OK, as I wrote last time (lab fab season 2), the past week’s agenda was to find the noise. First, I replaced the microscope bulb with a new one and saw the same fluctuating intensity. OK, not the bulb. Then, by opening the bulb-port door and removing the bulb, I could illuminate the stage with light from an IR diode array. Diodes are famous for emitting light steadily, so this source bypasses any flakey bulb business. But alas, same fluctuations; unlikely then to be the lights.
As an aside, “same fluctuations” is phony precision. I did not measure them. Instead, I took advantage of the live image window in the Micro-Manager software that runs the camera. This window provides a histogram of the gray levels of the image and updates it in real time (or as close to that as it can). The window also shows the min, max, and mean gray level, again updated continuously. The histogram was hopping sidewise by one or two gray levels, some times farther, every few seconds, and comparable hops were occurring in the min, max, and mean gray level values. So when I say “same” fluctuations, I mean the sizes and frequencies of the hops looked more or less similar. And definitely large enough to cause trouble when trying to do things based on quantifying intensity levels between a pair of images.
Now, back to the story. With a wonky light source ruled out, I turned my attention to the camera and also to the computer running it. Maybe running the camera through the Mac-Windows kluge is unstable? I could test this because along with root imaging workstation with the microscope, I have a different workstation for imaging larger objects (stems, maize roots) that includes an identical camera (connected to a macro lens, rather than a microscope), but in this case driven by an actual Windows laptop. I never looked for intensity flickers because, here, they would not matter. This workstation is in a darkroom, a few doors away. Letting that camera run on an empty scene, with a similar exposure time (20 msec) as in the microscope, but through the Windows laptop, I saw the same fluctuations. This rules out a problem with the Windows emulation on the Mac and makes unlikely a defective camera.
What is going on? I asked the folks who sold me the cameras. They suggested using a shorter exposure time. It seems that 20 msec is plenty long for that camera and if there are problems with noise in the camera electronics (“dark current” and other mysterious gremlins) then noise would get worse as the exposure lengthens. OK, well that is easy to test. Working in the darkroom, I moved the light source (the IR diode array) closer or farther and compared 2 and 20 msec exposures. To have a record, I took a frames every minute for 5 frames total and then played them back as a movie. This made the intensity fluctuations brutally clear. And left no doubt that contrary to the seller’s suggestion, shortening the exposure made the fluctuations worse. How about a longer exposure? By moving the IR source across the room, I could get a 200 msec exposure and it was more stable, though still bouncing a little.
This tends to rule out the camera and puts attention back on the light. I started thinking about the electricity in the building (“mains power”). Even though the IR array is powered from a DC adaptor, maybe noise in the mains power gets through? Why not try an LED flashlight, battery powered? Happily, one of the office staff had one, and hey presto! stable images. Well not perfect, but with a 20 msec exposure, far better than the wall driven IR LED array. Wow!
This is as far as I got. Noisy mains power would account for the fluctuations in both rooms and one hears it said that the power in Morrill is flakey. But the plausibility of a hypothesis is no guarantee of its truth. The next step is try a power conditioner, a device that is designed to minimize noise from mains power. Fortunately, George Drake in the Morrill IT headquarters, has one. I can test it out this coming week.