Provenance: This computer was assembled in 1971 or 1972, but never worked. So John Blankenbaker stored it in his attic until 2010, when he fixed it and put it up for auction. He couldn't locate any of the original incandescent lights, so he found a modern substitute, and replaced all the lights. He was also missing one of the aluminum side handles, so had a new one fabricated. Lastly, he reported that he found and fixed a problem with the clock driver circuitry, which is explained more below. Once he got it working, his daughter sold it on eBay in 5/2010 for $25,600 to a collector in Monza, Italy , who then re-sold it on eBay in February 2016. It is now in a private collection of computers in Massachusetts, and viewable on his website oldcomputermuseum.com.
I thank the current owner of this machine, for not only allowing use of his photos, but he painstakingly took some highly detailed blowups of the clock circuitry area, to try to investigate the clock circuitry mystery. He gave permission to use his photos in this site, but I noticed he doesn't list his name on his web site, and it took me some sleuthing to learn it, so I won't name him until I get permission to do so.
979 x 526 This is an enlargement of the third photo, showing the clock phase generator area.
6000 x 4000
6000 x 4000 This is the top of the PC board with markups showing areas of interest, where there's signs of modifications or repairs. The clean version is below.
6000 x 4000
500 x 334 This was from the eBay listing.
1023 x 684 from current owner, it shows the different replacement lights.
1023 x 684 The back serial number label, from the current owner.
750 x 368 likely from eBay listing.
In September 26, 2022, John Blankenbaker told me a surprising statement about a mistake in the production machine's printed circuit boards: "All of the machines except the very last one (Achim owns it now) had an undetected error in the memory drive circuits. The machines still worked with this error though when it was corrected the waveforms were much better. One of the drive transistors was wired incorrectly in the etch." John is essentially saying that all of the production Kenbak-1 computers have a wiring error in the memory clock signals, which still allowed the machines to function, but was none-the-less wrong. But he couldn't be talking about Achim's machine (the prototype, the first machine) which had a completely different circuit board. That had a completely different clock/driver circuit wiring, so would have little influence on the completely redesigned production machines. He was definitely talking about the serial number 183 computer, which he repaired in 2010, and which carried the same circuit board as the other production machines.
The possibility of a long un-noticed error intrigued me, as I and others had examined every portion of that circuit. I even recreated the circuit board with a schematic generated netlist, while looking at the original board to make sure traces were roughly in the same place. I found it puzzling that I, and others who fixed and debugged these computers could have missed such an error. I spent a great deal of time looking over the clock waveforms, and carefully tracing out the circuit connections. Yet everything appeared correct. Why couldn't I find the error he was talking of?
Pictured is the schematic of the CMOS shift register clock drivers. There are two of these, and each has 3 transistors. So I searched over the circuit boards to try to find out what's wrong. I could find nothing. Then I looked at the oscilloscope tracings for the clock signals (psi1 and psi2 in the schematic. ) It's tracing is below.
While the two phases are a bit bumpy/rough, this is because there's no filter capacitor anywhere near this circuit, and the power supply is showing the current surges. But both of the two clock signals are the same shape, which doesn't support anything like the reported error.
Unable to find anything wrong, I asked John again about the error for more details, and if he could recall which transistor he had to fix.
Unfortunately, John didn't recall details. He wrote:
"The circuit [he means the schematic] was correct. One of the transistors was wired incorrectly. I don't remember which one. I found the problem when I compared the the phase 1 circuit to the phase 2 circuit. In particular I believe it was the transistor that was the farthest (of the six) forward toward the front panel. To correct it, I cut the etch to one the three leads and installed a jump with the correct signal. That wasn't hard to do. My memory has gaps in it."
Now armed with a bit more definitive information that I was on the right track, I again searched over the circuit boards, yet could not figure out what he was talking about. My oscilloscope tracings of phi-1 and phi-2 match.
I put this problem on the "back burner" until several months later, when the current owner of Serial #183 was finally able to send me some photos of the "corrected" circuit board, and it in fact did show modifications of a transistor right where I thought it would be. These photos aren't great quality, but indeed shows there was some recent (dark flux) changes around the Q5 transistor.
The above red circles show the top and bottom of the circuit boards where the Q5 transistor is. Even with the low resolution and blurry photos, relatively fresh (dark) flux shows more recent soldering. And the Q5 transistor is not oriented like any other Kenbak-1. The flat side is pointing to the lower right. Compare this to Serial 197's orientation, where the flat side points to the upper left. This is definitely where he made his change, but it appears to me the circuit tracings were wired correctly in the first place, and the oscilloscope tracings seem to show the two phases match each other.
So what is going on here?
Serial Number 197's (Crosley's) clock driver area.
I'm beginning to suspect the problem may not be a true error. It's possible John replaced a faulty Q5 with a substitute transistor, then noticed it wasn't looking right on the oscilloscope. The 2N5449 transistors originally used aren't produced anymore, and substitute transistors often have different pinouts.
Here are several possible pinouts for a bipolar transistor in a TO-92 package.
If looking at a transistor with the flat side towards you, and the leads pointing down, the emitter, base, and collectors could be in any of the following orders:
EBC: Most common American pinout, as in picture.
ECB: Most common Japanese pinout.
CEB: uncommon, but known.
CBE: uncommon, mostly seen in European transistors (But THIS matches the Texas Instruments 2N5449 Pinout!!!
It does appear a substitute pinout could have flipped leads, and caused faulty clock signals in this level shifter circuit.
Until we get more information, I wonder if John replaced Q5 with a substitute transistor which had a different pinout, maybe reversing the collector and emitter. Most designers know the experience of getting "lost" when trying to follow printed circuit board traces, having to flip from front to back of a board. Like getting lost in a forest. Maybe he didn't notice his new transistor pinout was different, so he thought the circuit board was wrong. But he managed to fix it regardless. Until clearer photos, or the part number of the replacement transistor is available, this remains a mystery.