Bruce A. Wielicki
October 9, 1998
- The problem is definitely with the voltage converter and not with the voltage measurement (housekeeping data).
- The problem is not caused by radiation damage. Laboratory radiation tests of similar converters have shown
that it takes 100 times the radiation exposure that the TRMM spacecraft has seen to cause degradation of the
- The problem was not caused by the CERES instrument design, use, or mounting process at TRW.
- Of roughly 5000 of these military spec parts that the voltage manufacturer has sold, only 1 has come back
with a similar problem to our experience on CERES. That part was used in a high temperature environment
(>100C) in a commercial aircraft, and showed a problem with bleeding of a silver conductive epoxy during
the curing process. The silver epoxy is used to mount an optical coupler that controls the feedback gain
of the converter. Normally, this optical coupler can experience degradation in gain of as much as a factor
of six without affecting the performance of the converter. The bleeding of the silver epoxy, however, is
thought to have decreased the gain the optical coupler by acting as a resistor across two of the couplers
leads which should not be in contact. The theory is then that this gain loss still allows the converter
to operate normally, but without much margin for any further degradation. If later radiation exposure caused
an additional 1 or 2% gain loss of the optical coupler, then this might explain why the converter passed
extensive burn in, but later showed the anomaly on the CERES instrument. TRW is working up a detailed
theoretical model of the converter physics to test if this type of problem can explain the temperature
sensitivity seen in the CERES converter. Expect results in 2-4 weeks. While this is a candidate for the
anomaly, we still do not have a mechanism that fits all the observed behavior. This is currently, however,
the best candidate and it is being pursued further.
- The sensitivity of the voltage converter to instrument temperature change (variations in orbital heating
of the spacecraft and instrument) has remained constant at roughly 0.6V per degree C. What has changed is
the minimum temperature Tmin at which the sensitivity begins. Below Tmin, there is no sensitivity of the
voltage converter to temperature and it operates normally. From late July to late August the Tmin decreased
almost linearly with time until it reached about 17.3C and then stabilized over the last 4 days the
instrument ran in August (before we turned it off to investigate). If the converter has indeed stabilized
for the long term, then the maximum voltages we will see in the future will be less than 18V, which is
not a safety risk to the instrument.
- TRW has tested all 7 of the electrical components downstream of this converter to 30V. The maximum voltage
the converter will let through from the spacecraft in the event of total failure is estimated to be 29.4V.
At 30V, all 7 of the parts operated nominally, and TRW is readying a complete breadboard for long term
testing at 30V to test lifetime at this elevated voltage. This testing will begin next week and will
continue indefinitely. Unfortunately, there is no way to do an "accelerated life test" with this type
of problem. So it will take 2 months to know if the parts could survive a converter failure for 2 months.
This will, however, be a worst case scenario since the actual voltage fed through from the spacecraft will
vary from roughly 23 to 29V depending on orbital conditions.
- On October 13 at 8pm EST we will power up the CERES instrument on TRMM. It will slowly warm to operational
temperature, and we expect that around 8am on Wed Oct 14, that the temperature will increase to near the
17.3C value of Tmin that we saw in late August. The tiger team has planned a 6-day test sequence of all of
CERES operational modes to evaluate the converter condition and whether any further degradation has occurred.
This data will be evaluated, and along with the results of the other ongoing studies, a plan will be
developed for further instrument operation. We will be able to monitor throughout the 6 day period whether
any further instrument degradation occurs with further operation. We should know sometime on Wednesday if
it looks good to continue running the instrument. It looks like we should be able to continually monitor
the voltage converter sensitivity by using the temperature/voltage relationship: changes in this linear
relationship can be directly related to changes in Tmin: Voltage output = 15 + 0.6(T - Tmin). Given
values of voltage and T at any time, you can solve for Tmin. This allows a monitoring of Tmin without
powering down the instrument to get to temperatures below Tmin.
- If we can convince ourselves using the 29V lifetime tests that we have a high probability of the instrument
lasting at least 1 month after any failure of the converter, then I would propose that we continue to operate
the instrument in our normal 2 days of crosstrack followed by 1 day of rotating azimuth plane day while we
monitor the the voltage converter output for any further degradation. If it remains stable below 18V
(no additional risk) or perhaps even below 21V (design is 15V, so small risk) then we continue to operate
the instrument routinely. If the voltage converter continues to degrade and appears headed to a rapid
failure, then we safe the instrument, and save it for intercalibration with CERES data on EOS-AM. We
would run the instrument after the EOS-AM CERES instrument is operational and only for the two or three
day periods where the TRMM orbit swath and local time sampling is similar to that of EOS-AM CERES
instrument in order to verify intercalibration (this allows optimal time/space/angle matching of the
two data sets). This will allow extremely accurate ties of the TRMM data to the EOS-AM and PM data. Once
we get sufficient overlap data to tie the relative calibration of EOS-AM and TRMM CERES instruments to
one or two tenths of a percent, then we could operate the TRMM instrument continuously for as long as
it lasts. Note that the current best RUMORED date for EOS-AM launch is probably Fall, 99, which would
be two years after TRMM launch: with roughly 1.5 years of TRMM lifetime remaining before we run out of
fuel to boost it in the low orbit altitude of 350 km. The reason that overlap is critical is that the
CERES instruments calibration data within the first 8 months appear to be stable to 0.2% or better (all
3 channels). This indicates that rather like solar constant measurements: while you may not know the
absolute accuracy to better than 0.5 to 1.0%, you can measure the month to month and year to year changes
much more accurately: approaching 0.1 to 0.2%. We still need longer time series to prove this stability
with the new CERES instruments, but results so far look like this is the conclusion we will reach with
- We are checking the part lot numbers of the voltage converters in the other 5 CERES instruments on the
ground. If movement of the silver conductive epoxy is indeed the problem, then it turns out that this
process was changed about a year after the parts lot that was used in the TRMM CERES instrument. The
change eliminates the non-conductive epoxy layer because the voltage manufacturer changed suppliers of
the optical couplers, and the new optical couplers were lighter and did not need the mechanical support
of a non-conductive epoxy layer below the coupler. Potential cost and schedule impacts of changing the
voltage converters on the AM, PM, and FOO CERES instruments are being evaluated.
- We have looked at the data and in-orbit calibration data from August and it shows no sensitivity of the
CERES radiometric data to the elevated voltage of the converter (it shouldn't, but this has been confirmed).
- Bottom line: looks very encouraging so far for continued CERES operations throughout the TRMM lifetime,
but the acid test will be the results of the 6 days of operation starting October 13. A good target for
deciding to restart routine science operations would seem to be to restart routine data collection on
Nov. 1 or Dec 1 depending on the results of the 6-day test. Note that if we miss Sept/Oct/Nov data,
that we have basically missed one season of climate data. Fortunately the El Nino event was captured by
the first 8 months of data.
Return to CERES +15V Power Converter Anomaly Page