Read out and timing of the UVVIS Camera.

1/25/95 - Travis White and Darron Nielsen, Lawrence Livermore National Lab.

Imagine that the focal plane array (FPA) of the UVVIS camera is oriented vertically with the rows running horizontally and the 384 columns vertically. The top half of the FPA (290 rows) is the photoactive region and the bottom half (290 rows) is the frame buffer. The first and last rows of the photoactive region are transitional rows that separate the photoactive region from the rest of the ccd. Thus, the ccd has only 288 rows of photoactive pixels. Below the bottom row of the frame buffer is a single readout row. Except during integration, the ccd is continually passing charge downward along a column from one pixel to the next. The process resembles a bucket brigade in which one bucket sits in front of each pixel and one bucket sits between each adjacent pair of pixels. All pixels simultaneously pick up the bucket of charge above them and empty it into the bucket in front of them. Next each simultaneously empties the bucket in front of it into the bucket below. Then the process repeats. Ideally, there is no cross-talk from one column to the next. So, in the absence of "sloshing", the charge that finally arrives in a given slot in the readout row originates strictly in the column above that readout location. After each time the readout row is filled, it is swept out serially.

There are 5 distinct "phases" that the camera passes through in order to get an image. They are:

1. Prior to request for integration.

The camera is in what you might call a "rolling line" transfer mode. Charge is transferred downward columnarly. Each time the readout row is emptied, which is once every 94.4 microseconds, every pixel shifts downward simultaneously by 7 rows. At this rate, about 4 ms are required to transport one row of pixels from the topmost row past the bottommost row of the photoactive region. (The time to shift a row downward is very short compared to the time to read the readout row.) The 7 rows at the bottom of the frame buffer are summed into the readout row as they are shifted downward. All other rows are shifted intact. If the camera is not actively being read out, the contents of the readout row are simply lost, but the readout row is still flushed. If the camera sits quietly for more than a few frames, this default operation of the ccd essentially sweeps away any previous latent image; however, photons still arrive on the photoactive portion of the FPA and dark current still seeps into all pixels (active, buffer, and readout), producing some charge in every ccd location at all times.

2. After a REQUEST for integration but before the START of integration.

Eight consecutive frame transfers are executed in the following fashion: The bucket brigade speeds up by a factor of about 100, transferring one frame of 290 rows in about 235 microseconds and then waiting about 237 microseconds before resuming charge transfer, bringing the time required to execute one fast frame transfer and begin another to approximately 472 (=5x94.4) microseconds. In this mode, the charge transfer downward along a column continues while the readout row is being emptied. The frame transfer proceeds asynchronously with respect to the emptying of the readout row, so the frame transfer does not end on a 94.4 microsecond time hack for emptying the readout row. The 237 microsecond delay is used to guarantee that the readout row is flushed at least 2 times after the frame transfer completes. No data are transferred from the camera to the SASI output lines during these frame transfers.

The intent of the fast frame transfers is to flush the ccd (to execute what Carle Pieters has called a "refresh"), but what occurs is not a true flushing operation. Photons still arrive on the active portion of the ccd, and dark current still seeps into all locations in the ccd during each frame transfer. Although short compared to the dwell time during integration, the dwell time per pixel during fast frame transfer is not 0.0000ms, and SOME photocharge and dark current accumulate.

IMPORTANT: Although the camera has this 8-frame "flush" capability, we must confirm that the flight software used it. The camera can also operate so that no explicit flush occurs prior to integration. (I believe that the flush was indeed used, but we just need to double check.)

3. During integration.

The bucket brigade stops promptly at the end of the 8th frame transfer mentioned above, right at the start of the subsequent 237 microsecond delay. As soon as the bucket brigade stops, photocarriers begin to accumulate in each cell in the active portion of the ccd, and dark charge accumulates in all pixels --- both in the photoactive region and out of it. With the fast- frame-transfer capability enabled, the minimum integration time is 331 microseconds. In general, integration time is 331 plus an integer multiple of 94.4 microseconds. Although the integration does not begin precisely on a 94.4 ms time hack, it does end precisely on such a time hack.

4. During frame transfer, right after integration ends.

The bucket brigade resumes, again in the fast mode. 290 rows (1 frame) are transferred from the image portion of the ccd to the frame buffer portion in about 472 microseconds: 235 microseconds for frame transfer and about 237 for flushing the readout row before the true readout begins. Light still illuminates the active portion of the FPA. Even though the frame transfer is fast, it is not instantaneous. Consequently, bright points in the FOV tend to produce streaks along columns. This is because the bucket brigade procedure forces all charge packets that eventually reach a given column of the readout row to pass through the entire corresponding column of the image at some time --- either during the frame transfer prior to readout or during the last frame transfer prior to the start of integration. Presumably, during the short time required for frame transfer, the scene that is imaged on the FPA changes insignificantly.

5. Readout.

After the image has been transferred to the frame buffer, it is read out one row at a time. Every 94.4 microsecond, the entire frame buffer region shifts downward one row, and the data from the readout row are transferred to the SASI bus. It takes about 27ms to read all 288 rows of valid pixels. During this time, dark current in the frame buffer can contribute a small amount to the total signal. The first row to be read spends the least amount of time in the frame buffer and acquires the least amount of extra dark charge. The last row out acquires the most. The processes in steps 2 and 4 require fixed amounts of time and add a constant term to the output, independent of integration time. Except for thermal drift in the camera, the added (pixel- dependent) constant should be the same from image to image and orbit to orbit.