Recommendations

The obvious recommendation for all of the above is to obtain a 5 min sausage sequence every hour throughout the course of the night. In a typical observing run, slew and temperature drifts are slowly varying in time. Here are specific recommendations.

Temperature drift. It is unlikely that a universal temperature calibration exists for the TTF. If we derive the temperature vs. Z relation for both days (Fig. 2), we find that the curves are offset with respect to each other. The gradients may well be the same in practice since it was difficult to derive an accurate temperature slope for the second day. In any event, any form of temperature correction will be relative rather than absolute.

To sidestep this issue, we recommend the use of a small peltier cooler strapped to the side of the TTF to keep it fixed at one temperature to an accuracy better than 0.1$^{\circ}$. The TTF is a closed chamber and has a time constant of an hour or more. But this course of action requires some thought. After consulting with EJP and JP, the peltier device produces enough heat (18.5 W) on one surface that it would need a heat sink with an additional system (e.g. radiator) to prevent the heat from entering the chamber. We would need a 5V 10A supply - not presently in TAURUS - for a sensor controller in order to modulate the temperature.

Cooling the chamber might be the more difficult route: could we flush with a temperature controlled (cold) air/nitrogen flow? An easier approach may be to heat the chamber to 10$^{\circ}$ in winter and 20$^{\circ}$ in summer, and to control it at the raised temperature. We could do this by resistors mounted on the inside panels of the etalon chamber with a central sensor (or 4) to measure what the temperature is at each etalon position.

Slew drift. Until we establish the root cause (see Appendix A), it is difficult to recommend a general solution. In practice, TTF users already recalibrate when they slew to a new object. If one is doing long integrations ($\geq 30$ mins), the time derivative of slew drifts will be comparable to temperature drifts, so that once again, calibrations may be required between every exposure.

If the explanation in Appendix A is correct, the slew drifts for changing declination at a fixed R.A. which passes through the centre of the TTF and the Z-ref capacitor should exhibit a much larger amplitude than for changing declination orthogonal to this R.A. In which case, we can build a simple model for relative corrections which takes into account the $\alpha,\delta$ pointing of the telescope.

For future reference, since the TTF is placed in the pupil, we could use an additional reference capacitor and piezo at the centre of the plates. The pupil is the image of the telescope which has a 17% diameter hole in it. But this would require substantial development of the CS100 and TTF itself.