My first publicly published article is in the IEEE journal, IEEE Life Members Newsletter Dec 2018. This issue has a doubled or tripled section, 8 pages, of Tales from the Vault, which are articles by older IEEE members, important, amusing, or historical technical incidents from their engineering lives. My article wouldn't have appeared if they had had the normal three pages of Tales from the Vault. My article is on p. 9, A Logic Recorder Tolerant of Static Sparks. This is the first article I have had published in a public journal.
I submitted this article on August 11, 2018. I don't recall receiving an acknowledgement of the article being received, and IEEE didn't tell me that it was going to be published. When volunteers run the Life Members organization, things are less formal.
Here are two versions of the article, the one I submitted (which was edited slightly for the newsletter and picked up a few minor errors) and the longer one I did for my own recollection.
For Tales From the Vault word count 506
In 1982, during development of printers at IBM Austin Office Products Division for the Displaywriter, a word processor using the Intel 8086, I took on the job of finding EMC (electromagnetic compatibility) modifications for the printers. Good EMC was a key advantage of IBM products. A proprietary IBM ESD (electrostatic discharge) tester shocked office products at a 60Hz rate, simulating body static encountered in low-humidity environments, generated by nylon fabrics.
The 5215 Selectric™-based printer was not passing the ESD criterion. There was a rare machine state that was susceptible, once every 2000 sparks. Logic analysis during sparking could not be done; any test equipment would act as antennas, coupling RF impulses from the sparks right into the high-speed logic. The poor EMC engineer was blind to logic failures. The printer would hang, and there was no clue about why.
With a hope that the printer’s logic was, in the main, adequately tolerant of ESD, I determined to overcome the blindness. The digital design lab provided TTL counters and static, 4-bit RAM. I made a quick design for a custom logic waveform recorder that could reside inside the printer and record 1024 samples, triggering upon the static event and retaining addresses and 8-bit data, pre- and post-static. This might reveal the susceptible machine state if the recorder’s memory could be read out to a logic analyzer, after stopping the static.
But time was critical. I had a stock of Bishop Graphics prototyping copper patterns, with adhesive to apply them to perf board, so that I could do rapid prototyping. “Solder-through” insulated wire made hundreds of connections. In three days, I had a 13-integrated circuit, 5” x 6” recorder with 28-bit width (address + data) and 34 pins of interconnect to the printer logic. After laboriously connecting 34 wires into the printer logic, 5V, and clock, a trial showed that the noise margins of the custom recorder were adequate to record the printer’s logic during discharging. I was no longer blind!
Putting the static tester on one-shot mode, I dialed up the voltage, toward 20kV, until I saw the logic hang. Leaving the recorder inside the powered-on printer, I changed the 32 pins of signals over to an adapter for the logic analyzer and uploaded the recording. There it was, a tight, infinite loop that the logic had entered upon a critical shock.
It was lunch time but I found a logic designer who could look at the loop and the logic states leading up to the critical shock. He said little but went away for 40 minutes. Returning with an EPROM that had a proposed fix, probably multiple polls of a keyboard or paper-sensor signal, he changed out the EPROM. We set the tester to the 60Hz rate and were thrilled that the printer was now functioning through thousands of discharges at the required voltage. The fix was in EPROM, not by shielding, filtering, or logic changes. There would be no added product cost.
Based on this and other EMC innovations, I received an IBM corporate-level award.
Extra recollections that are beyond 506 words
For Tales From the Vault word count 506 submitted by e-mail Aug 11 2018
In 1982, during development of printers at IBM Austin Office Products Division for the Displaywriter, a word processor using the Intel 8086, I took on the job of finding EMC (electromagnetic compatibility) modifications for the printers. The previous engineer doing this work [Ralph Genz] had found a surprising electrical open (lack of continuity) in the zinc-flame-sprayed plastic partitions, qty 2, near the electronics card cage of the 5218 printer. The open was in the 90-degree inside corner. But this helped ESD a minor amount. It was the big improvements that were needed, like going from 13kV to 20kV. I had always had an interest in RF and I wasn’t too afraid of dinkering with tough problems. At my suggestion, my dept. spent money on a fine HP network analyzer that could go to 100MHz, amplitude and phase. The EMC lab had a HP vector impedance meter that was nice.
With these tools, we could look at ferrite beads and capacitors. One of my interests was showing that all capacitors, even rolled film-dielectric caps, had low impedance above resonance; they had the same impedance (inductance) as a wire of the same length. The so-called “high quality capacitor” was merely any capacitor, apart from high-stability or low-residual-charge capacitors. There was rarely a need for mica caps. [My first IBM circuit in a commercial product, in 1972, was a 1.875MHz crystal oscillator for the Mag Card II typewriter, which had a LSI CMOS processor (probably not really a microprocessor, there maybe wasn't an EPROM) and shift-register memory for the text processing.) The manuals I had reference to said you had to have high-quality capacitors in various places of a crystal oscillator, and I designed in a large mica capacitor for one of the caps. No one asked that it be changed to a cheap ceramic cap, but the cost and size of the mica cap were not justified for this circuit. I found out years later that cheap caps are almost always fine.]
Good EMC was a key advantage of IBM products. [A famous IBM example of this was a IBM account in Chicago, probably in 1965 during the winter, in a high-rise office building which probably had very low humidity. A woman secretary using a Selectric typewriter got a painful and surprising shock when she touched her typewriter. IBM personnel, probably including an IBM attorney, the account manager (marketing), and an engineer, visited the account to see the typewriter and probably check for any hot-to-frame short, though this electric typewriter was double insulated and has a 2-prong cord. No trouble was found, and the shock was probably static electricity, a bigger problem for women wearing nylon than men. Static can reach 30kV to 40kV, and body capacitance can store a good fraction of a joule. This was a big embarrassment for IBM. I don't know if they replaced her typewriter. When you get a big static shock, it causes fear of any electrical device you touched.] A proprietary IBM ESD (electrostatic discharge) tester shocked office products at a 60Hz rate, simulating body static encountered in low-humidity environments, generated by nylon fabrics. [The tester had four aluminum vanes, painted an IBM blue, a high-voltage power supply, and a high-power reed relay that discharged the vanes to ground at a 60Hz rate. The vanes had vinyl insulation along the edges of the vanes, but there was nothing to keep the user from touching a vane and getting a static-level shock. These testers were rated IBM Confidential and were available at each IBM EMC lab.]
The 5215 Selectric™-based printer was not passing the ESD criterion. There was a rare machine state that was susceptible, once every 2000 sparks. Logic analysis during sparking could not be done; any test equipment would act as antennas, coupling RF impulses from the sparks right into the high-speed logic. [Where the high voltage would burn out the logic.] The poor EMC engineer was blind to logic failures. The printer would hang, and there was no clue about why.
The 5218 daisy-wheel printer also had ESD problems. Chuck Linton’s leadscrew stepper-motor driver had susceptibility in the optical shaft encoder (it was a closed-loop motor setup). This was instrumented with my custom, 3-channel, fiber-optic, analog, 1.5MHz probe (it used photomultiplier tubes because I knew they would work and I had experience with them). It showed a weakness to static in the optical encoder, and Chuck was able to cure that.
My observation over several years of EMC work was that the shielding and filtering needed to suppress EMI, as the speed of logic clocks increased, was as a byproduct reducing ESD susceptibility. ESD had been the tougher problem but EMI became tougher as speeds increased.
The fiber-optic probe was a sensation among the EMC engineers, especially Adam Zelinski. I used no special optical parts, just the fastest LED I could find, filed down to get the best coupling into Edmund plastic fiber bundles, then go about 8’ to an aluminum box with PMTs, then oscilloscope. Only 8’ separation between ESD tester/DUT and oscilloscope was adequate to let the oscilloscope see the PMT signals. The cheap PMTs, probably 931A at about $40, had adequate red sensitivity for the red LEDs.
Chuck Linton had commented to me 6 months earlier that ESD was so frustrating because you couldn’t see what was going on. I let that cook in my mind for a while, then I made 1” square “ESD probes” with LED outputs, hearing-aid batteries, 3 orthogonal magnetic pickup loops (2 turns each) feeding into diode detectors, and a couple of transistors doing pulse stretching. You could maneuver these into nooks and crannies in your DUT and try to find hot spots that were conducting the amp-level ESD currents, the hot spots always being near wires. The probes had no external wires and did not influence the fields in the DUT. You could also make a dozen of them and lay them all over a DUT, looking for changes as you changed grounding and cabling, though we never went to that trouble of building so many. The general idea was to try to visualize where high electric field and standing waves might be, and visualize where magnetic fields might be concentrated around wires. These little probes didn’t get too much use, partly because the pulse stretching didn’t seem to do much. At a certain threshold of ESD current, the LED was flashing visibly, and increasing the ESD voltage didn’t seem to increase the visual brightness much. I never saw the sense of this. We could couple the little probes into fiber optics and then to the 3-channel photomultiplier optical receiver, but that was complex and we got the ESD work done without using the little probes very much. Someone else in IBM asked for a prototype of my little probes, which I had done as a rough PCB, and did a nice PCB. I don’t know if this person did more probing that I did.
I thought is was very neat to go from Chuck Linton’s observation about being blind to ESD, think about it, and come up with something that gave some visibility. This is the type of thinking that the famous physicist Feynman did in his little book about his early years. The chapter was titled, "He fixes radios by thinking about them."
The ESD with 1kohm in series, “personnel,” was less trouble than “furniture,” like from a wheeled metal cart pushed by a charged person, discharging to a metal table on which the DUT was. Furniture ESD had no series resistance and had higher current, maybe 20A peak. Personnel ESD peaked about 3A. I could see these currents using the Tektronix CT-1 current transformer, which had a hole to poke your wire through and 50ohm coax to 50ohm termination at the oscilloscope. I was the only engineer outside EMC lab who had an interest in the CT-1, and the only engineer who had an oscilloscope with 50ohm vertical plugin for maximum bandwidth.
While I had been at UT, I had my eyes open to experimental test setups in the Engineering Science Building. [ESB was demolished in 2012 and the site has a great new building.] Professors and grad students built up these fancy, expensive setups. One setup, with energy-storage capacitors weighing 200 pounds, had an oscilloscope (big, heavy, maybe tubes in it, back in 1970) inside a hardware-cloth (steel wire fabric) enclosure. That was mighty impressive and showed me that fancy setups were sometimes vital.
After the success of my very-homemade-looking, 3-channel, fiber-optic probe, I designed a 20-channel instrument that would have been quite useful. My manager, Mr. Davis, who had gone into EMC after being a product development manager purchased expensive glass fibers with, as I remember it, avalanche photodiodes for high-speed work, and they were hanging in a cabinet in the EMC lab, but in the spare time I had to work on this system I couldn’t get the receiver amplifier working, and that project languished. My plan had been to purchase Vector Inc. quality, EMI-gasketed cabinetry and put four receivers in one plug-in.
Mr. Davis contracted with a UT physicist to model the new, expensive, giant, 10-meter anechoic room in B. 045. The physicist did a computer model that ignored the floor gap around the turntable. The model accounted for about three bounces of waves. They came out with a prediction of the RF coupling from source to receiving antenna (called antenna factor), over frequency, that was pretty close to measurement. This was a big technical advance for EMI testing. They made a publication that was not IBM classified. This was the first non-military (nonclassified) antenna factor that was calculated from the physics!
With a hope that the printer’s logic was, in the main, adequately tolerant of ESD, I determined to overcome the blindness. (After all, the printer logic was doing OK for thousands of discharges.) The digital design lab provided TTL counters and static, 4-bit RAM. I made a quick design for a custom logic waveform recorder that could reside inside the printer and record 1024 samples, triggering upon the static event and retaining addresses and 8-bit data, pre- and post-static. This might reveal the susceptible machine state if the recorder’s memory could be read out to a logic analyzer, after stopping the static. I had to use binary counters to get sequential addresses for the SRAM, and the counter had to be clocked on the little board so it could read out to a logic analyzer. [The ESD tester had to be used in one-shot mode for the proposed testing. Triggering the recording from the ESD event meant that the data in the SRAM represented critical machine states before and after the static event, and I put a toggle switch on the board to prevent accidental re-recording after you thought you had interesting data.]
But time was critical. [EMC testing generally had to be done on early manufacturing prototypes of commercial products, with all metal parts at release level, and any hardware/software changes threatened product announcement or complicated manufacturing. EMC work was high pressure, managers were daily checking for progress] I had a stock (much more than any other engineer or technician had) of Bishop Graphics prototyping copper patterns, with adhesive to apply them to perf board, so that I could do rapid prototyping. This, and RF interest, were my specialties. “Solder-through” insulated wire made hundreds of connections. In three days, I had a 13-integrated circuit, 5” x 6” recorder with 28-bit width (address + data) and 34 pins of interconnect to the printer logic. After laboriously connecting 34 wires into the printer logic, 5V, and clock, a trial showed that the noise margins of the custom recorder were adequate to record the printer’s logic during discharging. I was no longer blind!
Putting the static tester on one-shot mode, I dialed up the voltage, toward 20kV, until I saw the logic hang. Leaving the recorder inside the powered-on printer, so that the SRAM remained powered, I carefully discharged any clothing static I might have and changed the 32 pins of signals over to an adapter for the logic analyzer (maybe a Biomation instrument, it could map many signals into some sort of XY space, and the infinite loop was very obvious) and uploaded the recording. There it was, a tight, infinite loop that the logic had entered upon a critical shock. There were about 20 machine clocks in the infinite loop.
It was lunch time but I found a logic designer who could look at the loop and the logic states leading up to the critical shock. He said little but went away for 40 minutes. Returning with an EPROM that had a proposed fix, probably multiple polls of a keyboard or paper-sensor signal, he changed out the EPROM. We set the tester to the 60Hz rate and were thrilled that the printer was now functioning through thousands of discharges at the required voltage. The fix was in EPROM, not by shielding, filtering, or logic changes. There would be no added product cost and no further delay. I never asked what the nature of the fix was.
I had to go on faith that I would get somewhere by making the little logic recorder. There were a lot of potential obstacles: would the logic clock from the DUT be useful in sampling the many logic signals [I was a novice logic designer], would ESD upset the little recorder, would the recorder act as extra antenna for fields and produce extraneous logic upsets unrelated to the upset I was pursuing, would the problem upset be attainable on one-shot spark mode, would the large number of connections to DUT, and the large number of logic analyzer connections, be unreliable and frustrating to use? Any one of these problems would have scuttled my effort, and would have been seen as wasted time on a hair-brained attempt. [EMC engineers were either circuit designers who could tolerate branching out into RF, or communication engineers with purely radio reciever/transmitter experience who could tolerate thinking about logic and baseband. I was the former. All EMC engineers are wild ducks.]
Based on this and other EMC innovations, I received an IBM corporate-level award, Outstanding Innovation Award. My later award for Industrial TEMPEST Program was Outstanding Technical Achievement Award, 1988. Both were signed by IBM CEO. Chester Crider submitted me for the first award and was mighty impressed that I could come up with minor miracles to solve inscrutable problems. Chester had come from IBM Huntsville where they made the Instrumentation Unit that flew below the manned module on the Saturn V.
John Engelbrecht Senior Member, retired life member IEEE member resident of Austin TX USA active with San Antonio Life Members Affinity Group
submitted by e-mail to Craig Causer lm-newsletter@ieee.org
file ieee life members from the vault logic recorder.odt GETTER computer Aug 11 2018 Craig, thank you for your service to IEEE members. Tales from the Vault are always interesting.