Data pre-processing

Understanding the contaminations in the shutterless mode of EIT

(David Berghmans, October 26th, 1998)

back to Shutterless main page

Introduction

This text discusses a few technical aspects of the shutterless mode of EIT. My discussion is based on the shutterless mode sequence of May 13, 1998. Specifically, I have a look at instrumenal effects such as the 'smearing' of the image due to reading out the CCD with the shutter open and skipping ccd clears. To conclude I give a few recommendations for new shutterless mode runs by EIT and for future EIT-like instruments.

The May 13, 1998 shutterless mode sequence (efz19980513.173209) had the following setup (see Fig. 1):

• a 128x96 subfield in the SW quadrant was choosen ( [512:639,352:447])
• read-out port B (bottom left)
• block east filter
• no ccd clears during the whole sequence to keep maximal cadence
• 230 images between 17:32 and 18:29, average cadence 14.9 s

For the following discussion we also note the following images on the same day:

filename image type filter efz19980513.165311 FOV test image Al sup efz19980513.170145 195 fdfr Al +1' efz19980513.183451 195 fdfr Al +1'

General procedure

After being exposed, a shutterless mode image of the subfield 'FOV' is transferred to the readout port. According to Delaboudiniere et al (SolPhys 1995, 162, p. 304) reading out a subfield takes the following three steps:

• dumping of all lines ahead of the fov. The lines preceeding the subfield are discarded at a rate of 0.480 ms per line. After this 'edge 1' has shifted to the left edge of the CCD and the charges collected in region FOV move to region A (see Fig. 1). Since the shutter remains open, the FOV image is contaminated by photons collected during the transfer through the non-occulted part of region B and through the remaining transfer path in the FOV itself.
• actual read out of the fov image now present in region A at a rate of 20.8 ms per line. Note that
  
  • thanks to the occulting mask the fov image is no further contaminated.
  • every time a column is read out, a 'black' column appears at the right hand side of the CCD (region D) which then further drifts leftwards when new lines are read out. In this process a smeared out image of the far western side of the sun is collected in region D.
• dumping of the remainig 1024-512-128=384 lines following the choosen subfield, again at a rate of 0.480 ms per line. Note that in this process the smear out image that was collected in region D is moved in the FOV image. Moreover this smear out image of the far western side of the sun is contaminated during the transfers through region 'C' and through the remaining transfer path in the FOV itself.

It is important to note that, since we do not do any 'ccd clears', this contaminated smear out image remains in the fov when the exposure of the next image starts.

Figure 1.

Estimation of the contamination

So, to summarise a subfield 'image' consists of the following contributions:

• smear out of region D with an effective exposure time of t1 = 128 x 20.8 ms = 2.6624 s
• smear out of the transfer through region C and the FOV with an effective exposure time of t2=384 x 0.480 ms =0.18432 s
• true solar image in the FOV region with an at this stage unkown exposure time of t3 s.
• smear out of the transfer through the FOV and the unocculted part of region B. The effective exposure equals the transfer time from the RHS of the FOV (512+128) to the RHS of the occulting mask (guess: 341) which gives: t4= (512+128-341)*0.480 ms = 0.14352 s

Finally, these 4 contributions are being transferred another 341-128=213 lines (which takes 213*0.480 ms = 0.10224 s) and are read out in 2.6624 s. So they spend T_o= 2.76464 s under the occulting mask. This value for T_o can be confirmed by noting (see Fig.2) that the average intensity of a shutterless mode image scales as the the time interval between the observation of that image and the preceding image plus an offset:

average_intensity(i) ~ obstime(i)-obstime(i-1) - ( 2.71 +/- 0.05 s)

Since we do not do any ccd clear the total exposure (t1+t2+t3+t4) time together with the time T_o spent under the occulting mask must equal the time interval between the observation of the preceeding image and the present image:

obstime(i)-obstime(i-1) = t1+t2+t3+t4 + T_o

This allows us to derive the effective exposure time t3 of the true solar image in the fov:

t3 = obstime(i)-obstime(i-1) - t1-t2-t4 - T_o
= obstime(i)-obstime(i-1) - 5.75488 s
= obstime(i)-obstime(i-1) - 5.76 s

Let us now, starting from the nearby full disc image efz19980513.183451 try to estimate what the combination of contaminations 1,2 and 4 look like. Taking the appropiate integrations from this full disc image and using the above deduced 'effective exposures' for each contamination, we sum up the

• contaminations 2 and 4 into a charge transfer contamination template with the following statistics in DN:
  Minimum = 3.9, Maximum = 9.1, Average = 6.4+/-1.3
• contamination 1, due to the read-out process of the previous image gives a read-out template with the following statistics in DN:
  Minimum = 0., Maximum = 52.3, Average = 13.0+/-11.0

Summing these two contributions gives the total contamination template which has the following statistics expressed in DN:
Minimum = 4.5, Maximum = 58.8, Average = 19.4+/- 11.0

Verification of the deduced total contamination template

Figure 3.

Let us now try to find observational evidence of the above scenario. Fig 3A shows a subfield extraction of the fdfr image efz19980513.170145 and Fig 3B shows the subfield test image efz19980513.165311. Both are taken with the normal operational mode of shutter opening and closing. Their difference, shown in Fig. 3c, are due to

• active region pattern due to solar variability between 16:53 and 17:01
• a bright line in the bottom of Fig. 3b due to 'subfield reading'
• vertical stripes due to the difference between the BLOCK_EAST filter and the Al + 1 filter

It's important to note that the difference in filter only produces vertical stripes, indicating that the two corresponding gridpatterns are well aligned in the y-direction but somewhat offset in the x-direction. A fortunate consequence is that although efz19980513.18345 has a different grid pattern than the shutterless mode sequence, we can still use it to estimate the smear out contamination since horizontal differences are absent and vertical ones get smoothed anyway when smearing.

In the next line, Fig. 3d shows a subfield extraction of the fdfr image efz19980513.183451 (Al + 1) taken with the normal operational mode of shutter opening and closing, while Fig. 3e shows the last but one image in the shutterless mode sequence (BLOCK_EAST). Their difference (Fig. 3f) shows all features allready seen in Fig 2c, but due to shutterless mode operation we now have an additional left-bright right-dark trend. We have enhanced this trend by applying horizontal smoothing to remove the grid modulation and median filtering to reduce the noise. This gives Fig. 4a which can be compared to the derived contamination template Fig. 4B. (shown in the same grey scaling). The bright white regions in Fig. 4a are regions where the shutterless mode image has a lower value than the fdfr subfield image, due to intrinsic solar changes.

On a global scale, Fig. 4a and Fig 4b show the same left to right dimming pattern. This becomes especially clear when plotting in Fig. 4c the vertical average of both Fig. 4a (solid line) and Fig. 4b (dotdashed line). Both curves have the same amplitude and the same trend. The largest differences (approx 5 DN, see Fig. 4d) are seen in between column 50 and 100, which is probably due to solar variability. Also several of the horizontal stripes in the derived contamination (Fig. 4b) are recognisable in Fig. 4a: in Fig. 4e we a vertical cut through Fig. 4a (solid line) and Fig. 4b (dotdashed line) averaged over the first 10 columns at the left. The profile of both curves is clearly similar, though the derived contamination apparently underestimates the observed difference by roughly 5 DN. Nevertheless this difference can be compensated by assuming that the exposure time mentioned in the fdfr image header was slightly too high (only 1.5 %).

By comparing the contamination template calculated from the 18:34 fdfr image with templates based on the 13:14, 17:01 and 19:13 fdfr image we conclude that the temporal variations of the total contamination template are typically less than 10% on the timescale of 1 hour which in absolute numbers means an uncertainty of only 2 DN.

Figure 4.

Conclusion and Recommendations

With the above analysis we have shown that due to the shutterless mode operation in combination with the absence of any ccd clears during the sequence:

• the images are contaminated by charges collected during the charge transfer phase and during the reading out of the preceeding image
• both contaminations add up to around 10 % of the DN-values in the darkest regions of the images and much less in the brighter ones
• by intrgrating along the transfer/read out paths in a nearby fdfr image, these contaminations can be relatively well estimated and can thus be subtracted from the shutterless mode images.
• after doing this the residual contamination is estimated to be less than 2 %.

Since this contamination can be removed relatively well, it is shown that skipping all CCD clears is an allowable option to enhance the observation cadence. For future shutterless mode runs on EIT, we recommend to add full disc images as close as possible before and after the run, both in the Al +1 filter as in the blockeast (or west) filter. This will allow an even better estimate of the contamination template and its evolution during the sequence. In case no ccd clear are programmed during a shutterless mode run, the occulting filter (block east or block west) should be choosen, not according to the selected read out port, but depending on the relative brightness of the far west or the far east of the full disc image.

For future EIT like instruments we propose to consider the implementation of 'partial ccd clears' which would allow to fine-tune the removal of contamination from the fov (and only there) without degrading excessively the observation cadence. In case of demanding mass/budget restrictions one could also consider to remove the occulting half masks completely, since our analysis shows that the resulting smear-out contamination can be well estimated and removed to within a few percent.