updated: 07/25/2006
When background levels are very low and/or the sky is very dark, A gradient can appear in color exposures when using the ST-10XME and CFW-8A. This didn't show up in the ST-8XE, apparently due to the lower sensitivity compared to the St-10XME. The mechanism is unknown at this point but anecdotal evidence seems to indicate the optical sensor for filter wheel positioning in the CFW-8 color filter wheel as the cause. ST-10XME's using other color filter wheels do not show this problem. The gradient appears only in the first exposure after the selection of a color filter. Subsequent exposures do not show this gradient, seeming to confirm the theory that as long as the filter wheel rotation sensor is not activated, the gradient does not appear. See Eddie Trimarchi's excellent detective work on this issue at http://astroshed.com/cfw8/cfw8.html
Some users have reported taking a throw-away 5 minute exposure (flush exposure) after each color filter change so that subsequent exposures will be free of this gradient. While this approach works, it does waste valuable imaging time. This can become a significant issue if exposure sequences of the from LLLRGB, LLLRGB, etc. are taken. Essentially 15 minutes is wasted for each series.
One way to deal with this lost imaging time is based on the assumption that the source of the fogging is reasonably constant. A fog correction frame is created for each filter and applied to subsequent exposures of that filter, after reduction and combining. This approach uses Mira AP to develop the correction frame.
1. In Mira AP, select CCD Proc | Correct Background. I had good luck with column terms = 2 and Row terms = 4. For the result select Create Image. Mira will create an image of the fogging with all stars removed. You will easily see the profile and magnitude of the fogging. Save this image as the correction frame for that specific filter.
2. Note the minimum levels of the correction frame. If the minimum is 300, subtract 290 using pixel math.
3. Repeat the above process for each filter. Save these images as references.
4. When you image and after you develop your master R, G, and B frames, subtract the appropriate correction frame from each one. The resultant image should have the fogging substantially removed.
I tried this with a sets of images taken 8 days apart with good results as illustrated below.
Figure 1: Calibration frame Figure 2: Correction Frame
Figure 3: Before applying Correction Frame Figure 4: After applying Correction Frame
All of the above images were taken with a 10 minute exposure with the blue filter at 2x2 binning and -10°C cooling. Figure 1 was taken on May 27 of NGC6027. Figure 2 is the correction frame that was developed in Mira from Figure 1, according to the above steps. Figure 3 is the combined image of 5 10-minute exposures of NGC5985 taken on June 8. There is a significant gradient from top to bottom. Figure 4 is the result of applying the correction of Figure 2 to Figure 3. Note that the gradient is substantially eliminated.
There may be other ways to do step 1 in other apps using appropriate low pass filtering and other processing with an appropriate star-poor field or even a capped OTA/camera. I found Mira AP ideal for this step. I have used the same approach to correct gradients that I know will appear when I shoot low to the south (over Tucson city lights).
Added 6/15/2003: I recently was processing someone else's image taken
with an ST-10XME/CFW-8A and noticed a gradient in his red master similar to the
above fogging. I applied one of my correction frames and it significantly
reduced the fog gradient! On a sample of two, this indicates a level of
repeatability for this mechanism. If you would like to experiment,
here is a link to correction frames for my R,G,B
set of filters taken with 10 minute exposures. You would probably need to
scale them for different duration exposures.
In an attempt to optimize the length of the flush exposure, an experiment was performed on a ST-10XME/CFW-8A of recent manufacture. The test was conducted in an office environment with the camera inside a cooled cabinet to minimize light exposure. The nosepiece was capped. Maxim was programmed to take a sequence of as follows:
| Filter | Exposure Time | Bin |
|---|---|---|
| Clear Red Green Blue Flush Blue |
10 sec. 10 sec. 10 sec. 600 sec. variable 600 sec. |
2x2 2x2 2x2 2x2 3x3 2x2 |
The sequence was an attempt to mirror a typical exposure sequence in terms of filter wheel movement. The Flush exposure was varied from 0 to 5 minutes. The second blue exposure was used to develop a background correction as described above. The difference between the maximum and minimum values of the correction frame were recorded as an indicator of gradient removal quality. A perfect removal would have a difference of 0.
Figure 5: First Blue image Figure 6: resultant Correction Frame
Figure 5 is an example of a typical blue frame without any flushing. Figure 6 is the correction frame that results from the frame in figure 5.
Figure 7: Flush Duration
Figure 7 shows the results of various flush exposures. Note that the unflushed peak value of the blue frame gradient was 32 ADU. This is a relatively low level and would not be seen except under dark conditions. As the flush duration is increased, the resultant size of the gradient reduces. Depending on your particular situation, you may be able to get by with less than 5 minutes. Indeed, depending on the sky conditions at your location, you may not even see the gradient.
Two methods are presented to deal with the gradient apparently resulting from the CFW-8A rotation sensor. One requires developing calibration frames and maximizes imaging time; the other prevents the problem from occurring at the expense of lost imaging time.
Examination of Figure 2 and Figure 6 show significantly different correction frames. These correction frames came from two different cameras. That of Figure 2 was manufactured last year and that of Figure 6 was manufactured last month. I don't know whether there has been an attempt on the part of the manufacturer to deal with this issue or this is a variation in the sensor. I do know that the peak value of the gradient was around 55 on the older camera and 32 on the more recent one.
Hopefully this note gives you some understanding of what is going on and perhaps explains those inaccurate flat fields or strange light pollution gradients. Ideally, SBIG would come up with a firmware or hardware fix to resolve this problem if possible.
7/22/2003 Update: SBIG upgraded filter wheel to the new design filter carousel, which replaces the indexing pins by flags that measure around 0.2" square. At the same time, I added flocking around the inside of the circumference of the CFW-8 housing in an attempt to attenuate any reflections from the emitter-detector pair. (Recall that an anodized surface is a great reflector of IR.) See below.
|
|
Also, all filter slots were populated, as recommended by SBIG. With the older index-pinned filter wheel, I measured fog gradients of 35-50 ADU, depending on filter and sequence. With the above modifications, I noticed a significant reduction of the gradients to 3 - 6 ADU. Both experiments were with 10 minute exposures at 2x2 binning, ST-10XME and at a temperature of -8°C. I think this is a significant improvement and is the result of the larger flags blocking the direct light from the opening to the camera as well as the flocking attenuating any reflected IR off the inside of the wall behind the emitter-detector pair.
