The first test was run using the UV-chrome window gun. The front face of the TED was about 2 inches from the window. Direction of the magnetic field relative to the detector was not measured. The TED skin was grounded. The chrome was biased at low negative voltages. The gse was programed to sweep the analyzer voltage over a 5V range in 256 steps while taking count readings on each step. A dc supply was used to offset the sweep center voltage.
The count rates were very low (less than 400Hz for 8mA lamp intensity) but with long accumulation times the data shows reasonably believable energy peaks down to 1eV. Runs were made with the chrome at 1V, 2V and 3V. The peaks in count rate were seen at approximately the chrome potential assuming the predicted analyzer selection factor of 1.26.
With 8V acceleration on the window, the peak is seen at 7.6eV. This peak appears lower in energy and is wider that the predicted response (to a mono-energetic isotropic flux). It also shows significant tails extending well away from the peak. There is no independent calibration the output energy spectrum of the source but some basic assumptions seem reasonable for this simple setup:
Any electron originating at the chrome window must fall through this acceleration potential to reach the detector. There is no possible electron source between the chrome and the detector.
Electrons are produced at the chrome by the photoelectric effect so they have a maximum energy that is the difference between the work function of chrome and the photon energy. The UV source is a mercury vapor lamp. It isn't monochromatic but it has a line spectra with some maximum frequency UV line. The maximum energy of an electron leaving the chrome is .4eV.
A source peak width of .4eV means the expected width will be wider than the model peak. The exact shape of the expected peak hasn't been calculated. The full width of the data peak is 1.07eV. If you subtract the .4eV source width you get .67eV. This is somewhat close to the .60eV of the model.
The biggest concern is the extended tails seen in the detector response. With the the chrome at 8V there are no 4eV source electrons. The data shows about 5% of the peak rate counts are seen at 4eV selection. If the assumptions about the source are correct, these tails must be due to detector response. Even for these weak 1KHz peaks this is about 50Hz and is 10 times higher than the background rate of 5Hz seen with the UV lamp on and chrome window at 0V. With the lamp off the dark rate is about 1Hz.
Some data was taken with the detector moved closer to the chrome window (about 0.15 inches from chrome to detector face vs 2.0 inches for previous runs). At 8V the flux increased by about a factor of 3 at the same lamp current. The count rate normalized peak from this run (green trace) is at the same selection voltage but is narrower and has lower tails. At this distance the angular spread of the source is probably narrow so the peak should in fact be narrower than the model for isotropic flux.
The tails are still significant but the decrease may indicate that the problem is related to particles coming in at large angles.
The setup was changed to try a filament electron gun. With this setup the filament (inside the gun assembly) is 3 inches or more away from the face. The gun assembly collimates the beam so the angular spread at the detector is probably small. A vague attempt to align the beam with an assumed magnetic field direction but no B field measurements were made.
The filament gun can produce much higher fluxes. An acceleration voltage of 6.5V gives a peak at the same selection voltage as the 8V peak seen with the UV gun. This gun showed output about 1eV higher than its acceleration voltage during the dust detector calibrations. It seems likely that at least 1V out of the 1.5eV difference is due to contact potential in this gun.
This gun produces similar looking peaks to the UV gun (offset by about 1eV). The off-peak tails are also similar. The high flux available allows looking at them in more detail. The tails have features that are reproducible and may give some information about what is causing them.
This plot shows data from three different gun energies normalized in both count rate and energy. Count rate values are divided by peak counts. Energy coordinates are divided by peak energy.
Off peak counts are a larger percentage at lower beam energies.
More leakage is seen on the high side than the low side.
There is a broad peak at about twice the beam energy in the high side leakage.
There is a change in slope at about half the beam energy the low side leakage.
Some of the differences at different beam energies may be due to the source energy spectrum. The output of the filament is expected to have about a .5eV high energy tail. At 1.4eV this is significant compared to the width of the analyzer.
The off-peak counts seem to scale with peak count rate. Three runs were done at different filament currents giving peak count rates differing by an order of magnitude. The off-peak counts are slightly higher percentage at high count rates. Some of the differences may be due to MCP gain change or errors in deadtime correction causing some loss of counts at the peak for high count rates.
One explanation for the off-peak counts is some sort of scattering inside the detector. The detector model assumes that electrons that hit a conductor will be stopped. If a significant percentage of them can scatter off a surface they may eventually find their way through the exit slit. The model was used to examine bounce trajectories
Runs were made with and without the brass wall. The high side off-peak counts were reduced by about a factor of 3 with the 2mm wall in place. (This plot is for a 3eV beam and is normalized to peak counts and peak energy).
The 2mm wall isn't a solution for flight since at lower energies it would interfere with the main peak. This test does indicate that scattering off the bottom plate is contributing to the off-peak counts.
A new piece with serrations was machined for the bottom plate. Runs made with the serrated bottom plate show a reduction in high side off-peak counts by about a factor of 1.4.
A higher wall is more effective at trapping bounce paths but it must be moved farther away from the exit slit to keep from interfering with the main peak. This means it should reduce bounce paths for high ratios but won't help near the peak energy. A wall 3mm high 15mm from the front face was tried. This high wall was very effective at ratios of 1.7 or more.
Attempts at modeling bounce paths show that a high percentage of the possible paths that can reach the exit slit won't enter at a steep enough angle to make it through the second slit. A screen was added to the first slit to see if small leakage fields near the slit were making it easier for bounce paths to make it through. The runs with and without the screen show no significant difference.