plugged it back in, verified that the short was gone, reconnected the cables to the computer and booted the whole system successfully!I got a round of applause and an undeserved reputation of walking on water.
As a follow-up experiment, I used a high-speed oscilloscope to measure the voltage between the DC ground and the frame ground at the memory and saw several volts of noise. This noise caused the normally quiet Reset wire in the cable to the computer to appear to have a pulse on it. Since the noise was random, the Reset pulse likewise occurred at random times and random widths. The short term fix with tape on the card edge allowed the project to continue; the long term fix was to refabricate the control card so that the ground plane was not exposed.
It’s amazing what you can do with a simple ohm-meter!
Contributing Writer and BSEE Andrew Berding is an inventor with nine patents and serial entrepreneur, having founded six companies including Advanced Memory Systems (Intersil) and was a major participant in the founding of two others. He has been a consultant to the industry for 30 years and will soon be celebrating his 50th wedding anniversary. Though unfortunately no documentation exists from his most famous investigation, Andrew sends a sketch he drew of the ohm-meter used: