Description: | Starting on 980619, periodic fluctuations in several of the measured AOS parameters (e.g.,
Aerosol light absorption coefficent, TSI nephelometer relative humidity, condensation
particle counts, etc.) were observed. The start of these fluctuations coincided exactly
with the shutdown of the AOS for power washing of the wooden stairs and stack platform.
Numerous attempts at troubleshooting were conducted via telephone with technicians at the site.
These concentrated on 1) trying to find an electrical noise problem or 2) determining
whether our stack heating unit was somehow producing particles. Tests were devised to check
both of these possibilities, and both were determined to be non-problems. The
troubleshooting of this signal fluctuation problem was compounded and delayed by the concurrent
unrelated electronic ball valve problem (see AOS DQR#980731.1).
Pat Sheridan and Jim Wendell from the NOAA Aerosols Group went to the SGP Central Facility
to diagnose and repair the problem. All electromagnetic noise possibilities, including
bad grounds and RF fields were checked. These were not causing the fluctuations, which
varied in period from ~15-30 minutes. Upon dismantling the insulated air intake assembly
inside the trailer, a minute crack was found in one of the spare sampling lines at its
connection to the manifold. This line was one of the original installed by the previous mentor.
Repair of this line break essentially repaired the fluctuation problem, although this
extensive troubleshooting revealed a slight temperature dependence for some of our reference
signals (more on this below).
Our explanation of this problem is as follows. When the AOS stairs were power washed,
either the activities associated with power washing (e.g. people or compressors on the
platform) or lowering the stack caused enough vibration to cause a hairline break in the spare
sampling line at the connection with the manifold. We believe this tubing is
high-density polyethylene, which is semi-rigid, and it has now been replaced with less brittle
material. The routine daily system zero checks established by the previous mentor would not
have detected this leak problem, because overpressurization of the manifold with
particle-free air showed up as good zero checks in all instruments and some of this clean air
would have simply flowed out of the spare line tubing crack. Since the crack was beneath at
least two layers of foam insulation which itself was covered with foam-backed metallic
tape, we believe that free air flow of trailer air to the crack was not likely. However,
during periods when the large air conditioning unit was on, pressurization of the trailer
caused air to slowly leak into the spare sampling line and back into the manifold and
other sampling lines. Upon arrival at the site, we found the fan for the A/C system
operating only when the A/C compressor came on. We were informed by Dan Nelson, Site Engineer,
that this fan is supposed to stay on all the time. Our data suggest that this fan has
been off since before the problem began. This is fortuitous, because the trailer was not
continually pressurized and room air was not sampled continuously over this period.
Thus, only a portion of the data are of questionable quality. This leak was not immediately
obvious (to AOS technicians or to us) because of its concealed location and because some
aerosol measurements (e.g., particle number concentrations, aerosol light absorption
coefficents, etc.) did not change dramatically during these periods (i.e., the aerosols
brought inside the trailer through the cooling system were often not much different in
concentration or optical properties than were aerosols sampled at the top of the stack). Thus
data during the A/C periods will be labeled as questionable.
We will institute a new leak check/zero check procedure which should catch leaks of this
type. The old system zero check devised by the previous mentor was insufficient to find
this leak.
The A/C unit in the AOS trailer is more than adequate to cool that space. The cold air
that comes out of the vents is probably >20F cooler than trailer air. This cold air causes
slight fluctuations in the lamp voltages and reference signals of several of our
instruments. We expect this problem to be more severe in the summertime because of increased
air conditioner use and indoor temperature swings. Fortunately, these signal fluctuations
are not observed in the final processed signals from these instruments. They do,
however, suggest that a gentler cooling of the trailer should be considered.
Possibilities for this could include:
1) Installation of a smaller air conditioning unit
2) Changing the ratio of outside air mixed in with the cooled air
3) Venting the exhaust from the cabinets into the air recirculating
system to preheat the cooled air.
We realize that these suggestions may not be in line with the notion of efficient cooling
of the trailer. However, we feel that the root cause of the observed temperature
fluctuations is that the air conditioning unit has far more capacity than is needed by the AOS
trailer. A smaller air conditioner, properly matched to the heat load of the trailer,
would probably be a more efficient solution.
Since the cooling of the trailer generally was faster than the warming-up period, a
majority of data should be valid. Also, not all measurements appeared questionable (i.e., the
data did not appear to change when the A/C unit came on). We have flagged all aerosol
data as "questionable" during A/C "on" periods because of the likelihood that at least some
mixing of trailer air with ambient air occurred. |
Suggestions: | An algorithm for flagging the questionable data during the periods the air
conditioner was running has been developed at CMDL. Basically, we use a filtering program to flag
data as questionable when the air conditoner was on. This is most easily determined by
looking at the room temperature sensor data (RoomTemp in sgpaosauxC1.a0 datastream). The
room temperature sensor is located directly under an air conditioning vent, and so it
should provide extremely reliable and fast-response data. When the first derivative of the
room temperature sensor data is negative, the trailer is being cooled. Data are flagged
until the room temperature first derivative changes sign. |