A comparison of some astronomical detectors

by Brian Skiff
October, 1996

As a major user of photomultiplier tubes for photometry of stars, I'll join the discussion here about eye vs. photo-tubes vs. CCDs:

Mike Richmond correctly pointed out the problem about fields of view when trying to observe bright stars with CCDs. The "obvious solution" of using a short focal length lens to get a wide field with comparison stars doesn't work, since it turns out the the precision of the measurements declines when you have a star image occupying only a single pixel. Even though scales around 2 arcsec/pixel are suggested for imaging of big deep-sky objects, you need something like 1 arcsec/pixel or better for doing photometry of stars, or more specifically, you want the star to cover at least several pixels.

For the problem of bright stars, the increased quantum efficiency of CCDs is of no help, since there are already plenty of photons. Indeed, the limiting factors in precision are things like scintillation (worse for small apertures), small tranparency variations (site dependent), and flat-fielding errors (acute problem for CCDs). Whereas there is plenty of traditional single-channel photometry published that's limited mainly by the precision of the standard stars, there has yet to be published a paper with CCD "all-sky" photometry that has an external accuracy better than about 0.01-0.015 mag. The best such CCD work I've seen is by a guy named Alistair Walker, a South African now working at CTIO. The best differential photometry (where comparison stars are on the same CCD frame as the target stars) has been done by Ron Gilliland, whose study of microvaribility in solar-type stars in M67 set the standard in this regard. He used 4-meter class telescopes on 13th magnitude stars, took literally tens of thousands of frames (he had to use a Cray machine to process them!), yet his rms errors per observation were about 0.008 mag. And this was despite the fact that he had not just a few but scores of comparison stars on the chip. In a set of 1-minute integrations on two stars using the Lowell 53cm photoelectric photometer, I get rms errors of 0.002-0.003 mag., and can "reduce" the results in about 20 seconds using a pocket calculator. So on bright stars, photomultiplier tubes win big, not only in sheer precision, but also in reducing the data to magnitudes.

Recently Robert Mutel (Univ. Iowa), who runs an excellent small telescope CCD facility (it's a 7-inch Starfire refractor with an ST-6), asked me about observing Algol for a project to look for some kind of relativistic effects. Despite having a 26 arcmin field of view on his camera, there are no stars less than _nine_ magnitudes fainter than Algol appearing in the field. Although he can observe mag. 12 stars without difficulty, getting enough photons for them without completely blasting out mag. 3 Algol (and the chip) is not possible, since the dynamic range of the typical chip is around seven magnitudes between saturation and not-enough signal.

So I see CCDs and photomultipliers as complementary: there are things one can do that the other cannot. If you want high precision, you go with tubes, if you need to go faint or do surface photometry, you go with CCDs. For similar dwell times and moderate precision, you gain about two magnitudes with a CCD in the limit compared to a first-rate tube: one magnitude comes from the raw "speed" (quantum efficiency) of the chip, the other comes from having taken the sky background readings at the same time and more accurately than can be done with a tube. Yes, small telescopes reach 20th magnitude or something with a CCD, but the noise level is about 1:1, and photometry of such stars is useless. Amateur observers rarely observe with filters, which are _required_ to make the measurements useful, but inevitably reduce one's limiting magnitude. On top of that, most variable stars (or asteroids, etc.) you'd want to study vary fast enough that exposures need to be no longer than about 10 minutes in order to sample the lightcurve adequately, which puts another crimp on the faint limit. (All this is part of the "no free lunch" principle.)

The eye and photography still have a place in astronomy: all the decently bright comets are discovered by amateurs either visually or photographically (a few CCD comet hunters are out there, too); following Mira variables is done almost entirely visually; supernovae are still discovered visually (one just last week, in fact). The best thing about the eye is convenience: you can be out there for quick scans much more often than is likely even with an automated telescope. The eye isn't real accurate for photometry, but the "image- processing" is done with amazing speed, and the knowledgable user can get reliable and consistent results.

Brian Skiff (bas@lowell.edu)

Back to Michael Richmond's home page