A Bit of Z80 History

If you used a Z80 chip back in the 1980’s, it almost certainly passed through this room. This photo is of the first Fairchild Sentry 610 test system at Mostek We tested every Z80 chip including Zilog branded ones in this room for many years. (photo from Mostek 1973 databook). We also tested Fairchild F-8 CPU chips, and exotic parts such as the Magnavox “Star” TV tuner chips. It was general purpose and could test RAMs, calculators, and ROMS, though we made our own custom testers for those. This room pretty much did all the heavy lifting on all the early CPU chips. Mostek had a second source license for many of the the popular CPU chips – the Fairchild F-8 series, the Z80, the Intel 8088/6, and the Motorola 68000 family. And we invented the multiplexed DRAM, too.

 

 

The machine was a beauty. Going clockwise from the upper left side is the disk drive, a 9 track tape, the Fairchild 24-bit CPU, a terminal and printer, the Sentry 610 Tester ( tall box on the right), and a pair of 64-pin test heads. The door to this temperature and humidity controlled room is on the right, in the back, down the ramp to the right.

*The disk* was a head-per-track unit, with a large platter rotating on its side. It was 26 inches in diameter and stored 75,000 bits. This particular disk developed flakies in 1974, so I was elected by the bosses to drape it with plastic, crawl under, and spend a shift carefully wiping the surface of it with alcohol. It worked, to my surprise. I remember that the surface looked like raw steel instead of the oxide that I expected..

*The CPU* was a Sentry 400 series CPU running “Factor”, their test language. I still have the test handbook for it. It was 24 bits wide at about 250 Khz. There were two boxes of cards, about 12″ X 18″, and 12 boards with 2-bit slices per board. You could access it by pulling the black latch on the left side, and swinging open the door. On the front of the CPU was a LED bank behind plexiglas to make das blinkenlights.

I remember this CPU had a “short” and “long” bus that was mapped to the various peripherals. One night it failed to boot, which would cost a few thousand dollars an hour. I assigned a technician in training to fix it and I went off to do something more important. An hour or so later he told me he had found it, but did not know what to do next. He toggled the switches to go to the short data bus, and after a bunch of switch flicking and looking through a file cabinet of documentation, he proved to me he was able to write the necessary data to access the disk via the long data bus and read back the disk register. Sure enough, there was a bit set, which the documents said was the “disk online” bit, and it was off.

“Oh”. I turned around, and the “Online” switch was flipped downward on the disk. “Click”. It was a great learning lesson in depth of how the CPU worked, for him and me both.

*The tester* had two doors that could swing open and had removable panels. Each door held three banks of cards of TTL and ECL logic, with a massive 5 foot tall wirewrapped backplane. I once tried to count the IC’s in the tester, and came up with around 60,000 chips. You can see the backside of it here. The black door latch for one of the doors can be on the upper left, and both the hinges are clearly visible running down the back. The center panel was filled with large rack mount (and very heavy) power supplies. At the bottom, in the black area, are three phase twist-n-lock 220V power connectors, a bank of 110V AC utility outlets, and cables running to the two test heads. The black panel running horizontally had some test BNC’s no one ever used, and a key to power it off and on.

We had to be careful about those AC utility jacks. The tester ground was at -11 volts at several hundred amps. If you clipped a scope to a ground, it would happily melt the insulation off the scope probe until it stunk up the room and pooled on the floor. We had to use a 3-prong adapter with the ground cut off.

This tester had a 100 Mhz internal clock and could test almost any type of IC at 10 Mhz speeds. The tester could DMA test patterns from the disk to the pins of the DUT at full speed, change power supply voltages at megahertz rates, and via microcode running out of ECL RAM, call test subroutines in hardware at the full clock rate. Each test vector looked like a binary 101011..001, 64 bits long, and each bit set an output high or low, or was an expected response. These bits basically were wired to the test pins. Somebody wrote a Z80 cross compiler that made “Set F 10101010101……100101” statements from assembly language, so we could test actual code on the Z80 at full speed. It was very exotic tech for us. We could only compare it to the assembly language PDP-11 testers we typically made. Mostek made all their own custom testers except for these.

*The test heads* ( one on the left has a chair in front of it) had a circular array of “Pin Electronics’ cards. PE cards had a lot of analog circuits for testing the device under test (DUT). I think this particular machine had 64 pin capacity. The DUT was plugged into a ‘top hat’ that had a “Zip Dip” socket which in turn was attached to the ‘Performance Board’ that had the loads and any special circuits for that part, which in turn plugged directly onto the circular array of PE cards. This is that square black hole in the middle of the test head. These IC-specific parts could be changed quickly, a new tape loaded by 9-track, and a new test set up in a few minutes.

These test heads were built to hold the very heavy Electroglas wafer testers, too.

*Controls* On top of the test head is a box with 4 lights and a “test start” button. There were two red and two green lamps. These are a go-no/go set for parametric failures such as continuity and leakage, and a second pair of Red/Green lamps for functional failures. I remember that a particular lot was failed by QC (this happened all the time) but I realized the same oddball part number had been failed a week before, and again before that. There was a pattern to it. A bit of investigation revealed that the new girl that ran the machine did not understand we needed the parts to have TWO green lights. It was a language and training issue. I was afraid that she would get fired when somebody realized that all her work for weeks was no good. But our shift supervisor was fantastic. After explaining to Robert what was wrong, he just told her we had a new procedure now. Two green lights goes in the good pile. Problem solved. And no one ever said another word.

*Failure modes* It was a beautiful machine, I loved it, and it broke every day. 100Mhz ECL chips got hot and failed often. But it was usually a Pin Electronic card (or 3 or 4) would just die. I would load a tape called ‘TVFY6’ and spot the problem in a few minutes. I eventually realized that a lot of bad PE cards happened just before or after lunch. It was caused by people sorting change on top of the machines, which would drop inside and Zzzzzzzt things. Thankfully, our warranty for this tester got us PE card repair for free. But one day Fairchild refused to send us any more. They claimed we had hundreds of them that were unaccounted for, and had never returned them for repair. We eventually found out that our boss had been hoarding them in his filing cabinet. He got fired that very same day. So me and my techs got to work a lot overtime to repair them. I ended up making more money as an hourly tech than a lot of engineers from all that overtime. I loved that machine, for lots of reasons.

*Circuit Breakers tripping* As was the group leader in the repair department from 1974 until 1979, I had to watch over everything about this machine. There had been mysterious catastrophic power supply circuit breaker trips on multiple weekends. One night, around 11 PM, I heard the (very loud) alarms go off. I headed up the ramp in the rear, and passed a janitor running down that ramp while dragging a floor buffer behind him. I finally realized he was plugging it into the tester utility AC outlet in that black area where cables come out and was tripping the circuit breakers. That incident got people’s attention. Facilities finally made us very nice laminated plywood custom covers for that bottom area and covered that problematic area around the DUT, too.

*I loved this machine* Mostek sent me out for factory training to repair them in September of 1975. I still have the certificates on my wall. It was quite thrilling to be on a company expense account (for an entire month!) in San Jose, the heart of Silicon Valley when I was just 21 years old. I had already worked on them for two years, so the class was a breeze and I had every afternoon off.

tl;dr This Fairchild Sentry 610 tester was a kick-ass Z80 tester.

from Hacker News https://ift.tt/TyjHVYd