OTA Logic

Each cell of an OTA is equipped with logic and pass transistors which can be set to three states: active, standby, and floating. In active mode, a cell's parallel gates are connected to dynamic parallel clocking signals provided on bond pads, and the cell's video output can be connected to an output bus common to all cells in that column. In standby mode a cell's parallel gates are connected to static parallel clock levels appropriate to hold the charge quiescent in each pixel, and the cell's video output is isolated from the column bus. In floating or disconnect mode, the cell's parallel gates and video output are isolated from the external world. Once a cell's logic is set it stays in that state until reset.

We have examples of OTAs in which a cell or two have massive defects spewing so much charge that they badly degrade their immediate neighbors and substantially degrade half of the device. A monolithic device with this problem could not be used in a focal plane. However, setting a bad cell to "disconnect" completely removes its noxious influence on its neighbors. It is unavailable for science, of course, but the remaining 98.4% of the OTA is uncontaminated.

We provide 8 row and 8 column select lines so any rectangular block of cells can have their state set simultaneously. To read out, therefore, all cells are set to standby, and then a row of cells are set to active. Application of the usual parallel and serial clocking signals brings the charge to the output amplifier and 8 channels worth of video appear. The output MOSFETs have a second stage follower on the cell prior to the pass transistor, and the column bus output is further buffered by the discrete JFET mounted on the package.

If a cell were designated for rapid readout of a subarray around a guide star, it would be the sole cell set to active and its readout would not affect any other cells. Distinct orthogonal transfer shifts can be applied to each cell by sequentially setting them to active, applying parallel clocks, and then returning them to standby. It should not take more than 10 ^sec for the logic to change state so we can ripple through the 64 cells quite rapidly.

It is an obvious concern that we might encounter cross talk problems between different columns in an OTA. The columns share signals such as reset drain, although each video column has its own drain supply, current source voltage, and return. We do see crosstalk between columns, sometimes positive and sometimes negative, at the 10-4 level. This is negligible in comparison, for example, with the cross-talk between the two channels of an SDSU video board which appears at 3* 10-4.

4.2 Orthogonal Transfer

Orthogonal transfer pixels are symmetric under 90 degree rotations, and use gates to define both pixel dimensions rather than a channel stop implant. This permits parallel clocking to occur horizontally as well as vertically, which provides a fast, noiseless way to shift the charge of a developing image to follow the motion of the optical light producing it. This is much more effective than telescope guiding because it is much faster and accurate, and can be done differentially over a focal plane if the focal plane is divided into individually addressable cells. These pixels are discussed at length by Tonry, et al. [5,6] and Burke, et al. [7].

The choice of 2.6 arcmin for the OTA cell was driven partly by the desire for a large fill factor and rapid readout, but also by the fact that the isokinetic angle over which atmospheric image motion is coherent is generally somewhat larger than that on Mauna Kea.

Detailed experiments with CCID28 orthogonal transfer devices reveal that photometry is not impaired at all compared to 3-phase pixels, and astrometry is actually somewhat better, because astrometric residuals with standard 3-phase CCDs appear to be worse in the direction across the channel stops. Other advantages are that OT pixels do not bloom up a column when they receive too much charge, rather the charge puddles in a circular blot.

One might worry about parallel CTE effects causing a blurring which offsets the advantage of OT shifting to remove image motion, but quantitatively it is not a problem. At a 10 Hz rate of shifting (fast enough to remove virtually all telescope shake and almost all of power spectrum of atmospheric image motion) over a 100 sec exposure, we are only performing about 103 shifts. A typical CTI of 10-5 means that we are therefore taking 1% of the charge of a given pixel and putting it into a skirt of adjacent pixels. This is completely negligible blurring compared to a PSF which is well sampled.

4.3 Charge Transfer Efficiency

An operational challenge in testing OTAs is that the cells are so small (600 pixels) that it is very hard to see any CTE degradation from a simple readout. Even Fe55 xray events will only lose about 0.5% of their charge from single pixel events at the top of a cell compared to the bottom. We therefore resort to using the OT property to exaggerate the effects of CTE so that we can properly quantify it.

The first step is to take a picture of a scene. In this case it is a picture of text on the LCD screen which the OTB images onto the detectors. The next step is to take an identical image, but then perform OT shifting before reading it out. In this case we shift it around a 32x32 pixel square 100 times, for a total of 12,800 pixel shifts.

Subtracting the two images is then a very sensitive test of how well the charge has been transferred. The difference image in Fig. 2 has been stretched by 20 to show the features better. A hot pixel shows the 32x32 shift box. Charge that is shifted off the edge of the CCD is simply lost to the scupper. Charge transfer inefficiency appears as slight shifts of the characters creating small positive-negative signatures. Analysis of this suggests that the typical CTI is about 2x10-5. Another way to quantify this is to ask how much worse the noise in this shifted pair of (bright) images is compared to a difference of unshifted images, and we typically see a noise which is 1.2 or 1.3 times worse than Poisson. Since this is approximately ~50 times more shifts than we would expect to perform during a 30 sec exposure, we regard this CTE performance as very good.

Figure 2. The difference between two LCD screen images shows very low level residuals.

The pixelization evident is the result of the LCD screen pixels - each LCD screen pixel maps to about 13 CCD pixels.

Figure 2. The difference between two LCD screen images shows very low level residuals.

The pixelization evident is the result of the LCD screen pixels - each LCD screen pixel maps to about 13 CCD pixels.

Was this article helpful?

0 0
Telescopes Mastery

Telescopes Mastery

Through this ebook, you are going to learn what you will need to know all about the telescopes that can provide a fun and rewarding hobby for you and your family!

Get My Free Ebook


Post a comment