Tilvi and I (and others) have been investigating the best and simplist way to do photometry with NSL data. We have now settled on a method similar to how photometry is done with a normal telescope: relative measurements. Here is an algorithm:
Pick the star you want to study.
Pick a comparison star. Now we recommend that this comparison star be the brightest star in that constellation. The high brightness would help keep statistical noise as low as possible. The same constellation is a quick and simple method of getting a "nearby" star, even though "nearby" might be several degrees away. Tilvi has verified that even for stars several degrees away and a completely different color, the expected systematic error in relative photometry is usually below the statistical error.
Take the ratio of picked star to the comparison star. Errors should be either added in quadrature or done with a "top high / bottom low" & "top low / bottom high" technique. This is simply done in Excel. Anyone can do it.
If all the stars in a constellation are showing similar variability, then the brightest star is likely the real variable and the light curves should be re-computed with a new comparison star.
We will build up the "first NSL catalog" like this, and determine which of the brightest stars are currently variable on the simplist time scale (one night) and on the level of the statistical errors. Anyone who wants can pick a star and play along. In another topic I estimated crudely that perhaps one third of all stars will show varaibility even on this level. Please report your results to this forum and we will keep an online catalog of results. Tilvi has started out and reports some preliminary results here.
As the project matures data can be combined to search for smaller amplitude variability on longer time scales.
- RJN
How to do photometry from NSL data
I now think the above analysis is NOT the best way to do photometry with NSL data.
I have been working with photometry files very closely over the past few weeks and my thoughts are continually changing. They have returned to something I have belived previously and even posted recently. (See thread on Pixelation error).
A better way to do NSL photometry of a given star is to compare C counts with those taken for the same star on a different night at the SAME SIDEREAL TIME. This is not usually possible or feasable with non-NSL data but is not hard to do with NSL data.
I have been able to show today that using comparison stars on the same frame will lead to a systematic error of perhaps larger than 0.1 magnitudes. This is because most NSL stars do not have a bright comparison star nearby and the relative photometry between stars separated on the order of degrees appears to give that amount of error.
Today I looked in detail at MK data taken last night for Alpha and Beta Cas. Taking the ratio of the two stars on the same night led to a drift of 0.1 magnitudes over the night.
However, taking the ratio of Alpha Cas on 040918 to Alpha Cas on 040917 gave very impresssive results. The ratio was much more constant over the entire night than using a comparison star.
Looking in more detail, the average standard deviation of the ratio of two measurement taken over two nights at the same sideral time was on the order of 0.01 magnitudes! What is possibly more impressive is comparing the average brightness of Alpha Cas over the two nights. The difference was only 0.0028 magnitudes! So Alpha Cas remained constant to 0.0028 magnitudes over the past two nights, including about 90 observations.
What I don't yet fully understand is computing the average standard error between the two data sets over the entire night. The standard error was only 0.00013 magnitudes. So the 0.0028 magnitudes was over 20 times greater than the expected standard error -- giving an indication that Alpha Cas appeared to vary between the two nights. I am more inclined to belive, however, that the opacity of the air varied by 0.0028 magnitudes, or that there is another source of systematic error, rather than attribute this right now to intrinsic variability of Alpha Cas.
Thoughts?
- RJN
I have been working with photometry files very closely over the past few weeks and my thoughts are continually changing. They have returned to something I have belived previously and even posted recently. (See thread on Pixelation error).
A better way to do NSL photometry of a given star is to compare C counts with those taken for the same star on a different night at the SAME SIDEREAL TIME. This is not usually possible or feasable with non-NSL data but is not hard to do with NSL data.
I have been able to show today that using comparison stars on the same frame will lead to a systematic error of perhaps larger than 0.1 magnitudes. This is because most NSL stars do not have a bright comparison star nearby and the relative photometry between stars separated on the order of degrees appears to give that amount of error.
Today I looked in detail at MK data taken last night for Alpha and Beta Cas. Taking the ratio of the two stars on the same night led to a drift of 0.1 magnitudes over the night.
However, taking the ratio of Alpha Cas on 040918 to Alpha Cas on 040917 gave very impresssive results. The ratio was much more constant over the entire night than using a comparison star.
Looking in more detail, the average standard deviation of the ratio of two measurement taken over two nights at the same sideral time was on the order of 0.01 magnitudes! What is possibly more impressive is comparing the average brightness of Alpha Cas over the two nights. The difference was only 0.0028 magnitudes! So Alpha Cas remained constant to 0.0028 magnitudes over the past two nights, including about 90 observations.
What I don't yet fully understand is computing the average standard error between the two data sets over the entire night. The standard error was only 0.00013 magnitudes. So the 0.0028 magnitudes was over 20 times greater than the expected standard error -- giving an indication that Alpha Cas appeared to vary between the two nights. I am more inclined to belive, however, that the opacity of the air varied by 0.0028 magnitudes, or that there is another source of systematic error, rather than attribute this right now to intrinsic variability of Alpha Cas.
Thoughts?
- RJN
I have now done Beta Cas and have similar conclusions to my last post: a really good way to do photometry with NSL data is to use, for a comparison star, the same star at the same sidereal time on a different night.
I do have one correction: the standard error quoted above for Alpha Cas was computed incorrectedly and should be 0.0012 magnitudes. This means that the summed photometry difference of 0.0028 magnitudes between the two nights is only significant at the two sigma level, which is to say not very significant.
Beta Cas also appeared quite constant between the same two nights (on the same MK NSL frames) although the star appears a bit dimmer. The summed photometry difference was 0.0038 magnitudes with a standard error of 0.0016 magnitudes, again different only by just over two-sigmas.
- RJN
I do have one correction: the standard error quoted above for Alpha Cas was computed incorrectedly and should be 0.0012 magnitudes. This means that the summed photometry difference of 0.0028 magnitudes between the two nights is only significant at the two sigma level, which is to say not very significant.
Beta Cas also appeared quite constant between the same two nights (on the same MK NSL frames) although the star appears a bit dimmer. The summed photometry difference was 0.0038 magnitudes with a standard error of 0.0016 magnitudes, again different only by just over two-sigmas.
- RJN
Dr RJNRJN wrote:I have now done Beta Cas and have similar conclusions to my last post: a really good way to do photometry with NSL data is to use, for a comparison star, the same star at the same sidereal time on a different night.
I do have one correction: the standard error quoted above for Alpha Cas was computed incorrectedly and should be 0.0012 magnitudes. This means that the summed photometry difference of 0.0028 magnitudes between the two nights is only significant at the two sigma level, which is to say not very significant.
Beta Cas also appeared quite constant between the same two nights (on the same MK NSL frames) although the star appears a bit dimmer. The summed photometry difference was 0.0038 magnitudes with a standard error of 0.0016 magnitudes, again different only by just over two-sigmas.
- RJN
This is very promising and one of the easiest way to do photometry. I think we can try for some more stars to find an average error we might expect in the photometry files and then we can follow and write this procedure for generating the catalogue.
Tilvi
Michigan Tech. University, MI.
Michigan Tech. University, MI.
Here is a more detailed example of how I did photometry with NSL data.
1. Browse the NSL archive pages to find a night with little moon and good weather.
Example: South Africa on 2004 May 16 looks pretty clear for much of the night: http://nightskylive.net/sa/sa040516.html
Note that many of the thumbnail JPGs appear mostly free of clouds
2. Scroll down and find a frame that shows your star. Alternatively, click on any "PHOTOMETRY" link and look for either your star or an interesting star. If your just choosing a star now, you might pick one that is hyperlinked. Following this hyperlink will take you to page where all the photometric measurements of that star on that night coexist.
Example: The photometry file http://nightskylive.net/sa/sa040516/sa0 ... 1420p.html
records star counts that occurred during the frame that started at Universal Time 20:14 and 20 seconds. Scrolling down that file one can find Beta Libra and click on it taking you to http://nightskylive.net/sa/sa040516/HD135742.html . This file records all the photometric measurements for Beta Libra for that night.
3. Copy this file into Excel. Just highlight the page starting with "date" on the left and highlight all the way down to the final zero on the lower left. I include the column labels which can help with future manipulations. In Excel, I copy this starting at the cell block A4, leaving the top 3 columns for my own notes and descriptors.
4. Keep this file only if the night appears clear. To determine this in Excel, highlight the entire "C5-B" column and the click on the plot icon. Then click on "Line" for your Chart Type and then click "Press and Hold to view sample". Does the plot look relatively smooth? If so, its likely a keeper. If it jumps all around all of the time, it was likely not a photometric night, clouds ruled, so dump it.
You will likely not find many completely clear nights. So if only a few points appear to jump around I would suggest keeping the file and deleting the rows with the jumpy points. This method will delete deep eclipses that occur during times of cloudiness, but smooth eclipses outside of cloudiness will be more ordered and not jump randomly around. Those we can find.
5. Delete all the columns except for "julian date", "alt", and "C5-B", assuming you want C5-B to be your photometric variable. "C5-B" means the average of the top five bins after subtracting the background.
6. Highlight the phometry file name and copy it to the first line above "julian date". This will ensure that you can find the original file if you need to. I also copy the star name and HD star number to the second line over "julian date" to remind me what star I am working on.
7. Find another night with good data for the same star and window copy its contents to the right of your old data. Repeat the above steps isolating only good data and the three data columns julian date, alt, and C5-B.
8. Find all the times when the star was visible on both nights at the same sidereal time. This means that the star appeared in exactly the same location in the sky and the frame. Therefore, it will have the same altitude. So match lines of altitude. I allow a tolerance of 0.015 degrees since CONCAMs actually can be slightly delayed in starting there exposures without large effect. If the second star does not have an altitude match with the first star, delete that data row. Keep nightly measurements at the same altitude on the same data row.
9. Take the ratio between "C5-B" columns for the star on the two different nights. If you set you Excel sheet up correctly, you can just divide one column by another. I do this by dividing the top line and then cascading the line down the entire column. Is the ratio near one? If so, this is a good sign that the star was constant between those two measurements. If it is near one, you can calculate delta magnitude between the two by taking dm = 2.5 * log(ratio). Astronomers more commonly report dm than ratios.
You should do at least one constant star to verify your technique before going on to make great photometric discoveries about stellar variability with NSL data.
An example Excel spreadsheet for Bet Libra from South Africa can be found here: http://nightskylive.net/temp/sa_play.xls
Please tell me of your personal experiences with this technique so that we can make it better and easier.
- RJN
1. Browse the NSL archive pages to find a night with little moon and good weather.
Example: South Africa on 2004 May 16 looks pretty clear for much of the night: http://nightskylive.net/sa/sa040516.html
Note that many of the thumbnail JPGs appear mostly free of clouds
2. Scroll down and find a frame that shows your star. Alternatively, click on any "PHOTOMETRY" link and look for either your star or an interesting star. If your just choosing a star now, you might pick one that is hyperlinked. Following this hyperlink will take you to page where all the photometric measurements of that star on that night coexist.
Example: The photometry file http://nightskylive.net/sa/sa040516/sa0 ... 1420p.html
records star counts that occurred during the frame that started at Universal Time 20:14 and 20 seconds. Scrolling down that file one can find Beta Libra and click on it taking you to http://nightskylive.net/sa/sa040516/HD135742.html . This file records all the photometric measurements for Beta Libra for that night.
3. Copy this file into Excel. Just highlight the page starting with "date" on the left and highlight all the way down to the final zero on the lower left. I include the column labels which can help with future manipulations. In Excel, I copy this starting at the cell block A4, leaving the top 3 columns for my own notes and descriptors.
4. Keep this file only if the night appears clear. To determine this in Excel, highlight the entire "C5-B" column and the click on the plot icon. Then click on "Line" for your Chart Type and then click "Press and Hold to view sample". Does the plot look relatively smooth? If so, its likely a keeper. If it jumps all around all of the time, it was likely not a photometric night, clouds ruled, so dump it.
You will likely not find many completely clear nights. So if only a few points appear to jump around I would suggest keeping the file and deleting the rows with the jumpy points. This method will delete deep eclipses that occur during times of cloudiness, but smooth eclipses outside of cloudiness will be more ordered and not jump randomly around. Those we can find.
5. Delete all the columns except for "julian date", "alt", and "C5-B", assuming you want C5-B to be your photometric variable. "C5-B" means the average of the top five bins after subtracting the background.
6. Highlight the phometry file name and copy it to the first line above "julian date". This will ensure that you can find the original file if you need to. I also copy the star name and HD star number to the second line over "julian date" to remind me what star I am working on.
7. Find another night with good data for the same star and window copy its contents to the right of your old data. Repeat the above steps isolating only good data and the three data columns julian date, alt, and C5-B.
8. Find all the times when the star was visible on both nights at the same sidereal time. This means that the star appeared in exactly the same location in the sky and the frame. Therefore, it will have the same altitude. So match lines of altitude. I allow a tolerance of 0.015 degrees since CONCAMs actually can be slightly delayed in starting there exposures without large effect. If the second star does not have an altitude match with the first star, delete that data row. Keep nightly measurements at the same altitude on the same data row.
9. Take the ratio between "C5-B" columns for the star on the two different nights. If you set you Excel sheet up correctly, you can just divide one column by another. I do this by dividing the top line and then cascading the line down the entire column. Is the ratio near one? If so, this is a good sign that the star was constant between those two measurements. If it is near one, you can calculate delta magnitude between the two by taking dm = 2.5 * log(ratio). Astronomers more commonly report dm than ratios.
You should do at least one constant star to verify your technique before going on to make great photometric discoveries about stellar variability with NSL data.
An example Excel spreadsheet for Bet Libra from South Africa can be found here: http://nightskylive.net/temp/sa_play.xls
Please tell me of your personal experiences with this technique so that we can make it better and easier.
- RJN