Hi,
I’m getting a series of image of NGC7790 for calculating the TG coefficients for the T5 telescope of iTelescope
I was wondering if images in BVI taken in three consecutive days could be enough for getting a good result.
Gianluca:
Determining whether three nights of images is adequate is what you will determine. The obvious goal is to get similar (statistically identical) transformation coefficients for each data set. If each night’s coefficients look similar, use the average that TG will calculate for you. With good images on each night, 3 sets of images are often reasonable. If you only take two nights data, it’s hard to decide which is true and if it is repeatable. If one of the night’s data looks poor (much larger error), you may need to reconsider this assumption. You might also consider taking replicate images in each filter and stacking them to improve SNR. Make sense. Clearly, this effort takes time and/or money! What precision/agreement are you willing to accept?
Ken
PS: There is not a ‘requirement’ that the images be collected on consecutive nights. Clouds do happen! A common practice is to re-check your coefficients every 6-12 months to assess whether the filters are changing. I have found a longer interval is reasonable but you should check that assumption.
Hi,
luckily the nice people at Itelescope are willing to sponsor the images for calculating the coefficients for T5.
I’m going to take three set of images in B, V and I, after I’ll calculate the coefficients.
If it’s OK I’m willing to share here the images and the data, I’d love to have feedback on this.
Hi,
this is the data calculated for T5 of iTelescope
[Setup]
description= TG - Version 6.9a, Telescope= T5, Time created (UT) = 2024_09_01_14:33:01
[Coefficients]
Tbv= 0.028
Tb_bv= 0.138
Tb_bi= 0.338
Tbi= 1.103
Tv_bv= 0.024
Tvi= 1.191
Tv_vi= 0.094
Ti_vi= 0.006
[Error]
Tbv= 1.715
Tb_bv= 0.046
Tb_bi= 0.367
Tbi= 0.050
Tv_bv= 0.207
Tvi= 0.355
Tv_vi= 0.208
Ti_vi= 0.032
[R Squared Values]
Tbv= 0.627
Tb_bv= 0.492
Tb_bi= 0.719
Tbi= 0.964
Tv_bv= 0.124
Tvi= 0.832
Tv_vi= 0.190
Ti_vi= 0.037
I’m attaching the files downloaded from Vphot for the three series and a screenshot of TG software with the data for the three series
Are they good? The errors are too big or OK?
Thanks
report TG analysis ngc7790 20240828.txt (2.4 KB)
report TG analysis ngc7790 20240830.txt (2.4 KB)
report TG analysis ngc7790 20240831.txt (2.4 KB)
Gianluca:
Some coeffs are just not reasonable. Look at Tbv=0.028! Should be close to 1 not 0.
Tbv error is 1.715. Looks like values are reversed? These values do not agree with TG screenshot shown.
Look at TG list of all 3 days. First day much larger for Tbv.
What do your plots look like? How many std stars in list? Did you delete data points as 2-sigma lines shrank inwards?
Try again.
Ken
Hi
First of all thanks for your help, very much appreciated.
I was so focused on doing the procedure right that I didn’t noticed the discrepancies in the values.
I’m going to redo the analysis from the raw data checking the values at every step and documenting everything.
By the way there is some source for the reasonable range of values for the various parameters?
Thanks again, have a nice sunday
Hello,
I repeated the analysis starting from the files download from Vphot.
The values for the same parameter now are in the same order of magnitude but I see big differences, in some case error bars are not overlapping.
Is this a signal of really bad observational data?
Using TG I noticed some low values for R^2 even when removing the outliers generated an apparently good fit, is it normal?
Tbv, Tbi and Tvi are the only values where I get a good R^2.
Any comment?
Gianluca:
The coefficients look much better! The b errors are a little larger and probably due to fainter comps? Probably better to take longer exposures to get more comps, especially redder. I note one possible (unexpected) discrepancy for mean Tbv. Could you calculate the Tbv mean by hand? Is it really 1.23 (math error in TG?)?
Your r^2 observation is due to a fact not known by most non-statisticians. IF the line you are fitting has a slope of 0, the r^2 actually also equals 0 not 1. Errors are more useful for comparison. 0.01-0.03 is typical.
Ken
PS: BTW, I normally use B exposures that are about 2x longer than for other filters.
Hi,
thanks for checking… the average of 1.23 is correct
This analysis was done with a set of B, V and I images with 30 secs of exposure on T5.
I have also images with 50 secs exposure, I’ll redo the analysis and post the numbers
Hi,
this is the analysis using the 50 secs exposures, some errors are lesser than the previous values but the Tbi error is still high.
I’m starting to think that the first set of images has some problem, even if at a visual inspection looks okay
Anyway I have planned two more sets at 30 and 50 secs for the next two days, if the new sets are aligned with the second and the third I’ll probably be tempted to ditch the first
Sorry I didn’t noticed before your note about the doubled exposure time for the B filter,
I’ll plan also a couple of run with the B exposure time doubled
Hi,
in the screenshot the fourth column on the right presents data with double exposure time on B filter (100 secs instead of 50), the list of starts used for the time-series analysis on Vphot is the same for all the images.
I was wondering if it’s normal to get values so different for the same parameter, I was expecting at least an overlapping of the error bars.
This variability (pun intended) could be linked to some flaw of the telescope-ccd ensemble? Or to different outlier selection from band to band (dropping single points manually from the various graphs in TG sounds a bit arbitrary to me)
Gianluca;
When you run the time series (BVRI) in vphot, what minimum snr settings do you use? I usually use SNR>20. that gets rid of the faint stars with larger errors.
I usually take long enough exposures that I get ~100 stars rather than <50.
Yes, the process is a bit subjective. The most important goal is to get some stars at color extremes. This most impacts the slope!
Your figure 6 above does show data deletion that I find reasonable. I assume your other plots were similar.
Don’t accept that errors of 0.01-0.03 are absolute requirements, just a guide. Use your judgement. Experience will help!
I still think your B coeffs are the most problematic. Thus, my comment about longer exposures.
Ken
Hi,
yes I use the default values of 20 for SNR.
I obtained 4 new sets of images BVI with 100/50/50 secs of exposure, I stacked them with the average method in Vphot and redo the analysis.
The numbers looks good?
Can I use the ini file for transforming my observations?
I put all the info about the process in the linked doc, including a link to a shared folder with all the files and the images.
Thanks again for your time, your help has been invaluable.
Hi,
I found out that transform coefficients were already available on Vphot (shame on me for not checking before…)
The values are different from mine but not too different, error bars instead are smaller than mine.
Statistically, they look pretty much the same to me. 
This is what you may expect. Good job!
Ken
Hi Ken,
thanks for your opinion, before launching myself in the transform jungle I wanted to be sure to have good foundations.
Looking forward to participate in your course on Vphot.
Have a nice weekend