Today, Holmes and I closed the loop and showed that a multi-epoch survey with random image centers and orientations crushes a multi-epoch survey done on a regular grid for precision of any self-calibration. Next week we plan to make the simulated data more realistic, add imperfections to the model (in real life the model can never exactly capture reality), and find out the scalings of photometric accuracy with density of standard stars, number of epochs, and survey strategy decisions.
Very interesting! So random beats grid, but might some well-chosen non-grid pattern beat random? This reminds me of how in radio synthesis imaging, regular polygons of antennas are worse than random placement for some purposes, but there are better choices still.
ReplyDeleteDave, I agree! In fact, if you are very, very insane, you could imagine optimizing the observing pattern to optimize calibration quality.
ReplyDeleteThat sounds like your sort of insanity.
ReplyDelete