MIT/Lincoln Labs CCID20 W6C2 CCD Test Results

This is a standard epi, thinned, backside MIT/Lincoln Labs 2Kx4K CCD. It is very similar to other standard epi CCID20s tested at Lick. Previously this device had a thermal problem. When the CCD was cooled to normal operating temperature only the first 2048 rows could be read out. Parallel clock phase 2 disconnected from the second half of the array. The device was returned to MIT/Lincoln where they were able to identify the problem (a break in several gold traces) and were able to repair it. The device has been temperature cycled several times at Lincoln and here at Lick. We have taken the device all the way down to -136°C. We believe the thermal problem in this device is permanently fixed.

A summary report (for the original working half which is virtually identical to the second half of the CCD) is available. Noteworthy points about this CCD include:

CTE. Near perfect CTE is achieved using the serial and parallel clocks which generate low spurious charge. This is typical of Lincoln CCDs.
Read Noise. Read noise is around 2 electrons. This is as good as or better than any thin CCD in the world.
Brick Wall. Images over a wide temperature range were made to test the sensitivity of the QE variations with temperature. A plot of the data is given below.
Cosmetics. There are no blocked or hot columns. Cosmetically this device is about as close to perfection as is possible.

Original postscript files are available from our anonymous ftp server and these provide better resolution and clarity than is usually possible on a web page. There are also additional postscript files showing CTE measurements, fringing, and the brick-wall pattern. Check the INDEX file for a description of what the other files contain. Here are a few figures to illustrate device highlights:

QE curve: This is a pretty typical QE curve for the CCID20s. The QE is an average over a fairly large portion of the CCD, so this result is an average over the brickwall variations.
Serial CTE: Near perfect CTE is achieved by the Lincoln CCD design.
Surface plot: This is one of the flattest of the CCID20 CCDs we've measured.
Brick wall: This is the pattern which results from the incomplete backside laser anneal. The change in sensitivity was measured from -40°C to -110°C.
Dark Current v. Temperature: We allowed our dewar to run out of liquid nitrogen and then we measured the dark current as the CCD was warming up.
Long Dark: This 1000s dark at -136°C was obtained after the CCD was repaired. The brick wall pattern shows up faintly in the dark. The raw data file is available via the ftp server.
Fringing: This 9000Å fringing pattern is seen in many CCID20s.

This plot shows the QE measurements made at two device temperatures. Little change occurs in the blue, although we know that the least sensitive areas in the brick wall pattern improve with increased temperature. (We think the -80°C point at 7000Å has a large measurement error for some reason.)

Graph of QE.

The Lincoln CCID20 design produces excellent charge transfer efficiency. This plot shows a test of serial CTE using Fe⁵⁵ xrays. Image showing serail charge transfer efficiency.

Image showing serail charge transfer efficiency.

The surface contour plot shown here is one of the flattest we've measured.

Surface plot showing P-P variation of a few microns.

The data for this curve was obtained by taking a flat field image at each of the indicated temperatures and measuring the amplitude of the quantum efficiency variations. The percents shown are derived by taking a single row cut through the image and computing a percentage as (MAX-MIN)/MEAN. Obviously this emphasizes maximum variations and different results will be obtained if a different row is selected. But the general trend is clear: To reduce the amplitude of the brick wall QE variations, operate the CCD at the highest practical temperature.

Curve showing increasing qe variations with lower temperature.

The dark current was found to rise rather rapidly, doubling every 3.2 degrees. This rapid rate of increase is probably due to surface states since the CCD can not be operated in MPP mode. Note that we typically run the parallel clocks down to only -6v, so we are not even running in partial-inversion mode.

Plot show dark current v. temperature.

This 9000Å fringe pattern is seen in many CCID20s. There is a circular pattern which is apparently centered on the wafer and a more random pattern on top of that. The circular pattern is only a fraction of a percent in amplitude and in this CCD the random pattern has about a 10% amplitude, which is quite reasonable. This image, showing the central section of the CCD where the illumination was most uniform, was obtained after the CCD was repaired.

Fringe pattern at 9000Å.

This is a 1000s dark at a temperature of about -136°C. This image was obtained after the thermal problem was repaired. The CCD was stable at this temperature for many hours. The very faint pattern is not similar to the epoxy bond pattern seen in W6C1. Instead it seems to be another result of the boron-implant/laser anneal problem that gives rise to the brick-wall pattern. This is not a surprising result. The amplitude of the pattern is less than 1 electron -aren't low-noise CCDs wonderful! The average dark level is about 2.6 electrons. This value does not have the "dark" in a 1s dark subtracted off, which would represent any spurious charge, so the actual thermal dark rate may be a bit lower.

1000s dark showing a faint pattern.

Back to Lincoln Labs Top Page