PinPoint Pre-Sales difficulty / issues

[Jeff Woods]

Hello Bob,

I've just downloaded the PinPoint trial, as I may want to upgrade to it for use with MaxPoint and/or CCD AutoPilot. I've run into some problems.

Short Version: I found an anomaly in scale sizes when solving with GSC, and could not get USNO A2 solves to work at all. I will buy the product if I can get USNO A2 solves working.

Details: Since I've had trouble "hitting the target" and want to get into automation of mosaics, I am testing Pinpoint full version on XP with 2 GB RAM, and (though not yet relevant), MaxImDL Suite (with MaxPoint), CCD Autopilot, Focus Max, all on a CGEM mount and a QHY8L color CCD that is 7.8um pixels square, an APS-C chip that is 23.65mm x 15.72mm, and 3032x2016 in pixels. The scope I used to take my test shots with is an EdgeHD 9.25 @ 2350mm @ f/10. That means my FOVI (according to TheSkyX) is 34.60 arcminutes x 23.00 arcminutes. This *should* be a scale of roughly 0.684 arc-seconds per pixel, in both X and Y (depending on where I have my primary locked down).

Here's what I have run into thus far, attempting to plate solve some raw subs near NGC 6888 from August 11. I know the actual location (in J2000) from TheSky.

Problem 1:

GSC 1.1 works. It solved an image (in 55 seconds -- it had to do a spiral, since there was some doubt as to the exact location) and gave me:

1179 image stars found
(doing spiral search)
444 catalog stars found
Solved using 46 of max 500, RMS residual is 0.01, order = 4
Solution took 55.4 seconds
Centerpoint RA = 20h 17m 22.589s Dec = 30° 17' 00.08"
WCS: Roll = -4.40 HScale = 0.554 VScale = 0.890
PA = 184.403°
FWHM = 1.18 arcsec
Zeropoint = 19.12 (1 sec.)

Now, I know this solve is correct (it nailed the RA/Dec of stars I inadvertently shot while trying to hit NGC 6888, the Crescent Nebula -- I missed by a couple fields of view, even with a Celestron Precise Goto!), and it also correctly nailed my position angle (which I know from a PEMPro run that session). However, why does it think my HScale and VScale are so clearly different than the known values? Why aren't they symmetrical, since I have square pixels?

Note that the frames I am using were inadvertently shot with a RED filter, but I still have "R" selected as Color Band in PinPoint Solver. Guiding was not great -- there is some 30-40% eccentricity in the stars according to CCD Inspector.

Problem 2:

I was doing the above in Test Mode (no WCS), and when I had to provide the RA and Dec (it wasn't in the header), I made sure "write data to FITS" was unchecked. However, it was altering my original FITS files, at the end of the solve (whether successful or not). I know it was the end, as I could watch the timestamp change in realtime, in Windows Explorer. Should it be modifying my masters, if I've asked it not to?

Problem 3:

I tried to switch to USNO A2.0 for epoch accuracy. I have not yet had a single successful solve on an image using A2, despite giving it many different additional guided frames from the same run, giving it the exact center as reported above by GSC-ACT (20.2974 RA, 38.2767 Dec). I have tried both scales of both 0.684, and using the scales as reported by the GSC run, to no avail. I've tried letting it do a 5x5 mosaic @ 120 seconds per field with 40% expansion. Took almost an hour. Nada.

I have used the Catalog checker to confirm that my USNO A2 catalog is in good order. It is.

Problem 4:

At some point during one of my USNO A2 attempts, I got a Runtime Error:

-2147220451 (8004041d)
The property has not been set: TargetRightAscension

I have a screen shot of this if you need it.

My primary concern is to get USNO A2 solves working. If I can do that, I'm buying the product. If I can't, I would imagine that PinPoint LE that comes with MaxPoint would be sufficient for my needs.

I can provide sample FITS images if you need them to test on your own, but for now, here is just one of the FITS headers (note that I appear to need to update my MaxIm scope definition -- the true Focal Length is 2350, and Aperture is 235 since I upgraded to an Edge HD recently).

SIMPLE = T
BITPIX = 16 /8 unsigned int, 16 & 32 int, -32 & -64 real
NAXIS = 2 /number of axes
NAXIS1 = 3040 /fastest changing axis
NAXIS2 = 2016 /next to fastest changing axis
BSCALE = 1.0000000000000000 /physical = BZERO + BSCALE*array_value
BZERO = 32768.000000000000 /physical = BZERO + BSCALE*array_value
DATE-OBS = '2011-08-11T06:57:25' /YYYY-MM-DDThh:mm:ss observation start, UT
EXPTIME = 300.00000000000000 /Exposure time in seconds
EXPOSURE = 300.00000000000000 /Exposure time in seconds
SET-TEMP = -10.000000000000000 /CCD temperature setpoint in C
CCD-TEMP = -10.446927826179945 /CCD temperature at start of exposure in C
XPIXSZ = 7.7999999999999998 /Pixel Width in microns (after binning)
YPIXSZ = 7.7999999999999998 /Pixel Height in microns (after binning)
XBINNING = 1 /Binning factor in width
YBINNING = 1 /Binning factor in height
XORGSUBF = 0 /Subframe X position in binned pixels
YORGSUBF = 0 /Subframe Y position in binned pixels
IMAGETYP = 'Light Frame' / Type of image
SITELAT = '40 20 22' / Latitude of the imaging location
SITELONG = '-75 59 36' / Longitude of the imaging location
JD = 2455784.7898726850 /Julian Date at start of exposure
FOCALLEN = 2032.0000000000000 /Focal length of telescope in mm
APTDIA = 203.19999694824219 /Aperture diameter of telescope in mm
APTAREA = 29252.214751229785 /Aperture area of telescope in mm^2
EGAIN = 0.50000000000000000 /Electronic gain in e-/ADU
SWCREATE = 'MaxIm DL Version 5.15' /Name of software that created the image
SBSTDVER = 'SBFITSEXT Version 1.0' /Version of SBFITSEXT standard in effect
OBJECT = ' '
TELESCOP = ' ' / telescope used to acquire this image
INSTRUME = 'QHYCCD QHY8L'
OBSERVER = ' '
NOTES = ' '
FLIPSTAT = ' '
SWOWNER = 'Jeffrey Woods' / Licensed owner of software
INPUTFMT = 'FITS ' / Format of file from which image was read

Thanks for any feedback you can provide.

Jeff Woods
Reading, PA

[Dick Berg]

Hi Jeff,

I'd certainly like to look at it, if you would zip up the image from Problem 1 and reply to this message. Given all the information you've provided separately, it should be easy (he said!) to see where the problem lies. I can also try the USNO A2 and other catalogs.

To attach an image, reply to this message, type a few words or a sentence, then hit the Go Advanced button below. When the new dialog opens you'll be able to attach your zip file with the Manage Attachments button.

Will your image solve if you open it in MaxIm and use the Pin Point Astrometry tool on the Analyze menu?

[Jeff Woods]

I'd certainly like to look at it, if you would zip up the image from Problem 1 and reply to this message. Given all the information you've provided separately, it should be easy (he said!) to see where the problem lies. I can also try the USNO A2 and other catalogs.

Will your image solve if you open it in MaxIm and use the Pin Point Astrometry tool on the Analyze menu?

Hi Dick,

No need to burn Bob's storage for all time. I've posted a couple files on my website that you can grab at your convenience:

http://www.adbjester.com/AUG11_MISSE...888_JWOODS.ZIP
http://www.adbjester.com/AUG23_MISSE...888_JWOODS.ZIP

AUG 11 is the run I accidentally did via a red filter with the color CCD, and also had worse guiding (35% star eccentricity). RA=20.2896 Dec=38.2934 is the approximate center.

AUG 23 is a much better run, through the correct (Hutech IDAS) filter, with often under 10% eccentricity. RA=20.2883 Dec =39.0626 is the approximate center.

Each has several subs from the respective runs, and each .ZIP is about 30 MB.

Yes, using Analyze -> PinPoint Astrometry from within MaxIM worked, but again it was completely wrong about some things (such as the position angle, which it missed by 35°, and the focal length, which it overstated by almost double, while understating the image scale at 0.38". It did this even after I edited the FITS header with corrected aperture dimensions.

Now to figure out what was different..... and where the misstatement comes from. I think I found the difference, but not the source of the misstated scales.

When I tried it from within MaxIm, I had to provide a few configuration parameters for it to pass to the engine -- it won't read the config from Visual Pinpoint. When I did that, I set the limiting mag down to 14, for speed reasons.

Since that was the only thing I knew was different from the Visual Pinpoint trials, I went back to Visual Pinpoint, lowered the limiting mag, and it worked twice in a row. The same image then failed when tried to solve with a limiting mag of 15, with interesting errors in the log file that didn't happen when it failed at 16th mag:

The calculated plate scale is extremely different from the configured plate scale. Solution is suspect.

I tried all the other images that failed last night, and they also solved with USNO A2, so long as the limiting mag was set at or below 14. Humorously, one told me the HScale was 1.220 and the VScale was -1.227. No, I'm not imagining numbers....

I seem to be able to get it reliably solve now, so long as I leave out the little stars. I have to wonder if my data is poor due to the red filter (acquired by an unbinned color camera) and/or the guiding eccentricity, is making the solves at deeper field depth fail?

A test found that, no, this wasn't the case. Even the "good data" from the Aug 23 run won't solve with the limiting mag of 16, even if I give the starting point as exactly what USNO A2 solved it as, with a limiting mag of 14.

Very interestingly, solves via USNO A2 on the SAME RAW FILE are giving different answers depending on the max magnitude! See here:

http://www.flickr.com/photos/6660017...ream/lightbox/

You can see that I solved the same image 5 times (never reloading it -- it was using the same internal scan of the image each time), and got three different answers, depending on the limiting magnitude. This test was using one of the best images from August 23's better-guided run, so the data ought to be pretty good.

Curiouser and curiouser.

Any thoughts after this new information? Thanks for your help!

Jeff

[Dick Berg]

Hi Jeff, Thanks for your detailed reply. I just wanted you to know that I saw this message - I won't be able to get around to looking at the files or thinking about what's happening until later this afternoon (EDT) or perhaps tomorrow. We're battening down our hatches here in DC.

[Bob Denny]

I just stopped by the office early this morning and noticed this. Your experiences are really crazy. What version of PinPoint are you using (Visual PinPoint, Help and Info tab)? Not that any version in the last 10 years needs to have the "limiting mag" (which I assume is the catalog max mag setting) set to 14. The erroneous solves were 8 or 9 star matches, right? By reducing the mag to 14 you cut catalog stars WAY back so it would match on very few stars. The reason it was taking so much time without doing that is that it was valiantly looking for a solution. Curiously it was solving correctly in MaxIm? Note that MaxIm uses the same PinPoint engine, yes the same one, not an "identical" one.

I'm going to guess that the pointing is more than a FOV off, in other words, the nominal RA/Dec are off by more than your field of view. I don't know your plate scale, I guessed 0.6 but PinPoint is super forgiving in that regard, and unconcerned about rotation (position angle). I have to go now, and will be back Monday morning. If you're still having problems I'll dig deeper than. Sorry, but it was a long week and do volunteer work on weekends so I gotta go.

[Jeff Woods]

Hi Jeff, Thanks for your detailed reply. I just wanted you to know that I saw this message - I won't be able to get around to looking at the files or thinking about what's happening until later this afternoon (EDT) or perhaps tomorrow. We're battening down our hatches here in DC.

No worries, Dick. We're not going to get hit as bad as you in inland PA, but we're going to catch it. I did our battening earlier today. Be safe! Thanks!

Jeff

[Jeff Woods]

I just stopped by the office early this morning and noticed this. Your experiences are really crazy. What version of PinPoint are you using (Visual PinPoint, Help and Info tab)? Not that any version in the last 10 years needs to have the "limiting mag" (which I assume is the catalog max mag setting) set to 14. The erroneous solves were 8 or 9 star matches, right? By reducing the mag to 14 you cut catalog stars WAY back so it would match on very few stars. The reason it was taking so much time without doing that is that it was valiantly looking for a solution. Curiously it was solving correctly in MaxIm? Note that MaxIm uses the same PinPoint engine, yes the same one, not an "identical" one.

I'm going to guess that the pointing is more than a FOV off, in other words, the nominal RA/Dec are off by more than your field of view. I don't know your plate scale, I guessed 0.6 but PinPoint is super forgiving in that regard, and unconcerned about rotation (position angle). I have to go now, and will be back Monday morning. If you're still having problems I'll dig deeper than. Sorry, but it was a long week and do volunteer work on weekends so I gotta go.

No worries, Bob. The trial is 60 days, and I caused a hurricane by buying new glass. (Sorry!). Not like I'm headed out to use it in the field any time soon.

When you say "nominal RA/DEC", not sure what you mean. The sample images were taken by MaxIm, without the scope connected, so there is no RA/Dec in the FITS header. I am supplying the coordinates manually when prompted, and I am giving it the coordinates AS PROVIDED BY PINPOINT, when solved with GSC.

When I refer to position angle and scale errors, what I'm saying is that the results that Pinpoint provides are telling me that the PA and scale are grossly different from that which I expected, sometimes even in negative numbers. 0.685 *is* the correct image scale -- I am 100% certain of that.

In my second message I arranged several solves of the same image on screen, and photographed it. You can see more precisely what I mean by looking there. Different results if I set the catalog max mag setting, etc. And no, the incorrect solutions were not based on just 8 stars. See the image here:

http://www.flickr.com/photos/6660017...ream/lightbox/

Same image (not even reloaded):

At max mag 14, expansion 40%, 332 catalog stars found, solved with 31
At max mag 13, expansion 35%, 80 catalog stars found, solved with 10
At max mag 12, expansion 35%, 158 catalog stars found, solved with 12
At max mag 11, expansion 35%, 30 catalog stars found, solved with 12
At max mag 14, expansion 35%, 291 catalog stars found, solved with 31

Even feeding it back the RA/Dec as found by the max mag 14 solve, no solution at 15 or higher.

The results of the max 14 solve were:

Centerpoint RA = 20h 18m 0.440s Dec = 39° 35' 37.55"

To convert that to decimal, I did the standard:

Convert arcmin to arcsec, i.e. 18 * 60, added in arcsec + 0.440, getting 1080.440 arcsec. Convert that to decimal arcmin by dividing by 3600, 1080.440 / 3600 = 0.30012. Thus, RA is 20.30012. Same conversion for Dec brings 39.59376.

It also estimated my position angle as 45.708° (I *know* this to be wrong), and my scale to be HScale 1.432, VScale -1.754. Why aren't these two both substantially close to 0.685?

In any case, I can feed 20.30012 and 39.59376 to Visual Pinpoint on the *same image*, and if the max mag is 15 or above, it will not solve, even though I told it precisely where it ought to be.

OK, now it gets even stranger. I just solved it in TheSkyX (into which I've compiled USNO A2), and got:

******** ASTROMETRIC SOLUTION RESULTS ******** Center RA: 20h 11m 43.5s
Center Dec: +38° 20' 15.6"
Scale: 0.66 arcseconds/pixel
Size (pixels): 3040 x 2016
Angular Size: 0° 33' 23" x 0° 22' 09"
Position Angle: 272° 52' from North
RMS: 0.79 (X: 0.51 Y: 0.60)
Number of Stars Used in Solution: 216 (100%)
FWHM: 6.55 pixels, 4.31 arcseconds
***********************************************


Now *that* looks correct. (It even overlaid it on the sky chart, so I know it was). The PA is correct, the scale is correct, and the angular size is correct. So why was/is Pinpoint giving such a different answer?


Just for grins, I am now going to feed that source image to Pinpoint at max mag 18:



RA 20.19541
Dec 38.33767


Uh, wow. It worked.



8354 image stars found
(doing spiral search...)
10736 catalog stars found
Solved using 167 of max 500, RMS residual is 0.31 arcsec, order =4
Solution took 39.4 seconds
Centerpoint RA = 20h 11m 43.428s Dec 38° 20' 15.75"
WCS: Roll = 87.15 HScale = -0.661 VScale =-0.661
PA = 272.850°
FWHM = 4.09 arcsec


Now I'm terribly pleased, and terribly confused.



Aside from the minor detail that it is reporting the scale in negative numbers, it got the PA correct. Why, then, was the solve returning positive results when given a starting point a couple FOV away?


I surmise that much of this would have been avoided had I had MaxIm connected to the scope, so that RA/DEC would have been auto-inserted to the FITS header at acquisition. I also suspect that much of this is all based on a visual estimation error of mine earlier. (This may be humorous to you, but irrelevant, so you can stop reading now if you want).


I was trying to shoot the Crescent, NGC 6888, but was having trouble with precise framing. I came up with this (from the Aug 23 run):


http://www.flickr.com/photos/4162274...in/photostream


I didn't think that looked terribly much like the Crescent, though, so I brought up someone else's shot as mentioned in the comments on Flickr:


http://www.flickr.com/photos/lizarra...os/6069816287/


Now, I know that that image is narrowband, and mine ought not look anything like that, but one thing leapt out at me. In the upper left of that other shot's narrowband, there's a square of stars with the very brightest in the upper right corner of the box, with a fifth bright star off to the left. That *seemed* to my eyeball to match the square that takes up most of image, so suddenly I thought my "Precise Goto" slew was off by at least a couple fields of view! Strange, since I am meticulous with polar alignment (within 1 arcminute) and with calibration stars.


Looks like I really did hit the Crescent, with only a tiny bit of framing error. But my processing only brought out wisps on the edge, and because I didn't see the oval shape, and I had the "matching squares", I found the RA/Dec of one of those stars in the (wrong) square (way off to the left of the Crescent), and was feeding THAT to Pinpoint as the approximate RA/Dec.


So, you were correct -- it was off by multiple fields of view. I'm still a bit taken aback by the false positives, but I think I can solve that by raising the minimum stars to match to ensure a higher level of certainty. (I also suspect that such a false hit would never happen at max mag 16 or higher, which is why I was not getting solutions there).


So, Bob and Dick, I think my problems were PEBKAC-generated. (Problem Exists Between Keyboard And Chair). I believe I'm likely to have solved the problem, so for now, no need to spend your time digging.


One good night out under the stars with this solving and correcting for me, and it's a sold product.


Thanks!


Jeff

[Dick Berg]

HI Jeff,

I looked at your images from August 11. Attached is a Solve.txt file from Visual PinPoint, to which I have added some comments.

I loaded up the four images in VPP, set the USNO A2.0 catalog as the one to use, and got the four "paragraphs" of data out of VPP solution. VPP ran on the original, unprocessed images. It took some time, because there were so many stars I presume, but all four images solved correctly and for all practical purposes identically.

The negative scale numbers are okay - it has to do with the relationship between the +x and +y directions of the camera axes and the +RA and +Dec directions in the sky. At least, that's how I think of it.

For the Maxim solutions that I added to the bottom of Solve.txt, first I loaded the original 0046888.fit image (see FullSize.png), and solved it using the USNO A2.0. Since the FITS header didn't contain the RA and Dec coordinates, I typed those into the PinPoint Astrometry window. It solved quickly enough, but unfortunately landed on the wrong solution - close but no cigar. The focal length is wrong along with the PA and scale, and RA/Dec for that matter.

Next I reloaded the original 046888 image, applied MaxIm's Process/HalfSize (see HalfSize.png), and then solved the plate. (That's the last paragraph in the Solve.txt file.) That came out correctly and nominally the same as the VPP solutions! The scale here is twice the original because the image is half-sized.

I've attached two .png image - the original .fit image and one of the "half-size" image, to show the difference between the two. I think that there must be some pre-processing step you have to apply to the raw images before you can analyze or combine them. My friend Andy Blanchard, who has a remote telescope in Chile, had a similar color camera (maybe the very same model) that produced raw images with this "screened" look.

So this is what I've done for now. I haven't thought about your subsequent messages yet.

[Bob Denny]

Jeff --

My head is spinning... Attach a couple of sample images (zipped) so I can see them. I understand your frustration and the ways in which you've tried to succeed but what you're experiencing is vastly different from the experience of others, so I have to guess that it's related to something about your images. Sometimes if the pointing is way off, with PinPoint operating at the hairy edge of the envelope, it will act weird like that - by lplaying with the parametrs you're creating "just the right environment" for it to find a solution. Anyway, the images will help. Include the plate scale (arcseconds/pixel) please.