March 98 Southern Ks Background Jump Problem

In the first southern pipeline runs a problem was found while examining the scan background plots. It appears that the dome tracking is not keeping far enough ahead of the telescope causing changes in backgrounds (primarily at K). The problem was most obvious on 3/21 when the dome drive motor failed and the dome moved clearly into view causing trememdous jumps in Ks background (from thermal emission) and corresponding drops at J & H (caused by vignetting).

HOWEVER, subtle hints of the problem, in the form of Ks jumps, appear on the 19th, and in subsequent hours and days after the drive motor was fixed.  This implies that even when operating without mechanical failures, there has been an ongoing dome tracking problem.

The following scan background plots illustrate the problem. The "Matching Range" plots show J, H & Ks at fixed Y ranges (though the range is allowed to float up and down), allowing absolute flux ranges to be compared scan to scan. The "Full Range" plots expand to the full background range in each band, showing maximum detail in each plot. The fixed ranges are 150, 300, & 600 counts at J, H, & K, respectively.

The abrupt drops in Ks background are presumably caused by the dome slowily drifting into view, slowly elevating the Ks background then having the dome motor kick in, moving it out of the way, and subsequently dropping the backgrounds within a frame or two. We note that in all of these nights the downward drops in Ks occur only in alternating scans (i.e. if you see the drops only in even [or odd] scans). Sometimes the intervening scans show Ks jumps (compare for instance the even/odd numbered scans on 3/19). This seems to support a mechanical nature for the origin of the jumps since the telescope alternates between north and south scanning from one tile to the next (and assuming one sense of scanning is more likely to pass in front of the dome than the other).

The major problems associated with the breakdown of the drive motor show up in scans 37-39. The fact that the glitches seen when there was definite dome vignetting appear similar in nature (though much larger than magnitude) suggest a similar origin for the smaller glitches as well. The smaller glitches are often only 10 counts or less, while the largest associated with the dome motor failure spans ~600 counts at K. In the largest glitches, anti-correlated behavior at H & J is seen, but at a much-reduced scale. The glitches shown for the other days are typical but far from exhaustive; roughly 1/5 to 1/10 scans seem to be affected by at least the small-scale jumps.
During the dome drive failure period, serious artifacts (similar to airglow backgrounds) are seen at H & Ks band.


Upward jumps seen on even scans, downward jumps on odd scans

Note that on scan 27, the first drop is ~30 counts at coadd ~108. In the coadd images one can arguably see a small step in the Ks background about 1/3 of the way down the coadd. The direction of this drop is opposite of the sense of the H background drift.

Scan 24   Matching Range  Full Range
Scan 27   Matching Range  Full Range   H Coadd 103  K Coadd 103
Scan 127   Matching Range  Full Range


Upward jumps seen on odd scans, downward jumps on even scans

Note that in scan 07, which has a ~20 count drop around coadd 97, the coadd shows no significant change in background, suggesting the effect was removed in processing. However, the similarity in form of the Ks jumps this night (in which the dome rotation is almost unquestionably the culprit) is very close to that seen on other evenings even though on other nights the Ks jumps are much smaller in magnitude.

Scan 07   Matching Range  Full Range   K Coadd 92
Scan 10   Matching Range  Full Range
Scan 11   Matching Range  Full Range
Scan 14   Matching Range  Full Range

Dome motor breaks:

In scan 36 it is clear in the coadds that now, when the dome is almost certainly the culprit for the background jumps, the coadds show dramatic background degradation at K, and to a much lesser extent and reversed sense at H.

Scan 36   Matching Range  Full Range   H Coadd 209  K Coadd 209
Scan 38   Matching Range  Full Range   H Coadd 115  K Coadd 115
Scan 39   Matching Range  Full Range

Scan 61   Matching Range  Full Range

Mainly downward jumps noted on these nights, only on even scans.


Scan 52   Matching Range  Full Range


Scan 112   Matching Range  Full Range


Jumps noted in almost every even scan for the entire night. Usually there are 2 jumps/scan, separated by ~100 frames, 5-10 counts in each jump. Typical larger plot follows. There is no evidence of the jump in the K coadd.

Scan 50  Matching Range  Full Range  K Coadd 092


Jumps noted in almost every even scan for 1st half of night; jumpy OH airglow baselines make it difficult to identify jumps later in evening. Jumps appear to be of order ~5 counts.

-R. Hurt   9/17/98

Problem fixed?

On 9/28/98 I examined 16 processed survey scans each from 980704s and 980909s, as well as all 74 of the survey scans taken on the 980801s. There was no evidence for Ks background jumps that might be associated with the dome. Somewhere along the line the problems seems to have been fixed.

On 9/21/98 Mike Skrutskie comments:

The Ks background jumps in the southern hemisphere are likely to be benign.

1) The level of the effect is about 20DN compared with a emissivity=1 signal of several thousand DN.  If the change in background were due to true occultation of source photons by the dome the bias would be a fraction of a percent.

2) I believe that the explanation of the jumps lies in:

This link shows out-of-focus images of a bright star from the Southern telescope.  The camera boresite is slightly off center of the secondary and the edge of the camera pupil image is close to being coincident with the edge of the secondary mirror. It is possible that the changing background is due to the illumination of the non-specular edge of the secondary by the dome radiation.  Varying amounts of dome radiation get scattered into the camera (at an optical location that produces uniform illumination of the entire array) depending on the orientation of the dome relative to the telescope.   We don't see the effect in the North because the Ks-band pupil is better centered  on the secondary.

One can test this hypothesis since real occultation by the dome will not illuminate the field uniformly (you see a very out of focus image of the dome).  Scattered light at the edge of the secondary mirror will illuminate the field uniformly.

If the offsets are uniform and tiny they will presumably be removed by the processing.  Are there artifacts in the images resulting from the Ks-band jumps? If not, we should probably live with the effect, since fixing it may introduce new problems.

P.S.  Rae is certain that the telscope points well clear of the dome slit -- at least as of his most recent visit there.  He notes that the system was  in a state of flux in late-March, so we should also be checking whether the most recent data shows similar jumps.

On 9/24/98 Rae Stiening comments (while at Mt. Hopkins):

Last night I saw several k band background increases that dropped after the dome moved. I looked at the dome and it didn't quite look centered. We initialized the dome "home" and I don't believe that there were any more k increases all night.

This morning I found an error in the telescope control system that explains these observations. I found that if the operator uses the paddle to move the dome while the dome is under automatic (tracking) control, the dome loses its reference by about the amount it moved.

The operator never uses the dome motion buttons so I believe that the K background jumps will probably be associated with my being on the mountain (either Hopkins or CTIO)!

I have contacted Dave Harvey (Comsoft) and we will prepare a fix for this problem.