Version vii 2000-08-02

Minutes of Holography Teleconference

2000-07-25

Minutes by Darrel Emerson, Jaap Baars & others

The meeting took place on Tuesday, July 25 2000, at 15:00 UTC.

Participants:

NRAO

ESO

IRAM

UK, MRAO Cambridge:

In advance of the meeting, participants were asked to review the following material:

1. The April 1999 Holography PDR material (HTML)

2. Chapter 5 of the ALMA Test Interferometer Project Book. (PDF)

Darrel Emerson apologized that some of this material had become out of date, but nevertheless was the best starting point in preparation for the meeting.

Agenda:

Meeting to discuss ALMA holography plans, primarily, but not exclusively, for the test interferometer.

1. Overview

Darrel Emerson gave an overview of plans, with reference to the April 1999 PDR, and the material in the Test Interferometer Project Book. Single dish holography using a coherent transmitter, with a frequency around 90 GHz, will be carried out on the prototype antennas soon after they are available at the VLA site. The holography receiver will be mounted at the prime focus, with a reference feed looking away from the dish built into the same package. The holography backend will perform realtime complex correlations, and these data, with calibration data and telescope coordinates will be passed to the realtime control system, at a rate of one set of about 10 numbers about once every 10 ms. The goal is to be able to complete a 256 x 256 on-the-fly raster map in significantly less than one hour. As recommended at the PDR, the specification is a surface measurement precision of 10 microns, but with a goal of 5 microns.

The same holography system will probably be used later in San Pedro in Chile for the initial adjustment of antennas arriving there. It is not clear whether this holography system would ever be used at the high site at Chajnantor.

Although the initial holographic measurements would be in single dish mode with the special holography receiver, at a later stage of the antenna evaluation, when reliable interferometric electronics and software was in place, further analysis would use interferometric holography with the 2 prototype antennas. These interferometric measurements would make use both of the terrestrial transmitter, and astronomical sources.

The transmitter will be mounted on a tower of 50 m height, some 300m from the antenna at the VLA site.

In the present planning, the holography receiver and backend will be built by NRAO. Europe is responsible for the holography tower, although it may be procurred within the US. Responsibility for the holography transmitter currently rests with Europe. Discussions for its production are underway with prospective instutes.

Michel Guelin questioned whether there really was a need for single-dish holography, given that 2 prototype antennas would be available in close succession. The reply was "yes," because in part of the tight schedule; it will be important to test the antennas as soon as possible after delivery to NRAO, which would be long before the necessary interferometer electronics would be available. The s/n on astronomical sources in interferometric mode would be good enough to map relatively large scale deformations of the dish, but would not give the desired signal to noise at very high resolution (e.g. 10 cm) on the surface of the dish; this resolution is helpful for accurate adjustment of individual panels. The high s/n available with the terrestrial transmitter would also allow some study of variations in dish properties with changes in ambient conditions such as temperature and wind, and to study the difference between daytime and night-time performance.

2. Site, tower:

Peter Napier summarised the arrangements at the VLA site, and for the transmitter tower. Details are available in Chapter 12 of the Test Interferometer Project Book. The tower is currently planned to be 50m high, at a range of 300 m from the antenna. The location and height are tradeoffs between the difficulty of the near-field correction, transmitter locational stability at the top of the tower, and having a reasonable elevation angle so that adequate size maps can be made without undue multipath interference.

Richard Hills reminded the group of the desirability of refocusing the telescope, to mitigate some of the near field difficulty. Michel Guelin mentioned the apparent ground reflection problems that had been encountered with the IRAM 30 Meter Telescope.
[See the Notes on Near Field Measurements from Richard Hills, received after the meeting, on the subject of near field measurements; these are attached at the end of these minutes.]

3. Transmitter, Receiver, Backend: (Some of the issues were already covered in the overview: see above.)

The issue of multiple frequency holography observations, or of swept frequency observations, was discussed at some length. The motivations are

For multipath effects, frequency sweeps or frequency steps of some tens of MHz are appropriate, while to study diffraction effects 2 or more frequencies differing by tens of per cent are needed. There is not at the moment a strong case for either, but someone should think about the implications and details. More complex schemes might require more complex communication with the transmitter on the tower, for example. For the moment, we accept as baseline a tuning range of about 100 MHz.

The most critical aspect of the holography is believed to be the measurement of the phase response of the feed used to illuminate the dish. Matt Carter agreed to look into whether the necessary 5 micron accuracy could be achieved on the IRAM test range.

4. Realtime software.

Mick Brooks and Brian Glendenning reported. ICDs for all parts need to be established, preliminary by October 2000, final by January 2001.

The Computing Groups plans at this stage are to use the M&C CAN bus for controlling the holography back-end and also for extracting the cross-correlated results from the back-end. There may, in fact, be two CAN nodes: one for the back-end and one for the front-end TBD. The interface type for M&C of the transmitter is also TBD. The distance to the tower (300m) is beyond a reasonable CAN run, so other alternatives need to be sought. It was noted that there is no perceived need for real-time synchronization between the transmitter and the back-end.

It was agreed that the transmitter should be remotely controllable; this control should be kept as simple as possible and could be held separate from the receiver M&C system. Larry D'Addario will summarize the necessary M&C functions for the transmitter. The question was asked whether alternate mapping schemes, other than on-the-fly raster scanning, were envisaged. If so, the software group needs details of this.

It was agreed that the baseline plan should be to support only conventional raster scanning. There was some interest in studying the map scanning issue further; for example, the rotating cloverleaf has a boresight reading every cycle, avoiding the need to start and stop the antenna to go to a boresight calibration every few rows. However, this is an "icing on the cake" issue and not a requirement.

5. Data Analysis.

IRAM has the responsibility for this. Robert Lucas reported that the plan was to base the ALMA holography analysis software on the existing Plateau de Bure system, with such additions as may be necessary for the single dish analysis, and in particular to give the near field corrections. A report of the planned software capabilities will be available in September 2000.

6. Choice of precise frequency or frequencies, and legal niceties.

Peter Napier reported:
The NTIA is the responsible body for frequency allocations in our case. There is considerable leeway for "Experimental Stations." Government Experimental Stations, of which the VLA is a designated site, can use any frequency without prior approval, subject to non-interference with other authorised service. This is not the case within the Radio Astronomy bands where transmission is not allowed, except that even in this case a low level of emission is permitted for the measurement of antenna characteristics.

Although it seems we can legally make our holography measurements with the terrestrial transmitter within a Radio Astronomy band, this would set a dangerous precedent for future protections of our bands from other services. Accordingly, this group strongly recommends that the holography frequency be chosen OUTSIDE of the approved Radio Astronomy bands. The University of Chile recommends that all negotiations about frequency allocations in Chile be made through ESO.

Morita reported on the Japanese progress:
The Japanese group will start holographic measurement of the new ASTE telescope at Nobeyama soon. Frequencies chosen for the experiments are 92 GHz and 105 GHz. For the holography experiments in Chile, contact will be made with SUBTEL through the University of Chile, although the Japanese are in contact also with ESO for the issue of frequency protection in the Atacama area.

The participants of this teleconference agreed that all contact with the Chilean authorities, from the US, Europe, or Japan, should go through ESO. The group again made the strong recommendation that, in order to avoid setting a dangerous precedent, and also to avoid mutual interference with other existing or future telescopes at the Chilean site, all holography beacon frequencies used in Chile should stay outside the protected radio astronomy bands.

7. Action items:

Appendix I: Notes from Richard Hills on near field measurements.

[These were received after the teleconference, so do not form part of the official minutes of the meeting, but are included here for convenient reference.] 

Subject: Near-field corrections
Date: Thu, 27 Jul 2000 15:00:17 +0100
From: Richard Hills 

...

One point from the telecon that I thought should perhaps be clarified was on the question of how close you can bring the source. I went through this in a JCMT memo (I regret unpublished) in circa 1986 and I think Dave Morris did too.

If you want to use Fourier transforms you can take out terms that depend only on the position in the antenna aperture. The first of these is the spherical term - r^2/(2*D) - where r is the radius in the aperture and D is the distance to the source. If you do only that, you are making what is, I think, called the Fresnel approximation. The size of this correction does indeed depend on dish diameter squared over source distance, as you were quoting. What is important however is the magnitude of the terms left out.

The next term in the relevant expansion is of the form r^4/(8*D^3), which corresponds to spherical aberration. You will see that for r = 6 m and D = 300 m this is 6 microns which is just significant for ALMA (for JCMT it is 1 micron). This term can in fact still be applied as an additional correction in the Fourier transform.

There is however another term that depends on position in the map as well as position in the aperture (i.e. u,v as well as x,y). The worst case amplitude of I think this comes out to be about (N^2*w^2)/(32*D) where N is the number of points across the map and w is the wavelength. (The wavelength comes in via the sampling interval.) For this one I get about 18 microns for N = 128, w =0.0032 and D = 300m. This can't be corrected if you use a Fourier transform. There is in fact a good deal of cancellation in real maps however and it is probably still just about OK at the level we need (numerical errors below, say, 3 microns rms). It might however be a problem for larger maps - N = 256, say. This, together with other factors like the obliquity factors and the depth of the dish, can be checked fairly easily by numerical comparisions of Fourier transform and full physical optics calculations. These should certainly be done.

Given that we can if necessary do the full calculations, then none of this is a real restriction on the minimum distance. The real problem is that in the Fresnel region the beam is spread out over a substantial diameter and one has to map all of this plus the surrounding diffraction pattern to get all the information and not distort the aperture map through truncation.

If the dish is not re-focussed then it is easy to see that this leads to a limit on the minimum number of points in the map of about d^2/(D*w), with d = 12 m (dish diameter) this is ~150 and you still have to add on the number of points required to give the desried resolution (perhaps in quadrature?). If you can re-focus it then this gets much better. How well you can do depends on the f-ratio of the dish primary and the distance to the source (both to high powers). Correction for the effects of truncation is possible by means of numerical modelling of the process but this gets somewhat complicated.

Having gone through all this I am reminded that we did worry about most of it at the time of the previous telecon and I did send a fax with some diagrams, which I have now found. These showed that truncation was significant on a 64 point map at 300 metres even after doing the best job on re-focussing, but it would be pretty minor on a 128 point map.

So the bottom line is that the choice of distance is trade-off involving several factors. The following favour large distance:

  • simplicity in processing data
  • can make small maps to check for large scale errors quickly (e.g. thermal effects)
these others favour short ones:
  • minimizing atmospheric propogation disturbances, especially near the ground
  • minimizing ground reflections
  • allowing large maps without running out of elevation range.

What have I left out?

If we had lots of time to make detailed studies we could no doubt find out more about these things and make an informed choice. Presumably we don't (and it is a bit of a shame really given that more than a year has gone by since we first talked about this, but I do realise that we are all stretched). I feel therefore that 300 metres is a reasonably conservative choice to make. You might be able to come down to 200 metres if Robert feels he can handle the modelling and you don't think that quick mapping is important.

Best Richard ...

Appendix II: Notes from Dave Morris & Robert Lucas

[These notes do not form part of the minutes of the teleconference, but are included here for convenience.] 


Subject: Holography of ALMA antennas
Date: Mon, 31 Jul 2000 19:04:15 +0200 (METDST)
From: morris@iram.fr (Dave Morris)
To: demerson@nrao.edu
CC: lucas@iram.fr, guelin@iram.fr, carter@iram.grenet.fr, morris@iram.fr,
     guillote@iram.fr

Dear Darrel,
                Here are some of our thoughts after the Teleconference last
week. The numbers really need checking !
                                        Dave

    ---------------------------------------------------------------------
                                HOLOGRAPHY OF ALMA ANTENNAS

 GOAL

  Root mean square surface measurement error 5 microns ( 1 degree phase
 error at a wavelength lambda of 3.3mm ).
 By near field phase coherent holography in 3mm band on a flat test range at
 VLA site.
 At prime focus with forward looking reference feed.

 GEOMETRY

 Diameter (D)= 12m
 Focal length (F)= 4.2m (F/D=0.35)
 Height of axes , about 13m
 Transmitter distance (R)= 300m, height (H)= 50m, elevation angle (E)= 8 degrees

 Angle subtended by data window (W) for N=128 pixel map at lambda/Diameter
 sampling= 2 degrees
 Focal motion for refocussing=F**2/R=59mm

  Path difference for ground reflections, for height h at antenna = 2*h*H/R = h/3
 therefore for height range 1-13m , path difference range= 0.3 - 4.3m.

  Frequency change for 180 degree phase difference, range 35-450 Mhz

  Distance of reflection point from antenna = h*R/(2*h + H), range 6-78m

 NEAR FIELD CORRECTIONS

 Assuming transformation from near field to aperture by FFT and Fresnel
 correction and with perfect refocussing.

  For a N= 128 pixel map Peak error= (N*lambda)**2/32/R
                       = 18 microns critical sampling, surface resolution=10cm
                       =  6 microns oversampled 2x,    surface resolution=20cm

 Initial simulations suggest that the error is confined to the edge region of
aperture and the rms will be much smaller.

 GROUND REFLECTIONS

  Will superpose a second distribution in the aperture plane with a strong phase
 gradient.
 Effect is a series of fast phase ripples, unresolved if the diameter of the
 near field data window is smaller than 4 times the elevation of the transmitter.
  If peak phase error is <= 1 degree (surface error 5 microns), the relative
 amplitude of the reflected signal <= 0.019 or in power -34 dB.
  Possible reduction of ripple (N=128 critically sampled map)
Reduction

  a) Reflection coefficient of the ground                              a few?
    - surface should be rough on wavelength scale.
  b) Polarization discrimination
    - circular polarization should be used (perhaps 5% residual)        -13dB
  c) Smoothing due to residual unresolved ripples <= W/4/E =1/32        -30dB
  d) Attenuation by transmitter feed polar diagram                        0dB
    - nearest point of reflection to telescope is at 6m
  e) Frequency scanning                                                 a few ?

                                                                Total <= -43dB ?


 ANGLE OF ARRIVAL FLUCTUATIONS phi (Seeing)

  The rms surface error is about 0.125*D*phi. For 5 micron surface error
 phi= 0".7 - probably possible over a 300m range at night.

 LINEAR MOTION OF TRANSMITTER

  <= R*phi = 1mm vibration in wind for example

 ASTIGMATISM

 Residual astigmatism of prime focus feed can be cancelled if receiver feed,
 or package, can be rotated 90 degrees.
 Less important for transmitter feed.

 PRIME FOCUS FEED

 Phase response needed to 1 degree at 3mm. Designed for small taper (difficult ?).

 SUMMARY

 The peak errors in the near field correction, using FFT and simple Fresnel term,
are predicted to be significant if high surface resolution  (<20cm) is needed, but
not a disaster. They can be removed by replacing the FFT by a "slow"
integration of the diffraction integral, for example in an itierative scheme.
 Ground reflections may be a limiting factor, particularly if the arguement based
on the "smoothing" is not justified. Some form of frequency scanning may be
needed.

 SUGGESTIONS

 Circular polarization be used and rotation of the prime focus feed be possible.
 Tests of additional reduction methods should be made.

                                                                D.Morris
                                                                R.Lucas
                                                                29 July 2000