[Aavso-photometry] GJ 436 Campaign Web Page

[arne]

Wolfgang Renz wrote:
> Hi Arne
> 
> So you recommend using 2 filters to be able to transform.
> But how high should be the "image ratio" between the two filters ?
> 50:50 ?
> Then one will get effectivly a B and a V lightcurve. But as the color
> of the target should not change, one will loos data in one filter that
> might be required to reach the needed precisition by appying a
> box-car average to the data of one filter.
> In real photometric nights, wouldn't it be better to image primarily
> in B and take a deep V band image just every 4th, every 6th or
> every 8th image and us a smoothed, interpolated V band mag
> for the transform ?
> 
The transits of GJ436b take about 50 minutes.  The important points
are ingress and egress if you want accurate timing of these flat-bottomed
events.  I'd say a time resolution of 30-60 seconds is adequate, concentrating
on high signal/noise at the expense of time resolution.

Few people will have enough signal/noise in a short exposure to also
be able to switch continuously between filters.  I think staying in one
filter or another (and I recommend either B or V for this), with a few
images with another filter during the series, is adequate.  This is
basically what Wolfgang is saying.

> In non-photometric nights (e.g. with slight cirrus, changing extinction,
> ..., or even worse) it might be better to use 50:50 to be able to reject
> outliers. But under non ideal conditions it might not be possible to
> reach the required precision at all.
> How good must the night be to make sense to start with GJ 436 ?
> 
Tough question.  I've done a lot of precision work under varying
conditions, and while differential photometry helps enormously to
remove the effect of clouds, it is not perfect.  The assumption is that
any clouds affect all stars equally in the field, and this assumption
cannot be perfect; there is structure in most clouds at the 0.01mag level
or worse.  In addition, clouds cut down the stellar flux so your Poisson
noise changes from image to image.  If I were to observe GJ436, I would
hesitate using anything but a nearly photometric night (light, scattered
cirrus perhaps).  However, I'm a perfectionist.

> How many bias/dark/flat frames do you recommend to take to not
> introduce to high errors by the calibration frames and to be able to
> reach the required precision in the few mmag range at all ?
> 
Depends a bit on how many pixels are contained in the stellar profile.
My usual rule of thumb is to get 1M electrons per pixel in the flats;
this corresponds to a 1mmag error per pixel from flatfielding.  Most
stars subtend 10 pixels or so, making the flatfielding error about
0.3mmag; relatively insignificant.  What precision you need in flats
depends on your project, and this one requires pretty deep flats.
Note that you are more concerned with variations from pixel-to-pixel
and less so about gradients across the frame.  As long as you keep
the stars tracking close to the same pixel all night, gradients
only introduce a zeropoint offset to the entire dataset.

Darks are a little harder to estimate, as getting a
high precision master dark is great to remove the mean dark level, but
the dark noise from the science frame will remain, and so errors in the
master dark are usually insignificant compared with the noise contained
in the science frame.  Still, darks are easy to do on cloudy nights, and
if your temperature regulation is reasonably good, master darks are
close to the correct adjustment.

The bottom line is that the noise in a measurement is dependent on dozens
of individual sources.  You have control over some of these sources,
such as getting high signal/noise in your flats, and you should do your
best job on those noise sources.
Arne