The system calibration is sensitive to the drifts between the transmitter
wavelength and the receiver bandpass transmission maximum. The measurements
show that the seedlaser drifts at 
100 MHz/h rate.
In order to achieve
a stable long term operation without frequent calibrations, the
wavelength of the transmitter laser is locked to the iodine absorption peak.
Wavelength locking to the Doppler-broadened iodine absorption line
was used because it requires much less power than a locking into
a hyperfine structure of the iodine absorption line with Doppler-free
technique [33].
Compared to the wavelength locking with a high resolution etalon,
the advantage of the locking to an iodine absorption peak is
that, iodine peak provides an absolute frequency reference.
Another technique
to lock the laser output wavelength to the Doppler-broadened
iodine absorption line was reported by Arie and Byer [35].
They use Fourier transformation spectroscopy to lock the
laser to the center of the Doppler-broadened peak. This
method does not require any dither of the laser frequency, but it
is more complex.
In the HSRL, the absorption spectrum of the 4 cm long iodine absorption cell
is used to provide information about the absorption peak maximum.
The absorption peak of the 43 long iodine absorption cell cannot
be used as a reference for the wavelength locking because the saturation
of the absorption at the peak causes the flat shape of the peak, and because
the signal at the peak is small due to the strong absorption. Therefore
it does not provide good photon counting statistics for the locking.
The length of the reference cell is chosen so that the
absorption is 
50%.
The cell transmission has to be high enough to provide a
good photon counting statistics within a short averaging time.
The locking of the laser wavelength to the iodine absorption peak is performed by using an automatic controlling program that works as follows. First, the location of the absorption maximum is detected during a calibration scan. After completing the scan, the program automatically sets the seedlaser temperature to the observed peak. In order to keep the laser wavelength locked to the maximum absorption wavelength, the seedlaser temperature is dithered around the optimum temperature and information about the ratio of signal from the second calibration fiber to the signal from the first calibration fiber detected with PMT 1 (see Figure 17) is gathered. The seedlaser temperature is kept at temperature that produces the minimum ratio. The basic idea of the tuning program is presented in the following.
The tuning procedure has four steps. First, information about
the ratio between calibration fibers is gathered for the
seedlaser temperature (T(peak)) that was detected to produce the minimum
ratio. Then, a temperature T=T(peak)+dT is applied and the change
in the ratio is observed. After gathering enough statistics (
30 s),
the seedlaser temperature is returned back to the temperature T(peak)
and a new value for the ratio at this temperature is measured.
After this, the optimum temperature is detected by finding the temperature
that produces the minimum ratio.
If the temperature T=T(peak)+dT produced a smaller ratio, then
that temperature becomes to the new optimum temperature T(peak).
If the ratio for temperature T=T(peak)+dT
was not better, the dithering to temperature T=T(peak)-dT is
performed and the procedure is repeated. The wavelength dither
corresponds to temperature change of 0.009 
C (0.052 pm).
A time history of the dither temperatures for a 9 hour
run shows that, the seedlaser temperature is dithered
between 3 temperatures under typical operating conditions.
Figure 17: Signal for the frequency locking of the laser.
An expanded view from the peak shows that as the temperature
is changed in either direction from the detected peak, a
change in the ratio is noticed.
Because the absorption peak of the 43 cm long absorption cell is
flat around the maximum absorption wavelength, the error due to tuning
to the amount of aerosol detected in the molecular channel (
) is
comparable to the photon counting statistics.
On the other hand, the error due to dithering to the amount of detected
molecular signal in the molecular channel (
) is a combination of
photon counting statistics and the error between the convoluted
signals at different dithering wavelengths. The errors in 
due to tuning
as a function of atmospheric temperature are presented in Figure 18.
Figure 18 shows the error in the determination of 
, when
the laser wavelength is tuned off by 
0.052 pm, but the calculation
of 
is made for the peak wavelength.
Also the error due to dithering is shown.
The asymmetry of the absorption spectrum makes the errors due to the
tuning asymmetric.
The total effect of the tuning procedure to the measured profiles
has to be calculated as a weighted average of the errors at
different dithering wavelengths, because the tuning program is
realized so that the laser spends 2/3 of the time at the
wavelength that produces the minimum ratio
and 1/3 of the time doing the dithering. Therefore, the total
error due to dithering is better than 0.1%, when measurement period
is long compared to the dither time.
In principle,
the error due to wavelength locking can be eliminated
by inverting the data by using different calibration coefficients
for different dithering wavelengths. This has not been accomplished yet.
Figure 18: The errors in 
due to the tuning.
Error when the seedlaser temperature is detuned from the
optimum temperature by 
one step (0.009 
C) and when
the seedlaser temperature is dithered, but the inversion
is performed by using the observed peak value.