A Detective Story

About to run the Final Qualifying Round for some second generation DIY Spectro, I placed the first blank into the cuvette holder, and pressed start (or rather ran python tryna.py; I’m tryna measure a spectrum, gosh!). The machine hummed into action, now that the motor control wire was plugged into slot 9, which the computer was communicating with rather than slot 2, which the computer was not.

A few minutes later I got the results: Nothing. The machine was not seeing any light. At all.

See, normally when I scan a blank sample, the detector (which has different sensitivities at different colors) shows a characteristic hill shape:

Some typical blank sample runs. Horizontal axis is motor position in degrees (0-180) and vertical axis is detector response. Black is the mean of the time series coming from the detector at a given motor position (1 sec data per degree); green is the standard deviation of the time series, and red denotes the maxima/minima. Time series were preprocessed to remove annoying serial communications glitches.

But when I ran this blank:

Oh noes! Graph and data conventions as above.

So, even though the motor is running and the flashlight is on, the detector thinks that it’s in the dark! Worse, it can’t seem to tell when the light source is shining on it, and when it isn’t!

Is the flashlight bright enough? It certainly is painful to look at directly. Is the detector broken? What about the diffraction grating?

Well, the multimeter said that the detector is okay. And I got the same issue when I replace the diffraction grating (which reflects relatively diffuse light) with a chip of mirror, which sweeps very intense light across the detector, like a lighthouse beam. Still nothin’.

Was it in the microcontroller board? The connections all looked good. Was it in the signal or data processing? Doubtful- why would a computer program work well yesterday but not today?

So I took my bike out. My bike is named Polished Steel, because they say that Pancho’s Horse was fast as Polished Steel, meaning that Polished Steel is actually faster. If Pancho’s Horse was faster, we’d say Polished Steel is fast as Pancho’s Horse, and maybe he wouldn’t have met his match in the Mexican desert. At any rate, the sky was dull gray and the weather swung back and forth between raining to raining hard. Rivers in the sidewalk with nowhere to go, itchy glass crystals of buildings poking through the skin of the earth, humidity rising.

And I thought about my days in the Carrboro ReCybery, pulling IR sensors out of teevees, defying the local home moaners association by saving children from malaria…

I got back and got back to business. First, I collected three lighthouse runs, and then I changed the batteries in the flashlight and ran three more. Bingo.

Here's six 'lighthouse' runs: three with questionable batteries (lower), and three with new batteries (upper). Graph and data conventions as above.

Although the flashlight seemed bright to me, the detector has a different outlook. I had the same problem experimenting with an IR sensor I was trying to calibrate. It looked for all the world like it was responding to visible light- but it wouldn’t respond to a white light LED! Drove me crazy. Finally I realized that it was only responding to visible light that’s verrrry close to the IR region, which were missing from the flashlight. The realization only came after perturbing a magazine subscription solicitor who found me on the porch frantically waving a 250 degree skillet around a box of electronics and cursing.)

The old batteries were pumping out ~3.1 volts, rather than ~4.8 from the fresh set. It didn’t look any different to me, but my eyes are made out of cytoplasm, not cadmium sulphide. To the machine, it was a world of difference.

Not only did I see the difference between the old batteries and the new, I was even able to see the old ones running down over the course of ~15 minutes! The three most recent runs are of diminishing size (and hence detector response), and the batteries afterwards gave only 4.4V.

case closed………………………………….