Your Camera Has a Shutter? How Quaint.

In the course of my duties as a faithful minion of the mad scientists I'm apprenticing under, I've been privileged to learn about a few things hands-on. A year ago this post would have been about adapting boat radar about as strong as an iPhone to track aircraft. (As you might guess, this endeavor failed.) Now it gets to be about something way cooler: cameras with no shutter.

First, some quick background on why I'm using this bizarre contraption. I'm currently tasked with analyzing data from the Air Force's CCD/Transit Instrument (CTI); it sounds cutting edge, but the data I'm working with was taken when I was 4. It was cutting edge at the time, however - the first shutterless telescope ever. Usually with a telescope you want a deep (long exposure) image of a single patch of sky. Not so with this. The CTI was meant to be an unblinking eye gathering a continuous image of a particular declination (latitude) of sky. It would operate the whole night, just looking up, unmoving, and produce a continuous strip a couple degrees high and a whole night wide. Watching without interruption, night after night, gave unparalleled data on variability of stars. (Why this is so important would take a whole blog post of its own).

The very high-tech CTI

Keep in mind, these continuous exposures aren't a video; they're stacked to form one still image of the whole stretch of sky observed. So how the hell do you do this? A nifty trick called time-delayed integration (TDI).

The exact details of what happens on the chip aren't important. Just picture it like this: on the right you have a single column of pixels that's exposed to the sky; that's where you get your image. But next to it you have a bunch of columns not exposed - here's why. Each time the chip's clock cycles, it kicks all the electrons (light values in CCDs) accumulated to the left, from the exposed column to the first unexposed column. The electron count in that imaging column is reset to zero then - it's blank and free to record another image. The unexposed columns can then be used to read out the pixel values without mucking up the exposure itself. So what you get is a single column of pixels taking a constant exposure, but the exposure gets reset and read out every n milliseconds.

(Alas, I can't seem to find a good image illustrating this on the interwebz.)

The individual images I'm looking at are in some ways single exposures, the exposure is just broken up into bite-sized strips a tiny fraction of a degree wide. Of course, it's more complicated than just pulling up the image and looking at it. Right now what I'm focused on doing if fixing errors in the calibration of these images. Probably errors. Well, we think there's something wrong but can't prove it yet. At least, it looks suspicious.

Naturally, this means put the new girl on it. Whatever, I get to science with really cool data from really cool hardware. For science!