RIXS-CCD Acceptance Testing

Initial tests to verify the 100nm positional stability requirement were carried out with the help of Georg Gassner.

STA provided a sample vibration spectrum measured with their confocal setup in the course of a flatness measurement:

This plot shows a peak at 30Hz near 60nm, though the units are not included on the axis.

Using Georg's system we measured this spectrum in the z (beam-parallel) dimension:

Similar measurements were made on SN31179 in three dimensions, results summarized here:

These measurements are represent differential movement of the sensor mount relative to the chamber.  "Background' measurements comparing adjacent spots on the chamber were of order 20-30nm rms.

Though the cross-dispersion result is larger than the 100nm requirement, the result is less than 1% of one pixel width in that direction and should not impact spectral resolution since it's in cross dispersion.

Note that significant settling time (>1h) is needed before the measured stability is achieved.  The active vibration control system measures and eventually cancels 10 harmonics of the cooler vibration and takes a fair amount of time to accomplish it.  Initial stability numbers can be x10 greater than the numbers quoted.  Please note that a 1um rms initial stability is one-tenth the minimum physical pixel dimension and the system should be quite usable from startup, except for experiments that demand the very highest interpolated spectral resolution.

To achieve these numbers, the cooler is set for constant-power cooling at its minimum level of 60.  To check the power, type status in the serial interface.  To set the power, type pwout=60.  Constant power mode is established with cooler=power.  In this mode the cold head settles at about 90K with the sensor set for 175K.

Typing cooler=on engages the cooler's own PID control loop.  Since we rely on the sensor's PID loop for temp control, having two operating simultaneously might lead to instability and is best avoided.  If the cold head's PID mode is needed, settings to control it are in the CryoTel manual referenced above. 

To verify STA's quoted read noise, SN31179 was operated in free-run, 2-d mode at different temperatures, spanning  173k to 233k.  In this test, read noise is approximated by the standard deviation along a single column, with the result averaged across all 4096 columns.  This estimate would include any contributions from non-uniformity along a column, but the result come out very close the STA's quoted noise so it seems a decent proxy for read noise.  The array average pixel dark signal (pedestal, averaged along a column and then averaged across all columns) increased nearly linearly, suggesting that thermal leakage is not dominant.  Over the same range the calculated read noise increased modestly, from 4.1e to 5e.

2d Read Noise vs. Temperature                                                                                1d Read Noise vs. Frame Rate

Switching to 1d mode and using the external trigger feature to set the frame rate, a sweep of read noise vs 1d frame rate was obtained.  In this measurement the standard deviation along a column gives the temporal variation in a summed column, since in 1d mode all columns are summed to a single value.  The measured pedestal decreases as rate increases, as might be expected if thermal leakage is a significant factor, but the 2d tests suggest that it is not.  The noise also decreases significantly as rate increases.  It's roughly equal to the 2d noise at the slowest measured rate of 10Hz, but drops by a factor of two by the time the frame rate reaches 50Hz.  This requires further study, as it's unlikely the read noise is as good as 2.5 e.