Sunday, January 19, 2014

Vertical stripes on images

I have seen a couple of times vertical stripes on short exposure images with the FLI camera. But last night when I took images of the Tadpole Nebula, I could see them on 20 min exposures (with an SII filter):


And what's worse, they didn't go away in the stacked image, but got even more accentuated:

And that even though I dithered the images. Sent a question to the fli_imaging_systems Yahoo! group.

I also processed the images with DeepSkyStacker and Pixinsight (finally had a real reason to try it out). With the same result. I then acquired the images with other programs (TheSkyX and FLIGrab) - they looked exactly the same.

So, it's not the acquisition and neither the stacking process.

Richard Crisp is suspecting that the stripes are FPN (Fixed Pattern Noise). Which should be eliminated by correct flats. But somehow they aren't. He recommended that I a) take more flat frames (20), and b) go up to 2/3 of full well. My camera testing report stated that saturation is 52266.1e- (with my gain of 0.81 that's almost exactly 64000ADU), so he recommends that I take my flats at 43000ADU.

This will be a problem with the Ha and SII flats. They already take 150 and 350 seconds to get up to 22000ADU. That would double the exposure time :-( Not only does it then take WAY long to get these (20 * 700 seconds = almost 4 hours!) but I also need to start taking darks for the flats. The only way to do this are dusk flats. Will try that out tomorrow morning.

---

I took 20 sky flats with CCDAutopilot. Exposure time ranged from 6.4 to 11.4 seconds. But the result was exactly the same. Richard Crips checked all my subs (lights, flats, darks, bias) for me - but he couldn't find anything either.

Next suspicion is the connection to my laptop. Right now, I run the USB cable from the camera to the USB switch on the mount. From there I run it to another USB switch and then through an (active) 15 feet USB cable to my laptop. I put my laptop next to the scope and connected the scope directly to the laptop.
... same result :-(

---

So, at this point, I know that I get these vertical stripes on my lights, but not in my flats. So, the issue is most likely with image acquisition. When I investigated last night, I found another issue: when acquiring bias frames they don't show any signal at all! Not sure it's related, but it's another indication that there is something wrong with acquiring or transferring the image.
  1. either capturing the photons in the camera
    Could be that the camera itself has an issue. I took some bias frames right after I received the camera which showed signal. So, at some point this worked. This indicates that this issue was recently introduced.
  2. transferring the images from the camera to the computer
    I tried various USB cables and connecting the camera directly to the computer. So, it could be that my laptop's USB port isn't working well.
---

I did more experiments and am convinced that this is an issue with the chip/camera itself. I took several flats last night with increasing exposure time (starting from 0.1 sec). And the vertical stripes are visible in the lower exposures, but they disappear in higher exposures when the signal gets stronger.

0.1 seconds:

2 seconds

40 seconds:

The lines are easily visible in the 0.1 sec exposure, less in the 2 sec and almost disappear in the 40 second exposure. Just to check that this is not an issues with stretching the image, I subtracted 44.000 from the 40 second exposure (it had a mean ADU of 45000, the 0.1 second exposure had a mean ADU of 1,800). But even then the stripes are not there:

Also, all these flats were acquired with the luminance filter, i.e. it's not related to the wavelength, but apparently purely to the signal strength.

---

After some more research, I found more references on the web about this issue with the 11002 chip, e.g.
CCDBand-Aid repairs vertical bars in Kodak KAI-11000M images or Files Not Being Calibrated Correctly. Bob Franke even developed CCD Band-Aid - a program to eliminate these vertical bands. I tried it but initially got an error message ("1x1 Phase is not an integer.") I selected a different image and now CCD-BandAid detected a phase of 23 (it handily saves this in its .ini file, so you have to run the analysis only once). I then tried to correct one of my subexposures:

Original Image:

Corrected Image:


I wanted to find out if and what CCD-BandAid corrected and created a diff of both files:

So, this only difference between the files are these vertical stripes. So, CCD-BandAid seems to have identified the phase correctly, but apparently didn't apply the correct amplitude?

I also contacted FLI to ask them for advice. They told me to send the camera in, but also cautioned me that it might not be possible to eliminate these vertical stripes 100% hardware-wise.

---

I was finally able to use CCD-BandAdi correctly and created narrowband images of the Tadpole Nebula and the Rosette Nebula. But I finally sent the camera back to FLI to correct the chip.