Week ending Dec 29, 2013


The output interface box was not working properly - when I hooked my Bus Pirate directly to the box and sent I2C commands, it became obvious that I had an error. I am rebuilding the board and ensuring it works correctly, before I put this back into service and resume debugging.

The new output board is working correctly, when driven by the Bus Pirate and hand entered commands. It is still not lighting when driven by the fpga, thus my suspicion will shift to the input box, which may be corrupting the signal, or the fpga logic driving the link to these chips.

After a bit more exploration, it became obvious that the fpga or bus pirate is not registering any zero levels on input coming from the MCP23017 chips over the serial cabling - it always sees the output of the chips as a logical 1. This seems like an analog/electrical issue once again. Oddly, it can detect a nack if a bad chip address is sent, but any data read appears to be all ones.

I had a suspicion is that the voltage divider formed by the pull up resistors which are at the fpga end and the series of resistances from the MCP chip across the cable are producing a voltage that is above the max threshold to recognize a logical 0. I relocated the pull up resistors to the input box.

There are other wiring problems inside the inputs box, caused by the waves of reworking and redesign applied to this pcb. There are a couple of areas that are obvious weak points, either causing the current problems or very likely to generate issues in the future. I will do some more rework here, eliminating those weak points before I get back to the debugging.

At worst case, I will build a new inputs board including the modification for the interrupt, yanking out the suspect board.I tried a new design for an input board - one board per MCP23017 supporting up to 16 input channels. I would use two such boards in the card reader box, to handle the 22 lines from the reader itself, the three input switches and the interrupt mechanism.

The old input board now smokes when I apply power. I can't see the exact part or trace but it convinced me that I need to replace the board with the new ones.

I was having some difficulties with board production - one side never completely clears, almost as if it is fogged with general light contamination. It is the first of the two sides I exposed, which further suggests that the 'safe light' I was using is not suitable. After I buy additional supplies at the local shop, I will adopt much more stringent measures to see if I can clear up the problem.

I picked up a used darkroom safelight and attempted fabrication of the board again. The process worked perfectly this time.The only improvement I can make to the process is to darken the toner on the transparency, as that would make circuit traces more solid. They appear tenuous but almost all passed signals well; a very minor touchup gave me a usable board.

I installed the three surface mount HC4050 level shifter chips and the surface mount MCP23017 main chip, wired up all the external connections and set it aside while I build the second board. The first board I produced handles 14 input signals thus needs all three level shifter chips. The second board will handle just eight signals, one of which is the interrupt A line from the MCP23017, which will require that just two of the level shifters be installed along with the main chip.

I have just enough stock on hand to complete the second board, but will have to do that work tomorrow. When it is prepared, I can cut the two apart, drill mounting holes and install it into the input box in place of the failed and smoking board.


I saw a deal I couldn't pass up, giving me two Tektronix 7306 mainframe oscilloscopes, 13 plug-in modules of various amplifier and timebase types, and a mobile Tektronix cart with storage space for four of the loose plugins. I could sell just part of this and come out ahead compared to the embarrassingly small amount the seller asked. One of the scopes has a problem with one power supply level, which I have begun to diagnose, a defect that was clear in the advertisement.

The low voltage power supply produces regulated power at -50V, +50V, -15V, +15V, +5V and +130V levels, but the +15V section outputs over 23V. Another anomaly is that the rectified but unregulated power for all the sections is too high. The schematic suggests value such as +/- 53V but I am seeing +/- 63. Similarly, the +/- 15V feeds are running about +/- 24V.

This scope has jumpers to configure the transformer primary windings for low, normal or high line voltages by adding or subtracting windings, but when Iooked my scope already had the jumper on the high setting to produce the lowest output voltages. This suggests that the the transformer has issues - probably there is a short that is cutting some of the primary windings out of the circuit, thus increasing the step up factor to yield this undesirably high voltages.

As long as the regulators can handle the power involved in dropping the excess voltage, it may be okay, but I suspect that the defect in the +15V section is a failure induced by the excessive input voltage to that section. If so, other sections are being strained and may fail. The ideal solution is a new transformer, although it may be economically infeasible. I can use my variac to pull the line voltage down far enough to get the power feeds to the target levels, although the variac is big, heavy and unsightly - not a long term solution if the scope would get regular use.

I bought a digital counter plug in on ebay, as this is a convenient tool to measure frequencies, time durations and event counts. The other plug-ins are single and dual input amplifiers of several speeds and timebase units with and without delayed trigger/sweep capability. The plug in needs a scope with a 'readout', a feature that displays characters on the CRT as the status and output of the counter. My OS 245 (the military version) does not have readout, so the plug in won't work in it, but this should be useful once the 7603 is repaired.

In addition, a curve tracer kit I ordered on Ebay came, I soldered it together, and I now have a curve tracer to plot the response curves of transistors. These are handy to deal with the pesky analog issues I will continue to face. The initial power supply I bought, that should be converting the power from a 15V AC wall brick to yield regulated +15 and -15V, is not working properly, giving me +15 but only -9V. The regulator came preassembled, not as a kit, but was very inexpensive.I suspect the 7915 regulator chip is not working properly, but don't have the time right now to chase this down and fix it as it is not a priority.


The only safe option given the current state is to remove the light panel assembly from the enclosure, reglue it together and make the changes needed for the plexiglass panel's fasteners to remain clear of the assembly. I will take this on during the next week.

No comments:

Post a Comment