PDA

View Full Version : Another Finished Kenbak-1, how many are functioning now?


Cappellanus
12-09-2007, 10:54 PM
Hello:

A week ago, I finished my Kenbak-1 Series 2, and am quite happy with it! I haven't had a chance to exhaustively check all instructions, like "set bit" and "bit test and jump" but the jump and mark instructions, rotates, conditionals, and other things work, so I suspect 96% of the circuitry has been tested out. Interesting: with microprocessors, if one instruction works, we suspect the whole computer will probably work. But with small scale integration processor design, a minor bug can hide in a little used instruction.

It "almost" worked when first turned on, but I noticed when I entered data to memory with the front switches, two bits would make neighboring bits also turn on before storage in memory. Review of the documents and a little reasoning said this should be near the "I" register, the only place where the suspect bits sit next to each other (parallel). Sure enough, I switched IC-6 and IC-20, and different bits became "sticky" suggesting it was not a soldering error. I replaced IC-6, and problem fixed.

A few comments:

1) Spray painting the front panel with clear enamel or laquer isn't really an optional step. The deep texture of the brushed metal will suck a finger print off the cleanest hands.

2) Wow! there were a lot of extra capacitors when I was done. I placed one everywhere I could, and still had lots left over. I guess having more than needed is better than too few. It's interesting that Erik's machine used *zero* decoupling capacitors, just two electrolytics on the board for +5V and one for +12V. I suspect John designed the board with lots of decoupling caps, but realized it ran fine without them. Standard design rules of the time, however, suggested lots of decoupling caps were needed.

3) The board scans and documentation never show c16 and c17, and how exactly they should be positioned. But a search of the internet showed documents at Erik Klein's web site and "compusaur" seem to show it. I notice you suggest c16 and c17 aren't necessary. Since Blankenbaker installed them on the back of the circuit board, I suspect they were a later modification to get rid of a little glitch which only rarely cause problems, maybe due to borderline spec IC's.

4) Interesting about the second ground connection from the power supply to the ground of IC-114 and IC-115. I think I know why Blankenbaker did that. The ground current from these two IC's could potentially be pretty high. You may notice that the open collector lamp driver IC's can source up to 15 volts, so higher voltage lamps (12-volt) could easilly have been used without modification. But also, each output can drain to ground up to 40 milliamps, or 240 milliamps/chip. I think he was worried that's a lot of current to go along a thin PC-board trace, and if not connected directly to power supply, a significant voltage drop may exist. I'm not sure of the state of PC board quality in 1971. He often used several plated through holes to carry power lines from one side of the board to another. Interestingly, my second ground inadvertently popped off my power supply, and the LED light brightness was just about the same. I haven't investigated why, but I can only guess the ground current dissipated through the the inputs of these two chips to the output of their drivers, but standard TTL can't sink much. I may investigate this more when I get time.

5) Blankenbaker had an amazing circuit, but has anyone noticed how he never tied an unused input directly to VCC? I can't understand why he always tied the /SET and /RESET inputs of the 7474's to a pull-up resistor, instead of directly to VCC? That's why we have several 1K resistors spread around the board. I wondered if he wanted the pull-ups for debugging so he could set or reset the values from time to time, but the pc-board was all-the-way production with no other debugging facilities. But he even tied the input of an XOR chip to a pull-up. The only answer I can figure, is he didn't know that TTL circuitry allows tying directly to VCC to keep an input high. Is that possible? Of course, John was a Physics and Mathematics major in college, who by chance got caught up with computers after his Junior year of college, and didn't have *formal* electrical engineering education. Besides, TTL chips were fairly new at the time, and many people with experience in other logic circuits may have become accustomed to tying inputs to high or low through a resistor. Don't yell at me, yes, I'm suggesting John's design was a bit naive in one technical aspect, but that's quite understandable to someone who's roughly "self-taught" at a time when TTL chips were fairly new, and few people even in digital electronics had used the 7400 series TTL. Does anyone else have another theory on why he used all the pull-ups?

6) I'm not sure if I'm supposed to use the pan-head screws that came with the case, or the smaller ones in the bag. Do you feel the smaller screws are more authentic? They don't look all that different to me.

7) A little piece of metal to cover up the "expansion" slot in the front panel wouldn't be a bad idea. I'll probably make one up. The only Kenbak-1 I've seen missing this is Erik's machine. I just noticed for the first time: The CTI labeled machines left off this slot completely.

8) I used automotive spade-connectors to attach to the power switch on the front panel. That allows completely removing the front panel.

9) In retrospect, I'd like to raise the power supply a bit. It's just a bit above the PC board, and hides too much of the Kenbak-1 pc board. If it was just 1/2 or 3/4 inch higher, you could see a bit more of the "guts". As it is designed, you can't even see the "kenbak" logo on the pc-board.

10 It's interesting that the transistors were opposite pin-out (ebc instead of cbe) which makes the transistor orientation look wrong, when compared to Erik Klein's machines. Did you use the same transistors as Erik's?

This was a lot of fun. I'm surprised with how complete the kit was. I figured I would have to supply a lot of miscellaneous wire and hardware, but didn't. Well done!!!

Cappellanus
12-09-2007, 10:58 PM
Ooops, forgot the picture.

No, not me. My little girl! She's actually less interested in vintage computers than you'd think by the photo.

Cappellanus
12-10-2007, 05:15 AM
I guess I talked too soon, and found the answer to why John Blankenbaker tied the inputs to his TTL chips through a resistor to the +5 volt supply. I appreciate that nobody has yet slammed me. I guess there was reason for all those extra resistors after all. Was this obvious to everyone except me? ;)

Grant Stockly
12-10-2007, 06:06 AM
I guess I talked too soon, and found the answer to why John Blankenbaker tied the inputs to his TTL chips through a resistor to the +5 volt supply. I appreciate that nobody has yet slammed me. I guess there was reason for all those extra resistors after all. Was this obvious to everyone except me? ;)

Is it to limit input current or prevent a latch-up? The Kenbak was designed in 70-71, so the TTL chips were pretty new. :)

Grant Stockly
12-10-2007, 06:24 AM
Hello:
But with small scale integration processor design, a minor bug can hide in a little used instruction.


If you make a program to test all instructions, share it. :) Something that would run all instructions and the result in a register would be predictable if everything went fine?


1) Spray painting the front panel with clear enamel or laquer isn't really an optional step. The deep texture of the brushed metal will suck a finger print off the cleanest hands.


Was it easy to do this without getting runs?


2) Wow! there were a lot of extra capacitors when I was done. I placed one everywhere I could, and still had lots left over. I guess having more than needed is better than too few. It's interesting that Erik's machine used *zero* decoupling capacitors, just two electrolytics on the board for +5V and one for +12V. I suspect John designed the board with lots of decoupling caps, but realized it ran fine without them. Standard design rules of the time, however, suggested lots of decoupling caps were needed.

My prototype used only 1 capacitor other than the two in the oscillator. I supplied a lot of extra ones because of alternative positioning. Erik's PCB actually had labels for some of the capacitors missing...

For the electrolytic caps I chose them based on what would fit (lead spacing). The newer caps are a lot smaller for the same capacity.


3) The board scans and documentation never show c16 and c17, and how exactly they should be positioned. But a search of the internet showed


I can only figure these were part of a fix to another problem, but not necessarily what ended up fixing the problem. My Kenbak ran for over a week and it doesn't have them. :)


4) Interesting about the second ground connection from the power supply to the ground of IC-114 and IC-115. I think I know why Blankenbaker did that. The ground current from these two IC's could potentially be pretty high. You may notice that the open collector lamp driver IC's can source up


Maybe, but his traces are pretty big. I think the traces used could handle it. It may be related to noise more than a current.


6) I'm not sure if I'm supposed to use the pan-head screws that came with the case, or the smaller ones in the bag. Do you feel the smaller screws are more authentic? They don't look all that different to me.


The smaller ones are 18-8 stainless. I wouldn't have even included the larger ones but OPTIMA tightened up the larger ones and in some cases it scuffed the aluminum. So I included both...


7) A little piece of metal to cover up the "expansion" slot in the front panel wouldn't be a bad idea. I'll probably make one up. The only Kenbak-1 I've seen missing this is Erik's machine. I just noticed for the first time: The CTI labeled machines left off this slot completely.


John suggested I leave it out too, but it didn't look the same without it. :) (Plus I have some ideas for it...) Are you thinking of covering it from the back? It might look good with copper or brass. :)


8) I used automotive spade-connectors to attach to the power switch on the front panel. That allows completely removing the front panel.


I did too. I tried to stay away from suggesting anything about AC. Without offending anyone I tried in the manual to suggest getting help from an electrician. I imagine most people doing the kit know enough, but I don't want to get in trouble for it. At least I can say "How did you get shocked? I didn't sell you a power cord!" ;)


9) In retrospect, I'd like to raise the power supply a bit. It's just a bit above the PC board, and hides too much of the Kenbak-1 pc board. If it was just 1/2 or 3/4 inch higher, you could see a bit more of the "guts". As it is designed, you can't even see the "kenbak" logo on the pc-board.


I would like to also. Sometimes things aren't obvious until its all together (and you have 20 in stock) I'm glad there were no show stoppers. I would have also made the standard fan spots capable of fitting a fan inside. Those are standard sizes for an 80mm fan I think. A smaller one could probably be fit on the inside with some strong double stick tape. I ran mine for a week in a 76F room with no problems. I do have heat sinks on a few of the chips.


10 It's interesting that the transistors were opposite pin-out (ebc instead of cbe) which makes the transistor orientation look wrong, when compared to Erik Klein's machines. Did you use the same transistors as Erik's?


The transistors I used are equivalent substitutions chosen from a cross reference. The pinout difference you mentioned is why my prototype didn't fire up the FIRST time I applied power. :) I'm lucky that the only thing I had to debug on the prototype was all 8 transistors being installed backwards. ;)


Vintage: Kit:
T1 2N5449 2N4401
T2 2N5447 MPS3702
T3 2N5449 2N4401
T4 2N5449 2N4401
T5 2N5449 2N4401
T6 2N5447 MPS3702
T7 2N5449 2N4401
T8 2N5449 2N4401



This was a lot of fun. I'm surprised with how complete the kit was. I figured I would have to supply a lot of miscellaneous wire and hardware, but didn't. Well done!!!

I'm glad you enjoyed it and I'm glad you received more than you expected. I think of myself buying the kit and what I would want to receive. I guess I succeeded with you. :)

Do you have any other pictures worth sharing? Anything special? I'm glad your daughter had fun working with you. You'll be able to show her as she gets older the 1s and 0s work inside a computer. Have you read the first part of the manual with her?

What are your thoughts on the manual? Any areas I can make it better?

Cappellanus
12-10-2007, 06:30 PM
I'll attach a picture of the inside, but not much to show. This looks just like your machine. I tried to neaten up the wiring by twisting the wires a bit, but it didn't do much. Maybe a nice braid next time? Still, the original machines had individual wires to the front panel - that made quite a jungle.

Regarding connecting TTL inputs to VCC through a resistor:

While I thought it was crazy to tie TTL inputs to VCC through a resistor, apparently some early TTL chips needed this. I found internet posts suggesting "antique" 7400 series TTL with "multiemiter input transistors" sometimes needed this. Apparently, power supply noise or transcients can allow an input to be 1/2 volt over VCC, and cause chip failure. Any data sheet stating max input of 5.5 volts were susceptable to this. 74LS and some other chips were immune because of slightly different inputs. I wondered why I hadn't heard of this before. I managed to look at a copy of a 1971 TTL data book here...
http://www.bitsavers.org/pdf/ti/_dataBooks/1971_TI_IntegratedCircuits/06.pdf
Looking on page 6-4, it gives 3 options for unused inputs of AND/NAND gates. You could tie them to an independent power supply of 2.4-3.5 volts, or tie them to another input, or pull them up to VCC through a 1K resistor. It seems to only mention NAND/AND gates, which use the "dual-emitters" so I'm not certain it's needed on the XOR gate or set/reset lines of the 7474's as the Kenbak-1 uses. None the less, there *is* a reason for what John did.

Regarding decoupling capacitors:

I seem to remember recommendations of a decoupling capacitor, 0.01 to 0.1 uF, one for every 4-5 7400 TTL chips. Interesting to note that Intel recommended 41 separate decoupling capacitors on a single Pentium II (all around the outline) and faster processors really need distributed capactor PC boards, where power and ground planes exist in the same layer, separated by a dielectric, which acts as a capacitor along the whole board. I'm surprised that the 131 -chip Kenbak-1 works so well with zero decoupling capacitors, just a couple big electrolytes.

Regarding heat of the machine:

I agree, my machine really doesn't get too warm. A couple chips and the memory gets hot, but the case doesn't get warm at all. Not sure why the older Kenbaks needed a fan and holes in the top of the case. My Mark-8 makes lots of heat, but that seems to be from the linear power supply. It's conceivable that the linear power supply in the original Kenbak-1 dissipated twice as much heat as the entire board. Also, I suspect the more vintage 7400 TTLs dissipated more heat than newer ones. Another theory I don't have time to test.