My concerns with ngVLA overview article in SPIE

Section 2.1: Performance

• Very high efficiencies (up to 0.86) are claimed for the lower bands where Ruze efficiency is inconsequential. I wonder if diffractive losses have been accounted for? Even if this high value is to be believed, the efficiency of 0.60 quoted for 93 GHz is inconsistently high, because the Ruze efficiency there is already 67.8%, and 0.86 times that value is 0.58. A more realistic value would be 0.54 (i.e. 0.8*Ruze), which means the total efficiency will be only 0.44 at the CO 1-0 line. That will lead some readers to underestimate the exposure time for CO science cases (if they use the center of band value) by almost a factor of 2! because (0.60/0.44)**2 = 1.86
  • RS (1/16/18): We have a good backup for these efficiency and T_sys estimates. Wes has a model of the front ends including all the appropriate efficiency factors with the shaped optical system. We also have an optical model in GRASP that has substantiated these estimates with additional analysis. Illumination efficiency is high - of order 0.95 for the optics and feeds, less Ruze losses and so on. Files here:
    • \\filehost\ngvla\Techdocs\30-Antenna_Electronics\05-Front_End\Band Configurations
    • \\filehost\ngvla\Techdocs\30-Antenna_Electronics\05-Front_End\Cascade Analysis
    • /home/ngvla/Techdocs/25-Antenna/NRC-Optics-Study/Report

Section 6.1: Calibrator availability

• "There should be a 25 mJy calibrator source [at 3mm] within 2° in 98% of observed fields, ensuring short slews." - This statement is rather optimistic, and is based on the estimate of 20,000 QSOs > 25 mJy in the whole sky (NGVLA memo 1, section 5). ALMA recently analyzed this topic (SCIREQ-1366), at 87 and 345 GHz. At 87 GHz, only 64% of the sky has a catalog source within 2 degree radius. The percentage rises to 87% if you increase to a 3 degree radius. Indeed, we often have to go out to 5 degrees at Band 3. The figures below were made by Akihiko Hirota (left) and Satoko Takahashi (right).

• • RS (1/16/2019): I'll defer to Carilli, Butler, etc., to substantiate these figures (or correct them). The good news is that the referenced pointing specification for the antenna, and the fast slew/settle spec are all based on 4-deg slews.
    • \\filehost\ngvla\Techdocs\25-Antenna\Antenna_Spec
  • TH (1/24/2019): I just checked the ALMA catalog and it currently has 11271 unique QSOs in it (well, objects starting with 'J', but there is some contamination of Galactic objects, since NGC7027 also has a J name). 3300 objects have ALMA measurements, for an approximate rate of 28%. No doubt, all of the bright ones are in this 28%, but we need to examine the cm flux density distribution of the remaining ones to estimate how many more might qualify at 3mm.
  • A. Hirota (1/26/2019): I checked the catalog of ATCA 20 GHz blind survey, which covers most of the sky are with declination below 0. According to the catalog description, it contains 5890 sources with the flux density limit of >40 mJy at 20 GHz. By assuming that 25 mJy at 90 GHz is the minimum flux density for an ALMA calibrator and also crudely assuming that all sources in the ATCA catalog have the spectral index of -0.7, QSOs brighter than 72 mJy at 20 GHz might considered to be candidates for ALMA calibrators. When I filtered the ATCA 20 GHz sources with the flux density threshold of 72 mJy, I got 4201 sources. If scaled to dec < 40 deg, then, N(>72 mJy@20GHz) is about 6900. Therefore, I tend to assume that about half (3300/6900) of the potentially usable QSOs have already measured with ALMA. The effort for increasing the number of sources with ALMA measurements is currently tracked in CSV-3123.

Section 6.2: Short Baseline Array (SBA) configuration

• It says the minimum baseline is 11 m for 6 m primaries. This rather large value is set by collision avoidance for the offset reflector design, but does it really need to be this large? The ratio (11/6=1.83) is inconsistently larger than for the 18m antenna whose minimum baseline is 30 m (30/18 = 1.67). Indeed, ngVLA memo 43 says the ratio only needs to be as large as 1.75. Every bit counts in this application. By comparison, ALMA's ACA minimum unprojected baseline is only 8.9 m for 7m primaries for a ratio of 1.27.
  • RS (1/16/18): The spec was for collision avoidance, which we can reconsider. Turns out it's proving hard to even get to 11m - present design is closer to 14m w/o risk of collision in any orientation. The reason for this being harder than the 18m is that the optics become more offset since the subreflector size can only shrink marginally. We can consider feed high, but the alternative then becomes more shadowing than an equivalent feed low design, so there is trade-space between minimum spacing vs shadowing angle. Memo in progress documenting this (Dunbar et. al. on the hook).
• A fixed configuration of 19 antennas will yield elliptical beams and heavy shadowing in the south. Are we considering making a handful of 6m antennas to be reconfigurable to a N/S elongated second configuration, like ALMA built for the ACA?
  • RS (1/16/18): B. Mason did a fixed configuration, but did spread them out in the N-S axis to partially mitigate shadowing and likely declination of sources. He's a good person to discuss this further with. You might also loop in Carilli and V. Rosero.

Section 6.3: Array Calibration

• "The system will model opacity based on barometric pressure and temperature monitored at the array core and each outlying station." - Barometric pressure and temperature can only provide the dry air opacity. The PWV will need to be modeled from the WVR data and added to the dry air term to get total opacity.
  • RS (1/16/18): WVR currently planned for phase cal and we can estimate PWV from it. Humidity sensors on weather stations are also a source of info, though at only one elevation. Good topic for the calibration group (Butler, Hales)

Section 6.4: Antenna

• Pointing error is mentioned, but dynamic tracking error is not. This needs to be a few times smaller than the DC pointing error.
  • RS (1/16/18): We treat 'pointing' as 'tracking' since nothing is actually static. We should clarify this though in the specifications. We should put this on Dunbar's radar (he's the owner of that doc nowadays)
• "...the optics will be shaped to optimize the illumination pattern of single pixel feeds" -- If both primary and secondary will be shaped, then the commensal low-frequency prime focus feed will be aberrated, so this needs to be taken into account for sensitivity and imaging performance.
  • RS (1/16/18): two points: (1) any commensal LF system is secondary to efficiency at secondary focus. (2) Defocus efficiency losses will scale with wavelength of course. This is a very similar trade-off to what's factor in to the P-band and 4-band systems on VLA (which is also a shaped system. It's actually worse for VLA since you can't place the commensal feeds at the apparent focus due to the Cassegrain optics.
• "Both performance and maintenance requirements favor antenna optical configurations where the feed support arm is on the “low side” of the reflector." - Sri Srikanth's 1989 GBT Memo 16 clearly shows (Fig 6) that the low side arm yields worse spillover for elevations < 30 deg. What new analysis claims otherwise?
  • RS (1/16/18): You can find the tipping curves in Lynn's optics report, with a supporting GRASP model (location below, Figures 10 and 19). Two clarifications - one, it's slightly better in Feed Hi at most frequencies, but the difference is less than 1K. over most elevations. It gets murky because you can improve on feed low with a spillover shield. The short version is that it's almost irrelevant with this particular optical design and at our operating frequencies. It's mostly a trade off between structural factors (favor hi), maintenace factors (favor low), and minimum spacing (favors high) / shadowing (favors low) ....
    • \\filehost\ngvla\Techdocs\25-Antenna\NRC-Optics-Study\Report