New CCD A-to-D Converters: More Dynamic Range and Faster Readout

The Kitt Peak CCD controllers have had 15-bit A-to-D converters for many years, yielding digital saturation at 32,767. This digital saturation requires some compromises between using all of the linearity range of the CCD and the sampling of the readout noise. For example, with T2KB, 0.1% linearity deviations occur around 220,000 , and the readout noise is 4 . Currently, if one wanted to use as much of the linearity range as possible, one would select a gain of 5.4 /ADU (gain index 3). However, this samples the readout noise poorly. Consequently, if readout noise was a key concern, one would choose a finer gain, but use less of the CCD's linearity range. Compromises of this sort are common. With a higher digitization limit, one would be able to achieve a better compromise, yielding a larger dynamic range with a better sampling of the readout noise.

This summer, we are installing new 16-bit A-to-D converters in all the Kitt Peak CCD controllers. (This does not include the controllers at WIYN, which are based on the ArCon technology and already use 16- bit A-to-Ds.) This conversion will be complete by the start of the fall semester. The 16-bit A-to-Ds double the digitization limit, allowing either twice the linearity range in electrons at constant sampling or two times better sampling at constant range. Users should consider the effects of the new digitization limit in selecting the best gain for their program. An added benefit, from the faster speed of the new converters, will be faster CCD readout. We conservatively estimate that 16 seconds will be trimmed from every 2048 x 2048 readout. From analysis of the Save the Bits archive, one of us estimates that the faster readout time will save 10 days per year of readout time!

In a previous Newsletter article, we described the effects of saturated stars on the CCD electronics and the resulting data. This overshooting phenomenon yields low-level trails to the right of saturated stars (see NOAO Newsletter No. 44, p. 33). Consequently, the video circuitry in the CCD electronics was adjusted to minimize the effects of overshooting for each of our direct imaging CCDs for a particular gain value (i.e., number of electrons/ADU). This optimal gain value for each CCD (the ICE default), was chosen based on the digitization limit of 32,767, the linearity limit of the CCD, and the readout noise. Since we are changing the limit, it makes sense to modify our overshooting-optimized gain values. The table lists the new optimal gains for each CCD used for direct imaging.

           Optimized Gain   electrons   0.1%linearity   1%linearity 
CCD       (electron/ADU)     @65KADU     (electrons)     (electrons)

S2KA            2.5          162,500        145,000       160,000
T2KA            3.2          208,000        180,000       190,000
T2KB            3.2          208,000        220,000       250,000
T1KA            2.7          175,500        250,000       250,000

We recommend that all users concerned with the effects of saturated stars on their imaging data use the optimal gains (which minimize the effects of overshooting; the overshooting will be significantly worse for other gain values). The ICE software has been updated to reflect these newly optimized gains. The default gains in ICE for these CCDs are the optimized gains. After an obsinit, even if one does not edit the "detpars" parameter set, one will by default be using the optimized gain.

We are updating the Direct Imaging Manual and the ICE manual to reflect the changes described above. Finally, we note that there is some effect on IRAF taping since the data are now 16-bit unsigned integers. If one simply uses the default wfits parameters, one will be safe.

Taft Armandroff, Rich Reed, Rob Seaman, Phil Massey