Four Level 4 Milestones for Fast Switching Tests ================================================ Mark Holdaway January 30, 2003 February 2003 0. Plan Prototype Antenna Tests of Fast Switching (This document.) July 2003 I. Perform Optical Pointing Telescope Tests of Fast Switching with the Prototype Antenna November 2003 II. Verify Fast Switching Functionality with Radiometric Single Dish Tests on the Prototype Antenna June 2004 III. Verify Fast Switching Interferometrically with Vertex and A/C/E Antennas The ALMA prototype antenna has detailed specifications designed to ensure that fast switching phase calibration is possible. In order to perform fast switching phase calibration, several conditions must be met: - The antenna must be able to accelerate and slew quickly - The antenna control and servo system must move the antenna smoothly so as to not excite vibrations in the antenna; or the antenna must settle down quickly. This and the previous requirement are summarized by the specification that the antennas must be able to move 1.5 degrees in 1.5 seconds with a residual pointing error of no more than 3~arcseconds. - The antennas need to maintain phase stability between themselves in time, going from the calibrator source to the target source - The antennas must detect the calibrator source with sufficient SNR to solve for the phase quickly. I. Perform Optical Pointing Telescope Tests of Fast Switching with the Prototype Antenna Radiometric tests of the prototype antennas will not occur until late spring, and poor atmospheric conditions will realistically push most radiometric tests into the autumn. Fortunately, we can perform a fair amount of the fast switching testing with the optical pointing telescope. The pointing derived from the optical telescope embedded in the backup structure will not be equal to the radiometric pointing. Effects dealing with the deformations of the backup structure, the feed legs, or the subreflector will affect the radio pointing but generally will not affect the optical pointing. However, the optical pointing telescope will reflect any pointing changes due to elements from the foundation up to the backup structure, including most importantly, the AZ and EL drives. Just because the optical pointing telescope indicates that fast switching appears to work doesn't mean it does work, but if the optical pointing telescope indicates a problem, that problem will be real. It is a necessary, but not a sufficient condition for fast switching success. The optical pointing telescope is able to measure the pointing with an accuracy of about 0.1 arcseconds at a rate of 20 Hz (though the 20 Hz rate has not been verified yet -- there may be a bottleneck somewhere in the system that prohibits this high time sampling). Radio pointing comes nowhere near this level of accuracy or this level of speed. The optical pointing telescope has a dynamic range problem and saturates for stars brighter than 5th magnitude. Over the entire sky, there are 600 pairs of stars between 5th and 7th magnitude which are between 15~arcminutes and 2~degrees apart from each other. Test I.A Optical verification servo system, slewing and settle down time for fast switching The basic specification on the prototype antennas is to slew to a source 1.5 degrees away and settle down to a 3~arcseond pointing error in 1.5~s. Fast switching in both azimuth and elevation axes should be tested. Optical verification should be straightforward. The fast frame rate and the high accuracy of the pointing measurements should be valuable in studying any problems or odd vibrations in the telescope. Test I.B Verify Fast Switching Performance Across Sky The antenna's fast switching performance is expected to depend upon elevation angle. Once we have verified that fast switching works, we can study how the slewing and settle down varies with elevation. Fast switching in both azimuth and elevation axes should be tested. Test I.C Build Fast Switching Model In order to more accurately calculate the efficiency of fast switching, it is important to develop a model for the antenna's slew and settle times as a function of: - source elevation - distance between calibrator and target sources - direction of slew (AZ or EL direction) - slew velocity Furthermore, with the addition of the sub-millimeter wavelengths, we will be interested in fast switching with a smaller pointing error than 3~arcseconds (ie, a longer settle time). It would be good to develop switching models which included a variety of end pointing errors. II. Verify Fast Switching Functionality with Radiometric Total Power Tests on the Prototype Antenna Once radiometric observations have been debugged, we can move to the next stage of fast switching testing. The radiometric pointing does not necessarily follow the optical pointing: pointing fluctuations due to feed leg motion, deformation of the backup structure, or subreflector motion will result in discrepancies between the optical and radio pointing. Turned around, we can use simultaneous radiometric and optical pointing observations to constrain where in the antenna the pointing fluctuations reside. The trick here will be to find a bright enough radio source with a bright enough optical source nearby. Planets would seem to be a good candidate, but they cannot be used optically as they result in saturation, and some modeling may be required to convert the flux variations of the beam wobbling over the extended planet into pointing variations. Radiometric fast switching tests can be made for any bright quasar or quasar pair. We don't really need a pair, as we can switch between a blank field and a quasar. We would actually specify a target position which placed the quasar at the half power point of the beam (offset in either AZ or EL) and then use variations in flux to infer pointing variations so we can determine how the antenna pointing is settling down at the end of the fast switching slew cycle. We will need to perform some sort of beam switching, and systematics in the beam switching over the short time scales we seek to measure may very well make this experiment unworkable. If we point the telescope at the half-power point in AZ to verify the pointing and antenna settle down, we could switch against the opposite half-power point. By takaing "ON" (ie, half power point on the left side) minus "OFF" (ie, half power point on the right side), any AZ pointing errors would manifest themselves as a positive increment from half power on one side and as a decrement from half power on the other, so ON - OFF will result in the atmosphere and the half source power cancelling, leaving only the AZ pointing signature. Since the subreflector nutates only in AZ, we cannot use this technique to sort out the EL component of the pointing errors. The brightest quasars at 90 GHz are usually like 10-20 Jy, or 5-10 Jy at the half power point. The system noise for the 3mm evaluation system is assumed to be about 80 mJy in 1~second. Hence, the peak SNR on the brightest quasars observed at half power would be like 60:1 to 120:1 in 1~second. At the half power point, the primary bea, gradient is about 0.025 per arcsecond at 90 GHz. So, if we observed a 20~Jy quasar, 1 arcsecond pointing variations in the direction of the beam gradient would result in power fluctuations of 0.5~Jy, or 3 sigma flux variations would permit the inference of 0.5~arcsecond pointing fluctuations. If we needed higher time resolution, we could get a 3 sigma detection of a 1~arcsecond pointing variation in about 0.25~seconds. So, with the very brightest quasars, we are able to get to interesting time scales and interesting pointing errors if we can actually realize the thermal noise limit. III. Verify Fast Switching Interferometrically with Vertex and A/C/E Antennas Assuming the test interferometer actually comes into being, we would like to verify that the phase determined on one quasar is actually applicable to a second quasar. To achieve this experiment, we actually need two bright quasars close to each other. We can probably find a pair of quasars that are about 500 Jy each at 90~GHz and about 3 or 4 degrees apart. (There are somewhat brighter source pairs at 8~GHz; at 90~GHz, they may be somewhat weaker.) Is this a reasonable experiment to perform? Typical fast switching at the ALMA will have cycle times of about 10 to 30 seconds, spending about 3 seconds on the slewing and detection of the calibrator. The target source need not be detected in a single switching cycle. However, in our test to verify fast switching, both sources need to be calibrators (to test the phase determined on one source on the other source), so we need to detect both sources in the switching cycle. Lets take a 35 second cycle. We assume 3 seconds is lost slewing back and forth, then we'll sit on each source for 16 seconds. The 1~sigma noise at 90~GHz for the interferometer will be like 20~mJy. With a 500~Jy source, 20~mJy corresponds to about 2 degrees. Hence, our preliminary estimates indicate that a full radiometric test of fast switching phase transfer should be possible. Combining data from the radio phase monitor, we will be able to predict the residual phase errors incurred while calibrating one source with the other source. This will not really be indicative of the ALMA fast switching case, as we will use all antennas in the gain solutions, but this single interferometer case will serve to test the antennas for phase stability and phase transfer as well as our understanding of the residual phase errors in fast switching phase calibration. Furthermore, we may find that it is impossible to perform the settle down test radiometrically with a single dish (test II above) due to the very short time scale beam switching systematic errors. Interferometrically, we do not have to do any sort of beam switching, and the flux fluctuations will lead to a superior inference of pointing errors during the settle down phase of the switching cycle. =========================================================================