Schedule June 2007

Wed June 06 - general

Sat June 09 - - 2nd Sat. - Volunteer Day

• Presemt were Ron Williams, Bob Erickson, Grant Saviers, Tim Coslet, Dan McInnis, Ed Thelen.

• Ron Williams and Bob Erickson worked on 1406 addressing problems

• Then when Dan McInnis came in to keypunch his object deck, we found that Salim's keypunch didn't work. Bob Erickson found a relay (related to punching alpha (not numbers) was getting very hot. Then he found a relay that had been swapped out of the 3rd 026 keypunch (with out a note or a tag - naughty naughty).

  Swapping back the relay seemed to help - in any case we now have two good working 026 keypunches again. Dan used one, and I used the other to make patches to change the operation of the Big Print - Banner customer demo ;-))

• Grant Saviers came by and had an extensive conversation with Bob Erickson.

• William Caulli (Brisbane, Australia) came in and said that he was involved with a similar 1401 configuration as ours except they had 6 tape drives and only 12,000 characters of memory. He was quite excited about seeing our system and seemed *very* conversant with many details of hardware and software.
  He started talking about code breaking with the aid of a 036 sorter with a count printer attached. Soon he and Bob Erickson were talking about the breaking of the Japanese code called J26 using IBM unit record equipment. Mr. Caulli seems involved with the MacArthur Museum Brisbane, especially the upcoming code breaking section. We made a Banner printout with his name.

• Tim Coslet said the left board was made by IBM and the right board by Transco. They are different - two resistors in the upper left quadrant are replaced by wires. The total resistance is 680 ohms on IBM board and 250 ohms on the Transco board.

  Schematic of the board

  Test jig showing both circuits working well on the IBM board. The Transco board was in spec, but did not light the second light.

Wed June 13 - general

About 729 tape drives, from Van Gardner

On the subject of the Card to Tape diagnostic getting tape errors.

The fact that this appears to only happen if reading the card takes enough time that the go line to the tape unit is dropped then raised may be a clue that the prolay is not getting the tape up to full speed before writing starts. If this happens the characters will be too close together on tape. This can be caused by too large a go gap between the prolay idler and drive capstan on the right side.

Another cause is the left side prolay "glitching" caused by overthrow when the arm moves from the stop capstan to the neutral position. If the left side idler to capstan gap is too close or there is too much slop in the idler, arm and armature pivots it will momentarily go past neutral and hit the drive capstan slowing the tape.

When you worked on prolay machines all the time you could walk into a room and recognize the "clacking" sound when they were glitching. I used to demonstrate this to CEs in field schools I was teaching by setting the off-line tester to read an all bits tape with a short go down time. Then I would move the left prolay closer to the drive capstan until we could hear the glitching. Looking at the preamp output on a scope you could see the dip in the start waveform caused by this.

It's going to be hard to get the system up to error free performance with the lack of parts. We used an enormous amount of prolay nylon idlers and their shafts. The original idlers were white nylon that was almost translucent, then a solid white, then black. In a shop that ran 24/7 they would get a grove worn by the tape along with bushing wear. The idler shaft, arm shaft and armature shaft are probably standard drill rod sizes that can be bought at an industrial supply house.

Before the navy I was a machinist that made parts for sock knitting machines and I used to order it all the time. Now I have gone to rambling again.

One of the best things you can do to help the TAU debugging is to get a second 729 working. That's the best way to tell if the trouble is in the TAU.

Van Gardner

Sat June 16 - Grant's card testing proposal

---

On Saturday June 16th, Grant Saviers sent the following:
I think the goal of substantially increasing the reliability of the machine ought to be our first priority, and encourage us to develop some plans to do this. Cleaning and checking cards is one approach. Another is to use the margin capability in the machine to isolate marginal circuits. Thoroughly testing most SMS cards and weeding out the weak semiconductors is a third.

We can combine all three approaches into one plan:

I would suggest testing each card during the cleaning/inspection removal process, or removing cards of the same type in a bay and replacing them with known good, tested cards as a way to maximize up time for overall system margin testing. In many cases we have enough spares to do this on at least on a partial by bay approach. So the challenge is to have known good cards and that requires volume testing. Finding some more spare SMS cards is also desirable where we have few or none (50% of card types).

With the conclusion of each round of module testing, inspection, and cleaning, run the machine diagnostics and use the margin test supply. I tested the portable margin power supply with negative results and concluded that its design needs to be documented as the test conditions I used may have not been appropriate (load resistor only). Assuming we gain confidence in this supply we still only have the ability to test the entire machine under margin which I think may be unproductive (too many simultaneous bugs to shoot, but this assumption is easy to test by testing the whole machine under margins). I inspected the power terminal blocks for each gate and I think it is feasible to install a reversible modification (plug) that would allow margining by gate.

I have been considering different designs for an SMS card tester and have settled on the following ideas:

1. Test each card at max design frequency (triggers) and/or minimum input pulse widths (gates & triggers)
2. The logic test should use input signals at the specified minimum input voltage swings.
3. The margin supply should be able to be set at three levels (max, nominal, & min) and the logic function test performed at each margin level.
4. The logic function should be checked hot (125 F?) and cold (50 F?).

After considering a number of tester design alternatives, I've rejected the following:

1. Anything that is programmable. (programming has stopped the TAU Analyzer for over a year)
2. An "obsolete" analog IC tester, since the speed and pin count is about right. (how would we maintain it? who would program it? where would we put it?)
3. A "universal" tester that is connector programmable. For a while I thought a 96 pin wire wrap connector could be used as a crossbar for test signals and logic inputs, but this got to be too complex considering the number of SMS circuit families (P, N, etc) and number of logic test signals needed for all types of cards.

In other words, I propose build simple testers that most any engineer can understand, use, and fix. Never have to worry about the next release of Windows.

What I think might work as an approach is as follows:

1. The tester approach is to provide a few number of mildly customizable testers that are self contained (power & test signals). The test output is observation of the card section outputs on a scope as all states of each logic section on a card are free running cycled. Every input of each logic section would be tested. Scope sync is provided for the free run test loop.
2. For each logic family, build
   - A trigger/latch/flip flop tester
   - A gate tester
3. I've looked at all of the schematics we have and my simple tester approach seems to cover most of the card quantities in the 1401. (The 729 lists ought to be developed)
4. For special cards, eg 1403 hammer drivers, core sense amps & drivers, 729 read amps, either don't test them, or build special purpose testers.

The tester build and card test priority is by two sorts:
- Higher performance gates and registers/triggers first
- Quantity of card type in the machine

I would make each tester complete with it's own self contained +12, +6, -6, -12 and both margin supplies with high, nominal, and low outputs. This is a little more work but yields redundancy of test capacity. Volume card testing can commence as soon as the first tester is built.

A 4 bit synchronous counter and a few gates can produce the necessary logic test signals. All of the trigger inputs and states (and set up times met) can be tested with 4 bits. Most of the gates are 4 or less inputs (plus expander) wide and back to my idea of a plug crossbar, a 25 pin or so connector can personalize the gate testers for each card type. (test signals to inputs plus loads as needed).

One part of the design still pending is the level shifters from 74HC to logic family. I'm partial to using high speed video/op amps because:

1. short proof parts are readily available with plenty of drive current and more than enough slew rates,
2. able to set output levels,
3. able to set rise/fall times with a capacitor.

The double width card test box that I made is the mechanical model, but it only needs to be one card connector wide. I've got most of the parts (connectors, voltage regulators, logic, boxes) to build 3 or 4 testers. The hot/cold temperature chamber is a Peltier device with heat sinks and mini muffin fans both sides in back to back recirculating chambers. The bottom is open so either hot or cold air can be circulated over the card under test. A variable current source controls the temperature differentials.

A next step would be to debate these ideas, get a team together to design & build card testers, and to develop a plan for cycling inspected, cleaned and tested cards through the machine. I'm out of cycles to finish the design and build testers on my own. My shop is available to the team.

Regards,

Grant

---

Tue June 19 - cleaning

From Mike Cheponis

I cleaned, vacuumed, and deoxit Row F of Module B, Gate 1 this evening. Machine still seemed to work when I was done, so I probably didn't break anything. Attached is a photo of a typical card as removed before cleaning. The rest of the photos are in http://web.1401.org/Module_B/Gate_1/1401__Module_B__Gate_1__Row_F.zip - This file is 76 MB so I didn't want to attach it to email. I've also included in that .zip file a .txt file with the card types for that row. It's interesting looking at these. Card 5, type KA, has that burnt-looking resistor; I could not find an identical type KA card in the spares, so I did not replace this card. Card 25, type QC, has 2 power transistors that had a piece of insulation glued on top of the TO-3 cases, presumably to prevent it shorting out the adjacent card (or maybe for thermal?). Anyway, the one remaining dark disc of insulation fell off the transistor, and so, insulation must be placed on this card, on both transistors!!! I re-installed this card without insulators (I did not have any insulating material handy) so that I could test the machine. There is clearance, but I think either this card should be replaced or some sort of insulator needs to be installed on this card. ---------------------------------- I plan on cleaning the rest of the rows in this gate, then moving to another of the 8 gates in Module B (the lower "quadrant", just below the 1401 control panel; Gates 1-4 are in the front, and 5-8 are in the back.) Then perhaps to Module A, the one on top, and then to the modules on the right, modules C and D. It took 2 hours to remove, photograph, clean, vacuum, dexoit on contacts, and deoxit gold on gold contacts for these 23 boards (and another hour to tweak-up the photos) . Therefore, if the rest of the modules are similarly fully-loaded with cards (and I know they're not), the total number of rows in the machine is: 6 rows per gate x 8 gates per module x 4 modules = 192 rows of cards, meaning it will take 384 hours to clean the CPU box. If we figure I can get it down to 1.5 hours per row, and that about a module's worth of power supplies exists, then that's around 220 hours. At 4 hours per week devoted to this task, it'll take about a year to finish; and the work really needs to be done when others aren't using the machine. -Mike

Wed June 20 - general

Thurs June 20 - tapes

e-mail from Bob Feretich

If a diag deck used to load, but won't now, I'd check the order of the cards in the deck. They may have gotten shuffled. I loaded 5040 today and it read fine. We determined that the I/O channel terminator shoe that has black tape wrapped on its cam lever is intermittent. The drive select operations worked fine using the other terminators. We found a broken ground wire in the 729 read amplifier gate. It didn't seem to be related to any known problem symptoms. I believe that we are having tape drive port select relay problems again. The 729 we are using has a dual port feature. When we access the drive from port "A", the C-bit in data bytes get dropped. When we access the drive from port "B", the A-bit in data bytes get dropped between the final read amplifier and the channel cable. There are only wires and the relays in between these points. The tape diagnostics like still 5040 act as a good demo. The card reader reads, the tape spins, and the printer prints (a lot of error messages). Maybe we should leave these dropped bit bugs on the machine, if we fixed them running the deck would be less interesting. ;^) I'll contact Andy and see if he wants to and can help. Regards, Bob

Sat June 23 - 4th Sat

• Present were: Ron Williams, Bob Erickson, Bill Flora, Tim Coslet, Ed Thelen. Judith Haemmerle joined us after 11:30. Mike Cheponis came in after lunch.

• Doron Swade, CHM Guest Curator and force behind getting Babbage's Difference Engine constructed after 150 years, is lecturing some docents interested in committing to a year demonstrating, cranking, maintaining a loaner Engine (all 5 tons) for one year.

  There are slots for 4 maintainers to be trained by the British experts. (I think "they" should set up bleachers so "the rest of us" can watch ;-)) "They" could probably charge admission ;-))

• Judith Haemmerle signed up as a candidate maintainer, having worked on slot machines, pin ball machines, clocks, and other physical things - but worried that her lack of working on big stuff would hurt her chances of being selected. I said come to the 1401 room and enjoy working on restoring a 513 reproducing punch.

  She seemed to fit right in.

• I bet there is no way to guess what these guys are trying to fix.

  They are examining the print out of a tape drive test. What can be causing this type of data garble???? (Blanks are printing as @, everything else seems OK.)

• Tim Coslet found that this diode took 1.17 volts to conduct a little, instead of the normal (for germanium) of 0.3 volts. (Silicon diodes take about 0.7 volts to conduct a little.) This high bias, coupled with a slow weak circuit driving it, caused a killing delay :-((

• We successfully reproduced the Big Print, Banner deck, using the 1402 Card Reader Card Punch. (Well, the object deck and text deck ran just fine :-))

  There were frequent reader checks and punch checks, but the final result was OK -
  AFTER I found out that the 1402 punch cards are loaded 12 edge in !! (The reader side is loaded 9 edge in !!) The rational for this strange situation seems to be that both reader and punch cards can go into the center pocket of the 1402, and this makes that pocket have a reasonable output deck :-|

• re: the Werner IBM unit record equipment - Mike Zahares came by and said that his contact in Portland discovered that the Werner house has a septic drain field under the back yard. Mike thinks that the required sized fork lift might crush the tile pipes under the yard, and/or get stuck in the wet soil. Mike is thinking of getting a crane to lift the heavy items (especially the 407 accounting machine) from the back basement door to the van. I told Mike that Paul Pierce thinks he is loaning his IBM 709 to Dag Spicer, and is waiting for the truck to arrive. That is not currently part of Mike's task.

Wed June 27 - general