The new imaging camera for MRO has been completed and will be installed on the telescope in May 2003. The new chip is a thinned, backside illuminated chip with 13 µm pixels in a 1056 x 1027 format. Unlike our earlier front-side illuminated Ford and Loral imaging chips, this chip is not an MPP device. It will therefore have slightly higher dark noise from our earlier chips, but at the temperatures that we run at you will never be able to see this. At the image scale of the focal reducing optics on the telescope, the 13 µm pixels of the new imager correspond to 0.507 arcseconds/pixel and provide a total field of view of 8.92 x 8.68 arcminutes.
The new chip is cosmetically perfect with very good QE. Figure 1 shows a flat field taken with the new chip which is essentially featureless. Close inspection will reveal approximately 5 points of low QE. These points are only a few percent low in their response compared to surrounding pixels. Figure 2 shows the first lab test image taken with this device which is, appropriately enough, an image of a photograph of the galaxies M51 and M104 which was taped to the ceiling above the camera dewar. Figures 3, 4, 5, 6, and 7 show the first astronomical images taken with the camera at MRO. These figures show M27 (The "Dumbell Nebula") through U, B, V, R, and I filters, respectively. Figure 8 shows a color composite of the Dumbell Nebula constructed from the U, V, and R images (Thanks go to Don Brownlee for putting this together!).
Technical information about the camera is given below in the following figures.
The factory test sheet gives the QE curve shown in Figure 9. The peak QE of this chip is quite a bit higher than what we have been used to (the old chips have about 50% peak QE). The blue response (below 4000 A) of this new chip is also much better than what we were getting with the old detectors.
Measurements of the gain, read-noise, and linearity of the imager as a function of the "cgain" software setting in the CC200 the controller are given in Figures 10, 11, and 12, respectively. These measurements were taken while the camera was unbinned (1x1 binning). In these figures the polynomial fits to the data are shown along with the data error bars. The numbers in the tables are the parameters to the polynomial fits. The error bars for the 1x1 binned gain data are smaller than the circles used to plot the data points. No error bars are shown for the linearity plot of Figure 12, but you can estimate the errors for each point from the deviation of the points from the linear fit below 10,000 ADU. The device appears to be perfectly linear (to within 2.4%) up to 114,000 e-.
The read-noise in this new Marconi chip is quite impressive with a value of only 2.8 electrons at the higher cgain settings. You should be aware that the price that you pay for higher cgain settings is slower image acquisition, since the camera takes longer to read out at these settings. In Figure 11, note that the read-noise near a gain setting of 20 is at least as small as it is at the higher cgain settings. This is a real effect, as indicated by the plotted error bars, and if one wishes to have both high dynamic range and low read-noise, a setting of cgain=20 is a very good choice for 1x1 binned imagery. At cgain=20, Figure 10 shows the system gain to be 7.18 e-/ADU so that the ADU limit of 16384 corresponds to 117,637 e-. As mentioned previously, the device is quite linear up to these data levels. Just a little beyond this signal level (~a few thousand more electrons) the chip begins to go non-linear.
Users will find that the default cgain setting is 90. At this setting you have a system gain that is low enough for you to get a good sample of the read-noise. This is what you will want for imaging the faintest objects possible. As discussed earlier, for high dynamic range, a cgain setting of 20 is optimal for 1x1 binning.
Under normal seeing conditions at MRO, most users will want to bin the detector 2x2. The gain, read-noise, and linearity of the imager under these conditions are given in Figures 13, 14, and 15, respectively. By comparison with earlier figures you can see that the gain does change slightly between the two binning modes. Also, the optimal cgain setting for high dynamic range changes from 20 (for 1x1 binning) to 15 (for 2x2 binning). The lowest read-noise is still available at a high cgain setting of 90 and is still a very impressive 2.8 electrons. In 2x2 binning mode the chip is quite linear clear up to the ADU limit of 16384 with a gain setting of 15. At this setting the system gain is 7.5 e-/ADU and the ADU limit corresponds to 122,896 e-. This is beyond the limit at which the chip is linear when using 1x1 binning. So, by using 2x2 binning you also gain in dynamic range. There is no gain setting for which the chip is non-linear with 2x2 binning. Figure 16 shows a linearity curve for a gain setting of 0 while binning 2x2. The ADU limit in this plot corresponds to a signal level of 148,964 e- and it is clear that the chip is quite linear even at this setting. The system gain at this setting is 11.13 e-/ADU.
So, in summary, if you compare our new Marconi chip with our earlier Ford and Loral chips you will discover that we have made very impressive improvements in the QE (the new detector is over 40% higher in QE than the old one and has much better blue response) and read-noise (2.8 e- vs. about 8!) at the expense of a little field of view (our FOV went down by only 12%).
This new camera represents a very significant improvement for the observatory. Once again, we would like to thank the many people whose private contributions to MRO have made this work possible. Among these we would specifically like to thank Jim Naiden, George Best, Larry Shea, and Gerald West of the Seattle Astronomical Society. In addition, within the Astronomy Department here at UW we would like to thank George Wallerstein, Julie Lutz, Karl-Heinz Bohm and Erica Bohm-Vitense for their generous contributions. Without your generosity and encouragement MRO would have long ago become a closed institution!
