file: Calibration.txt version: 29Jun99 contents: Description of calibration and quality assurance procedures David C. Rogers Colorado State University, Department of Atmospheric Science This document describes the calibration and quality assurance procedures that were used in support of ice nuclei and aerosol measurements during the NASA FIRE Arctic Cloud Experiment, May 1998. Diagrams and photographs showing the equipment and air sampling arrangement can be found at http://lamar.colostate.edu/~dcrogers/Arctic.html CSU equipment was installed in a C-130 instrument rack obtained from NCAR. The CSU equipment is designated "CFD," which stands for "Continuous Flow Diffusion" chamber. It includes the chamber, refrigeration system, computer for displaying and recording data, air pumps and associated valves and tubing, plus electronic interface and sensors for measuring pressure, temperature and air flow. Also, a condensation nucleus counter (TSI Model 3760), on loan from NCAR, was mounted in the rack, and its counts were recorded by the CFD data system. Two test flights were conducted (April 24 and 30), and the results were examined to evaluate the performance of all components of the CFD system and to inspect comparable measurements from other investigators. I. CALIBRATIONS AND QUALITY ASSURANCE Basic calibrations were performed on the electronics and sensors in March and June 1998. The calibrations included voltage, temperature, air pressure, air flow, and aerosol particle size. a. VOLTAGE A precision voltage source was used to calibrate the electronics. In addition, a precision voltage regulator is built into the CFD electronics. This voltage is used as the reference for precision thermistors and is measured every time the data system starts, as a check on consistency and to look for stability and drift. No significant drift occurred during the field project. b. TEMPERATURE Temperatures are measured with two thermistors and eleven type T thermo- couples attached to various parts of the CFD. Thermocouple temperature is derived from a cold junction (CJ) reference in the electronics. The CJ reference was calibrated against a platinum resistance standard. There are also two precision thermistors (YSI 4403a) attached with thermal epoxy adhesive to the two chamber walls, adjacent to thermocouples. Overall temperature accuracy is +/- 0.5C over the range +40 to -60C. The real time CFD data system displays temperatures from both the thermocouples and the thermistors. c. AIR PRESSURE Air pressure of the chamber and aircraft cabin was measured with factory calibrated Motorola MPX4115 piezo-resistive sensors. Additional calibrations were done at CSU using a Wallace and Tiernan aneroid cell, which we use as a standard in our laboratory. At the start of each flight, before the aircraft was moving, pressures from the CFD system were compared with those from the aircraft data system. They were usually within 1 to 2 mb. d. AIR FLOW Mass air flow through the CFD chamber is measured with two Honeywell Microbridge mass airflow sensors, type AWM5104vn. Mass flow through the CN counter (TSI Model 3760) was measured with a similar sensor, type AWM5101vn. All of the mass flow sensors were factory calibrated with dry nitrogen. Additional calibrations were performed at CSU with air, using a Gilibrator (volume measurement); local pressure and temperature were used to convert volume flow to mass flow. Coefficients derived from the CSU calibrations were used in the real-time CFD data system. e. AEROSOL PARTICLE SIZE In the CFD chamber, ice crystals nucleate on ice nuclei and grow to several micrometers in diameter. The primary ice nuclei count is based on an optical particle counter (Climet) at the chamber outlet that detects all particles larger than about 0.5um. Ice nuclei are identified on the basis of particle size as follows. An impactor at the chamber inlet removes all particles larger than 2um, and then particles larger than 3um are identified as crystals (nuclei) at the chamber outlet. High resolution measurements of particle size are obtained by sending the Climet analog voltage pulses to a 256 channel multichannel analyzer (MCA). The efficiency of the inlet impactor, as a function of particle size, was measured by sampling ambient air with and without the impactor. Before and after the field project, polystyrene latex (PSL) spheres 2.0um diameter were used to calibrate the Climet. Response of the Climet to a larger range of particle sizes was evaluated in earlier calibration tests in the laboratory, with PSL 0.4 to 2um and glycerin particles 1 to 12um. f. SAMPLING FILTERED AIR There are no such things as reference standard ice nuclei, so there is no way to perform ice nuclei calibrations. There is, however, a standard for zero particles. The zero reference in the CFD system is a HEPA filter that can be selected with a valve to remove all particles from the air sample. This filter was used many times for brief periods in all the flights to estimate what the zero response was when the air contained no particles. The position of the valve ("Filter" or "Sample") was automatically monitored by the CFD data system. The filtered air could be routed to the CFD or to the CN counter with another valve. g. PARTICLE DILUTER FOR CN MEASUREMENTS CN number concentrations were calculated by counting the number of pulses from the CN counter and dividing by the air mass flow rate. The counter chip in the CFD data system is a 16-bit counter (0-65535). It was interrogated approximately 5 times per second, giving the CN concentration in a volume of ~3.3cm^3. A particle diluter was used upstream of the CN counter to reduce the CN concentration for two reasons. First, the counter can roll over for CN >~20,000/cm^3, producing ambiguity in the derived concentration. Second, coincidence errors become significant in the TSI 3076 for concentrations >10,000/cm^3. The particle diluter was a cartridge filter punctured through the center with a small brass capillary tube. Most of the air flow (75%) went through the filter media, where particles were trapped. The remaining 25% of the flow, along with the particles, went through the capillary tube. The dilution ratio, 4.0, was measured before, during and after the project. CN data in the archive files are corrected for this dilution. h. RESPONSE TIME CORRECTION The ambient air was sampled with a 0.5 inch forward facing orifice. This expanded to 1 inch metal tubing which was brought into the C-130 cabin, expanded again to 2 inch diameter, and then was exhausted overboard. The ventilation rate was ~1000 LPM to achieve approximately isokinetic sampling at the 0.5 inch inlet. The residence time was <0.1s, based on the tubing volume and flow rate. The CFD and CN system sampled from the 2 inch tube at the location of the CFD instrument rack through 0.25 inch i.d. tubing. This smaller tubing was ~150cm length, including various branches and valves. Before and during the field project, travel times from the start of the small tubing to the CN counter were estimated by sampling cabin air or particle free air. Travel times through the CFD were estimated in the same manner. The response time was ~4 to 6 seconds. The archive data have been corrected by subtracting this time. Thus, the clock times of CFD and CN data correspond to the air sample entering the 0.5 inch inlet. i. PARTICLE LOSSES IN TUBING Diffusion particle loss in the tubing, including tees and valves, was estimated with the CN counter by sampling cabin air with and without the tubing. This correction factor was 1.40, based on several tests before and during the field project. The archive CN data were corrected by multiplying the indicated concentration by 1.40. The greatest diffusion losses probably occur for particles ~10 to 100nm in size, which are also usually the most abundant. The factor 1.40 is only a first order empirical correction. A more accurate estimate could include particle size, pressure, temperature, electrical charge, and other effects. For the ice nucleating particles, however, no information is available on their size, and no corrections were applied to create the archive data. j. DATA SYSTEM CLOCK CORRECTION The C-130 aircraft data system time was distributed through a real-time RS-232 data stream. At the start of each flight, the CFD data system clock was set 2 seconds ahead of the RS-232 data stream time, based on advice from NCAR that the serial data were delayed about 2 seconds. A more nearly correct "real time" clock appeared on the aircraft video camera displays, and several times during each flight, the CFD clock time was compared visually with that display. Differences were entered into the CFD log book. There were systematic differences between the aircraft and CFD clocks. The CFD gained ~10 s during a 10 hour flight. This clock difference was corrected by software that generated the CFD archive files. II. DATA REDUCTION Calibrated data were recorded as binary records on the CFD data system. The data reduction process involves unpacking these records, calculating derived parameters, applying correction factors and writing the results to a sequential ASCII text file. Over a period of time, different versions of data logging and unpacking software have been developed. The unpacking software checks the consistency of these programs. Plots are then created and reviewed by an analyst. By looking at these and referring to the log book notes, the analyst selects time periods when the data are of acceptable quality. Time periods with filtered air samples are automatically excluded from the archive data. The multichannel analyzer data contain high resolution measurements of particle size at the outlet of the CFD. These are recorded at 10s intervals and can be accessed for additional studies. The MCA data files are separate from the continuous CFD records, although they have a common time stamp. Data files were re-processed in June 1999 to subtract the background or blank count. The "blank" response is the average count obtained when sampling filtered air. It can be a fractional value such as 0.25. During data reduction, this blank value is updated whenever a filtered air sample is obtained. The blank value is subtracted from the total integer count of subsequent data, and the net value is rounded off and listed in the data files. Thus, if this net value is less than 0.5, zero is reported. For calculating the average IN concentration, ten second accumulations were made, using values before any rounding is done. The concentration is the total net counts divided by the total sample volume. Due to the rounding, some of the numbers may appear inconsistent. For example, during a ten second period, there could be net 0.25 counts each second, and this would be reported as ten integer zeroes. The concentration, however, would be based on 2.5 counts, and when it is divided by the typical 10s sample volume, 0.16 liters, the average concentration is 15.63 per liter. Another caveat concerns the raw IN count. The value in the data archives is based on particles (crystals) that produce an analog voltage pulse larger than the fixed threshold. This thresold corresponds to 3um diameter when the total flow rate is 10LPM. However, the pulse height varies with flow rate, hence the threshold "size" also varies with flow rate. There is a way to correct for this dependency, by deriving the crystal count from the MCA pulse height distribution data, combined with flow rate data. This correction has not been done yet. III. KNOWN PROBLEMS Several problems with the instrumentation were encountered during the field season. First, a leak was discovered in the CSU CN counter, as indicated by its failure to show zero concentration when a HEPA filter was placed on the inlet tubing. The amount of leak according to the pressure difference between the aircraft cabin and the CN counter, which was approximately at ambient pressure. Thus, the leak was worse at high altitudes. At low altitudes, less than about 500m, the leak was not observed, and the CN counter would zero properly on filtered air. In the CSU archive data set, CN values are reported throughout all flights. Data users should be aware that the CSU CN data are probably okay at low altitudes but not at high altitudes. The CFD data system hard disk failed on three flights, May 4, 11 and 18. When this occurred, the data logging program halted. When the program was restarted, the hard disk usually operated okay, although there were sometimes additional failures during the same flight. On one occasion, the hard disk did not recover. A new disk was installed on May 9, but it also failed, and we surmised the vibration levels in the instrumentation rack were the cause. These failures occurred during ferry flight periods enroute from Fairbanks to the SHEBA ice camp. Gaps in the data could be as long as several hours. This problem did not cause us to lose all data because we also had a backup data logging procedure that used diskettes, and it did not suffer from the vibration failure problem. Overall, some data were lost on three flights enroute to the SHEBA ice camp, but no data were lost in the vicinity of the ice camp.