Don’t trust the machine

By Marcus Wilson 28/04/2014


Back to blogging, after a nice holiday in Taranaki dodging the rain showers (and, as it turned out, the volcano, which we never even got a glimpse of) and a frantic week of lab work while the undergraduates were away. 

Both were very interesting, but it's the lab work I'll talk about here. 

Something that I've learned over the years is that if something looks dodgy, it probably is. Obviously, when doing experimental work, we don't know what results we are going to get. (If we did, we wouldn't bother doing the research). It is true that sometimes results can surprise us. Sometimes this is the start of a discovery of a new phenomenon, which will make the experimenter famous. But more usually, much more usually, it's because you've stuffed something up in your method or analysis. If your data just looks wrong, it probably is.

We had this with our conductivity measurements in the lab two weeks ago. We were using a moderately high-tech (approx 10k NZD or so) piece of equipment to measure electrical impedance of our samples of biological tissue. The results were odd – we had an unexpected jump in conductivity as we changed frequency. It took a while to track down what the problem was. First, I talked to a colleague who used to work for the company that made the equipment we were using. He hadn't seen anything like it before, and offered a few suggestions as to what we might do to track down what was going on. There was the suggestion that it might even be a calibration failure in one of the machine's internal circuits. 

We did a few tests, and were still puzzled. We tested progressively simpler and simpler things, trying to isolate the problem. It was a good exercise in troubleshooting, really, and it took a while. We ended up with testing the impedance of just a single 10 ohm resistor. We didn't expect this to be an issue. But, when the machine told us that it's impedance was 8 ohms for frequencies lower than 121 Hz and 6 ohms for frequencies above 121 Hz we knew something was terribly wrong somewhere. Then the machine refused to work altogether. At this point the thought of ten thousand New Zealand dollars going up in smoke in front of our eyes did cross my mind, but only momentarily, since suddenly it kicked into life again and started reading 10 ohms. Then I just touched the front and it was 8 ohms. A bit more experimenting quickly narrowed down the problem to a dodgy lead.

That was all it was. One of our coaxial cables had a dodgy connection on it. We replaced the lead, and suddenly the results are all perfectly believeable again. Only two days of work for two people completely wasted by a cable worth only a few dollars.

The moral of the story: It pays to do some really simple tests of the equipment every time you use it. Don't just blindly trust the readout on the machine. Check it's working first. I recall now the words of one of our (now retired) technicians here – "If you have a perplexing problem with something electronic, it's a fair bet it's simply a dodgy connection." How true.