July 26: Geoff Marcy alerts us to a change in the QE of the
middle chip of the HIRES mosaic. He supplies this
figure. showing large areas of depressed QE. The bright streak
is the well known scattered light feature known as the meteor and is
unrelated. Note however the wedge shaped area of lower signal
level extending upward from the lower left. Flats taken immediately
upon receipt of the alert confirm the problem, and estimates by eye of
the severity vary between 15 and 30 percent.
July 27: A first attempt at a quantitative estimate of the severity
is performed. Flat fields taken with the cross disperser cover
closed create a quasi uniform wash of light over the whole detector.
Such flats, taken in October 2004 were extracted from KOA and new ones
were taken with an identical setup and exposure times. After bias
subtracting each set of images, ratios were formed in the sense July/October.
These ratios are remarkably flat, with values near 1.0 for CCD's
1 and 3. The ratio for CCD 2 looks like this
though. In this plot, the blue and
black curves are cuts through the ratio images of CCD's 2 and 1 respectively.
They are taken through adjoining sections of those images and should
be the most similar. The green and red curves represent cuts
through the middle of CCD 2 and near the edge adjoining CCD 3. Based
on this plot, the changes in QE since last October seems to only be on the
order of a few percent.
July 28: Bob Kibrick notices that humidity in HIRES is higher
than normal. Mike Wagner investigates and finds that the dry air
purge valve is not open (reason unknown). He opens it and humidity
levels drop. An email discusssion generates a number of ideas and
tests to explain the almost order of magnitude difference in estimates of
severity.
July 29: A picture is taken looking into
the HIRES detector. The apparent features on CCD's 1 and 3 are artefacts
of the camera flash. Something is visible on or near the surface
of CCD 2. This drawing is very useful
in understanding what one is seeing.
July 31: Flats taken in October 2004 by the California-Carnegie
planet search team were extracted from KOA. These are conventional
(i.e. cross dispersed) flats. Identical flats were taken using an
identical setup and exposure times. Orders were extracted from the
July and October flats. Ratios were formed in the sense July/October.
In this plot the blue, green and red curves
are the ratios (July/October) of orders with approximate wavelengths of
4900, 5600, and 6400 Angstroms. The black curve is a reference ratio
formed using orders on CCD's 1 and 3 adjoining CCD 2. The variations
with a spatial scale of roughly 100 pixels are due to shifts in the position
of the meteor. The broader variations are due to QE changes in CCD
2. If one assumes that QE changes can only be negative then
they reach about 20 percent in places. Based on the deviation from
the reference order they reach about 15 percent in places. A reasonable
assumption is that the QE variations are less severe for whatever wavelength
is predominantly reflected off the closed cross disperser cover.
July 31: A spectrum of BD+28d4211 taken on July 4 (UT) is reduced
using the new MAKEE. This plot implies
that although the problem is present at that time, the effects are removed
via flat fielding.
August 1: Since problem appears to flat field out, and it would
take two to three weeks to warm, pump, oxygen soak and re-cool the decision
is made to plan this action for the Keck I shut down.
August 2: This web page created.
Work Plan.
A Word document presenting a first pass at the work schedule and the
supplies needed can be found here and in pdf format
here.
Action items for both Lick and Keck is available as a Word document
here and as a pdf file here.
Actual Work.
Week of Aug 22: Pump, plumbing, warming straps, and oxygen soak
equipment cleaned and stored, ready to be used following week.
August 29: Started plumbing in pump, RGA, and O2 soak apparatus. Had to fabricate a connector for the O2 to KF25 flange. This connector and flange cleaned and baked overnight in preparation for warming and pumping next day.
August 30 (mid-day): Autofill disabled, valve from 50-liter to 3.8-liter closed and tagged out. Finished plumbing in pump, O2 and RGA.
August 30 (2:30 PM) Started pumping line to dewar. 3.8-liter dewar ran dry shortly after. Turned ion pump off.
August 30 (5:30 PM): Line to dewar at 8E-6 Torr and dewar temperature at -80C. Opened dewar and pressure shot up by a factor of 100. RGA shows a large spike at 18 AMU (water). Moderate spikes at 17, 28 and 44 AMU (water, N2 and CO2). Many minor spikes at 43, 2, 32, 45, 16, 38, 41, 15, 30, 27, 14, 19, 28, 40, 20, 1, and 12 AMU in descending order.
August 30 (7:00 PM): A visual inspection of the mosiac is encouraging.
Already, no contamination is visible on it. Here are
before and after shots.
Note the time stamps!
August 30 (7:30 PM): Over previous 40 minutes pressure had shown a slow
rise but turned over at this point at 1.8E-3 and temperature -42C. Within
15 minutes, pressure had dropped to 1.0E-4 Torr. At this point the CCD
electronics were powered down and pump left to run over night. Note:
warming straps not yet put on.
In this picture taken from near the Hextek
looking back toward the horse collar, one sees the 4-way junction mounted to
the horse collar mount. Connected at 3, 6, 9, and 12 O'clock respectively are
the lines to the 3.8-liter dewar, the RGA, the pump, and the O2.
This picture looks a little bit more toward the door
so that the pump and O2 bottle are more visible.
This picture is taken from the other direction,
i.e. looking back toward the Hextek from behind the horse collar. Note the O2
bottle in the foreground. Here is
slightly different angle showing the pump a bit more.
August 31: Warming straps put around 3.8-liter dewar.
September 6: Vacuum at 3E-6 Torr.
September 8: Intended start of backfilling with small quantities of dry N2
to facilitate thermal conductivity between warming straps and coldfinger and
aid in removal of water. It was found that exposing any of the plumbing to
atmosphere though resulted in the RGA detecting water. Based on this finding,
it was decided to use O2 as the backfill gas since its bottle was plumbed in
already. Backfill delayed as a result.
September 12: Small quantity (approx 0.1 Torr) of O2 bled into 3.8-liter
dewar and left in overnight with valve closed.
September 13: O2 pumped out in AM. Water detectable by RGA.
After pumping all day 3.8-liter backfilled with O2 again at end of day.
September 14 (AM): Second backfill of O2 pumped out in AM.
Water still detectable by RGA. Ideally, further backfill/pumpout cycles are
desirable but possibility of Keck I shutdown ending soon require proceeding to
O2 soak. This picture shows the setup with the heating
lamp now in place.
Here is Lick Observatory drawing H-9611 which
includes a schematic of the setup and some instruction.
September 14 (3:00 PM): O2 carefully bled into 3.8-liter dewar and valve
closed.
Delicate manipulation of the O2 regulator was required, along with switching
between high and low vac gauges to ensure dewar was not over-pressured.
Just under 1 ATM was bled in. Lamp turned on. Pump left running.
September 14 (7:00 PM): Lamp turned off and plumbing as far as 3.8-liter
dewar gradually filled with O2 to about 0.5 ATM, thus preloading pump.
Valve to 3.8-liter dewar
gradually opened and then valve to O2 closed. Dewar rapidly pumps down and
within minutes is below 1E-5 Torr.
Within 5 to 10 minutes of opening the
valve to the 3.8-liter dewar a manual fill is performed.
Just before filling, the RGA indicates that N2, H20, O2, CO, CO2, H2 have
partial pressures of 1E-6, 8E-7, 3E-7, 3E-7, 2E-7, 1E-7 respectively.
September 14 (10:00 PM): By this time, ion pump is running, CCD electronics
turned back on and crate restarted. Dewar topped off.
September 15 (AM): All equipment removed from HIRES in preparation for
testing.
September 15 (PM): Very preliminary throughput exposures using quartz lamp
indicate that QE is uniform over all three chips and is at or near delivery
values.
September 19: Preliminary testing indicates best focus position of camera
may have shifted slightly.
Status.
September 15 (AM): All equipment removed from HIRES in preparation for
testing.
September 15 (PM): Very preliminary throughput exposures using quartz lamp
indicate that QE is uniform over all three chips and is at or near delivery
values.
September 19: Preliminary testing indicates best focus position of camera
may have shifted slightly.
September 22: Bias levels appear to have changed slightly on CCD numbers
1 and 2 (the "blue" and "green" chips) and significantly on CCD number 3
(the "red" chip).
Bias levels so far appear to be on the order of 20 electrons lower for the blue
and green chips. For the red chip the bias level is on the order of 350
electrons lower.
