Skip to main content

Meridian flip photometry offset

GFB
GFB's picture
Offline
Joined: 2010-07-08

 

I've always understood that when a GEM flips during a time series run one potential problem is that errors in a flat may show as an offset in photometry readings.  The target remains near mid frame but the comp or check stars will flip sides in some orientations.  A flat that isn't flat will cause altered readings on the comp and check causing an offset in the time series.

 

While I haven't normally seen this, recently I shifted to using another reduction package(I'll call M) and spotted an offset quite clearly.  It took awhile to figure it out but it clearly happens at the flip.  I checked the frames carefully and tried different combinations of check and comps, same problem.  

 

As a test, ran the same images thru my other reduction package(I'll call C) and low and behold, no offset!  I made several passes with each package matching the aperture sizes and comp stars, same result, one shows offset, the other doesn't.  Some parts of the comparison I don't control of course.  One package uses the Pinpoint plate solution to place the apertures, the other solves the plates during the reduction run to place apertures.  The packages use diffent set up for apertures, M gives a finer control of the aperture sizes.

 

I then loaded 20 of the images to VPHOT, 10 on each side of the flip and reduced them there.  VPHOt also does not seem to show an offset.  The M package seems to show about 0.4 mag offset and that's just not there in C or VPHOT.

 

So the dilemma.  Do I just pick the package with the results I like the best and blow off the other?  I guess I'd prefer to understand why they are different, and if it's still some cockpit error I'm making, fix it.  Is there anyother way to check these flats?  I seem to remember moving a target star thru a grid pattern on the frame and making measurements at each point for comparison.  Maybe that's next.  Any other ideas?

 

Bill

Hi Bill It may be that one
jji
jji's picture
Offline
Joined: 2010-07-24

Hi Bill

It may be that one of the programs isn't handling the flat properly.  That is, it isn't keeping the flat properly orientiated to the image after the flip.

I have been fighting an offset after the flip all summer.  It only happens occasionally and usually when I am pointing to the south at a fairly low altitude.  I am about convinced it is caused by scattered light from a neighbor's yard light from hell.  I've never taken a look at the situation with another program.  Maybe I should.

I use MaxIm BTW.

Jim Jones

 

 

 

Probably more than one variable at play here
KTC
KTC's picture
Offline
Joined: 2010-12-08

Probably more than one variable at play here...and probably more than one way to fix, or at least mitigate problems.

Is a flat truly flat?  We have run AAVSONet scopes through a 5 x 5 raster scan pattern...putting the same star field at 25 locations across the CCD FOV (getting close to the edges/corners, but not so close that you can't do photometry on your star(s) of choice)...and at the beginning and end of the 5 x 5 raster scan, adding another image with the star(s) of choice at the center...so you have a sequnce of 27 images.  We took the data on photometric nights and looked at instrumental mags.  If there was an error on these rigs, it was radial...the farther from frame center...the larger the error.  Better flocking/baffling (in these cases) fixed the problems.

A scripted set of commands, or set of various targets in a fixed-order series can do this...and then you have the plan in the can if you need to run it a second and third time. 

Attached is a spreadsheet in Open Office and XL format...a crude tool to generate some RA/Dec coords for a raster test.  (Remember that 1 arcsec in Dec is 15 times smaller than 1 arcsec of RA at the equator...projected on the sky.)  You can adjust it for your FOV and center coords...and get a list of coords to import into planning/scripting software.

But your cause of non-flat flats (if it exists...not yet proven) could be different.  It could be how you take the flats, or it could be flop/flexure...which tends to be worse in bigger scopes.

And different results from different software?...on the order of 0.4 mag?  That large?  Four tenths of a mag?  Or did you mean 0.04?  (The worst I've seen in my rigs here is about 0.02 mag.)

You may want to look at offsets with various comps scatterd about the frame.  If the comps are close to the target...is the offset smaller?  If the comps are far, is the offset bigger?  If the comps are N/S of the target...is the offset different than if the comps are E/W?  Answer to these questions may give us more insight.

Feel free to share images to my VPHOT account at KTC, or make some available for download (Google Drive?  other?  do a search on "large file transfer service" and find one).  We may need to dig down and look at flats, raw frames, calibrated frames, etc.

I hope this helps.

Meridian flip offset
WBY
WBY's picture
Offline
Joined: 2010-07-24

Bill,

 I assume that the camera remains really fixed on the telescope during the flip. No rotation occurs. Have you noticed the offset on more than one set of images?

 Are you flipping the images, even if only to display them on the screen,  when calibrating them so that the target and comps have the same orientation before and after the flip? If you are, try calibrating them again with package M without reorienting and see if it eliminates the offset.

 

Brad Walter

Meridian flip...
GFB
GFB's picture
Offline
Joined: 2010-07-08

jji wrote:

Hi Bill

It may be that one of the programs isn't handling the flat properly.  That is, it isn't keeping the flat properly orientiated to the image after the flip.

I have been fighting an offset after the flip all summer.  It only happens occasionally and usually when I am pointing to the south at a fairly low altitude.  I am about convinced it is caused by scattered light from a neighbor's yard light from hell.  I've never taken a look at the situation with another program.  Maybe I should.

I use MaxIm BTW.

Jim Jones

 

 

 

Yes, I suspect something like that as well.  I've noticed a couple of other things that I need to chase down tho.  The tough part is I have a pile of images that need processing ;-(

Bill

Probably more than one variable...
GFB
GFB's picture
Offline
Joined: 2010-07-08

Tom,

Well, that's the challenge isn't it, figuring out what they may be.  I noticed today that if I plot the target and check in Peranso, the target shows the offset and the check doesn't .  I like the raster idea and will probably end up doing that just to insure to myself it's working as it should.  It won't help explain why one package shows the offest and another doesn't, which is still a puzzle to me.

Bill

Meridian flip offset
GFB
GFB's picture
Offline
Joined: 2010-07-08

WBY wrote:

Bill,

 I assume that the camera remains really fixed on the telescope during the flip. No rotation occurs. Have you noticed the offset on more than one set of images?

 Are you flipping the images, even if only to display them on the screen,  when calibrating them so that the target and comps have the same orientation before and after the flip? If you are, try calibrating them again with package M without reorienting and see if it eliminates the offset.

 

Brad Walter

Hi Brad,

Yes, the camera is secure.  I know that some programs will flp images for display, but in this case all the calibration was done with one program(M) in a process that doesn't open them.   Then the calbrated images were opened and reduced with M and C.  One shows the offset, the other doesn't.  I'm gonna make some more tests with comp and check star, choosing different pairs.  I'm also going to load a larger set to VPHOT.  I noticed that VPHOT plots all the comps and checks so it might be another opportunity to see if the offset is seen in the comps and check.

 

Bill

Look at instrumental magnitudes
KTC
KTC's picture
Offline
Joined: 2010-12-08

It may be helpful to examine the instrumental magnitudes from the various software packages.  Differential mags involve two star measurements, but instrumental mags of individual stars avoid that problem.

meridian flip offset
HQA
HQA's picture
Offline
Joined: 2010-05-10

Hi Bill,

Like Tom K. said: do you really mean 0.4mag difference, or 0.04mag difference?  My reply will be quite different depending on the magnitude of the shift.

Arne

meridian flip offset
GFB
GFB's picture
Offline
Joined: 2010-07-08

 

Hi Arne,

Well, this has proved to be quite a challenge and as Tom suggested, I've been digging deeper.  A couple of things have started to flesh out.

1.  The initial offset was 0.4 as I had stated.    This offset is on the variable, not the check, so I suspect I would never have noticed a 0.04 offset.  The check star line showed none of this offset, pretty much a straight line.

2. I now see, most of this 0.4 offset error was mine.   Greatly magnifying the images and apertures I found that there were two small background stars that were in the annulus.   Enlarging the gap and annulus to straddle these stars really helped.  The offset reduced to about 0.15 mags but didn't go away.  C or Canopus also performed much better but still doesn't show the offset.

4.  M or Mira, which shows the offset, has a 'scaled' aperture tool where the apertures are adjusted frame by frame based on FWHM.  I'm not fully comfortable with this tool yet, I occasionally get odd measurements that are way off series, but I decided to try it.  In a series of 150 frames I'll get 8 or 10 that are clearly errors, all before the flip.  But this scaled aperture series otherwise really looked good, the check star is almost flat lined and the offset was not visible.  This added to my suspicion that the centroiding system is probably showing this difference in the offset.

5.  Mira also produces a detailed photometry output table with dozens of columns.  I've noticed that the check and comp star magnitudes are very consistent across the flip.  The target stars magnitudes do show the offset across the flip.  I also noticed that the net counts and SNR of the variable vary across the flip.  The net counts are in the 5500 range and SNR about 20 before the flip but jump to 7500 and 29 or so after the flip.  That leads me to believe that there is a stray light change happening.  But it seems to be happening to the target not the comp or check stars.

6.  In CCD School we talked about using lots of darks and biases to make masters.  I've discovered that using 150 darks produces a master that has faint dark columns in it.  On the calibrated target frames these dark columns start to show up.  Using the magnified image I've found that before the flip one of these columns runs right thru the aperture of the target.  After the flip, it does not.  So now I'm wondering if this dark column is pushing  the values down.  I may recalibrate with fewer darks to and try it again.

Other than that, I'm out of ideas.  I guess part of this is looking at the results of every reduction to see if you catch some oddity like this.

I might add that Peranso has been a great help showing the target and check star light curves easily.

Bill

meridian flip offset
HQA
HQA's picture
Offline
Joined: 2010-05-10

Hi Bill,

If Canopus does not show an offset after the meridian flip, but Mira does, then my guess is that you are not using Mira correctly.  You should double-check in Canopus and make sure that the peak counts in the star match the values you obtained from Mira before/after the flip, just to ensure that nothing strange is going on at that basic level.

Regarding the faint dark column - does it just pass through the measuring aperture, or does it actually pass through the star?  Note the difference in the signal level between the dark column and its neighbors in the master dark - my guess is that the difference is extremely small and only shows up in the 150 dark master because of its better signal/noise.  Remember that the dark is subtracted, so a difference of a fraction of an ADU between two columns would not make any difference in the photometry, especially a 0.4mag difference (that is a factor of two difference in the sum total counts in your aperture.  My guess is that is a "red herring" - spend your time understanding how Mira handles companions in the aperture, and whether you can just use a more normal fixed aperture method with Mira instead of the resizing option.

Arne

Sample images or screen shots?
KTC
KTC's picture
Offline
Joined: 2010-12-08

Don't be shy.  Post sample images, or screen shots showing (for example) the measurement aperture and annulus on the target, or a profile plot to get an idea of FWHM, etc.

...I suspect I would never have noticed a 0.04 offset.

If you do time series data...and you expand the vertical scale of your magnitude plots...it's often easy to see a post-meridian offset of 0.01 magnitude.

The check star line showed none of this offset, pretty much a straight line.

This can be quantified with basic statistics, and perhaps a moving average, of pre and post-flip data.  No need to be vague.

I found that there were two small background stars that were in the annulus.

Most photometry packages handle two stars in the annulus with ease.  Do you mean annulus, or measurement aperture?  Sample images or screen shots are worth 1000's of words.

Mira, which shows the offset, has a 'scaled' aperture tool where the apertures are adjusted frame by frame based on FWHM.  I'm not fully comfortable with this tool yet, I occasionally get odd measurements that are way off series....

Don't be shy with data.  How about reporting the average FWHM measure in each image?  Is there a change after meridian flip?  Are there images that have grossly different FWHM values compared to the others?  (That may indicate drive/guider problems, or wind gusts, etc.)

I also noticed that the net counts and SNR of the variable vary across the flip.  The net counts are in the 5500 range and SNR about 20 before the flip but jump to 7500 and 29 or so after the flip.

Do you refocus post-flip?  Do you plate solve/reslew/recenter after the flip?  (If the mount pointing/centering changes (yes, there's a 180 deg. rotation with the flip)...and/or focus changes because of the flp...those are variables that can be controlled with automation software...tell it to refocus after the flip, or plate solve after the flip.

In CCD School we talked about using lots of darks and biases to make masters.  I've discovered that using 150 darks produces a master that has faint dark columns in it.

Maybe that's an accurate representation of your dark frame behavior?

If you compare master darks made with various software...identical results?...close?...or out to lunch?

Don't be shy about making available your master (and some individual) calibration frames.  We may learn some things from examining them.

Using the magnified image I've found that before the flip one of these columns runs right thru the aperture of the target.

With automation software you can slightly alter the desired pointing/coords...and shift the target/comps away from potential trouble spots in your frame.

I might add that Peranso has been a great help showing the target and check star light curves easily.

If you are not adept at taking the output file of a photometry program and importing it (as a delimited file, or copy/paste operation) into a spreadsheet...I recommend you do so.  I do that all the time...and then I do it for focuser logs....plotting time versus temperature, or temperature versus position, or time versus FWHM, etc. etc.  Each time you look at data (or logs) in a different way, there is a potential to learn new things...find problems, and then fix them.

meridian flip, again
GFB
GFB's picture
Offline
Joined: 2010-07-08

 

Arne,

I agree, the faint column was just something I noticed that seemed different, but it was so faint it's hard to imagine any real difference.  In looking at the frames magnified I've found that conditions change thru the sequence and right at the flip.  At one point, I suspected that a focus change added to the problem, but other nights flips don't show a focus change so I'm less sure that added much.  And a focus change effects all objects on the frame so I don't think that's the problem.

I've looked at outputs of Mira and Canopus and what's reported is different.  Mira gives a report with 'net counts', Canopus will give instrumental mags on comps and checks but not counts.  There might be other tools I haven't found.

After running every combination I could think of with no resolution and recognizing that it may be the frames are just not good enough I shifted gears and double checked images of a target with no nearby field stars(V455 And).  No offset.  I still suspect that differences in centroid routines is causing this.

I also worked thru a different nights images on the problem target (NSV 25966), again this is the one with the nearby field stars.  I've attached some images here.

#1 shows the Mira and Canopus aperture settings.  The gap is large to straddle the field stars.  Mira places the apertures based on the plate solution.  Clicking thru several frames with the apertures in place I've noticed that the target apertures wonders somewhat from image to image as it includes some of the field stars pixels.  The comp star apertures don't move.  The Canopus apertures are only set here by me clicking on the target just to show the relative aperture sizes.  Canopus also sets apertures based on plate solution.  Anyway, it's just to show that the apertures are as close to the same as I can get them.

#2 is a Peranso screen of the output files from these two runs.  The flip occurs at about the 1/4 mark or -0.2 on the graph.  The upper graph is the Mira output.  The upper dots are the check star, lower dots the target.  The lower graph is the Canopus output.  No offset, but also none of the other peaks that Mira seems to be showing. 

#3 Now the excitement really begins!  I loaded a 40 frame subset to VPHOT.  Due to the upload time, I uploaded 20 frames before the flip and 20 frames after the flip.  And presto! it's different that either of the other two.  VPHOT aperture setting are different again.  It's not clear whether the aperture and annulus you set on a single frame is used in a time series.  On the time series setup you only can set the aperture size and not the gap and annulus, so I'm not sure how its done there.  I tried to match the aperture to the same appearance as Mira, 4.5 pixels.  With VPHOT the upward peak after the flip that Mira shows is there, but the magnitudes before the flip are much more like Canopus.

VPHOT has some nice reports for the series, FWHM, comp star light curves, skyglow, etc.  I don't see anything that jumps out as an error.

So there you are.  Three programs, three different results, admittedly a tough target.  I'm thinking of taking up competitive frog-jumping.  Calaveras County is only a few miles away!

At the risk of being presumptuous, I wonder if you might consider running the small subset of 40 frames thru your IRAF pipeline?  The frames are on VPHOT if there's a way to get them.

Bill

Still many variables at play
KTC
KTC's picture
Offline
Joined: 2010-12-08

#1 shows the Mira and Canopus aperture settings.  The gap is large to straddle the field stars.

The gap may be large, but....

- the actual annulus is rather narrow...not many pixels in it for the statistical analysis/math.  You may want to try a wider annulus.

- this is a rather dense star field...and I see faint stars....several of them in the annulus.  I recommend you expand the outer annulus radius so that you capture more sky pixels...and hopefully not too many more faint stars.

The target star has a near neighbor that is not faint.  Depending on how the various software is set up...centroid may drift toward that neighbor somewhat...spoiling the photometry.  Are there options you can configure...to alter the centroiding routine/logic?  This could be very important.

Plese feel free to share some of your VPHOT images to my account at KTC.  I'd like to see the data/numbers for FWHM of the target star, and how near the neighbor is.

It looks like your meridian flip caused about a 10-minute gap in images/data?  Is this typical for your rig?  Why is the gap so large in time?  Why is the gap not smaller, like 90, or 120 seconds?

meridian flip, again
GFB
GFB's picture
Offline
Joined: 2010-07-08

Tom,

I shared the small subset of images on VPHOT to your account.  The AAVSO sequence is too bright for this object so I used three other field stars for my runs that are much closer to the targets magnitude.  I've attached a text file with their coordinates and magnitudes.  The sixth column from the left is the V mags. which I used.

I don't know of any parameter setting to alter the centroiding routines.  There are some related to tracking but not to how the centroiding works, that I know of.  Yes, the neaby field star does cause the aperture to wonder somewhat in Mira where the effect can be seen, whereas the comp star apertures remain fixed.  Setting the apertures smaller seems to make the overall photometry worsen, the comp star measurements become more erratic and noisy.

Aperture-gap-annulus is not really adjustable in VPHOT for a time series, as I see it.  You can set any sizes you want on a single image, but that doesn't seem to be used on a time series.  The setup form only allows for a scaled or fixed aperture choice and setting the aperture size only, not the gap and annulus, at least that I've seen.  Maybe someone can point out where it is, it there.

Bill 

You can also share sequences in VPHOT
KTC
KTC's picture
Offline
Joined: 2010-12-08

You can also share sequences in VPHOT.  If you created any, please share them.

Various software have various ways to adjust centroiding and annulus settings...some are easy, some complex, and some are rather limited.

More comments later.

meridian flip, again
WGR
WGR's picture
Offline
Joined: 2010-08-17

Hello Bill

I am not a frequent user of VPHOT, but I remember that the aperture can be set either in pixels, which gets used on all frames of a time series, or you can set it in terms of FWHM.  If you set it this way, it measures the FWHM of each image in a time series, and adjusts the aperture accordingly.  Don't remember exactly where that is done.  Perhaps a regular user can point this out.

 

Gary

Yes, this is a tough target
KTC
KTC's picture
Offline
Joined: 2010-12-08

I downloaded and analyzed your too-sharp/undersampled image (FWHM near 1.7).  According to AIP4WIN, with a measurement aperture radius of 3 I can get most of the star's light in the aperture...but light from the neighbor stars is only about 1 pixel outside of that aperture.

If the FWHM increases to something like 3.5 (oops! your images meet that criteria), or the software centroid is not spot-on...then neighbor star light will be present in the aperture...spoiling the photometry.

You could try a smaller measurement aperture, such as 2.5 or 2.0 pixels...but that has problems of its own...depending on how various software deals with 'partial' pixels at the edge of the measurement aperture.

This is a tough target.  Keep focus good, but not too sharp.  Avoid breezy nights.  Or choose easier targets.

Too-tough of a target
KTC
KTC's picture
Offline
Joined: 2010-12-08

Just looked at an image with FWHM = 2.9.

Measurement aperture of 3.0...light from neighbor is blurred/spread...into the target's aperture ... now it's spoiling the photometry.

Can you cover other targets?  I recommend you do that.

AAVSO 49 Bay State Rd. Cambridge, MA 02138 aavso@aavso.org 617-354-0484