The goal of last night was to carry out extensive nulling sequences on bright and then faint stars. While we had clear conditions, the PWV increased during the night, and we found it essentially made our attempts to record calibrated null sequences useless. Consequently, we switched to the LEECH program for the second half of the night.
We started by carrying out further testing of the phase loop. We saw similar behavior to the previous night (350-450 nm RMS). In checking out the system, we uncovered a problem with the fringe "hopping" code. This is intended to bring the phase loop to the correct OPD, by jumping one wavelength whenever it detects an asymmetry in the fringe envelope. After fixing the problem, we noticed much smoother behavior in our phase loop. Indeed, this may have solved the majority of the phase loop "break" issues. The rest of the night the loop behaved much smoother than it had any previous run. Conditions were average to mediocre for seeing (1-1.3 ") so this was encouraging. At this point, the PWV was 3.8 mm.
We carried out 8.7 and 11.1 nulling sequences to understand the phase stability in each passband. We noted that the tip-tilt stepping for 8.7 needs to be set differently.
We then moved to a HOSTS target to carry out nulling sequences. Setting the best set-point was significantly more difficult than our attempts the previous night. Indeed, when we looked at the data we were seeing nulling variations of 10-15% over 10s timescales. The cause appeared to be the 5 mm of PWV . Simultaneous LMIRcam dispersed fringe data were acquired, so we have data to learn whether this can be used to modify the NIR set-point.
There are two problems with high PWV. The first is that it adds phase variations to the null sequence. However, this should be random, and average down, even if it decreases our precision. The second problem is that it becomes difficult or impossible to measure the correct set-point for obtaining a good 10 micron null. This results in the sequence having few frames that measure the true null depth. Essentially, we can end up taking junk sequences in bad PWV conditions.
Looking back at our null stability over our past several attempts we see a clear trend. Here's the rough numbers:
||null stability (PPM)
This suggests we really need <4 mm (<3 mm preferred) for usable nulling data. This is reasonably common. The SMT, next door tracks these statistics. You can find there statistics at:
Note their data is in tau_225. The conversion to PWV is PWV=20*T_225. Looking at their data we can expect < 3mm 30-50% of the time.
Going forward, I think we need a two prong strategy:
- Develop queue scheduling that allows HOSTS to use dryer conditions.
- Develop an algorithm that uses the LMIRcam fringes to track the water vapor dispersion.
Both of these are likely doable, although timescale and effort needed are still open questions.