Campbell performs calibrations (Roy Sorensen, pers. comm. 95) by operating the unit in room conditions for 24 hours to get some scaling, calibrating [we think] in the air over various salt solutions at room temperature, cleaning the lenses with distilled water, and repeating the calibration over the salt solutions. They measure the optical path length, and provide regression equations for both scaled and cleaned cases over the whole range (2-20 g/m3), over a dry range (2-10 g/m3), and over a wet range (8-20 g/m3). The equation they use is:
ln V = (X Kw) rho_v + ln VoThe coefficients X and Vo from the whole range regression are shown in the first two lines of Table 1. Although the offset, Vo, is quite different (lower) when the sensors are scaled, the gain, Kw, changed by less than 4%.
Table 1: Summary of calibrations done 9/94-11/94 K1101 K1133 Kw Vo (mv) Kw Vo (mv) CSI scaled, 15pts -0.133 4462 -0.129 4191 CSI cleaned, 15pts -0.131 5122 -0.124 4527 +25, l->h, 12pts -0.141 6634 -0.126 4673 +25, h->l, 10pts -0.135 5427 -0.121 4025 +10, l->h, 7pts -0.133 4729 -0.123 3816 ~-4, l->h, 3pts -0.138 4156 -0.124 3511 +40, l->h, 6pts -0.137 3282 -0.134 3166 +25, l->h, 10pts -0.142 3123 -0.133 2531 +25, l->h, cleaned, 10 -0.131 6681 -0.118 5302 mean, NCAR -0.137 -0.126We used our Thunder Scientific temperature/humidity chamber which sets the relative humidity by the ratio of the vapor pressure in the chamber to the vapor pressure of air saturated with water vapor. The calibration of the temperature sensor was determined in an oil bath using a platinum resistance thermometer as a reference. Thus, this chamber is close to a "first-principles" device.
The desired range of temperatures and humidities used in the Thunder chamber calibrations is shown in Table 2. In fact, it was not possible to get to all of these low humidities, and the chamber did not get down to -5 C due to insufficient settling time. (The chamber was commanded to settle for 80 minutes after changes in temperature and 20 minutes after changes in humidity at constant temperature.) Most of the calibrations were performed changing from low to high humidities, except for the first run at 25. A plot of these data is shown in Figure 1 and the regression coeficients to all of the data at each temperature are presented in Table 1. The last line of Table 1 gives the average value of the gain, Kw, for all of the chamber tests.
Table 2: Chamber tests of Krypton hygrometers
Temperature: Humidity (g/m3; RH)
+40: 2 6 10 14 18 20
4% 11% 19% 27% 34% 42%
+25: 2 4 6 8 10 12 14 16 18 20
9% 17% 26% 34% 43% 52% 60% 69% 78% 87%
+10: 1 2 4 6 8
11% 21% 42% 64% 85%
-5: 1 2 3
29% 59% 88%
Table 1 also shows large changes in the offset, Vo, presumably due to scaling. Vo dropped during all of the chamber tests - including the first two runs at +25. Note that scaling during the calibration also causes the gains for these two runs to differ. The scaling rate appears to be worst at high humidities (as Tanner had predicted), as seen by the 1.2 V drop on #1101 between the first two +25 C runs and the additional 0.9 drop after the +40 C run (both of which included points at 18 and 20 g/m3). The last chamber run, performed after cleaning the lenses, shows a return of Vo to levels even higher than originally seen by Campbell.
Unfortunately, later field tests (during PAMIII94 at the NCAR Marshall field site) still showed drifts in the mean values of one of these units which were unexplained by simple drift (and did not appear to be simply related to temperature). Figure 2 shows an example of this, where the difference between 5-minute average humidity values from the krypton and from Vaisalla capacitance probes (the bottom purple line) appears to be unrelated to temperature (the top black line). The bottom black line in this figure shows that the mean humidity from the other unit was off by a constant 2 g/m3, presumably due to scaling. In this example, the drift was large enough to cause 5-minute average variances to be up to an order of magnitude too large and fluxes to be up to 100% too large - clearly a problem for turbulence measurements. We also determined that the drift was not related to external radiation by changing the array oriention between vertical and horizontal. This problem finally was solved by replacing the "weak" detector tube. Thus, there are cases when an instrument can give a signal, and even calibrate properly, and still not work properly in the field.
We also have seen source tube failures in cold and/or wet conditions. However, these normally are characterized by a total lack of signal. The only possible solution we know of to prevent these failures is to cover the tubes with bags in heavy precipitation.