ECL/TTL translator board diagnosis 4/16/05
The box experienced a total failure in Switzerland. All channels stopped passing signals, and the two translator boards within had overheating problems. All of the chips on the boards were swapped out at PSI. Since the rest of the system was left in Switzerland, I was told to find the problem and fix it. The system would need to function indefinitely without attention.
I used a digital oscilliscope, an arbitrary waveform generator, and about five multimeters. The boards were left plugged into the original power supply. The generator was connected to the ‘scope and to a plug that would fit in any of the 34 pin inputs. An ECL(Emitter Coupled Logic) signal and its inverse can be sent to any of the sixteen channels on the board. There was some distortion of the signal, but it worked.
The incoming ECL signals were square waves, which went from –2 to –1 volts. This signal is sent to the base of a transistor. The signals are sent with their inverses through twisted pair wiring, and triggered when the two signals crossed. This prevented the signal from being affected by outside noise. Since the base of the transistor is always on, this the transistor can switch faster. The incoming ECL waves are 10 ns long.
TTL (Transistor Transistor Logic) signals are square waves from ground to 5 V. Off the board the signal is sent with its inverse for the same protection from outside noise. This signal turns on a transistor connected to an LED and a capacitor, which then discharges through the LED, firing off some light, which is then detected like a decay event. This signal switches transistors slower than ECL signals, since the difference is greater and the transistor is all the way off.
The first problem was overheating of the four ECL/TTL (Transistor Transistol Logic) translator chips(74_125). If any one or two of the chips was on the board, those signals would work just fine. Current through the ground to the power supply was about 0.08 amps, current through the +5V power supply was about 0.16 amps, and –5V was 0.08 amps. If three or four were plugged in, all four would overheat. Three chips would cause currents 0.56, 0.64, 0.08 amps with some variation depending on which of the chips was missing. Four chips about 0.96, 1.04, and 0.08. These currents were erratic and could change by about a tenth of an amp, except the –5V supply. The only chips which used the –5V supply were the translator chips, which were overheating. Everything else drew exclusively from positive voltage source but was not overheating. The circuitry would start to smell like it was burning in about thirty seconds. All tests were thus done in twenty second increments.
The second problem was that the signals would no long get translated and passed. The signal would be translated correctly by the three overheating chips and then fail at the next(74_02). One of the 74_125 chips was not translating a signal at all. All translators were swapped out again and all still overheated but translated the signal. None of the second or third line of chips had over had a noticeable heating problem. If the input signal at the 74_02 (the second in the path of the signal) was crossed with the +5V supply once, primed, the board would then pass the signal just fine until it had to be turned off to avoid burning out the translators. One of the All channels were checked for output, and every channel on the first board had the same error. The channels of the second board were checked, many same problem, but one channel was not passing the signal at all. All translator chips were working in the same way as the first four. All were replaced, and the problem persisted.
All of the soddering was checked on both boards. Something was fixed in the channel that did not work at all. Now all channels were failing in the same way, on both boards.
Power depends of current squared, so the extra current could easily cause the heat. But, I thought, the chip was functioning, and if the power supply was failing, it would probably put out less voltage, which would lower the current, this lowering the power, and should not cause overheating. Instead, the board would just fail, being unable to create a TTL signal.
The power supply was replaced. The overheating problem stopped. I was wrong.
But the signal at the second line of chips, the 74_02, still needed to be jumped to pass the signal on all but two channels in the second board. Both this chip and the next were there to make sure the TTL signal went out for the right length of time, about 100 ns. The first had two flip-flops on it, which, when fed the inverse of the TTL signal, would turn the signal off. The other created the TTL inverse signal which it did when it detected the real signal. It fed the signal back, which told the flip flops to turn off. This took about 100 ns, which was the same length of time as the desired output signal.
One of the 74_02s was replaced with a new chip. Same problem. One of the 10_123s was replaced. The problem disappeared. Both channels which passed through. The old 10_123 was reinserted. The problem reappeared. The new chip was put in again. It burned out. Another 10_123 was tried. The problem again disappeared.
On the new power supply, the board was left to run for three hours. The chip did not burn out. The signal went through fine, without being primed. The translators did not overheat.
Current theory: the power supply started fluctuating when connected to a high impedance load, which caused chips to burn. The changing current made the multimeters give multiple readings, and overloaded some components. With a new power supply and all functioning components, the board works fine. This would only be a problem with a bad power supply, so the six boards that worked fine probably had fully functional power supplies.
There are two things left to do. The first is replace all burned out components on both translator boards. If the signal requires priming, the appropriate 10_123 should be replaced. The second is to determine the reason for the power supply’s failure. The voltage from the supply is fine in an open circuit. The power supply was connected to a potentiometer, capable of the 0.8 to 30 ohm region, and the voltage remained stable for about ten minutes and then suddenly fluctuated for half a minute before going to zero. Nothing gave off a burning smell. Afterwards multiple trials did not get any voltage at all.
At PSI, this problem can be recognized by high currents in the supply voltage and the signal failing at the 74_02 chip. The best way to fix it seems to be to replace the power supply and any 10_123 chips on channels that need priming. This would probably take less than a couple hours with an oscilliscope for an entire box. The 10_123s may be difficult to get, as they are listed as obsolete/discontinued by DIGIkey. Mouser Electronics claims to have them in stock at a cost of $2 per, more or less depending on quantity, made by Fairchild. No products of this company were tested because I do not have any. The websites of most of these people want more information then the IRS just for a price quote, but some of them do claim to have the chip in stock.