We have continued experimenting with the atmospheric phase compensation under various atmospheric conditions.
Figure: Model based phase correction at 91.2 and 230.5 GHz on
October 21, baseline 176 m. First two rows: Amplitude,
uncorrected and corrected phases for 91.2 and 230.5 GHz,
respectively. Third row: Phase noise histograms of uncorrected
phase (shaded) and corrected phase (thick line). One scan
corresponds to 120 s of integration. Last two rows: amplitude
and phase at 230.5 GHz have been vector-averaged over time bins of one
minute. The bars mark the 
error of the average.
At the beginning of October, an analog/digital converter in the
detector chain was found to dominate the noise of the total power
signal. This was limiting the overall relative stability of the
detection system to about 10
, thereby limiting the accuracy of
the phase compensation to about 20-30 degrees rms in the 84-116 GHz
band. Proper electronic grounding of the A/D converter has reduced
its noise contribution by a factor 6, and the stability of the
complete total power acquisition chain is now about 
.
As a consequence, the rms of the corrected phase has been improved sufficiently to be of interest in the 210-248 GHz band. An example is given in Fig. 5, which shows a sequence observed on October 21st, after the passage of clouds, while the phase stabilized over a time interval of about 1.5 hours for a baseline of 176 m (the correction is independent of baseline length). Opacity was low, with a total amount of precipitable water around 3 mm.
The time resolution of the correction is one second, and it is
exclusively based on the optical path predicted by the atmospheric
model (M.Bremer & J.Cernicharo) from the calibrated total power.
Phase noise histograms have been calculated to illustrate the effect
of the correction. The scan based distributions peak in the
and 
rms bins for 91.2 and 230.5 GHz,
respectively.
The observed phase at 1.3mm moves too fast to be continuously traced;
the atmospheric phase predicted from the total power fluctuations has
indeed an r.m.s. of 
.
For an integration time of one minute, the gain in coherence of the 230 GHz continuum signal is spectacular.
Implementation of the phase correction software in the CLIC calibration program is under way. A full correction, with phase continuity from one source to another, is not possible because of slight, but significant, receiver gain variations for large telescope motions in elevation. Accordingly, the phase correction will be made with a zero mean. This can be done a posteriori for the continuum data which are sampled every second. This is not possible for the spectral line data which have an averaging interval of 1 or 2 minutes. Since the correction breaks down in the presence of clouds, we are currently working on a scheme which will provide both corrected and uncorrected spectral line data in real time.
Michael BREMER, Stéphane GUILLOTEAU,
Robert LUCAS
