Some answers from a Facebook post:

“If we consider only one system, then keeping everything constant is the most important choice for lightcurve accuracy. Any change will hurt, partly due to the non-linearity of the chip, but also due to nasty secondary effects for example colour dependence in extinction. Unfortunately some things won’t stay constant, particularly the weather. And with heterogenuous lightcurves from lots of systems we don’t have a choice, we have to take out most of this in processing, as much as possible. But with one single system lightcurve precision of 0.01 mag is very realistic over the course of a few days or weeks.” AS

“If you know your detector well, I would avoid having any interesting stars in the range beyond linearity, which is worth finding out (and an interesting exercise). We usually. need an ensemble of stars, so, you can’t optimise for one star. But you can avoid the worst ranges of the chip and set the exposure time accordingly. Excellent tracking and autoguiding is a good idea – don’t dither and avoid drift, keep the stars on the same spot of the detector.” AS

“Good calibration frames are vital. Darks and Bias frames – simply take a large number of those and median average them. The most import are the flatfields. Taking good flatfields is difficult. But if they have gradients in them (from the sky itself, esp. if you have a large field of view) then this will introduce noise we currently cannot correct for, even in the post processing. Make sure you have up to date flatfields as well as things change over time. Make sure your flatfields are flat. If you have two sets then divide them by each other. If you get systematic differences if the result that are not 1.00 then you don’t have a good flatfield! A value of 1.01 or 0.99 mean you are causing 1% errors in the photometry.” DF

“Magnitude ranges: Most of our data typically has stars brighter than 10mag saturated. Most images then range down to between 16 or 18mag. This works quite well for the targets and science we are doing. We are interested in getting light curves for as many young stars in the target regions. Most regions are closer than 3000 light years, Typically a bit closer. Stars with Solar luminosity at 3000 light years distance are about 15 mag. So our typical data generates good light curves for stars in those clusters that are up to 100 times as bright as then Sun, and we also get good detection of stars with as low as 10% of the solar luminosity. That covers almost anything there is. Yes on the faint end we miss the very low mass stars, but we knew this from the start. On the bright end: Massive stars that bright are quite rare hence we only miss a few objects – but yes, they are interesting.” DF

“Exposure times: As a rule of thumb: A 40cm telescope with standard CCD/CMOS/DSLR will typically saturate a 10mag star in 2min if your seeing is about 3arcsec and you have 1arcsec pixels. It is actually quite difficult not to saturate star brighter than 8-9mag with such a large telescope. Our calibration procedure will show you which range of magnitudes the stars in your images have. You can then adjust exposure times if needed. The calibration automatically flags up stars that are saturated or in the non-linear regime. Such measurements can then be easily excluded from any analysis.” DF

“Consistency is everything… Simply not changing anything (other than up to date calibration frames) for years is the key. We can then extract such subsets of data and improve the calibration. Thus, if you have limited observing time (as we all have), it is generally a good idea to stick to the same target (and multiple filters) to build up a good subset of data – this is what we aim to do with the LCO sub-project.” DF

“Accuracy: Our detailed analysis of some example stars (e.g. V1490Cyg, V1598Cyg) shows that our post-processing and calibration results in typical uncertainties of a few percent (0.02-0.05mag) for stars of roughly 15-16mag. This is good enough to analyse outbursts, typical eclipses by binary companies or disk material, and rotation periods. For brighter stars the noise is of course typically lower.” DF