[Aavso-photometry] Re: Dithering Techniques

[Wolfgang Renz]

Hello Matt

> It will be interesting to see how this performs.  I know CCDSoft and
> possibly SBIG's CCDOps guide with the centriod on the corner of 4
> pixels. The thought being this approach is most sensitive to guide
> star movement (certainly more-so in undersampled conditions).

If I remember right then the guide star position can not be specified
to sub-pixel accuracy in these two !?

Is it possible to choose between pixel corner and pixel centroid
guiding in these two ?
This would be basically sufficient to do a simple 2x diagonal dithering.
But without scripting it would be a pain to do for many short exposure
time images.


> So, if MaxIm can remain sensitive to guide star movement with
> sub-pixel accuracy it could have a leg up.  But how do you maintain a
> centroid of a star to sub-pixel accuracy when 90% of the star's light
> is on a single pixel?
That might become a real problem if its really 90% or more. But
actually it should not be so bad as long as the pixel size of the guiding
chip is small enough and its operated in 1x1 binning.
As the guiding chip is far off axis, the resolution camera lenses at low
f number is already worse and the FWHM of the guiding star might be
up to twice as large as on-axis on the imaging chip.
I wouldn't mind if the guide star centroid would temporaryly move in
a range of +-0.5 pixel as long as the mean position stays where it is
supposed to be for the 'dithered' exposure. Remember, I don't want
to get an enhanced resolution but just want to distribute the starlight
evenly on the more or less sensitive areas of the CCD's inter and
intra pixel structures.



The software, that I'm looking for, must not support dithering explicitly.
Having the possibility to specify the guide star position to better than
a whole pixel should be sufficient. Having some kind of scripting (to
not have to make the setting of the guide star centroid manually) would
be very helpful.



> It appears MaxIm allows sub-pixel dithering, but I am only just
> starting to use it and don't know its full capabilities yet.

I don't own MaxImDL/CDD. But I had a quick looked at the Diffraction
Limited homepage a couple of weeks ago. It didn't say anything about
dithering in the 'CCD Imaging' section. And in the 'Autoguiding - AO-7'
section it stated that it supports AO-7 dithered images via autoguider.
So I thought that it works just with an AO-7. And that cann't be used
with camera lenses.

Now I looked again and found in the 'CCD Camera Control' section
that the new version can 'Automatically dither each exposure position
to smooth out effects of hot pixels. Works via telescope mount or via
autoguider.'

In the Online Manual it gives the following infos on dithering:
*** Dark Frame Calibration
...
Since hot pixels have extra noise, you may still find your image has
“speckles” after calibration; they are just smaller, and now some are
dark and some are bright. There are several techniques for removing
these noisy pixels. One way is to simply replace them with the average
of the surrounding pixels (in MaxIm DL both Kernel Filter and Remove
Bad Pixel commands can do this). A better way is to “dither” the
pointing of the camera slightly between exposures, thus distributing the
noise contribution of each hot pixel to a different position on the
image, and then combine a number of images together using a
median or sigma clip algorithm, which will reject the hot pixel
contributions altogether. 
...
*** SBIG AO-7 Control
...
The four AO-7 Control tabs are:
...
Setup    Select the track box size, view or edit AO-7 tip/tilt mirror calibration settings, exercise the tip/tilt mirror system to check operation, and dither exposures to compare guided and unguided star images. 
...
*** AO-7 Setup Tab
...
Enabling the Dithering option causes the AO-7 tracking position to change by a small amount after each main chip exposure. This is used to produce a sequence of images centered on a particular location, but with each one offset by a small random amount. The maximum number of pixels that the tracking position will be offset is minimum of the Amplitude setting and one half of the size of the Track Box. For instance, if an 8x8 track box is used, and the Amplitude is set to 5, then dithering will offset each exposure by up to 4 pixels.
...
*** Sequence Tab
...In some cases, it is desirable that the images in a sequence not all be centred at exactly the same point in the sky, but instead cluster around the specified position. For example, you might want this if your CCD camera had a significant defect that could not be remedied by the usual image calibration procedures. The Sequence tab provides two ways of doing this: Dither Images via Guider and Dither Images via Mount. The first method requires that an autoguider is available and running (see the Guide Tab). In effect, MaxIm CCD automatically adjusts the coordinates of the guide star between exposures of the sequence such that the centers of the images are randomly distributed near the initial position. The second method requires that your mount be controllable by the computer, and accomplishes the same thing by slewing the telescope to randomly-picked nearby locations between exposures. The two methods are mutually exclusive, so that selecting one deselects the other. You can, of course, turn dithering off altogether by clicking on the dither command in the menu that is currently checked.
In either case, the size of the random excursions from the nominal center position is controlled by the Maximum Dither Deviation command, which brings up a small dialog allowing you to enter the desired maximum excursion, in pixels. Enter a value up to 10 pixels and click OK, or click the X (Cancel) button to dismiss the dialog without changing the dither amplitude.
...
NOTE
With the exception of modes where its usage is either required by or incompatible with the sequence options, autoguider operation is independent of sequence execution and can be turned off and on at any time. Normally, if the autoguider is active at the start of a sequence, it will continue to run throughout. This is true even for cameras with integrated autoguider capability. Autoguider imaging is paused during main camera readout, with the guide star being automatically reacquired after the readout is complete. A delay of five seconds is imposed in order to allow time for the guider to settle, the value of which you can increase by setting Delay between on the Setup Sequence dialog to more than five. A longer delay is advisable if Dither Images via Guider is in effect since the guide star coordinates are also being changed between exposures.
...
*** Special-Purpose Filters
...
The best method is to take a series of exposures, and “dither” the camera position slightly between each exposure (MaxIm DL can automatically dither using the autoguider, telescope control, or AO-7 control). You then combine the images using Median, Sigma Clip, or SD Mask combine. This very effectively suppresses hot pixels without artificially substituting pixels.
- Taking Exposures – Advanced
...
Make sure none of Dither Images via Guider, Dither Images via Mount, Capture Mosaic, Continue Mosaic or Astrometric Resynchronization is checked in the Options menu. These are advanced features described in the reference section for the Sequence Tab.
...
*** What's New in Version 4?
...
CCD Imaging
...
Sequences - Setup wizard for quick and easy setup of common situations such as LRGB. Includes 16 slots, save/load sequence setups, optional seperate per-slot repeat counts, and run a different post-exposure script per slot. Dither images using mount, autoguider, or AO-7. Automatically capture image mosaics, with complete sequence executed for each mosaic tile. Periodically run PinPoint to detect drift and recenter image using GOTO mount—great for long unguided runs. Displays estimated total running time for sequences. View settings while sequence is running.
...

Sounds all as the exact dithering positions are not controllable in
MaxImDL/CCD and this kind of dithering is aimed to move the
single images by a couple of (full) pixel.


But it has a nice feature called 'Offset Tracking ... to specify a series of positions for the guider to track to. This can be used to track solar system objects such as asteroids and comets, or to compensate for flexure in your system.'
This could be used to change the position in relative or absolute pixel guide star pixel coordinates during a longer exposure.


As the guide star pixel positions can be specified in MaxImDL/CCD
to a fraction of a pixel, it could be changed in between several short
exposures. As it also has some scripting, I could give it a try.


Clear skies
  Wolfgang

-- 
Wolfgang Renz, Karlsruhe, Germany
Rz.BAV = WRe.vsnet = RWG.AAVSO




----- Original Message ----- 
From: "Matt Thomas" <groupmail at astromatt.com>
To: <ccd at listserv.wwa.com>
Sent: Friday, February 18, 2005 2:03 AM
Subject: Re: [CCD] Re: Dithering Techniques


> Wolfgang,
> 
> It appears MaxIm allows sub-pixel dithering, but I am only just
> starting to use it and don't know its full capabilities yet.
> 
> It will be interesting to see how this performs.  I know CCDSoft and
> possibly SBIG's CCDOps guide with the centriod on the corner of 4
> pixels.  The thought being this approach is most sensitive to guide
> star movement (certainly more-so in undersampled conditions).
> 
> So, if MaxIm can remain sensitive to guide star movement with
> sub-pixel accuracy it could have a leg up.  But how to you maintain a
> centroid of a star to sub-pixel accuracy when 90% of the star's light
> is on a single pixel?
> 
> -Matt Thomas
> http://www.astromatt.com/
> CCD Commander: multi-target unattended imaging
> http://ccdcommander.astromatt.com/
> 
> Thursday, February 17, 2005, 4:31:45 PM, Wolfgang wrote:
> 
> > Hello
> 
> > There is an other application for dithering in photometry.
> 
> > In photometry we do no image processing (except of bias, dark &
> > flat frames and sometimes image stacking) to not influence the 
> > measurements. So in photometry single hot and cold pixels are
> > not 'smeared' over several pixel due to image processing.
> > To get high internal precession within a differential magnitude time
> > series, we usually try to keep the stars at the same pixel. So any
> > remaining anomalies stay constant and produce just a constant
> > offset in the data.
> 
> > To get high accuracy magnitude values in the centi (and especially
> > the milli) mag range for all stars in a single image, the remaining
> > anomalies show up significantly.
> > One of the sources is flatfielding. Due to internal reflextions in the
> > scope etc., an evenly flatfielded background doesn't imply that the
> > image is also well corrected photometrically. To improve accuracy
> > one can apply an additional photometric flat field correction
> > (photometric super flat) after flatfielding. Bruce Gary is doing some
> > research on this issue currently.
> 
> > An other major source of inaccuracies in photometry on UNDER-
> > SAMPLED images are the insensitive inter pixel structures (with
> > full- frame CCD chips like Kodak KAF chips and especially with
> > inter-line CCD chips like Kodak KAI and Sony chips) and the
> > different blue sensitivity of the two gates of the Kodak blue
> > enhanced E and ME chip pixel.
> > Microlensed chips try to focus most of the light onto the most
> > sensitive area of the pixel. At high f numbers that will for sure work
> > in the center of the image, but that might also fail or even make it
> > worse at low f numbers near the edges and in the corners.
> > The solution to overcome these inaccuracies would be to make
> > several undersampled images and dither these images on a sub
> > pixel scale (2x diagonal, 2x2, 3x3). As the aim of this dithering is
> > not to improve the resolution, the images might be added by a bi-
> >linear sub-pixel shift and add algorithm (so the count totals stay the
> > same and the stars are enlarged by just one pixel). As severely
> > undersampled images just occure at short FLs, the targets would
> > be bright ones that would require multiple short exposures to not
> > saturate the CCD and to beat the scintillation noise in such short
> > exposures below 10 seconds.
> 
> > An average camera lens has in the center a resolution of ~ "30
> > LP/mm that equates to 16.7 microns per line". A good camera
> > lens has in the center a resolution of ~ "50 LP/m that is 10 microns
> > per line" (Source: email by Paul Howell). 
> > Many CCD chips have a pixel size of 20 microns. So it is no
> > wonder that the light of a star might fall just on 1, 2 or 4 adjacent
> > pixel. Thats pretty badly undersampled. But even with the 6.8
> > micron pixel of a Kodak KAF- 3200 chip and an average camera
> > lense its still undersampled without defocusing.
> 
> 
> > Is sub-pixel dithering already possible and supported by any
> > existing image aquisition software  ?
> 
> 
> > Clear skies
> >   Wolfgang