Making your own Kenbak-1 Reproduction.
Back in 2005, I wanted to learn how this 1971 machine worked, and I figured making my own would be the best way to understand it. I was inspired by Erik Klein's website which showed photos of his Kenbak-1's circuit board, which seemed pretty simple. Just 7400 IC's, on a single board. I managed to get a very poor copy of a copy of the schematics and state diagrams through Herbert Eisengruber at the Computer Museum of Nova Scotia.
Reproducing the PC board wasn't easy, but after a lot of work, I had fun and had a working computer.
This page shows how you can make your own Kenbak-1 , not just an emulator, but using the same circuitry as the original 1971 computers.
Step 1: Get a PC Board, and Circuit Diagrams:
This is a very hard thing to make yourself from scratch. I recommend downloading these GERBER format files and send it off to your favorite PCB fabricator. It's a large board, but doesn't need any high-precision fabrication, or any silk-screen labeling or soldering masks. In fact, just a plane board with 2 layers is all you need. I found an online company who made up 8 boards for about $25 each.
In the above images, right click on them to open the image in a separate window, to see details. Or directly click on the link to download the fabricatable GERBER and drill files. Note that these aren't exactly like the original boards (I'm not trying to fool anyone) but they're close, and convey the spirit of the originals.
If anyone finds problems in these PC boards, I can give you the FreePCB project files, which can be read from the FreePCB program, or maybe you can find software that can edit an existing Gerber file.
You'll also need to get a copy of the Schematics, State Diagrams, Signal Definitions, and explanation of the internal operating states. This is in the KENBAK-1 Theory of Operation manual, which can be found clicking on the link in this sentence. You'll likely get very acquainted with this information before it's working.
Step 2: Collect all the Components
There are about 131 7400 series TTL, a handful of 1N914 diodes, several capacitors, many resistors, and a few bipolar transistors. All of these are readily available, and can be acquired from internet dealers, such as unicorn electronics (www.unicornelectronics.com) for under $40.00. But there are two other integrated circuits which are a bit hard to find. The memory is implemented as Intel 1404A dynamic shift registers. I have purchased several of these on Ebay for $25 or less, but they are becoming scarce. I found several vintage IC dealers on the network who promised to get me large quantities at $30 each, plus some expensive shipping. But luckily, I found a source of AMD second-source components out of some old medical equipment. These shift registers may be difficult to find. If you have more money than time, I may supply a few, but may request around $100 each to discourage depleting my supply, depending on how many I have and if I find a new source.
I thought about substituting 74LS or other TTL families here, but someone told me the newer faster chips may not work. I don't know why, and I doubt this. I ended up buying NOS vintage IC's from one of the many obsolete part dealers on the internet, but I think substitution of 74LS chips may be an interesting idea.
These are the parts, listed with quantities followed by the part number
45x 7400 Quad 2-input NAND
2x 7403 Quad 2-input NAND, Open Collector
3x 7404 Hex Inverter
1x 7408 Quad 2-input AND
15x 7410 Triple 3-input NAND
3x 7416 Hex Inverter Buffer
9x 7420 Dual 4-input NAND
10x 7430 8-input NAND
4x 7442 BCD to Decimal decoder (16-pin)
1x 7451 Dual 2-wide 2-input AND-OR-INVERT
10x 7454 4-wide 2-input AND-OR-INVERT
10x 7474 Dual Type-D Flip-Flop
3x 7486 Quad 2-input XOR
13x 7495 4-bit shift reg
2x 1401A 1024 bit shift register (8-pin metal can best.)
6x NPN transistors (original computers used TI branded 2N5449, Grant Stockly 2N4401's for this. Anything similar with a TO-92 case may work, but pay attention to pinouts which vary between types.
2x PNP transistors (original computers used TI branded 2N5447, Grant Stockly used MPS3702. Use a TO-92 case for authentic look, but pay attention to pinouts, as some similar transistors have different pinouts.
26x 1K ohm resistor 1/4 watt (0.09" x 0.25" size)
4x 3K ohm resistor 1/4 watt
4x 270 ohm resistor 1/4 watt
2x 220 ohm resistor 1/4 watt
4x 120 ohm resistor 1/4 watt
2x 68 ohm resistor 1/4 watt
1x 56 ohm ( this is a larger 0.7" x 0.33" wide 2-Watt Composite)
1x 3K ohm resistor 1/4 watt (R15's value, arrived by experimentation), Grant used 1K)
8x 1N914 fast switch diodes, DO-35
1x 5.1V Zener diode 0.75 watt min (123 mA) DO-41 (Grant used IN7433)
1x 100 uF 16V electrolytic, 0.6" tall, 0.35" diameter
1x 33 uF 6.3V electrolytic 0.35" tall 0.22" diameter
2x 0.1 uF, 16V disc, 0.6 inch diameter preferred
2x 0.0015 uF disc used in oscillator. Different original Kenbak-1's apparently used slightly different caps, the prototype used one mica and one ceramic, some used larger ceramic, but most used small ceramics. May need to adjust to get 1 MHz rate.***
2x mica-dipped capacitors, needs to be about 0.20-0.23 inches between leads. Mounted on underside of board. Approximately 10-6800 pF. One on Nielsen 3 is 10 pF, Grant used 6800 pF in his reproduction Series 2 kit, which I think was at my poor advice. I have heard it seems to run without these. ***
15x normally open push-button momentary switches. Ability to attach authentic-appearing knobs are preferred.
8x lights for bits, Yellow lite, but white/clear diffused lens preferred, LED or incandescent < 40 mA. Incandescents which operate on 5V and < 40 mA are difficult to locate, so LEDs may be easiest. Panel mounts also needed.
4x lights for control, Yellow lite, with yellow diffused lens preferred, LED or incandescent < 40 mA. See above.
1x mini toggle switch to lock-out the "store" button.
1x large toggle for main power switch.
2x ??? uF electrolytic capacitor to debounce store memory and read memory switches. These are soldered onto the back of the front panel. I've run the computer without these, and it seems to work fine.
1x Case, or at least front-panel on which to mount the switches and lights. I've done some experimentation with reproducing the original case style, contact me if interested.
1x +5V 3A, -12V 0.5A power supply: while a linear power supply could be created as per the original Kenbaks and schematic, it's quite reasonable to purchase a small modern switching supply, which would possibly be more stable and protective of the delicate electronics. On my computer, the measured currents were 2.3 Amps on the 5V line, and 0.2 amps on the -12V line.
Making it look Authentic:
Component selection can be as simple or complex as you want. Using vintage carbon composition resistors looks a lot more 70's than the metal film resistors you commonly find today. These older resistors are available online with a little searching.
Also, there's a lot of light and pushbutton options. The original used 5 volt miniature incandescent lamps, but LED's are much more reliable. Finding an LED bezel which matches the look of the originals is hard.
For the pushbuttons, I used some modern push-buttons, but removed the plastic button top that came on them, and glued on some old calculator keys I spray painted with epoxy paint. They're more reliable and durable than the Raytheon keys which had to be "glued on" to the front panel in the originals. Key bounce has never been an issue.
Step 3: Assemble the Board and Wire Components.
Sockets on the IC's make it easier to troubleshoot and replace components, but the original machines didn't use any sockets. The layout of the board is show to the left. Left click on it, to open in a new window, and get a high resolution look.
Note there is one jumper needed on the board. Double check with the schematics if any questions on diode orientation.
The original Kenbak-1's just point-to-point wired up the pushbuttons and the lights to the respective holes in the board.
layout of compoents and jumper on Kenbak-1 computer PC board.
You will notice that some components, like two capacitors, are mounted on the bottom side of the board.
And while not 100% authentic, it really does help with the "rats nest" of wires on the front panel to install headers so that the front panel can be unplugged from the pc board. This isn't too modern. The "Dupont" connectors (also known as Berg connectors) were actually invented long before the Kenbak-1, so were available at the time, but I don't think the little black pull-apart blocks on the pins were available back then. You could slide the black plastic blocks off the pins once they are soldered.
Step 4: Get A Power Supply:
modern switching power supply installed in Kenbak-1 case
If I recall correctly, my reproduction Kenbak-1, with all standard vintage 7400 IC's, requires about 2.5 amps at +5 volt DC, and under 0.3 amps on the -12 Volts line. There's usable many modern switching power supplies for well under $20. I don't think anyone has recreated the original Kenbak-1 power supply, as it takes two heavy transformers, some high power transistors, and a large metal plate to act as a heatsink and dissipate the heat. On my original Kenbak-1, with original power supply, it draws 1.3 amps at 120 volts totaling 156 watts. This means the power supply is accounting for about 90% of the power, but one transformer gets very hot, so I suspect leakage on the large capacitor, or I may have a bad power supply diode. I hope to track that down soon.
This photo shows my modern switching power supply mounted to the back panel. It also shows my "modern" pushbutton switches, and wiring up of the front panel. No fan was needed, as only the shift register memories get a little warm.
Step 5: Make up a nice case and front panel.
I was a little limited in my metal fabrication abilities, but I managed to find a metal shear I could use, and a bending break, and folded up the top and bottom pieces pretty easily. The back panel where the power supply attaches was a bit of a problem. I didn't have access to a spot welder, so I cut the holes for the metal screens, and then just glued with epoxy glue the screens in place. If you clean the metal well, I think this will last a lifetime.
To paint, cleaning all metal very well with degreaser, give one coat of metal primer, and then several good coats of a high quality enamel. The original case, the Bud Industries Grand Prix is long gone, but Grant Stockly managed to get it's original specifications, and put them on his Kenbakkit web site, to help with dimensions. Of course, I didn' thave access to that, so spent a great deal of time measuring photos, comparing distances with known dimensions, and came up with a pretty close result (within 5 millimeters in width.
To the right are authentic original Bud Industries cases, to show exactly how the metal was folded. I wasn't aware the little tabs were just on the top and bottom, so I went through the extra work of folding the tabs on the sides, which was not necessary.
Above are the top and bottom of genuine Bud Industries case.
The rubber "feet" are new reproductions, as I think every photo
I've ever seen of the Kenbak-1 rubber feet shows that they
dried out, flattening to millimeter thickness, and
fragmenting to powder, as old foam rubber tends to do.
My reproduction case is at the left. For the aluminum side handles, at the time I didn't have access to a machine shop, so I just made them out of Oak wood, and sprayed them with a chrome enamel. Again, a little time on a machine shop could fashion up some perfect reproductions.
Make the Front Panel Nice:
This is the part everyone will see. There's a number of methods for transferring an image to a metal plate, but if worse comes to worse, try rub on lettering, and pinstripe black drafting tape, then spray on clear enamel. A sheet metal hole punch ($40 Amazon) makes the holes perfect right where you want them..
Click on the above picture of the front panel design to get a high-resolution image, that if printed on 11x17 inch paper, would be full size. The edges of the panel are not on the high resolution image, but there are little "dots" where the corners are, so you can align this with a plate.
Note that the number "4" under the number 2 light is offset to the left a little. This was a mistake on the Original Kenbak-1, but corrected when CTI made up their own face plates with their name on it.
Step 6: Debug it:
Surely you didn't think it would work first time? I've assembled 5 of these, and not one of them worked when I turned it on. This may be because I used a bunch of old vintage 7400 series IC's and other components from the 1970's and 80's, but it's also probably because there's a lot of solder connections to make, and even with a magnifier, it's hard to see cold solder joints (that's when the solder melted, but didn't completely melt both sides of the connection.)
John Blankenbaker assured me many times that he's built and fixed all of his machines with nothing more than a multimeter. I don't see how, but he probably has the logic states and logic signal definitions memorized.
When Grant Stockly sold his Kenbak Series 2 kit, I posted to his Forum to see how many working kits were finished. I'm not sure, but I suspect only 5-6 out of 40 kits were successfully built at the time I was asking. I'd ask Grant, but he stopped responding to the Forum or questions about 14 years ago.
The only way I've ever been able to debug is with a logic analyzer and lots of patience. Above is the big old HP Logic Analyzer I had in 2004 when I was first trying to build my first one. Now I have a nice "pocket sized" one. I'd recommend looking at the current state, and seeing what states the machine is cycling through, to figure out what is not working as expected. With a logic analyzer or even an oscilloscope, you can watch the memory cycling through the shift registers and find out if flaky memory chips are flipping bits.
If you don't have access to a logic analyzer, a logic probe or a multimeter to check the state machine's output lines would be possible, but I suspect it would be terribly hard to figure out the sequence of states when the states change quickly. A two channel oscilloscope may help.
Step 7: Enjoy your new Computer
Sit back and consider just how amazing it was when John Blankenbaker got his machine working. But remember he didn't only assemble the computer. He had to envision the whole architecture, the serial memory and ALU, and design a state machine which could execute instructions while keeping track of all those bits as circling through the serial memory chips. Then he had to optimize the logic diagrams and schematics to use the fewest IC's, and make it all fit on one board, with little room left for traces. His feat is nothing short of incredible. He essentially did it all himself, and just had a little help with his brother in laying out the PC board.
Pause for A Sanity Check:
I really don't expect anyone reading this to make up a Kenbak-1 from scratch. It's possible, but a lot of work. But I do hope readers gain a appreciation of the design of this computer, and the amount of work to just assemble it.
Definitely, if anyone is going to embark on this project, go to the Kenbakkit.com web site, and there is great information on Grant's choice of components, and even the assembly manual of his kit. You're going to need as much help as you can get.