The calibration session at ALMA Week 2003 was very productive, giving 
a snapshot of our current thinking on the major items of the 
calibration plan and some discussion on still unresolved issues.
I give here an outline of the talks and discussions which occurred.

Based on the discussions preceding and during ALMA week, and 
specifically during this session, we derive the following calibration 
requirements for the main calibration components of ALMA:

 - Amplitude calibration to 1% at f<300 GHz; 3% at f>300 GHz has
   been specified, which implies:
   . A calibration system in front of the receivers to track 
     fluctuations in system temperature.  This system must allow 
     the receivers to be coupled with the following input signals:
     o "hot" load
     o "ambient" load
     o sky
     o combination of 50 % "hot" and 50 % sky
     o combination of 50 % "ambient" and 50 % sky
     The temperatures of the loads remain TBD in detail, but the 
     temperature of the "hot" load should be 370 K or above, and the 
     temperature of the "ambient" load should be 283 K or lower.  This 
     calibration system must apply to all receiver bands except band 1.
     Different technical solutions exist for the combinations of 
     loads (semi-transparent vane; wire grid; dielectric beamsplitter),
     but we cannot choose between them at this time.  We can abandon 
     the dual-load in the subreflector.
   . A system of measuring calibrator flux densities in an absolute 
     sense.  The details of such a system are TBD, but one such system 
     involves the use of standard horns which have absolutely 
     calculated or measured gain, used interferometrically with the 
     ALMA (or ACA) antennas.
   . Correction for decorrelation must be quite accurate (we expect 
     to do this with a combination of WVR and fast switching).

 - Phase calibration of the fluctuating atmospheric component of 
   delay to 10*(1.1 + PWV) microns of path length, where PWV is the 
   precipitable water vapor in mm, which implies:
   . Ability to use the fast switching by itself to reach this 
     requirement when atmospheric conditions and observing frequency 
   . Ability to track the atmospheric fluctuations with a WVR system.
   . Ability to calibrate that WVR system using a combination of 
     internal loads (for calibrating the temperature scale of the 
     radiometers) and observations of astronomical sources to 
     calibrate the conversion from WVR observable to interferometric
     phase.  We anticipate using the fast switching capability of the 
     antennas, where the switching timescale is of order 10's of 
     seconds to minutes (note that the electronic or instrumental 
     phase must be separately calibrated - it should be stable on many 
     minutes timescales).

 - Bandpass calibration to 1 part in 1000 for all cases, 1 part in 
   10000 in select situations, which implies:
   . The subreflector must be equipped with a scattering (or "tangent") 
     cone, tangent to the hyperboloid at a radius TBD (1.2 to 1.5 times 
     the radius of the blockage zone) from the apex.
   . This scattering cone must be removable.
   . The possibility for a broad-band, coherent emitter (some type of 
     photonic, likely) to be installed in the subreflector must be 
     retained - implying also fiber optic cabling to the subreflector.
   . The total mass for either the scattering cone or photonic emitter 
     shall not exceed 5 kg.

 - Polarization calibration to 0.1% in amplitude, 6 deg in polarization
   angle.  In addition:
   . A quarter wave plate should be available for Band 7.
   . The calibration load system should be useable in combination with 
     the quarter wave plate (we note that a wire grid could perhaps be 
     rotatable, allowing to improve the calibration accuracy in 
     polarization - but have not developed a requirement in this 
     respect yet).
   . The software support for polarization calibration is not trivial, 
     and needs to be addressed as soon as possible.
A number of other requirements have been developed or adopted for 
other calibration components of ALMA (pointing, antenna location, 
etc...), and will be elaborated in a forthcoming document describing 
in much more detail all aspects of calibration of ALMA.

Now, more details on the individual presentations.

Jack Welch and Jesus Martin-Pintado gave good summaries of the test 
results on the dual-load and semi-transparent vane amplitude calibration
devices.  Jack made an unequivocal statement that the dual load device 
in the subreflector should be abandoned.  Tests show that the device 
has strong standing waves (peak to peak of a few % [20% of a few K])
which are variable with time and frequency.  It is not understood what 
the origin of these is (they should be at least an order of magnitude 
weaker), but the measurements show them clearly.  Jesus described the 
second set of measurements of the S-T vane system on the IRAM 30-m 
telescope.  See the report for full details.  5% overall accuracy seems 
attainable with this system, 2-3% with herculean effort.  Polarization 
effects were not measured precisely (< 0.5% is preliminary result, but
not much work has gone into that yet).  The ageing of the vane may be 
a potential problem.  Standing waves are clearly a problem - similar to 
the dual-load system.  Instead of continuing the tests on the S-T vane 
system at the 30-m, they may instead attempt to test a wire grid 

Stephane Guilloteau and James Lamb then discussed a newly proposed 
system which replaces the S-T vane with a wire grid.  Stephane has 
calculated that you need to measure the grid transmittivity extremely 
accurately (to 0.1%) because the slope of the gain fluctuation vs. 
coupling factor is large, if you couple to two loads.  If you use the 
sky instead of one of the loads, this factor is needed to 1%.  He 
showed a schematic diagram that Matt Carter and Mark Harmon have come 
up with - the device is big, but not too heavy (few kg).  The best 
load temperatures still need to be derived - Stephane thinks the 
ambient load should be cooled slightly, while the hot load should be 
around 370 K.  The load temperatures should be maintained to 0.5 K or 
better, which may be difficult given the level of temperature control 
in the receiver cabin.  James then described some practical experience 
with wire grids from OVRO (along with some theoretical underpinnings).
25 micron wires are typical for wire grids in mm systems.  There is a 
need to characterize the receiver polarization characteristics very 
well, since the wave is naturally polarized by passed through the grid.
James presented a new (well, it's not totally new, since Dave Woody and 
Richard Hills have been talking about this for some time - see the 
Calibration Group Telecon minutes) idea on replacing the S-T vane or 
grid with a dielectric film.  50 microns or so thick.  He then presented
a nice table with pros and cons of S-T vane, wire grid, and dielectric 
beamsplitter.  He likes the dielectric beamsplitter, because they are 
relatively simple and cheap.  Wire grids are ~$1000 each, and are 
relatively fragile.  Dielectric beam splitters are used at OVRO and BIMA
for LO injection.

John Payne and Darrel Emerson then described briefly a device they have 
been looking at recently as a photonic calibration device (in the 
subreflector).  These are "super luminescent LEDs".  They emit 
broad-band emission at 5 mW over 30 THz (!).  The stability might be an 
issue, however.  One nice thing is that the frequency response is 
controlled almost completely by the photoemitter.  They will continue 
to look into these devices.

Jack and Stephane then described a technique for determining the 
absolute flux density of astronomical sources.  The idea is to strap a 
standard gain horn (of precise manufacture, so that the gain can be 
calculated from theory, rather than measurement) onto the side of one 
of the interferometer dishes, then use the correlations between this 
gain horn, the antenna it's strapped onto, and the other antennas, to 
bootstrap the absolute gain of the large antennas, then use the 
correlations of the large antennas to deduce the source flux density.  
Jack and collaborators have used such a system at 28 GHz on the BIMA 
antennas to measure Jupiter quite accurately (145.1 +- 2 K total, 
143.7 +- 2 K thermal).  They have also done similar measurements for 
Mars and MWC 349, but those data are yet to be reduced.  They plan to 
repeat this with a 90 GHz horn next winter, measuring Venus and Mars 
(but note that resolution may start to be a problem here).  The trick 
for them was the transfer switches, getting the atmospheric opacity 
right (they used a combination of TIPs, sonde measurements from a 
network of four surrounding launch sites, and atmospheric models to 
calculate and check this), and the fact that they had to do *lots* of 
ON/OFFs.  So, the scheme for setting the flux density scale is:
 . measure planets as often as necessary
 . transfer planetary flux densities to phase cals (use 2 or 3)
   perhaps each day - looking ahead to the needed ones
 . accurate measurement of opacity is still needed
and the work to be done is:
 . transfer Jupiter to MWC 349 and Mars at 28 GHz
 . do the 90 GHz experiment, with Mars and Venus (at least)
 . generate a scale for MWC 349
 . do some accuracy tests for the determination of opacity
 . test Stephane's idea at 90 GHz
Stephane then discussed his modification to Jack's idea, which is 
essentially to not strap the gain horn onto the side of another antenna,
but just let it sit on top of an antenna mount, then use all the short 
spacings available (the ACA is probably key for this).  Stephane notes 
that there is a possibility of getting some funding for research into 
this idea in Europe - as much as 10% of total project cost is available 
in principle (5-10M Euros).  "Value adding" to the project must be 

Larry D'Addario then discussed his recent ALMA memo 466 which details 
some issues related to gain stability.  Cooled IF amplifiers will be 
the dominant source of gain fluctuations.  A few mK stability of the 
entire system is needed for dG/G=10^-4 (as required).  The current spec 
is 1.5 mK on the 4K stage (from Charles Cunningham).  James Lamb 
pointed out that lab measurements on SIS mixers showed fluctuations 
that were clearly not related to temperature fluctuations.  This puzzled

Al Wootten and Mark Holdaway then discussed phase calibration.  Al 
gave an introduction to the problem - reminding all that without proper 
phase calibration, ALMA will be limited to baselines of a few hundred 
meters at most - not very satisfying.  He discussed the various elements
of the phase, and current ideas on techniques to measure the fluctuating
atmospheric component (fast switching and WVR).  He pointed out that 
different combinations of configuration, frequency, conditions, and 
experiments will require different combinations of these two: fast 
switching only, fast switching with frequency switch, either of those 
combined with WVR.  He noted that the WVR systems would be discussed 
in detail the following afternoon in a separate session.  Mark then 
discussed fast switching.  Memos 403 & 404 are still the state of the 
art.  Complications:
 . calibrators often too weak at high frequency, so may have to
   observe them at 90 GHz, but then need to extrapolate in frequency
   (especially troublesome in the submm because of dispersion from 
   the strong lines), and the electronics need to be stable and 
   calibrated separately
 . dry fluctuations may be a problem
 . decorrelation must be dealt with, unless it is corrected by the WVR
Typical cycle time is ~ 20 seconds.  Typical flux density of calibrators
is ~ 50 mJy.  He discussed briefly the test plan for testing fast 
switching on the ATF interferometer.

Stephane then described the bandpass calibration.  He described a 
calculation in which it is shown that the total bandwidth over which 
accurate bandpass calibration can be done is only of order 100's of 
MHz in the submm, due to elevation differences between source and 
calibrator, and errors in opacity (or PWV).  Precision is also limited
to 0.1%.  We therefore may not get the required accuracy by using only 
astronomical sources.  One way around this is to used an astronomical 
source in combination with a single load calibration.  Then use the 
autocorrelations to calculate the bandpass.  This fails if the line 
or continuum being observed is strong, or in the presence of RFI.
Standing waves are still an issue.  The thinking is that a tangent 
cone will be required to reduce them to a reasonable level.  There is 
a question on whether a photonic load can shine through a small hole 
in the tangent cone.

Bryan Butler then described the polarization calibration.  He discussed
requirements, both at high level and in the FE instrumental specs, and 
the fact that some of these are not very well thought out.  He then 
discussed the theory of polarization calibration.  The main point is 
that it can be described as calculating the elements of two Jones 
matrices - a gain and a leakage.  In the case of circular polarization 
systems, the linear polarization is contaminated by a leakage times the 
Stokes I.  In the case of linear polarization systems, the 
contamination is by a gain times Stokes I.  He described some 
implications of the choice to use linear systems on ALMA (this choice 
made because wider bandwidth ratios can be achieved on linear systems):
 . need very stable receivers, since gain instabilities couple Stokes I 
   into the linear polarization
 . no polarization snapshots, unless gain and leakage are extremely 
   stable (need variation of parallactic angle, X, to break apart the
 . source circular polarization and instrumental polarization are not 
   cleanly separated, so observation of a source of known polarization 
   properties is required (astronomical or injected signal)
 . the XY phase offset is difficult to measure - you need a strongly 
   polarizaed astronomical source or injected signal
He then noted that we could possibly get around some of this by 
rotating some feeds (half of them?) by 45 deg. with respect to each 
other, but this might be problematic operationally (and in software).
In addition, as pointed out by Stephane, this requires that the 
correlator be run at full polarization mode with all four correlations 
at all times, limiting maximum bandwidth even when only Stokes I is 
desired.  This effectively limits sensitivity.  He then discussed the 
possibility of using an injected signal instead of astronomical 
sources.  The difficulty is generating a signal with precisely known 
polarization properties that is stable enough.  More research is needed 
in this area.  He mentioned that a 1/4 wave plate will be used for 
precise polarization work, but only on band 7.  He then summarized 
 . problems with specifications - poorly defined or nonexistent
 . scattering from atmospheric particles may be a problem in the
 . wide-field imaging and mosaics have special polarization
 . there is a small issue on pad perpendicularity
 . there is no manpower in the project to work on this issue currently

Tetsuo Hasegawa then discussed some issues related to the ACA.  The 
four large 12-m antennas are required for calibration of the smaller 
7-m antennas.  12-m antennas will also operate in total power mode,
so transfer from total-power to interferometric flux densities is 
possible with ACA.  It may be possible to use the ACA as a secondary 
flux density calibrator monitor - depends on pressure to use it for 
total power measurements.  The polarization is a problem with the ACA 
and has not been thought out very well (similar to the array overall).  
If polarization rotators are needed, is there room?

bjb - 2003jun13