PiDP-11: RECREATING THE PDP-11/70
Work in progress. This page will grow as builders report problems they've had, and the fixes they've applied.
This PiDP-8 troubleshooting document may be of help, as the circuits of the PiDP-8 and PiDP-11 are similar. It contains some useful debugging techniques and common mistakes.
Out of 2200 kits, all PiDPs with problems have been fixed as far as I know, so do not worry. Except one, due to a lack of interest from the builder... If you are reading this upfront: 95%+ of kits work straight away without troubleshooting. But actually, troubleshooting can be instructive, even fun. As long as you realise you will get it sorted. On your own, in the Google Group, or by emailing me.
Most problems are related to a group of LEDs not lighting up, or a group of switches not working. Based on the experiences with the PiDP-8, here is the list of potential problems. In practically all cases, they are due to problems 1-3 on the list below:
a forgotten solder point. Easy to spot and fix,
a bad solder spot. Harder to spot, the solder is there but it does not make an electrical connection, just reflow the solder,
a broken trace due to overheating. Impossible to see, best solution is to solder a small wire to replace the broken trace,
a short caused by too much solder. Generally visible, only a handful of reported cases though - not likely.
Other problems that have occurred so far:
IC placed upside down (or did you solder it on the back of the PCB?)
a switch got overheated during soldering (check whether it works with a multimeter, but you'll feel and hear it clicks different from other switches as well).
a LED is soldered in the wrong way round. Easy to spot from close up: LEDs have a flag-shaped internal structure, check all LEDs have their flag pointing the same way. Assuming you did not put all of them in the wrong way: check the flag direction against a spare LED from the kit and the little drawing on the PCB that tells you the long leg needs to go in on the left side.
Problem 4 is very unlikely, if you do not see a short after a thorough visual inspection, discard that possibility.
Problems 2 and 3 are often not visible, but obviously the symptoms are: LEDs that do not light up or switches that do not respond. If visual inspection does not lead you to the problem spot, you need to deduce its location. Which is not all that hard, because the circuitry is just simple traces from point to point.
So, starting with the relevant pin on the GPIO connector, use a multimeter to see if there is an electrical connection from point to point leading up to the LED/switch that does not work. (Note that you can't multimeter your way through a diode!)
The picture below is a high-res PDF, that can be enlarged (and downloaded) to show all traces in great detail. Use that with your multimeter to find the bad spot:
To go on a proper bug hunt, you'll save yourself a lot of time to understand how the switches are wired.
groups of 12 LEDs get their 5V from a led line that comes out of the GPIO connector, goes through the IC 'for amplification', before it reaches the group of 12 LEDs. Individual LEDs in the group are then switched on or off through 12 column lines that end up at the 12 resistors before going back in the GPIO connector.
The switches too, have row lines that come out of the GPIO, go through a 1K resistor, connect a group of 12 switches. And the same column lines used for the LEDs then hook up to each of the 12 individual switches on the other end.
See the pictures to the right: the led lines come out of the IC. They are just 'amplified' versions of the xled lines leading back into the GPIO connector. The chip does nothing but buffer the signals out of the Pi's GPIO, in other words.
It makes sense to read the Technical Details page for more background, but the above explains all the wiring on the PCB. This PiDP-8 explanation will also be useful. To the right are the pinouts of the GPIO connector and the IC, click to zoom. Assumes you are looking at them from the front of the PCB.
Theoretically, the IC can be bad. Has not happened in the last batch of 1500 IC's I have used (since I started sourcing Japanese chips from Mouser, instead of cheap chips from you-know-where) so discard the possibility for now.
You might have the IC mounted with its notch facing downwards though - in which case, only one row of 12 LEDs lights up during the Lamp Test. The IC will not have been damaged, just put it in the right way.
Theoretically, a switch can be bad out of the factory. Only happened twice in 50,000+ switches, so unlikely. Except... if you grossly overheated it. That has happened a few times since I started recommending the 'solder only 1 pin, reheat&reseat if necessary'. Hmm. Check the switch with a multimeter, or just short its pins with a bit of metal to see if the PiDP responds to that. But you will get confirmation from the different way it clicks. Spare parts available, they're just $1.
All SR switches work, but not all at the same time
In other words, if you flip up all SR switches, one or two remain undetected. Individually, they all work fine.
Aha! This has happened a few times (out of 2200 kits) on the PiDP-8. Probably, there is a high-resistance (partial) short on your PCB. But we never found where, and you do not need to bother finding it. Just replace the 1K resistor (one of the three, that caters for the particular group of switches that is affected) with a resistor anywhere between 500 - 900 ohms. The 1K value was chosen for minimum current use, and should be sufficient. But if you have a small 'leak' somewhere, the value is too tight. Don't overanalyse (we already did), just use a lower-value resistor. And please let me know.
Bad Raspberry Pi
It is possible (but unlikely) that one or more of the GPIO pins on the Pi itself are bad. It has happened, but it is so unlikely that you should discard the possibility until the end. If, however, you are at the end, either try another Pi or rig up a test without the Pi.
Test without the Pi:
Take out the IC, and bridge the six led lines from the left-hand side of the IC to the right-hand side with pieces of wire or just a staple... By now, you'll have checked the high-res PDF on this page (and the schematic PDF from the Technical Details page) and you'll know what I mean. The IC just acts as an 'amplifier' for 6 signals from the GPIO, coming in from the left, and sends them through on its right-hand side.
Use a 5V power source, any will do. Put the + end on a led line coming out of the gpio connector. Put the - end in the relevant column pin of the GPIO connector. The LED corresponding to that led/column combination must light up. If it does, and it does not when you use the Pi - you know the Pi is bad. You'd be the first, except for me. But I did something dumb to fry that Pi.
Word of caution
Almost always, you suspect your problem is a bad IC or a bad Pi.
Almost always, it's not.
It's a bad solder spot or overheated&broken trace. Problems 2 and 3 at the top of this page will be your problem. And using your multimeter from point to point using the PDF is your solution.