[Aavso-photometry] Errors on AIP

Wed Jul 14 15:55:21 EDT 2004

I think you found a typo. I was using 'ADU Comp' instead of 'ADU Sky' 
in the calculation of the noise, which is totally wrong. Thank you. I 
will update my spreadsheet online shortly.

In regards to the SNR formula, I know there are different ways to do 
it. I used an equation I found on the net. I tried to get Newberry's 
paper when I was doing this but NASA ADS was unavailable at the time. I 
would welcome suggestions on the best SNR equation to use.

Finally, yeah, I'm not sure how to integrate the read noise into the 
noise equation. I'm just adding it in quadrature now. Otherwise the 
noise equation makes sense to me.

Cheers,
Michael Koppelman
http://www.lolife.com/astronomy/

On Jul 14, 2004, at 2:45 PM, Pertti Pääkkönen wrote:

> Michael,
>
> I have been working with the same problem. AIP does not save the sky 
> backgound STD
> in
> the photometry, which is needed for uncertainty estimates. 
> Furthermore, we should
> speak about uncertainty here, not error. When we reduce our 
> measurements into
> standard system, (where the Comp stars are measured) we can speak 
> about measurement
> error. And this reduction introduces another uncertainties.
>
> I looked through your file, and these are my comments (please, don't 
> take it as
> ultimate truth :-)
>
> - Your uncertainty formula seems reasonable, provided that SNR is 
> correct.
>
> - Why do you add K-C STD when you calculate the final uncertainty for 
> V-C? It may
> give correct results, if C and K are of same brightness. However, it 
> DOES give you a
> good indication of your uncertainty for V-C provided that V and K are 
> of same
> brightness.
>
> - There is something weird in your SNR formula. Under square root you 
> coadd
> StarCount
> times Gain, AverageSkyBackgound times StarApertureArea, 
> StarApertureRadius squared
> times StarApertureArea (=Pi times StarAperture^4), and finally 
> ReadOutNoise. This
> does not make sense to me...
>
> I have found a reasonable formula, which is discussed in the 
> references of AAVSO CCD
> Observing Manual. Please find the Newberry article "Signal-to-Noise 
> Considerations
> for Sky-Subtracted CCD Data" (1991), or alternatively the article 
> "Measuring the
> Signal-to-Noise Ratio S/N of the CCD Image of a Star or Nebula", found 
> at:
> http://www.phys.vt.edu/~jhs/phys3154/snr20040108.pdf . In this article 
> there is a
> nice explanation how SNR is calculated. It also gives a formula to 
> calculate the
> SNR.
>
> I have applied your excel sheet to my series of U Gem in case of 
> bright and dim V
> star. In AIP_Err_bright C is star #1 and V is star #2 in ugem.seq by 
> Henden and
> Sumner. In AIP_Err_dim C star is #1 and V is star 11. Thus, in the 
> first run B-V is
> about 0.5 and in the second B-V is about 4.1. For SNR calculation I 
> add measured STD
> of the Sky Annulus in column P (measured by my own software), SNR 
> calculated using
> Newberry formula in column Q, and finally uncertainty 
> 2.5*log10(1+1/SNR).
>
> In AIP_Err_bright there is practically no difference bitween yours and 
> my results.
> This is because SNR is dominated by photon noise sqrt(StarCount). In 
> AIP_Err_dim
> there is some difference between SNR's. But still, if you look mean 
> values of
> calculated uncertainties (Err), they are equal.
>
> I suspect that there is something wrong with your SNR formula. It 
> takes account
> ReadOutNoise, but not actual noise in the image SkyBackground. This 
> means that it
> will not take into account changing observing conditions nor random 
> effects like
> gamma rays etc. If the STD of SkyBackground is used, then photon 
> noise, thermal
> noise, ReadOut noise and random effects are taken into account. But 
> this is
> practically significant only with poor SNR (<50 or so).
>
> When playing with uncertainties, it is adviced to compare your 
> estimates to STD of
> your measurements. With reasonably large amount of measurements the 
> uncertainty and
> STD should be equal regardless of star SNR. However, in my studies the 
> ratio DeltaV
> / STD(V) appears to be close to 0.6 (cross-checked with Arto's 
> MaximDL).
>
> I wanted to start discussion about what is correct way to give error 
> esitmates in
> AAVSO WebObs. I have learned that the same discussion was there just 
> one year
> earlier.
>
> I hope this helps. Please find attached zipped two files with my 
> photometry.
>
> Well, still 6 weeks when we can think about observing because of 
> summer here in
> Finland...
>
> Cheers,
>
> Pertti
>
>
> Lainaus Michael Koppelman <lolife at bitstream.net>:
>
>> This is an experiment. I know a lot of people use AIP for photometry
>> and it doesn't calculate the signal-to-noise ratio (SNR). I have
>> created a spreadsheet where you simply paste your AIP data into it and
>> it calculates the "total" error in your photometry using the equation:
>>
>> Err = sqrt( (1/snr)^2 + comp_sigma^2 )
>>
>> It also calculates your comp star (K-C) sigma for you and tells you 
>> the
>> SNR and "total" error for each data point.
>>
>> I would appreciate it is people who use AIP could test it out and let
>> me know how it works. I would also appreciate it if those people who
>> have a good understanding of SNR could double-check my math.
>>
>> The spreadsheet is here:
>>
>> http://www.lolife.com/astronomy/AIP_Err.xls
>>
Cheers,
Michael Koppelman
http://www.lolife.com/astronomy/

>> _______________________________________________
>> Aavso-photometry mailing list
>> Aavso-photometry at mira.aavso.org
>> http://www.aavso.org/mailman/listinfo/aavso-photometry
>>
>
>
> -- 
> Pertti Pääkkönen, PhD, Laboratory engineer
> University of Joensuu, Department of Physics
> Tel: +358 13 351 3238
> Fax: +358 13 251 3290<AIP_Err.zip>