DARKS HISTORY

Bill Wheaton
13 July 2000

Introduction

In discussions with Roc & Mike on 11 July, it became clear that the operational complexities of the pixel threshold implementation are heavily dependent on variability of the camera offsets O, as defined in a previous document (Ref 1). If T is the saturation threshold for a pixel, which the RDFRAME sub-system uses to set the pixel saturation flag, we expect the saturation level T', measured above the offset O (or DARK1 or DARK2, which are all approximately equal; see below):

T = T' + O

to be more stable with time. Thus we expect to compute T' infrequently with something closely related to the current analysis software, and then compare READ2 by

R2 < T' + O,

assuming a suitably stable proxy for O can be found.

The offset O can be calculated for every READ2, READ1 frame pair, but a sufficient approximation for each night is the mean dark offset, <O>:

<O>	=	D1 - _Dt₁
	=	D1 - D

where D1 is the pipeline DARK1, D is the pipeline DARK = (DARK2 - DARK1), and

= t₁/(t₂ - t₁),

and

_D = (D2 - D1)/(t₂-t₁),

as discussed in ref 1. Since is small (typically about 0.04) and D is also small (a few hundred), the matter boils down to the variability of DARK1, since the value of DARK1 is large, typically of the order of 10,000.

Here are shown plots of the median DARK1 for each available night.

Data

These plots have been made by averaging the medians of the even and odd columns of the arrays, taken separately, to reduce variability due to the odd/even structure of the dark frames. (Since the even and odd medians are separately well-defined, I hope that the average will be stable also.) Differences between the odd and even medians are on the order of 1000-2000. Abscissas are survey day number.

The J cameras (blue) are plotted as "+", H (green) as diamonds, and K (blue) as "X".

North Cameras:

A PostScript version is here.

All three cameras show sudden jumps, H of up to 2500 DN, with intervals of stability between, during which variation is confined within a range of 500 or less. Most of the major changes are well correlated with known survey events, but a few, notably in the H camera near days 650 and 950, are not.

South Cameras:

A PostScript version is here.

Again as in the north, most changes can be identified with known survey events, especially the switch to Leach electronics at survey day 729. The most visible exception is an upward step of 200 DN in the J camera between survey day 507 and 513.

Remarks

Basic conclusion is that the threshold arrays will need to be changed about three or four times in the north, depending on the camera, and two or three times in the south. In particular, it is clearly not necessary to re-compute them every night or change them every week.

Note that the stability of T', hypothesized above, has yet to be fully verified.

Last modified: 18 January 2001, by WAW.