[Aavso-photometry] Multi-color photometry

[Justin Pryzby]

On Thu, Nov 03, 2005 at 03:37:41AM -0600, Brian C. Barnes wrote:
> I'm just starting to get into doing photometry with filters other than "V",
I'm new too, lets see if I can say anything useful..

> 1) Assume I have B, V, and R filtered images of M67 along with B, V, and R
> images of a target. I can create transformation coefficients for B-V, and
> using those, I can calculate B and V. I can also create a transformation
> coefficient for V-R, and I can then calculate R by calculating V-R and using
> the V value from the B-V calculation, or I can calculate V-R and then V and
> R individually just like I did with B and V. This would give me one B value,
> 2 V values, and 1 R value. I can then average the 2 V values to get a final
> V value. Is there an advantage to calculating this extra V value and doing
> the averaging?
It seems to me that if you *have* B, V, and R data, then there is no
reason to compute transformation to those bands.  Transforming from
one band to another is an approximation, of course, and real data is
always better.  Also, transformations assume blackbody-type emission,
an assumption which some objects violate.

The only reason one transforms from one band to another is if you
don't have data in that band, and you want to compare with values from
an observation or star list which does have data in that band (but no
data from the bands for which you do have data).

It is my understanding that even when you do observe with a filter,
you still have to do some correction to put yourself on a some
standard system.  Of course, the 1st order correction is just an
offset ("zeropoint").  You necessarily have to add *something* to the
log(object_flux/exposure_time) to go from instrumental units to any
semblance of standard units.  Then you might make some small
correction to convert from "your filter" magnitude to "standard
filter" mag (which is ideally zero, but never happens).

As best as I can tell, all of photometric calibration is iteratively
figuring out the source of the largest remaining trend between your
predicted mag value and the accepted values, and then fitting a linear
least-square slope to a line such as to optimally correct for that
trend.  Stop iterating when the slope of the line is sufficiently
small such that the largest correction you would make for your dataset
is smaller than your tolerance for systematic error.

The transformation formulas are either empirical observations specific
to someone else's instrument, or derived from some theoretical
calculation.  Instead of computing data points for which you already
have *real* data, it would make much more sense to compute empirical
transformation formulas for your own equipment (if for nothing else,
than to figure out how to do it).

I'll stop now, hopefully someone with experience will jump in.

-- 
Clear skies,
Justin