2129 RIQ FCa : Discussion about the DIT NDIT convention

[astronomyk]

https://jira.eso.org/browse/MET-2129

As far as I know, each instrument has its own convention for the use of DITxNDITxNexp = Nframes. In the case of METIS, does the master cube referes to the total observing sequence (i.e. Nframes) or is there a difference wit Nexp - is it possible to clarify ?
What is the minimum DIT allowed considering the read speed of the detector ?
Is there any other consideration to take into such as the possibility of temporal binning and spatial windowing ?
Also another minor question in the same frame: is there any smearing effect that could appear in the so-called "ADI sequence" (in pupil tracking) ?

@Rumpelstil @gotten @roy are you able to answer any of these questions? Preferably directly in Jira (with a note here).

[hugobuddel]

I accidentally submitted an answer to this RIX on Jira, but that was supposed to go to https://jira.eso.org/browse/MET-2158 . I've asked the RIX to be reopened

[Rumpelstil]

Minimum DITs for focal plane arrays with analog read-out circuits (non-GeoSnap) will be determined once NGC2 has been tested by ESO with a range of cable lengths. Details on potential temporal and spatial windowing will have to await test results with actual NGC2 units.

[hugobuddel]

Answer from Roy added as comment to Jira:

(1) We may need a short discussion about terminology, perhaps what we adopted has slightly different names in ESO speech.

The minimum data unit we will store is DIT*NDIT, we called this a “super-DIT” but maybe in ESO speech this is an “EXP"(osure). We will recored NsDIT of those per offset position (where offsets can be made with the chopper and/or the telescope).

So the total amount of integration time of an observation would be
T = DITNDITNsDIT*nOff
where nOff is the total number of offsets made (can be be back and forth between the same set of positions many times)

This results in NsDITnOff stored detector images, each of which is the average of NDIT individual integrations. So those NsDITnOff planes are the input for PIP to create the data product from.

(2) for the fastest frame rate we currently assume op to 90 Hz (DIT=11 ms) frame rate for GeoSnap, and 25 Hz (DIT=40 ms) for the h2rg in the IMG, and 1 Hz (DIT1s) for the h2rg in the LMS. If NGC2 fails to meet those speeds by a substantial amount then we will need to consider mitigation strategies (particularly for M-band imaging where the well depth is most of an issue).

(3) spatial windowing is not considered, also it doesn’t bring as much as one would like because we can only speed up according to the field reduction in 1 of the 2 dimensions. For example, if we would go to 1024^2 pixels instead of 2048^2 pixels on the h2rg, we would gain only a factor of 2 in speed even though we read only 1/4th of the pixels.

As temporal binning we currently plan to bin the N-band data down to ~100ms sampling, i.e. reducing the data volume by ~1 order of magnitude compared to “burst” mode (=storing each individual DIT) for broad-band imaging. This should be enough to sample atmospheric variations.

(4) The fastest field rotation rate in pupil-tracking observations does not occur at +40 deg of declination, but near zenith, i.e. for a target whose declination differs by + or -1.5 deg from the observatory geographic latitude; see appendix A and B of this draft document. The 1.5 deg comes from the ESO top level requirement that we should be able to observe targets at zenith distances >=1.5 deg. In the extreme case, the field rotation rate reaches 0.145 deg/s = 2.53 mrad/s. So if we have an on-axis guide star and a source in the edge of the IMG-LM FoV, so at sqrt(2)*10.5/2 = 7.4 arcsecond, then the field rotation causes a translational motion of ~18.8 mas/s, which is almost a PSF FWHM at L-band. That means in such an extreme case, the DIT should be limited to ~0.1s (and NDIT set to 1) if the PSF smearing in the corners of the field is to be limited to 10% of the perfect PSF width. This is an extreme case, though. The simplest mitigation would be to do the observation a little bit later, I think we will want to avoid observing observing so close to zenith if we can anyway, and also the hypothetical target described here is not the typical METIS target. The latter will be of much smaller extent and the corresponding smearing will be smaller (provided the science target is also the AO NGS, which is usually the case, but not always).