[Aavso-photometry] Banding in flats

[Wolfgang Renz]

Hi

>> Your master flat shows a top to buttom gradient of ~ +550 ADU and
>> a left to right gradient of ~ +150 ADU. This is probably due to an
>> uneven illumination of the flats.
>> Did you take sky flats ?
>> With a diffusor, or without one ?
> 
> This is a twilight flat with no diffuser.

Then you might try it with a relative strong diffusor (~ 33% transmission).
This should get rid of all star trails and gradients pretty well. But it might
increase issues due to scattered light a bit. So you might add a shorter,
well blackened "dew shield" and place the diffusor on top of it as a cap.
Then it would be best to use that "dew shield" for imaging too.

>> A top to buttom gradient might be explained with not dark subtracted
>> images also by the accumulated dark current during read-out time.
>> But such a gradient should disappear when applying appropriate darks.
>> Darks might not be appropriate e.g. when the CCD temperature was not
>> the same as with the flats or lights, or when the read-out times dif-
>> ferered (e.g. due to different CPU or I/O loads). This might be true
>> too when taking exposures of less than a few seconds and using scaled
>> darks due to not considering a shutter correction (= true shutter
>> opening time doesn't equal stated exposure time).
> 
> The flats where properly bias and dark subtracted (see the .fits header).
> All frames were made at the same temperature.  

That was just to explain in general where a gradient might come from else.

>> Aside of this, the master flat is really pretty evenly illuminated.
>> So I guess that your 8" f/5 Newton is optimized for photography and
>> has a larger secondary mirror than if optimized for visual observ-
>> ing, right ?
> 
> Yes, it has a 2.14" minor axis secondary.  I get 100% illumination over
> most of the CCD in the ST-7.

>> It has a dark band of ~ 12 pixel at the top that is probably caused
>> by the shadow of the guiding chip mirror of your ST-7XME when using
>> it at a fast scope as you do. This usually leaves no dark band in
>> a flat calibrated images and should therefore not affect photometry
>> much.
> 
> The shadow at the top and the reason for the larger gradient top-to-
> bottom vs. side-to-side is the shadow of the autoguider pick-off mirror
> from the ST-7XME.

Yup, but just for the pixel at the very edge, not for the entire height gradient.

>> The master flat gets brighter towards all 4 edges within a range of
>> ~ 20 to 35 pixel. I have not seen this for quite some time. I don't
>> know the reason for this for sure, but if I would have to guess, I
>> would say its due to a reflection off the bright golden package areas
>> sourrounding the CCD chip. So these areas are effectivly lost for
>> photometry due to inappropriate flat fielding.
>> The master flat shows 5 a bit darker bands in the top half, a band
>> free area in the middle, and 4 a bit darker bands in the buttom half
>> on the very left. If the above is the true cause for the bright ed-
>> ges,  the bands are probably due to the shading of the bonding wires
>> to the CCD chip. The are 6 of them on both sides in the top half and
>> 5 of them in the buttom half. If one doesn't count the ones going to
>> the very corners of the CCD chip and therefore can show at most half
>> a band, the expected number of full shadows is 5 and 4. So this would
>> fit the observed shadows.
>> If this is the true cause, you might get rid of the bright edges
>> and the banding by placing a mask directly over the CCD chip. The
>> mask would have to be a bit larger than the active area of the
>> CCD chip to consider the f# of the used scope (chip width/height
>> + distance of chip to mask / f#).
> 
> If the banding is due to the bonding wires, then why are they only seen
> on the left side?  The bands are tilted a bit too, not exactly horizontal
> as I would expect if they were due to the bonding wires.

This question I asked myself too. As you use a Newton, maybe due to
uneven off-axis illumination of the focal plane. This might be e.g. pos-
sible when the CCD side with the bands was towards the primary mir-
ror and there was light falling directly through the tube opening into the
eyepiece tube. You can check this by making two flat sets and rotating
the camera by 180° inbetween. If thats the actual cause, using a  short
"dew shield" with the Newton should help with this issue too. If it helps
well, adding a mask close to the focal plane might not be necessary,
but still usefull if you use the camera with other setups too.

If its not due to the shadows of the bonding wires, it might be due to re-
flections of the golden package bonding pads. There are 12 of them to
the left and to the right. But just ~ 9 or 10 of them might fall on the chip
as they are spaced more widely.

If its due to neither of both, you might have a finned surface somewhere
that gets illuminated somehow (either shadows or reflections of it). Some-
thing like a screw, threaded rod or other thing with windings that is orien-
tated about parallel to the shorter side. If the banding is due to a collimat-
ed beam, it should have a pitch of ~ 512 * 9 um / 11.5(+/-0.25) = 0.40 +/-
0.01 mm = 0.0157 +/- 0.0004 in = ~ 63.5 +/- 1.4 tpi (teeth/turns/threads
per inch). So, assuminmg its within the accuracy), it might be a ISO1502/
DIN13 M2 (1.51 mm = 0.0594 in diameter), a BA 9 (0.0748 in = 1.90 mm
diameter), a coarse UNC #1 (0.0730 in = 1.8542 mm major diameter) or
a fine UNC #2 (0.0860 in = 2.1844 mm major diameter). If its something
like this its probably something that is tangentially through the eyepiece/
focuser tube.
If its something in a covergent beam, the pitch can of course be larger.

>> The calibrated master flat shows also some kind of 1 pixel wide,
>> a bit darker grid with a period of ~ 22.5 pixel in y. Something
>> like this could easily be explained for an interline chip as a
>> possibly present block structure. But I have not seen something
>> like this with a full frame chip yet. So it might be due to an
>> incorrect calibration of the flat.
>> Is this pattern also present in your master bias, master dark
>> (without bias subtraction), and mast flat frames (without bias
>> and dark subtraction) ?
> 
> No.  This faint grid pattern can only be seen in master flats that
> are stretched quite a bit.  Probably just structure on the CCD.

Yup, then its probably caused by production tolerances that affect the
area of the pixel and/or their QE (e.g. due to slightly misplaced micro
lenses, ...).

>> If I look at the FITS header, it says:
>>   EXPTIME 1.0000000000000000 Exposure time in seconds
>>     HISTORY Changed data type to 32 bit real 
>>     PROCESS '1001' Processing flags
>>     HISTORY Subtracted bias frame mbias6-1-20 MinMaxClip.fts 
>>     HISTORY Subtracted dark, fdkth20s-1-10MMC-20C031409.fts,
>>                       scaled by 0.05 
>>     HISTORY Combine Images: Min/max clipped mean of 10 images 
>> So it looks like as you used a scaled dark which rises the question if
>> the subtracted scaled dark frame was bias subtracted too. If not, you
>> might have ended up subracting the bias twice (once with the bias sub-
>> traction and once with the dark subtraction) which would introduce the
>> basic bias structures into the dark calibrated flats again.
> 
> The subtracted dark frame is actually a thermal frame (already bias 
> subtracted).  I use many different exposures for both flats and light
> frames and it is not practical to take matching darks for everything.

Yup, with many different exposure times, taking matching darks can get
real work. But if one has a camera that gives issues due to scaled darks,
one doesn't have much of a choice. In this case it helps to choose a set
of fixed exposure times (e.g. exposure times with 2^(1/2) = sqrt(2) =
1.414 or 2^(1/3) = 1.260 as step factor). This will not allow to optimize
the exposure times 100% for the upper end of the linear range (for flats
or bright variables), but with a sufficient small step factor one can get
close enough to not get a significant loss and one also gets easily used
to them even for taking flats.

> Darks and bias frames are not the source of the banding since they are
> visible on raw flat frames that have not been bias or dark subtracted.

That comment was meant on the grid issue and not on the banding.

> I've posted a .jpg of a V master twilight flat here that shows the banding
> a little better:
> http://chagrinvalleyastronomy.org/images/mflatsky2s-1-10V%20MinMaxClip.jpg

In this V band master flat the bands are better visible than in the in the Rc
band flat. Also the two very faint bands in the "gap".

> I'm still thinking that the banding is caused by the cover slip.

How should that work ? Interference patterns due to cancelation in reflec-
tions like in a Halpha filter ?
The difference is ~ 50 ADU at a signel of ~ 32k ADU. This is a full 1.5 % !
Thats is more typical for a reflection.

The bands are at the very same position, have the same orientation, and
the same spacing in the V and in the Rc band flat.

> Unfortunately, the cover slip is epoxied to the chip and I cannot remove it

Yup, that would be pretty dangerous by using force as if the glas breaks,
it might damage the CCD surface or cut a bonding wire.
If one would know how the cover glas was glued to the package, one
could search if there exists a solvent to "unglue" it. But if one exists, I
wouldn't bet on it that it doesn't disintegrate the plastic of the package.

> to test my theory.

But you could test if it makes a difference if you rotate the camera by 90°
or by 180°.

Clear skies
 Wolfgang

-- 
Wolfgang Renz, Karlsruhe, Germany