[Aavso-photometry] Differential photometry for dummies

[Michael Newberry]

Hi Brian,

Sorry, but I don't think you quite have the whole picture yet. Combining 
images is done both for pretty picture and for images that are to be used 
for science ---  to be inspected or measured, such as for photometry. But 
photometry just adds a potential issue regarding calculation of the 
uncertainties, and you can make that one go away. It is FINE to combine 
images for doing photometry. You may need to or you may not need to.

By registering and combining images, you'll get higher S/N, smaller 
magnitude uncertainties at a given brightness, and you'll be able to go 
fainter. Conversely, if your project involves rapid variations, the 
combining of images may  not give you the time resolution you need. My point 
was simply that you have to be careful to use the right software resampling 
method when you do it, else you'll not get the right magnitude sigma's.

The photmetric uncertainty (the sigma, or random error) is estimated by the 
software. Depending on the software, it can be all over the place (meaning 
more nearly correct or less nearly correct). It is not that using 1/SNR (or 
actually 1.0857 / SNR) is a bad approximation,  but rather the devil is in 
the details of how the SNR is determined.

Michael Newberry

----- Original Message ----- 
From: "R. Brian Potter" <potterrb at comcast.net>
To: <aavso-photometry at aavso.org>
Sent: Sunday, June 17, 2007 2:58 PM
Subject: Re: [Aavso-photometry] Differential photometry for dummies


> So I totally understand that I should not combine images, but rather 
> analyze
> the individual images after subtracting a master dark created from at 
> least
> 7+ images.
>
> My questions remaining are
> 1) How do I measure signal-to-noise (specifically in CCDSoft if someone is
> doing this)
> 2) How do I provide an error estimate for my results? (Std deviation on
> multiple measurements, 1/SNR, ???)
> 3) The CCD Observing Manual states that the observed K-C should be equal 
> to
> the known K-C.  Using the 9.8 mag star in the field for R Uma as my comp
> star gives a value of 8.04 for my check star, vs a known value of 8.2. 
> How
> close is close enough? If its not exactly equal to the known value, do I
> make an adjustment to the magnitude of my target star to account for this?
>
> Sorry for all the questions, but I want to make sure I do things correctly
> for my first CCD submission.
>
> Thanks,
> Brian
>
> -----Original Message-----
> From: Michael Newberry [mailto:mnewberry at mirametrics.com]
> Sent: Sunday, June 17, 2007 12:58 PM
> To: Michael Koppelman; R. Brian Potter; aavso-photometry at aavso.org
> Subject: Re: [Aavso-photometry] Differential photometry for dummies
>
> I totally agree with Michael. And I would add one more critical point:
>
> Do not allow resampling of the pixel values for the combined image. If you
> do, then the magnitude error estimates will be wrong, Not noly does
> resampling decrease the noise by square root of N image that are combined,
> but it also will smooth or otherwise corrupt the noise structure of the
> combined image. The result is that the magnitude uncertainties determined
> from the noise in the image will be underestimated. For that reason, we
> included in Mira a "nearest neighbor" option for the pixel resampling. The
> software computes and applies the transformation to register the images 
> but
> then combines the "whole" pixel values from the image set that are nearest
> each pixel location. In other words, there is no splitting of pixel 
> values.
> The result is that the gross adjustments are applied but there remains 
> about
> a 0.3 pixel "jitter" in the pixels that are combined at each location. For
> photometry, this inflates the FWHM by a quarter of a pixel or so, on
> average, across the frame. But the benefit is that you perfectly preserve
> the ransdom noise structure of the images you combined, hence your
> photometric errors are good.
>
> Michael Newberry
>
> ----- Original Message -----
> From: "Michael Koppelman" <lolife at bitstream.net>
> To: "R. Brian Potter" <potterrb at comcast.net>
> Cc: <aavso-photometry at aavso.org>
> Sent: Sunday, June 17, 2007 9:44 AM
> Subject: Re: [Aavso-photometry] Differential photometry for dummies
>
>
>> Most people would say you should not align or combine the images.
>> Like Tom said, just do photometry on the individual images and then
>> combine those mathematically. The exception is if the object is so
>> dim that useful measures are not obtained from individual images. In
>> that case, you could align and combine, but you should make sure your
>> alignment routine is "flux conserving".
>>
>> M.
>>
>>
>> On Jun 16, 2007, at 11:33 PM, R. Brian Potter wrote:
>>
>>> 3) Align individual images after dark subtraction
>>> 4) Median combine images to create a single processed image of the
>>> field
>>
>> _______________________________________________
>>
>> Aavso-photometry mailing list
>> Aavso-photometry at mira.aavso.org
>> http://www.aavso.org/mailman/listinfo/aavso-photometry
>>
>>
>
>
> _______________________________________________
>
> Aavso-photometry mailing list
> Aavso-photometry at mira.aavso.org
> http://www.aavso.org/mailman/listinfo/aavso-photometry
>
>