return to main page

Maintenance
primarily electrical and logic

Table of Contents - and in chronological order
- Background
- Thoughts about computer logic maintenance
- No need to get fancy
- 1401 Card Test
-

Background
In November 2012 Susan Sherwood. director of Center for Technology & Innovation sent two DVDs which had IBM Training Movies (converted from 16 mm film). These explained the advantages of transistors over vacuum tubes in data processing (computing) equipment, a version of how diodes and transistors worked, and how to use transistors as logic elements..

The thought was that these might be helpful in training new people to maintain our old equipment. Center for Technology & Innovation has a working IBM 1441 and we (CHM) have two working IBM 1401s.

the following comments ensued ;-))

Thoughts about computer logic maintenance, from Ed Thelen, Nov 2012
You mentioned that these videos might be useful in training new people in maintaining/trouble_shooting our archaic discrete transistor computing machines.

They are a nice introduction to semi-conductors, but the down and dirty of
- finding the bad circuit card, connection, ...
- fixing the above
is a MUCH larger world :-((

Not only does the student need to be much more proficient in logic, adders, flipflops, core drivers, sense amplifiers, ...
the student also needs to know the
- circuit set of the machine in question such as the SMS family of logic
- theory and practice of the machine in question including details of
- - - how to read the schematics, ALDs, ...
- - - cabling and environment
- - - door and component interlocks ;-))
- - - power supplies, and bringing up power
- - - control logic, clocks, and data flow
- - - operation of the control panel and instructions
- theory and practice of diagnostic tools
- - - scopes, meters, test decks, ...
- tricks
- - - voltage margins
- - - temperature margins
- - - vibration, run pencil along cards while running diagnostics
- - - changing the clock frequency ( a little, carefully )
- and I've likely forgotten several as I escaped to software after fixing computers for a several years :-)))

The above fixes computers to the circuit card level -
The person fixing the circuit card needs a totally different set of skills -
- gross tests of components -
- - - "passive" resistive, capacitive, inductive,
- - - "active" components such as diodes, transistors ( of various sorts)
- - - make test jigs and circuits to exercise items under test
- spec sheet tests of active components
- - - gain, leakage, voltage capability, maybe speed ;-))
- - - curve tracers, in circuit gain tests, temperature range, ...
- - - http://ibm-1401.info/WeirdParts.html
- how to remove suspect components w/o destroying the circuit card
- how to install "good" components w/o destroying the components
and likely more 'cause this is not my specialty ;-))

Actually, the boundary between the
- bad circuit card finder
and the
- bad circuit card fixer
is not as sharp as one might imagine.

Say a circuit card is suspect, computer works with a "good" card replacing the suspect -
but the circuit card fixer can't find anything wrong with the suspect - :-||
There should be cordial, effective cooperation between the finder and fixer -

like the system location of the "failed" card should be logged and associated with the physical card. In tough cases, an extender card may be needed to scope out the trouble element by element.
Field people are suspicious of cards that are returned from Repair as "fixed" or "no fault found".
There are tales of field people crushing chronic troublesome cards.

"We" spent considerable time thinking of making a circuit card tester.
Indeed - but there are so many types of SMS cards in a 1401 system that the effort seemed overwhelming - http://ibm-1401.info/SMS-Card-Usage.html
a major engineering project -
but it seemed/seems desirable - but "non-trivial" ;-))

My summary, the movies might be the first hour of an intensive three? month course.
Preceding?? another intensive three? months on peripherals -

How to maintain these machines after the/we old geezers become inactive is an enormous and worrisome challenge !!
The old geezers have already been trained for months/years, and have time and memories associated with these toys
Young geezers have their own fish to fry,
- chase the other sex
- pay for the car, house, dental care, college, boat, ...
- jostle for position to better pay for the above, or stay employed -

I'm CCing some people more skilled in maintenance reality -

Best Regards ( and much sympathy, empathy )
Ed Thelen

No need to get fancy, from Stan Paddock, Nov 2012
Ed,
 
1. A transistor has three connections:
    1.1 Emitter
    1.2 Base
    1.3 Collector
2. Current wants to flow from the Emitter through the Base to the collector.
3. The Base controls the Emitter to Collector current by current injected into the Base.
4. The ratio of Emitter to Collector current to Base current is 10 to 40 normally.
    This is the amplification of a transistor.
5. Transistors can be built where the collector is positive in relation to the Emitter (PNP) or
    built where the collector is negative in relation to the Emitter (NPN)
 
    A Diode is a transistor without a base(NP).
    If the N end is negative in relationship to the P end, the diode will conduct.
    If the N end is positive in relationship to the P end, the diode will not conduct.
 
If your circuit does not work, more than likely there is a bad transistor or diode to blame.
 
 
What more needs to be said?
 
Stan

1401 Card Test, from Grant Saviers , Nov 2012
Ed,

What the 1401 experience demonstrates is that "fix when broken" repair of SMS cards is no big deal, as blown/short diodes and transistors are readily found with an ohm-meter.

First principles say to make a machine of the 1401 era really reliable, one should margin check it on a regular basis. This is the result of the MIT work re Whirlwind and TX-0 and what was incorporated into all early DEC machines, following what the DEC founders learned from building those machines. Of course, *really* reliable is debatable, my definition is to compile and execute about the largest FORTRAN program the machine can handle and run it for 24 hours without faults.

Later work even showed that semiconductors usually show degradation before failure and that predictive failure tools could find circuits about to fail with about the same success rates as they could predict the future failure of mechanical structures.

It's the second approach that requires something other than an ohm-meter to "fix" "bad" cards. i.e. the detection of slow, weak gain, leaky, temperature sensitive, etc. semiconductors. Having worked on a big machine (10x the 1401) with "morning sickness" I"m also well aware of the frustration in chasing marginal bugs.

A SMS card tester is really no big deal, I've scoped it out rather completely for the variety of 1401 cards. However, there is no point to building it unless there is some reason to do so and for it to be used.

Grant



return to main page