Diffuse illumination at close quarters
The BTTF is largely used at
small plate spacings in order to achieve low resolving powers. At such
small spacings, their are reflectance phase effects which must be calibrated
carefully at some stage. One particular concern is the need to bring the
calibration lamps closer to the focal plane since they are so very faint.
At one stage, we placed the lamps on the guide probe in order to get enough
signal. With S J Maddox, one could clearly see changes in the
long-slit spectra as the guide probe was moved. This is due to uneven illumination
by a structured source. We obtained 80-shuffle images on 18 Sep 1999. The
deuterium lamp at the top of the chimney shows no such effect since the
light is properly (?) diffused down the chimney from a much larger
Charge transfer inefficiencies at low light levels
There is always the concern that
shuffling at low light levels will lead to serious charge transfer losses.
In the red bands, the sky is sufficiently bright to provide a "fat zero" or
pedestal. JBH was originally concerned that this would be a problem
in the U and B bands. However, tests with P J Francis on 11 Sep
1999 seemed to indicate that there was no such problem. In other words,
shuffling 40 times over 40 minutes gave the same sky background as a single
exposure in the same amount of time. This requires more work.
Gain and time constant
A flaw of the new TAURUS control system is that, while it provides the experimentalist with complete control over the hardware, in a curious way, it detaches the experimentalist from the hardware as the control panel is not located in the Cass cage. There are several immediate consequences.
First, we used to set most control parameters through analogue switches, e.g. the gainand time constant. In so doing, it was obvious what impact these parameters had on the action of the plates. For both TTFs, we run the gain low and time constant high. This was in order to avoid overload problems (see below). With the push to faster and faster TTF calibrations and switching techniques, one must be wary now that the plate settle time is accomodated for in the charge shuffling. It used to be that the shutter time imposed the strongest constraint, but now that we are able to keep the shutter open while dragging the charge, the plate settle time is expected to dominate in most cases. A quick experiment demonstrates the hazard involved. Perform a shuffle arc with increasing "z" values, and then try decreasing "z" values over the same range. Note the phase shift:
This got me into trouble
in the very first TTF science run in 1996. That's why sausage cubes are
always to be relied on because the shutter is used for each of the exposures,
and there is an additional delay of 1 sec while the small image is being
no time to finish this now.
We can avoid these altogether with the protection switch off, but the consequences of this have not been looked at carefully.
The overload problem appears to be due to the large range that we typically scan through. The error signals will change dramatically for such a long rapid scan. The CS100 may be detecting a rapid change in resistive component as an overload signal and jumping out of lock.
There are several things to try:
With the new broad-narrow shuffle, jumping from 4um to 12um may require slightly different sets of (X,Y,Z) settings, in addition to different quadrature and resistive offset settings. New procedures are in place for producing very fast shuffle arcs at many different plate spacings in order to measure the multi-parameter space trajectory.
- ramp through the range in discrete steps
- reduce the time constant of the system and make it a little less responsive
- switch off the protection circuitry which cause lock to be lost if an overload or oscillation is detected; scan up from small to high gaps
- simply change the front panel dc gap value and ensure that the resistive component is well balanced at both extremes of the scan.
Minimum gap (site engineers only)
When you power up for the first time you will balance the system in the usual way. Since you have long range piezos you can then set Z_c = -3 or higher before closing the loop.This will ensure that the servo takes the gap to a larger value when the loops are closed. You can then scan down in gap until the minute, invisible dust particles or coating sputter marks (or whatever is on or in the coating) start to deform the surface locally and show up as Newton ring-like patterns. Do not go below these gaps. The minimum optical gap that I could safely set to was 2.5um.
You can use the front
panel controls to set the mean gap for the digital range as all these signals
are added together.