In early 2022 (between January 29 and February 2), Ben introduced an empirically determined pointing correction that changed the zero offset of the RA and DEC with the goal to remove the slow drift due to SPAR flexure. The SPAR flexure is not sensed by our short guiding telescope and causes a slow motion of the Sun under the occulter for both instruments. Historically, this drift had been addressed by manually repointing the telescopes during the observing day.
In its present and past form, the drift correction algorithm introduces a large unwanted jitter which is particularly noticeable in the UCOMP polarization data. There were multiple version of the algorithm. Currently, the algorithm finds the 5000 brightest pixels in a KCor image and uses the center of mass of these points to supposedly find the centre of the Sun under the KCor occulter (i.e. all 5000 points are weighted equally). This center is then used to determine the zero offset correction. I noticed UCOMP data to be noisier after the correction was implemented. In response Ben changed the algorithm to use the average center of 5 KCor images instead of the center of one KCor image and used smaller steps to implement the adjustments. The zero offset adjustment is currently implemented in smaller steps with a 21Hz frequency. The offset is updated every KCor image, i.e. uses a rolling average of 5 KCor images.
The plots below show the zero offset change over the day (top plot) and the 15s adjustments (bottom plot). We see that the adjustments in the zero point are very noisy: they can be positive or negative and are much larger than the expected drift due to spar flexure. The plot in February is when only one KCor image was used to compute the adjustment. The June plot corresponds to a newer version of the algorithm. The jumps in the adjustments are smaller for the June case, but still much (~20 times) larger than the slow and smooth drift expected from the spar flexure. We need to average these bumpy up-and-down adjustments over 100-120 KCor images to tease out a slow drift of about 0.01-0.02 arcsec every 15s. Is this the SPAR drift we are after?
The large and rapid oscillatory motion introduced by the zero offset correction that started in January-February 2022 completely changed the way we point the SPAR and had unintended consequences. It introduces noise and degrades the quality of the data: it reduces spatial resolution, makes 5m difference images noisier, and results in intensity induced spurious polarization. This is a more severe issue in UCOMP because UCOMP has smaller pixels and these up-and-down adjustments can happen within a single file. The UCOMP pipeline assumes all frames within a UCOMP file are co-aligned, which is not true anymore because of this correction. Note that aligning individual frames within a single UCOMP file is not an easy task. Single frames are noisy and we do not have a centering algorithm for science data as the center is found in the off-band images only. Properly aligning raw UCOMP frames may not be feasible and would significantly slow down the UCOMP pipeline.
It is high priority to avoid this unwanted jitter when we reopen the observatory. The zero offset adjustment should be computed using 100-120 KCor images, not 5 KCor images. The correction should be applied with a lower frequency of ~1Hz (see comments below) to avoid moving the SPAR faster than its internal frequency.
Questions to be answered are:
-
Is the slow drift of ~0.01-0.02 arcsec every 15s, seen after the adjustments are averaged, the drift due to the SPAR flexure? In other words, is this correction working? If so, in spite of the short-term bumpiness introduced in KCor data, KCor should not require pointing adjustment during the observing day after February 2022.
-
The drift due to spar flexure that caused the Sun to slowly move under the COMP occulter was estimated of the order of ~10arcsec per hour, i.e. twice as large as the slow trend seen in the KCor adjustments after averaging. Is this reasonable? COMP and KCor are mounted on different parts of the SPAR, but a factor of two difference seems quite large. Note that if the same drift is seen by COMP and UCOMP, then UCOMP drift is not corrected by these zero point adjustments and still requires multiple manual re-pointings over the observing day to keep the Sun well centered under the occulter.
-
What is the cause(s) of the noisy up-and-down adjustments found by Ben's algorithms? There are several possible explanations:
a) the bumpiness could be solely or mainly due to the inaccuracy of the center-of-mass algorithm used to find the center. This can be easily tested. If the center is correct, extracting an annulus around the center in KCor images would give a uniform intensity distribution with position angle. This test can be done for random images on random days to check how well the centering routine works.
b) seeing could be a factor, however, the bumpiness in the adjustments does not seem to be lower early in the morning and to increase as seeing degrades
c) by constantly sending large adjustments to the SPAR, we could induce an unwanted SPAR oscillation. According to Joan, Greg Card estimated that the natural frequency of the spar was 0.5-1Hz. Thus, slowing down the adjustments seems wise. We can also test this by applying Ben's algorithm on data taken on days when the zero point correction was not applied, i.e. before January 29 2022, on a selected time of 1-2h when there was no manual repointing and the zero offset was stable. If the difference between 15s adjustments found by Ben's algorithm is smaller and more stable than what we see in 2022, it would be an indication that the correction applied since February 2022 induces an unwanted oscillatory behaviour by forcing the SPAR to move faster than it is capable of. If the adjustments are just as noisy as in 2022, it would suggest the jitter comes from the inaccuracy of the center finding algorithm.
In early 2022 (between January 29 and February 2), Ben introduced an empirically determined pointing correction that changed the zero offset of the RA and DEC with the goal to remove the slow drift due to SPAR flexure. The SPAR flexure is not sensed by our short guiding telescope and causes a slow motion of the Sun under the occulter for both instruments. Historically, this drift had been addressed by manually repointing the telescopes during the observing day.
In its present and past form, the drift correction algorithm introduces a large unwanted jitter which is particularly noticeable in the UCOMP polarization data. There were multiple version of the algorithm. Currently, the algorithm finds the 5000 brightest pixels in a KCor image and uses the center of mass of these points to supposedly find the centre of the Sun under the KCor occulter (i.e. all 5000 points are weighted equally). This center is then used to determine the zero offset correction. I noticed UCOMP data to be noisier after the correction was implemented. In response Ben changed the algorithm to use the average center of 5 KCor images instead of the center of one KCor image and used smaller steps to implement the adjustments. The zero offset adjustment is currently implemented in smaller steps with a 21Hz frequency. The offset is updated every KCor image, i.e. uses a rolling average of 5 KCor images.
The plots below show the zero offset change over the day (top plot) and the 15s adjustments (bottom plot). We see that the adjustments in the zero point are very noisy: they can be positive or negative and are much larger than the expected drift due to spar flexure. The plot in February is when only one KCor image was used to compute the adjustment. The June plot corresponds to a newer version of the algorithm. The jumps in the adjustments are smaller for the June case, but still much (~20 times) larger than the slow and smooth drift expected from the spar flexure. We need to average these bumpy up-and-down adjustments over 100-120 KCor images to tease out a slow drift of about 0.01-0.02 arcsec every 15s. Is this the SPAR drift we are after?
The large and rapid oscillatory motion introduced by the zero offset correction that started in January-February 2022 completely changed the way we point the SPAR and had unintended consequences. It introduces noise and degrades the quality of the data: it reduces spatial resolution, makes 5m difference images noisier, and results in intensity induced spurious polarization. This is a more severe issue in UCOMP because UCOMP has smaller pixels and these up-and-down adjustments can happen within a single file. The UCOMP pipeline assumes all frames within a UCOMP file are co-aligned, which is not true anymore because of this correction. Note that aligning individual frames within a single UCOMP file is not an easy task. Single frames are noisy and we do not have a centering algorithm for science data as the center is found in the off-band images only. Properly aligning raw UCOMP frames may not be feasible and would significantly slow down the UCOMP pipeline.
It is high priority to avoid this unwanted jitter when we reopen the observatory. The zero offset adjustment should be computed using 100-120 KCor images, not 5 KCor images. The correction should be applied with a lower frequency of ~1Hz (see comments below) to avoid moving the SPAR faster than its internal frequency.
Questions to be answered are:
Is the slow drift of ~0.01-0.02 arcsec every 15s, seen after the adjustments are averaged, the drift due to the SPAR flexure? In other words, is this correction working? If so, in spite of the short-term bumpiness introduced in KCor data, KCor should not require pointing adjustment during the observing day after February 2022.
The drift due to spar flexure that caused the Sun to slowly move under the COMP occulter was estimated of the order of ~10arcsec per hour, i.e. twice as large as the slow trend seen in the KCor adjustments after averaging. Is this reasonable? COMP and KCor are mounted on different parts of the SPAR, but a factor of two difference seems quite large. Note that if the same drift is seen by COMP and UCOMP, then UCOMP drift is not corrected by these zero point adjustments and still requires multiple manual re-pointings over the observing day to keep the Sun well centered under the occulter.
What is the cause(s) of the noisy up-and-down adjustments found by Ben's algorithms? There are several possible explanations:
a) the bumpiness could be solely or mainly due to the inaccuracy of the center-of-mass algorithm used to find the center. This can be easily tested. If the center is correct, extracting an annulus around the center in KCor images would give a uniform intensity distribution with position angle. This test can be done for random images on random days to check how well the centering routine works.
b) seeing could be a factor, however, the bumpiness in the adjustments does not seem to be lower early in the morning and to increase as seeing degrades
c) by constantly sending large adjustments to the SPAR, we could induce an unwanted SPAR oscillation. According to Joan, Greg Card estimated that the natural frequency of the spar was 0.5-1Hz. Thus, slowing down the adjustments seems wise. We can also test this by applying Ben's algorithm on data taken on days when the zero point correction was not applied, i.e. before January 29 2022, on a selected time of 1-2h when there was no manual repointing and the zero offset was stable. If the difference between 15s adjustments found by Ben's algorithm is smaller and more stable than what we see in 2022, it would be an indication that the correction applied since February 2022 induces an unwanted oscillatory behaviour by forcing the SPAR to move faster than it is capable of. If the adjustments are just as noisy as in 2022, it would suggest the jitter comes from the inaccuracy of the center finding algorithm.