The Case of the Good and the Bad Boards

By John Powell

I work at a company that makes drivers for liquid crystal variable retarders. I’m an EE and I am in charge of our electronics, including the controllers. When I’d first started, I encountered a perplexing issue with our basic controller. This controller had no microcontroller and had knobs to control the envelope on the 2KHz AC drive voltage to the liquid crystal (LC). This basic controller had been made for probably 12 years at this point, and the current revision for several years. With LC controllers, one critical specification is that the DC component of the LC drive voltage must be minimized. A typical specification for the controllers produced by my company is that the DC voltage’s magnitude not exceed 5mV. The issue with the controllers was that a batch of them were showing high DC offset when they came in from the contract manufacturer. I started to investigate the 2HKz generating circuit. The 2KHz signal was first generated by a ring oscillator circuit (an odd number of inverters with feedback and an RC circuit to control frequency), then the frequency halved with a flip-flop. Both the inverter chip and flipflop were through-hole 74HC chips (’04 for the inverter and ‘109 for the flip-flop.)  My investigation went as follows:

First, upon visual inspection, the solder on the ground pins of the chips looked dull, so I tried reflowing those pins as well as the grounds on the bypass capacitors. No change. The capacitors appeared different between a good and a bad board, so I switched them. This appeared to fix the problem, so a couple units were left on to check stability. Unfortunately, the problem re-appeared.

Next, I thought, “Do I have a bad lot of ICs?” So I replaced the logic chips. Still no change. The frequency was not drifting, but I tried reflowing and replacing the frequency selection pot and cap. That didn’t help either. What about a bad run of boards, since the board films are getting old? I checked resistance of various traces on bad and good boards, and found no significant difference. I noticed one of my good boards had a different style pot on the voltage adjustment knobs, but switching it out changed nothing. I also tried numerous other tests, including using 74HC14 schmitt trigger inverters (which have hysterisis, in contrast to the standard inverters in the 74HC04) to no avail.

By this point, a couple months had elapsed since we received the boards, and we needed to ship some.  I finally located as many “good” units as I could (we use some of these controllers internally), and opened them up to see what was different between the “good” and “bad” units.  To my amazement, the only thing I saw different was the brands of the 7400 series logic chips.  The good boards had brand “A” 74HC04s and brand “B” HC109s. The bad units, however, had brand “C” ’04s and brand “A” ‘109s. This got me thinking, so I perused datasheets. There were only slight differences in parameters, but there were differences, most notably in the propagation delay.  Therefore, I made a test matrix with all the brands of both 74HC04s and 74HC109s we had for spares.  That matrix had a lot of combinations. but was simplified because although the units had two 74HC04s each, I assumed that we’d only have one brand of 74HC04s installed at any given time.

The test matrix confirmed my suspicion that the brand of the chips made the difference. For most circuits, a 74HC04 is a 74HC04, and the brand name makes no difference. Unfortunately, here, the specs for the circuit are so tight that brand really was important. Fortunately, the test matrix yielded several good combinations of ‘04 and ‘109 chips. Replacement chips were ordered to fix the bad boards, and the BOM was updated and sent to the manufacturer.  Best of all, sales was happy again, as they could sell the product again.

The most important things I learned are that: 1. it is invaluable to be able to compare a “good” and “bad” unit when troubleshooting, and 2. that when specs are extremely tight, even parts that should be direct replacements have enough difference to cause issues.

John Powell has been interested in electronics since high school. He graduated from Colorado State University in 2004 with a BS in EE, emphasis in optoelectronics. He has been the electronics engineer for Meadowlark Optics for 6 years.