Calibration Strategy PDR Cambridge England 21-22 June 2001

Attendees:

I've arranged these notes in several sections. First, extracts from notes at the meeting. Second, comments on talks given at the meeting (or not given). Lastly, recommendations from the Review Committee.

The ALMA Science Advisory Committee has suggested:

The ALMA calibration specification of 1% for absolute intensity is adequate scientifically perhaps even a bit aggressive.

Among the items to be considered at the review are:

* What can we reasonably expect to achieve?

* Should this differ among frequency bands?

* What does this mean in the submillimeter, where a stand-alone ALMA will be challenged to recover all source flux?

While it is good to have an aggressive strategy, it should be one which is achievable.

Notes

Matt Carter began with a talk on optics. He pointed out that since the Front End PDR some improvements had occurred, including a novel design to minimize cross polarization. Investigation continues on employing this at Band 7. A WVR mirror has also been included in the 'widget space' since the PDR. Interest has increased in semi-transparent vanes since they were first considered at the Berkeley ASAC meeting. A scheme which employs vanes as follows was presented:

Bands 3-6 Ambient plus semi-transparent vanes

Bands 1-2 Ambient plus 1 semi-transparent vanes

Bands 7-10 Ambient vane only

with the vanes to be inserted once every several (5-20) minutes for several seconds with the proviso that Band 3 remain available at all times, and that the WVR be used with all measurements except those with an ambient load inserted.

These requirements present a number of technical and engineering challenges. Foams for possible use as the semi-transparent vanes have been measured at IRAM-typical numbers are 1.5-2.5 db loss 205-255 Ghz; phase also measured. The requirements suggest a pattern of loads moved in one plane such that proper windows lie over proper receivers at the appropriate times. Carter presented one such solution. Development on this system needs personnel and monetary resources desparately. Testing should proceed on prototype antennas.

Several potential problems needing consideration were noted. The flat plate of loads could bring about problems with standing waves. The temperatures of the foam loads are not stabilized and difficult to control. An alignment specification on receivers needs to be worked out so that the 1% error budget is not affected by placement of the loads; one does not want antenna-dependent offsets for each antenna.

Radford reported on the site. The NRAO and NRO tippers were run side by side in 2001 April showing the cross-calibration to be excellent confirming 10-15% better transparency at Chajnantor than at Pampa La Bola. Phase-wind speed studies have shown the wind speed is higher in the winter, but phase stability is better, suggesting a more laminar flow. In an article shortly to appear in PASP, Giovanelli et al present results of seeing measurements from the site. The 50^th percentile is 1".09 at the ALMA site and 0".7 at Cerro Chico, a small hill nearby.

Pardo reported on atmospheric modeling. Many mechanisms affecting transparency are now included in the model including line broadening mechanisms, pseudo continuum terms, and minor atmospheric constituents. The ATM model represents the atmosphere up to a frequency of 10 THz currently. For water, measurements in the millimeter windows have shown excess absorption compared to expectation. Explanations include: incorrect modeling of far wings, failure of the impact approximation in the far wings, or the influence of water dimers, unlikely at high dry cold sites. Dry terms include the Debye spectrum of O below 10 Ghz and pressure induced O2 and N2 terms. The models are compared to FTS spectra above 100 Ghz for which a full resolution scan takes 100 s round trip times 2 pairs for about 20 minutes total (including calibration scans on hot and cold loads). Derivation of the wet and dry continuum suggests an excess 29% dry continuum in the submillimeter. A fictitious H2O line at 1.7 THz provides a pseudo continuum which can match observations. Cirrus has some effect in the high submillimeter windows. To model this one needs to know the particle size (50 microns? 30 microns?) So that to model the atmosphere well one needs at least one submillimeter measurement in addition to that at 183 Ghz. With this one can constrain particle size and equivalent water path and build a good model of the atmosphere.

Guilloteau presented studies described in his new memo 372 on receiver calibration. 1% precision appears possible at millimeter frequencies but 3% is more achievable at submillimeter frequencies. Atmospheric modeling may limit calibration to this level. A specification on standing waves of <2% power was suggested; it will be hard to avoid variations between antennas. Even keeping them low, accurate calibration is hard. Given the same antennas, the spacings of reflectors will be about the same; since amplitude is diffraction related and goes inversely as frequency this will pose a lesser problem at higher frequencies. An antenna gain model would be useful using holography, surface errors, and illumination patterns. This would be quite hard at submillimeter wavelengths-Baars thought 5% would be the best achievable at 3mm. Woody pointed out the problems of non-heterodyne response of SIS receivers-sometimes IR photons are detected when loads are inserted, confounding calibration. Two receiver calibration systems were discussed-the dual load in the secondary and a system of semi-transparent and ambient vanes. Problems with the former include measuring the coupling efficiency accurately (it is small for ALMA antennas), length of integration needed in the submillimeter, and sky stability. Welch reported on studies of the BIMA prototype of this system briefly; studies continue. Semi-transparent vanes offer an alternative below 300 GHz-an engineering plan was discussed earlier by Carter. Welch discussed a BIMA implementation for solar observations-one might consider lossy foam, which may pose scattering problems, or a beamsplitter, which has other problems (polarization); Hills suggested a sheet of plastic may be better than foam. At BIMA a 350 micron thick mylar sheet mounted at a 35 degree angle in PVC tube lined with absorber provides a 30% absorption. Mangum presented his work from Memo 318.

Welch proposed a scheme for absolute flux calibration used at 1cm at BIMA. A strong source, e.g. Jupiter, could be measured to better than one per cent using a standard gain horn, measured in the lab, mounted on the side of one of the ACA antennas. Alternating between observations with this horn and the antenna feed, the gain of the antenna could be calibrated. The calibration could be transferred to other antennas in the array cross correlated with this standard antenna. Fluxes for the standard sources (phase calibrators also) during the next period of observations could then be tied to the absolute standard for use by observers over that period. In the BIMA implementation, a clever switch design is used to measure and match delays. It was suggested that one might avoid the switch by just correlating the horn with the other antennas. Effects to be corrected include the delay, resolution of the source (a smaller correction with the ACA antennas) and stability of the system. Stiffness of the antenna, a hallmark of the ALMA designs can ensure the latter; the overbuilt ACA antennas should be best in the array. Welch noted that the Sloan survey employs a special optical telescope, smaller than the primary one, to ensure flux calibration to 1% in a somewhat analogous fashion.

Holdaway discussed beam calibration, based on his yet-unpublished memo. Important here are those errors which tie the calibration to the imaging. Observations at several frequencies and elevations are needed to provide 2D beam models-are there bright enough sources at the high frequencies? For higher dynamic range images of simple sources one might employ direction dependent self calibration. Recommended: creating software to do 2D beam applications (parallactic angle corrections), establishing methodology for measuring the 2D beam as a function of frequency and elevation, studying the availability of suitable sources, research on the direction-dependent self calibration method, and simulations to guide choice of method. It was discussed whether the antennas are stiff enough to calculate deformations to the needed accuracy. High frequency corrections will present a problem.

Mangum and Welch presented optical pointing discussions. Mangum discussed plans for the prototype antennas to use high quality optical telescopes in the near infrared. Welch noted that if one can only point to 1/10 beam or so images will be severely restricted in dynamic range. The Keck telescopes, for instance, only point absolutely to 20" or so but achieve accuracy by using guide stars and the subarcsecond seeing available on the telescope. A BIMA experiment was described for a night time observation seeing 4".5 aperture 4" achieving a SNR~15 on a 12m star in 1 sec; such a star will exist in a 10' square field. In daytime, the aperture should be smaller, perhaps 1".8 using a near infrared filter in seeing 6".5 SNR~20 is reached on a 3m.5 star in 1s (F and G stars). With a polarizer, one might penetrate an order of magnitude deeper. At Chajnantor, seeing is better; one might reach V=14m shrinking the necessary field; during the day with an IR filter one might reach 6.7m stars, typically 1.6 degrees to a star of this brightness for 1.5" seeing. Questions include how well the camera can hold to the radio axis-observations of stars with SiO masers would help to determine this. Woody noted that the wet component of pointing can be non-trivial-10"-but that the dry site should help here.

Wright reviewed his draft memo on VLBI-related calibrations noting that it could get ten microarcsecond resolution enabling ALMA to see a flea in Australia; sensitivity could be improved by 20x. ALMA would be well placed for inclusion in an array. Four or more antennas should be included. We'd need to correlate the phased array with a reference antenna, provide the same sense polarization as the rest of the array, use the local correlator without phase switching and provide the appropriate number of summing circuits (two for the polarizations, and n for the n subarrays).

Hills reviewed the 183 Ghz WVR options. The goal is to measure ten microns of vapor; at 183 Ghz we should measure about 1 K per 100 microns of path, hence we need to measure to better than 0.1 K. Problems in doing so include temperature and pressure profiles as a function of height, beam alignment, and the contribution of the dry component,which could amount to 50 microns rms on a 1km baseline. The WVR is the remedy of choice for phase correction on smaller scales. At some scale, fast switching remedies phase fluctuations also. Lastly, of course, clouds pose problems-water clouds very much so, ice clouds less so. The instrument now on the site suffers as we compare the signal from the horns, sensitive to atmospheric emissivity, with the signal from the 12 Ghz interferometer, sensitive to coherent amplitude as the satellite signal is modified by passage through the atmosphere.

Vaccari presented the photonic calibration system. The radiator at the subreflector caused some concern. In its first implementation, this system would be used primarily to correct for polarization but many other calibration uses have been proposed.

Holdaway presented his new memo evaluating fast switching on ALMA antennas. He reminded us that the site statistics suggest that the phase variations will be withing 20 degrees at 300 Ghz on 300m baselines only 20% of the time. Tradeoffs between fast switching and WVR for phase corrections need to be studied for the observer trying to decide how to obtain the best images.

Matsuo proposed using total power at 450 Ghz to provide phase correction, in the fashion employed at IRAM now. Advantages include observations through the same patch of sky. The system would require good weather though, and simultaneous dual band operation.

Pardo described simulations of phase corrections and transfer from one frequency to another using a DEMIRM Array Simulator developed by Viallefond, Badia and Pardo. Here one can simulate fast switching and radiometric phase correction for evaluation. This is a very interesting machine, under construction.

Holdaway presented his third new memo, on dispersive phase effects. These should not pose a large challenge.

The committee met after recess was called for the day and before the presentations continued on Friday 22 June.

Woody discussed WVR-based radio isplanatic patch measurements. How should one optimize the use of the fast phase calibration and WVR data? He noted that the former is analogous to an optical guide star. The critical parameter is the isoplanatic angle-that angle over which the sky is planar and coherent. This can be estimated from the structure function and the altitude of the screen. At OVRO, measurements of the altitude have been tried and sometimes work, breaking down when multiple layers exist, for example. One could compare phase power versus time lag-which depends upon the orientation of the baseline to the wind direction at altitude. With many telescopes, this could be measured at the expense of lost telescope time. If water is the source of the delay, a single WVR might perform this measurement, measuring as a function of offset and time. Woody proposed such an instrument, scanning conically at constant elevation over a period of about 10s. This instrument could serve as an indicator whether to use fast switching or WVR. Such instruments, when available, often prove indispensible-note for instance the crutch the CSO tipper provides many observatories on Mauna Kea.

Woody then discussed the OVRO WVR. At 22 Ghz, two continuum channels straddling the line are compared to the line channel. Cooled 15 K amplifiers can get to the needed 3mK sensitivity in a few to tens of seconds. The scale factor is 10mm/K, constant to 10%. On long baselines at OVRO, the WVR always improves the data by some amount 5-40%. It typically gets to a measurement of 100-200 microns, good enough for 3mm work but not quite for 1.3mm. The path to images involves lots of astronomer intervention. The system needs further evaluation of the cooled systems, and software effort.

Wright discussed the 22 Ghz swept frequency system at BIMA. About half the time it improves but rarely to better than twice the expected level. Experience suggests it follows broad fluctuations well but smaller ones less well. Discussion centered on whether small scale fluctuations exist at a larger than expected level. Woody was optimistic, feeling that physics was right and his instrument wasn't up to it. Welch noted that a Kolmogorov spectrum is a grand average and may not apply at a given instant. Schilke noted that some fluctuations on small scales seemed to be needed to explain bolometer array sky noise.

Lucas described the WVR system planned for use at IRAM. Three 1.1 Ghz channels at 19, 22 and 25.2 Ghz are employed; three radiometers (uncooled) are in the lab in the system being developed by Mattioco and Bremer.

Sault presented the 22 Ghz system planned for the ATCA-3 3mm baselines should be available imminently. This system employs four channels, uncooled, between 16 and 25 Ghz. The WVR is 10' offset to be used with 3mm and Kband horns. Limitation always seems to be systematics of various sources at the moment.

Wiedner brought the state of 183 Ghz radiometry up to date with a discussion of her work Although the instrument will be improved to meet ALMA specs, already observations show it is close, consistent with expectations.

Avery described IRMA I and II, which operate in a band at 462-505 cm-1 covering about two dozen water lines, some saturated and some not below 1.5mm pwv; all are saturated above that level. IRMA I was deployed at the JCMT in 1999 and showed that it was most sensitive for <1mm pwv, and could achieve measurements of 10 microns at 1mm pwv. IRMA II was deployed August 2000 to achieve 5-10x better stability, which it realized. It has been compared to SCUBA skydips at the telescope; it is designed to match the beam to the 183 GHZ radiometer at 500m altitude. It provides 10 Hz sampling and was discussed in conjunction with a tip tilt implementation to provide anomalous refraction corrections. Questions which remain are 1) what are the consequences of divergent beams? 2) How important are cirrus and water droplets? The plan is to deploy two IRMA III devices to the site in February 2002 for phase measurements. The actual H2O lines will also be modeled. Also multibands may be employed to separate continuum from lines. This may be useful for a measurement of the isoplanatic patch also. Brad Gom took over to discuss IRMA III.. He noted that 20 microns is near the Planck function peak for atmospheric temperatures, maximizing sensitivity, and IRMA of course emits no RFI. It will sit in a closed cycle 77K cooler, mechanically sealed and built robustly and toughly. The size will be no more than a sheet of legal paper with an 8" height, so it could be placed in many positions on the telescope unobtrusively. Two might be cross correlated to provide an estimate of the height of the turbulent layer; it is planned to use IRMA to correct SCUBA data. The group thought that the use of this instrument for correction of anomalous refraction held some promise. Its stability is 1 part in 10000 in 10 minutes. Hills thought that to get the information from a cone one might have to assume a single turbulent layer at a known height.

Welch described an atmospheric temperature sounder operating at 60 Ghz on the edge of the oxygen band. By frequency sweeping or tipping one can produce T(h) through the atmosphere; this can be bought now off of the shelf at Radiometrics Inc. for $150K to provide 1 K sensitivity below 2km.

Matsuo described an FTS and all sky monitor. The latter operates at ten microns and images the whole sky (11-65 zenith angle) in a weatherproof maintenance free package at night only. It is uncooled; fabrication for ASTE is being studied. He also described the wide band fast scan low resolution multibeam FTS to provide T(h) but not to the accuracy mentioned above (5K here). Operating 100 Ghz - 1,5THz in a 2 s scan it provides a resolution of 3 Ghz with a 10 degree beam. Several horns can provide 3-5 beams. It is cooled using a 4K pulse tube cooler and is now just in a proposal stage.

Guilloteau summarized ALMA and ACA calibration in the table above, based on Memo No. 372. For ACA pointing above, it is assumed in the first case the calibrator must be within 7 degrees; for the second, 4 degrees (hence fainter), and that the strongest available QSO is used. The ACA could use stronger sources. Wright noted that borrowing all array antennas cut the integration time. Hills noted that submillimeter sources may provide some calibrators. Welch noted that what one needed was good day to day stability, for which planets are suitable. He noted that the ACA antennas may be the best performing antennas also. Matsuo noted that FIRST, PLANCK etc will provide a source list also. Calibrating the ACA needs further thought.

Hills then reviewed WVR plans for the new devices to be finished by the end of next year.

Kohno presented a review of the ASTE plans.

Hills closed the conference with his sketch of an anomalous refraction corrector, a large rotating mirror illuminating half of the telescope. The output is the slope of the incoming wave, of the pointing correction. The uncooled 183 Ghz WVR provides almost enough SNR for this.