Soon after I powered on the 11/44 the first time, I noticed that one of the LEDs on the front panel began blinking every once a while. It was the POWER LED, and the blinking means: “POWER NOT GOOD”.



Hmmm, good power is important. I measured all voltage and found some to be a bit off limit, but not much, so I decided to ignore the blinking. Typically it stopped after a few minutes, so I imagined some old stuff in the power supply that got into tolerances again at operating temperature. And the machine run fine ... well ... at least, none of its problems could be traced back to bad power.

Grounding, thermo shutdown

Months later, the day after I build in SMD and SDI disc controllers and the DEUNA network card, the POWER LED began blinking continuously, even if the machine was warm. I “corrected” this problem by grounding the separate SMD drive to the 11/44 chassis ground: the blinking stopped. Well, grounding is important, no wonder it solved this problem. But read on!

Later the same day the machine began to power off spontaneously. I recognized a bizarre pattern:

The machine powered sometimes down, only if I began transferring data between my PDP11GUI and a MACRO-11 disc driver. It did not power down if PDP11GUI exchanged data with the serial console processor! So not the physical connection triggered the fault, but the software controlling the serial port. This were facts, but they made no sense.

I was in process of installing 2.11BSD UNIX onto the SMD disk and needed a week, until I accepted that destiny wanted me to understand this mysterious big old H7140 switching power supply RIGHT NOW. Then, after a flash of insight, I suspected this might be an over -temperature shut-down. All manuals are on the web, for H7140 I used “EK-BA11A-TM -003 BA11-A Mounting Box and Power System Technical Manual” and “MP00837 H7140 unit assy Field Maintenance Print Set”.  The manuals indeed told me that the H7140 power supply has a temperature monitoring circuit.



So I removed the H7140 from the BA11 mounting box. It looks not so big on the picture, but the long side is 40cm, and this thing must have a mass of 25 kgs. I opened it (yes, it contains 300VDC and 5V at 120A ... real men don’t care!), pulled out the “BIAS AND INTERFACE BOARD” and put a scope onto the over-temperature trigger output. Bingo! If the machine powered down, this signal went active. Only problem: the heat sink with the thermo sensor on it feeled not hot at all.



So I drilled an hole into the big heat sink of the +5V rail, pushed an own thermo sensor into it and powered the machine on. The temperature of the heat sink did not go over 32°C, so it certainly was not a regular thermo-shutdown. I had to stop further research then, since suddendly no power-downs occured anymore ...

+15V now failing

But the red POWER LED now blinked all the time.



I raised the Console Interface Module (CIM M7090), which checks voltages and controls the power LED, out of the box and measured power directly: +15V were dead, the main +5 .1V were at 5.3V, other voltages were OK. Not good. It seemed while dealing with the thermo circuit, I ruined something other?

I try to be careful ...

+15V are generated on the LOGIC INVERTER CONTROL BOARD. I pulled it out, made a flat ribbon extender so that I could run it outside the H7140 power supply and started measuring. To isulate the error, I disconnected the +15V from the power supplys mother board (my error #1!). I also noted a strange odor in the card cage, so I removed all logic cards from the card cage to protect them from the defective power supply (my error #2).

The +15V were dead on the LOGIC INVERTER CONTROL BOARD too, so for two weeks I measured different op-amps, removed power transistors etc. etc. I made another error #3: I tried to get the voltage drop of a power transistor and connected the Ground strap of my oscilloscope to a wrong point in the circuit. Since the scope is grounded and the 11/44 is grounded too, scope ground has logic voltage ground potential and made quite a short cut. I burnt a power diode! Waiting another week until replacement arrives ...

Finally I found that the +15V was actively shut off by a voltage monitoring circuit shortly after power on. The reason: +15V is stabilized by a capacitor on the power supply motherboard, but I had cut the connection between +15V and the capacitor. So after power-on voltage raised too fast and with high ripple and a protective circuit (“crow bar”) switched it off. It was not a bug, it was a feature. Well, I shamed a lot, but was thankful to that DEC engineer who must have anticipated my try-and-error procedure long time ago.

+15V get lost between regulator and card cage

After I reconnected the LOGIC INVERTER BOARD to the mother board, +15V were still dead on both the CIM module and on the UNIBUS backplane. But as I measured them inside the power supply again, I found them to be OK now! So some internal connection was open.

I read still more manual pages, because there are lot of connectors on the H7140 mother board:



I tried to call this an “interesting experience”, but it was not the best time in my live. I disassembled the whole power supply and traced the +15V from the LOGIC INVERTER CONTROL BOARD over the MOTHERBOARD, and POWER DISTRIBUTOR board to the KD11-Z backplane. The connection between Power Distribution Board and UNIBUS -backplane is made with 3 so called “flex print cables”, but they are not very flexible: they resemble more thin copper plates than ususal flat cables.



The upper two connectors are +5V and GND for 120 Amperes. Do not put anything metallic into the small gap between them! The connector on the bottom (labeld “Pins” by me) carries other voltages including my +15V, “power good” signals and line time clock. It is called “P1”. In the picture it’s already resoldered.

Since +15V is routed over connector P1, it seems P1 was not connected to the UNIBUS backplane anymore. To check this I had to disassembled the backplane: BINGO! 15 of the 20 wires in P1 were ripped off from the backplane.



The broken wires still sat in their holes, making sometimes contact and sometimes not. This explained all the intermittant errors, the failures in the thermo circuit, the failures in +15V, the faint strange odor in the card cage (heating contacts slightly in touch) ...

I had to solder P1 directly onto the backplane. Easy to say, but soldering this big copper connector was like soldering a heat sink: It absorbed all heat, to get it hot enough I should have used a bigger iron, but the leads were very narrow .. well, I managed it somehow.

Don’t throw your 11/44 onto the floor

One last question arose: What ripped that sturdy connector P1 off its board? Obviously my 11/44 fell onto one edge of the power supply once ago, the edge is quite bent. The BA11 mounting box contains a big heavy power supply on one end, and a very rigid steel frame containing the backplane in most of its volume. In case of a crash these both elements will remain stable, but the BA11 case itself is made of thin steel plates which will twist and bend then. So I think at the moment of impact the power supply moved against the backplane, and P1 got ripped off.

I did not understand how you could throw such an expensive device onto the floor ... until I almost did the same with a PDP-8/e, as I slided her out of the rack with loosened slide hold levers. Oops.

Don’t test an idle power supply

I was quite happy, but the story was still not over at this point: The POWER LED on the 11/44 console was still blinking! I measured again the voltage on the CIM module: +/- 15V was floating around +/-12V, +5.1V was still +5.3V. Was God playing games? Would this never end?

Solution was easy this time: the H7140 power supply needs at least a load of 6A on its +5V rail. But I had an empty card cage. So I plugged in all CPU and memory cards, and now THE BLINKING POWER LED WAS STEADY! And it never blinked since then!

What I learned:

  • Bad power can cause a lot of phenomena
  • Naive testing can cause still more phenomenas
  • Don’t give up!