Campbell Scientific EC155 CO2 and H2O Closed-Path Gas Analyzer and EC100 Electronics with Optional CSAT3A 3D Sonic Anemometer - Manual (Page 55)

Appendix A. Filter Bandwidth and Time
Delay
The EC100 measures CO
2
and H
2
O from the EC155 gas analyzer head (as well
as wind velocity and sonic temperature from the optional CSAT3A sonic head)
at 100 Hz and then applies a user-selectable low-pass filter. The available filter
bandwidths are 5, 10, 12.5, 20, and 25 Hz. FIGURE A-1 shows the amplitude
response of these filters. The EC100 filters provide a flat pass band, a steep
transition from pass band to stop band, and a well-attenuated stop band.
FIGURE A-2 compares the EC100 10-Hz filter to a 50-msec moving average
filter with approximately the same bandwidth.
The ideal eddy-covariance filter is one that is wide enough to preserve the low-
frequency signal variations that transport flux and narrow enough to attenuate
high-frequency noise. In addition, to minimize aliasing (the misinterpretation
of high-frequency variation as lower-frequency variation) the measurement
bandwidth must be less than half of the sample rate, or the datalogger scan rate.
Two factors complicate choosing the ideal eddy-covariance bandwidth. First,
the flux signal bandwidth varies from one installation to another, and the flux
signal bandwidth varies with mean wind speed at a given installation. Second,
the fast sample rate required to anti-alias a desired signal bandwidth may result
in large, unwieldy data sets.
Fortunately, the covariance calculation itself relaxes the need for the ideal
bandwidth. First, the time-averaged (typically thirty-minute) covariance
calculations inherently reduce noise, and second, aliasing does not degrade the
accuracy of covariance calculations. Therefore, the factory default for the
EC100 bandwidth (20 Hz) is rather wide to preserve the signal variations that
transport flux, and that bandwidth is suitable for most flux applications.
Additional bandwidths are available for experimenters desiring to match the
EC100 filter bandwidth to their data acquisition sample rate to avoid aliasing.
In this case, the selected bandwidth should be one-half of the sample rate (or
datalogger scan rate), and experimenters should be careful to avoid attenuation
of flux-carrying signals.
The EC100 electronics synchronously sample the gas in the EC155 sample cell
and the CSAT3A sonic head. However, delays induced by the intake assembly
must be accounted for. The exact delay will depend on the length and size of
the intake tubing and the pump flow rate. This delay needs to be
experimentally determined by shifting the time delay until the covariance of
the vertical wind and the gas concentrations are maximized.
Experimenters wishing to synchronize their EC100 data with other
measurements in the data acquisition system must account for the time delay of
the EC100 filter. TABLE A-1 shows the delay for each of the filter
bandwidths. The EC100 provides a constant time delay for all spectral
components within each filter’s pass band.
The following examples show how to use TABLE A-1. To synchronize
EC100 data to other datalogger measurements when the datalogger scan rate is
25 Hz and the EC100 bandwidth is set to 20 Hz (a 200-msec time delay from
TABLE A-1), delay the non-EC100 data by five datalogger scans. Similarly,
for a 10-Hz datalogger scan rate and the same 20-Hz EC100 bandwidth, delay
A-1