Author: Tom Droege Date: 960104 Revision: #0 960104 Key Words: instrumentation, CCD
Tests of the Mark III indicate a noise level near the expected value from the data sheet. Dark current measurements track the published curve.
Since it has been snowing and I have stayed home, I have been running some tests on the Mark III. I patiently wait for clear sky.
Since it takes a long time, 469 seconds, to read out a complete drift scan we have to run long term tests and project back to get the actual dark current.
The scheme is to first clear the CCD of all the charge. This is not so easy. If we just run a bunch of vertical shift pulses, this saturates the horizontal (output) charge shift register. The scheme used here is to first send 512 vertical shifts, followed by a few times 768 horizontal shifts. Sending a string of vertical shifts always saturates the horizontal register for some reason, but after a few of these sequences it quickly clears.
After the last such sequence the time is recorded. Then 100 vertical lines are read out, and one particular pixel is measured and recorded for each line. The mean and RMS value of this pixel is computed.
Below, Run (1) is a sample data run:
(1) Seconds TEC CCD Water Mean RMS
After Current Temp C Temp C e- e-
Reset Amps
107 0.5 -1.8 18.5 1282 52
200 0.5 -1.7 18.5 1338 51
294 0.5 -1.7 18.5 1379 52
398 0.5 -1.7 18.5 1392 52
487 0.5 -1.7 18.5 1430 50
576 0.5 -1.7 18.5 1439 45
Plotting these values and projecting back to time zero indicates a drift of 189 electrons in 469 seconds. This is 0.4 e- per second and is very close to the curve given in the KAF-0400 data sheet.
Run (2) is made at a different temperature:
(2) Seconds TEC CCD Water Mean RMS
After Current Temp C Temp C e- e-
Reset Amps
110 1.0 -11.4 19.3 291 24
486 1.0 -11.8 18.9 313 24
768 1.0 -11.9 18.9 311 30
Some data points omitted from table (2) to save typing. Result here is 0.05 e\ per second, matches the Kodak data sheet, and is consistent with a doubling of dark current for each 5 degree C temperature increase.
Some points point at a lower temperature (3):
(3) Seconds TEC CCD Water Mean RMS
After Current Temp C Temp C e- e-
Reset Amps
120 2.0 -17.9 24.1 56 20
308 2.0 -18.5 23.7 60 21
590 2.0 -18.7 23.5 60 22
A curious thing can be noticed. The measured mean value increases with temperature. For the steady state case, more than 469 seconds after a reset, we expect an increase due to the accumulated dark current. If case (3) is near zero dark current, then case (2) should be larger by 0.05*469 = 23 electrons. Case (1) should be larger by 0.4*469 = 187 electrons. In each case the mean output is much larger than would be expected, and further the RMS values have increased beyond that expected from the number of dark current electrons. Something else is going on.
To investigate we took the data (4):
(4) Seconds TEC CCD Water Mean RMS
After Current Temp C Temp C e- e-
Reset Amps
Any 2.0 -19.5 22.1 35930 2 Chan. 1
Any 2.0 -18.9 22.9 35930 2 Chan. 3
The Mark III is a three channel camera. The electronics is designed to read out three CCD chips separated by 15 degrees in RA. Each channel has a 2.2K resistor to ground to load the output amplifier of the CCD. This is followed by an emitter follower and capacitor coupling to a double correlated sample circuit for each channel. One software parameter switches the channel address with everything else remaining the same.
By switching to a channel that does not have a CCD connected, we can compare the performance of everything outside of the CCD chip, keeping everything else the same. We note that the noise level is much lower when we are not connected to the CCD. Experience with 9 billion channel hours indicates that the circuits following the CCD are well understood and are not likely to be doing anything funny.
Note that the difference in mean value between the CCD connected and the dummy channels is well understood. The CCD channel has been offset from zero to make use of the full range of the ADC. The dummy channels are looking at zero volts, but have an offset added in the software.
First the measured noise level. The 22 e- measured in (3) is encouraging but it is at the high end of the Kodak specification which indicates 13 e- as the nominal value and 20 as the maximum value. Note that the noise level computation depends on the output amplifier scale factor of 10 uv per electron. This and a full well capacity of 85,000 e- would indicate a full scale signal of 0.85 volts. This unit has a full scale signal closer to two volts. Either the well capacity is much larger than 85K electrons wich is not considered to be probable, or the amplifier scale factor is more than 10 uv per electron. If it is more, then the noise is less than that indicated. This is considered to be the most probable answer. If the well capacity is near 85K electrons, then the noise level is near 10 e-.
It is noted that the measured mean value increases with temperature. This is curious. This means that there is a larger signal left after the Reset Clock when the temperature is higher. A very long, 4 us, Reset Clock is used. The spec sheet lists a minimum value of 10 ns. It would be expected that the 4 us reset would surely drain all the charge to a stable value.
It is as if there is a large positive temperature coefficient resistance in series with the output. This then accounts for the larger offset and the higher noise level with temperature. Comments by experts would be appreciated.