Hows of ALMA

The Atacama Large Millimeter/submillimeter Array (ALMA), will be a single research instrument composed of up to 66 high-precision antennas, located on the Chajnantor plain of the Chilean Andes in the District of San Pedro de Atacama, 5000 m above sea level. ALMA will enable transformational research into the physics of the cold Universe, regions that are optically dark but shine brightly in the millimeter portion of the electromagnetic spectrum. Providing astronomers a new window on celestial origins, ALMA will probe the first stars and galaxies, and directly image the formation of planets.

ALMA will operate at wavelengths of 0.3 to 9.6 millimeters, where the Earth's atmosphere above a high, dry site is largely transparent, and will provide astronomers unprecedented sensitivity and resolution. The up to sixty-four antennas of the 12 m Array will have reconfigurable baselines ranging from 150 m to 18 km. Resolutions as fine as 0.005" will be achieved at the highest frequencies, a factor of ten better than the Hubble Space Telescope.

This project has come a long way and includes:

  • Up to s 12-meter antennas located at an elevation of 16,400 feet in Llano de Chajnantor, Chile
  • Imaging instrument in all atmospheric windows between 10 mm and 350 microns
  • Array configurations from approximately 150 meters to 10 km
  • Spatial resolution of 10 milliarcseconds, 10 times better than the VLA and the Hubble Space Telescope
  • Able to image sources arcminutes to degrees across at one arcsecond resolution
  • Velocity resolution under 0.05 km/s
  • Faster and more flexible imaging instrument than the VLA
  • Largest and most sensitive instrument in the world at millimeter and submillimeter wavelengths
  • Point source detection sensitivity 20 times better than the VLA

How it all works...The Very Basics

From the incoming light signal from Space, down to your computer screen, several steps are necessary in order to make ALMA possible. In a very schematic way, the problem can be reduced to use 2 antennas to make a simultaneous measure of an incoming signal from sky. For this the light hitting each antenna needs to be gathered, and transmitted from each antenna to a main computer that will combine the signal from both antennas and produce what astronomers call scientific data.

After the sky signal enters the Earth's atmosphere, we first need to collect the incoming photons. This is done using a paraboloidic antenna. A bigger antenna mean more photons/signal can be collected. More antenna collecting simultaneously, also mean more photons.

After as much signal as possible has been gathered in the focal plane of the parabola, the incoming signal needs to be transferred to a central computer located in a central building which may be up to 18.5 km away. The light had been traveling mostly unperturbed through space, and suffered considerably attenuation due to its encounter with the Earth's atmosphere. In order to produce accurate measurements of the characteristics of the space signal, as in any science, it is our main interest to keep this incoming signal as pure as possible from any changes that running thought our instrument may cause.

In order to take the signal from the focus of the antenna to the central computer, the light will have to be taken through instruments that will guide until it reaches a fiber-optics cable that will take it in an underground journey of 0.3 to 18 km down to the central computer.

Once the signal from the 2 antennas has reached the central computer, they have to be combined in a certain fashion in order to produce the data. However, prior to combination, the differences in path taken by the two different light-beams (from each antennas) needs to be corrected. The introduction of these so called delay corrections ensures that we are combining light coming from the same region of the sky, emitted at exactly the same time.

Only after proper delays have been applied, the signal from the different antennas enter the correlator, which will combine the data from the 2 antennas producing the main science measurement, called visibility. Visibilities will be computed simultaneously for all pairs of antennas every 16 milliseconds.

This sets of visibilities correspond to a basic Measurement Set, which is the raw data astronomers will receive and process in order to perform their science. From these huge data-files to Nobel-Prize winning images ,it is a long way of data processing. State-of-the-art astronomical software is required to make this task feasible.

How ALMA Works : The Real Deal

As described briefly above, the process producing interferometric astronomical observations can be decomposed in 5 different stages:

  1. Collect photons from astronomical source
  2. Guide the incoming signal from the Antenna to the Main Correlator
  3. Correlate (multiply) the signal from the 2 antennas
  4. Analise the data from multiple set of antennas and produce scientific data

In ALMA, 1 is done by the antenna, where photons hit the paraboloidic surface and are all sent toward the antenna Focus. The ALMA antennas are the highest performance antennas ever built, and their optimal operation is ensured by the Antenna IPT group. Process 2 is very delicate, and is actually done in two stages. In order to transmit the sky signal from the antenna focus to the correlator, the signal needs to be transmitted throughout waveguides and optical fibers. Signals at frequencies above 100 GHz are very difficult to process, and phase stability is particularly difficult: the path-length of the signal sent to the correlator needs to be known up to a fraction of the signal's wavelength. Due to these instrumental reasons, the incoming sky signal is usually converted from sky signal's frequency (RF) to an intermediate frequency (IF) (the term RF stands for Radio Frequency, and was originally coined during the early days of radio astronomy but is still used to denote any incoming electromagnetic signal from radio to submillimiter wavelengths). This process is called downconversion, and is carried out by different pieces of instrumentation located inside the antenna called the FrontEnd. After leaving the Front End, the original sky signal has been down-converted in frequency from its original frequency down to frequencies <15GHz, which are less hard to handle. The amplitude and phase information of the sky signal are preserved (as much as possible), and so is the physical information from the astronomical source.

BackEnd gets the down converted signal and sends it to the correlator. The signal is actually digitized, and is not analog anymore, BackEnd also sends a reference laser signal to the Local Oscillator in each antenna, that allows to synchronize the phase in the array.

( (which could be for instance from 345-353 GHz) to a 8 GHz-wide window ranging from

main building called Front End and Back End. Front End takes the collimated flux of photons reflected, which we will also refer as Sky Signal, and processes it so that it can be

Signal enters the Atmosphere

What is the closest object we can see interferomerically (does the software do any near-field correction?). 1) Near Field 2) Refraction ( - correct pointing of each telescope, - and then effects on the interferometer: If atmosphere and earth were flat, the plane wave coming from sky will still be a plane wave arriving on the 2 antennas, so no correction needed if antennas have same height. However atmosphere is curved, so need to correct. For correcting the extra-path due to different antenna heights need to know (T, p, humidity) close to the antenna. For correcting for diffraction due to curved atmosphere need to know (t,p,humidity) all the way up to the top of atmosphere.

We will have 7 weather stations, 1 central and 2 in each arm.

Water Scale height is about 2km. The order of the correction is proportional to WaterScaleHeight/EarthRadius~2/6400. One still has to look up and measure T atmosphere as function of height. For that we use a Oxygen Sounder, which uses a the O2 line at 50-60 GHz and uses it to measure the Temperature of the O2 (by measuring accurately the shape of the pressure broadened O2 line). Bojan

3) Absorption (correction due offline).

Are the online corrections stored so one can remove them if they were not accurate ?

4) Fluctuations (introduce phase).

More on atmosphere:

how much water in a cubic meter:

K_H2o~ 0.5 -10 g/m3 (at sea level). at 10% humidity and 5C degrees K_H2o~0.5

If you integrate: K_H2o*dH = 2km * 0.5 gm/m3 = 1Kgm/m2 = 1 mm thick layer of water, which we call 1 mm of precipitable water vapor.

n=1+K_H2o * refractivity, where refractivity comes in units of N

0.5 * 6.3

In atmosphere, N2, O2, Arg, CO2 are well mixed. Water is not. Lots of fluctuations.

Also, changes in T are reflected in changes in density which propagate at the speed of sound in order to keep pressure equal. So even if N2 and O2 have little refractive indexes, they still must be taken in account when T changes: 'dry air fluctuations'.

Refractivity H2o goes down with frequency , that is why optical interferometers work...

Fluctuations in water in the atmosphere may cause refraction of a few arcseconds, causing the source to 'move'. First radio astronomers called this anomalous refraction. Inversion layers (warm dry air on top of cold wet air) is where most fluctuations happen (adiabatic lapse rate defines how steep can the T gradient be in order to get a stable atmosphere.)

Plot Atmosphere emission (inverse of absorption) with peaks at 22, 183, 325, 500, 700 GHz) on top of refraction versus T.

n=n'+in'' n= refraction n' path n'' absorption, kramers=kronig relation shows n is the derivative of the absorption.

We need to be careful when there are changes in the absorption as there will be changes in the refractive index.

After each peak of absorption, n goes down a bit (from the nominal 6.3 to )


CALC does the calculations. Originally made for VLBI, adapted to ALMA. It should be doing all calculations (including the atmosphere ?).

ALMA Configurations:

  • Conway: 100m to 5 km
  • Holdaway: 5 km to 15 km
  • Morita-san: ACA

Reconfigurations: 4 antennas per day.

Intermediate configs do not have a very good UV-coverage (compared to compact versus most extended configurations). Need 10 more pads to make it better, but the improvement (only a few parts in a 1000) probably does not justify the extra 4 million dollars.

Shadowing between antennas + transporter constraints impose more difficulties for Site management and construction.


Melco No1 Moves; photo by Morita

- Accurate surface - Pointing: foundation good ,azimuth bearing, , optical pointing (god about basic structure), path errors (interferometry). Rely on design (stiffness, etc).

Shadowing: no more than 0.4 within a plane. ACA: 10m sphere within antennas. It actually drove the new design for the ACA 7m antennas. They are Steel and not carbon fiber, which has about 10 times more thermal conductivity. This means thermal stability must be 10 times better in the ACAs than in the ALMA in order to get the same accuracy.So they blow air through all parts of the antenna, even the quads!

ALMA: 15m sphere within antennas

Interface between the antennas and other systems are different (ICD, Interface Control Document)

AEM Antenna : *A*(lcatel Aenia Space) *E*(uropean Industrial Engineering) *M*(T Aerospace): Section panels are solid, and put together to make the dish structure. There are 5 rings of panels made of nickel (0.3 mm), separated by Aluminum. Since it has a very poor electric reflectivity, they have been coated with Rhodium (Noble metal, gone very expensive!). Elevation drive is a set of moving magnets and coils (+/- alternate magnets, and the coils change polarity in order to drive the antenna).

Antennas have a 3 mounting points. Their positions should be repeatable within 50 microns. 3 Mounting points introduces flexing of the antenna when heavy parts move between those support points.

Subreflector: 6 degrees of freedom (Focus + tilt to align with corresponding receiver beam). Different sub-reflector support feed will produce different illumination patterns. It is not clear if having them different is good or not, but in big mosaic images they will have to be corrected for.

Optical analysis: Legs have metal cone that scatters stuff out to the sky rather than to the ground (in order to reduce the amount of ground emission pick-up)


Backing structure is Carbon fiber on the front and covers on the back. Panels are made of solid aluminium (1mm width). 7 rings of panels. The part that joints the dish structure with the Antenna are made of InVa (In Variable, which has a very low thermal conductivity ~ 1 ppm/K versus 12 of steel). Most of the thermal gradient between the Cabin and the Carbon Fiber happens within the InVa. Thermal changes make the telescope focus change.

Nutator: present in the 4 total power antennas. Can nutate up to 4 arcmin in the sky. Mirror rotations is compensated by rotation of actuators in order to keep angular momentum change 0 within the antenna.

Many things change as the antenna moves, and are different from one antenna type to the other. Most corrections will be implemented in the software.


- Surface accuracy 25 um rms. - Pointing Absolute 2" rms ( which must be good for up to a month time, i.e. the same model must be good up to 2" RMS in day/night for whole month). - Track for 15mins, move 2 degreees and have a pointing accuracy of 0.6" rms (at a maximum of 9 m/s wind speed).

Pointing model (encoders offset, tilt of the base, misalignment of the axis) should remain fixed while Two offsets, dAz, dEl are applied periodically too.


- Temp and wind are the main blockers for meeting these specks


- Temp deformations can be predicted by metrology systems within the antenna (based on physical conditions). Melco has about 16 T sensors. - Wind deformations are harder to predict. Melco chose to use a separate structure within the main structure and put sensors to measure the difference between the positions of both structures (reference frame metrology).


- Tiltmeters and linear displacement sensors. 3 Tiltmeters on the first antennas (inclinometer in the azimuth bearing (bubble inside liquid with electric sensors, take time to settle and get right measurement).


Pointing model can be applied by ALMA software or within each antenna ACU. Metrology offsets usually are sent directly within the ACU, but we should be able to monitor the terms it is inputing.

Holography, OPT, Radiomteric Pointing, Nutators, RF membrane.

References (3 documents of each antenna design).

Shutter and Membrane:

Membrane to prevent airflow from cabin to outside. Current design 0.5mm thick goretex. It works as a dielectric reflective layer. Being reflective, it scatters emission from the cabin back into the Rx, adding to the overall Tsys. Assessment done for an alternative Foam thin membrane.


We hope to correct atmospheric fluctuations by measuring the 183 GHz water vapor line. The line intensity goes up and down depending on the amount of water. One can use the effective T measured on the side of the lines to

Using 2 stable Rx

An arm picks up the sky emission and sends it to the WVP.



Adjustment on the pointing model in order to get the Rx we of wanted band to look at the focus of the telescope.

Pointing offset d : d=> dTheta_sky~d/Feff , d= delta in pointing, Feff = Effective focal length, Feff=96m (f=4.8m, magnification ratio 20, so Feff=96m).

F_eff=f-ratiox12, d=1mm -> Theta~2.1".


Most FE have optics, usually 2 mirrors before the feed horn. Filters to stop the IR. Band 3 and 4 have the optics outside the Rx. Bands 1 & 2 just look at the sub-reflector using a lens. Bands 3 & 4 have the 2 optic elements outside the cryostat. The other ones have the optics within the Rx structure (cooled).

Lens in Bands 1 and 2 may introduce losses (dielectric lens: reflections on the surface (can be reduced by cutting the material in order to make a smooth transition for the incoming beams ; material absorbs too, means you loose light and also re-emits adding Tsys).

Calibration Device:

Assuming a linear system, with only 2 points one could extrapolate Tsys (knowing Tamb and Pamb and Thot and Phot). Hot load coming from europe, studies being made to check their blackbody behavior. Without Cal device you cannot associate the number coming from the correlator to a physical quantity (Antenna Temperature). It is a basic way of measuring the gain of the instrument.

Quarter-Wave plate:

Optimized for 345 GHZ,

Solar Filter:

Solar Temperature is 5000K at 1mm (it is a B down to 1 cm, the goes up again in the radio up to several million). We have a filer to attenuate 15db.

Optical power from the sun arriving to the focal plane should be less than 3kW/m2, but that would actually melt the subreflctor (even at 10% absorption). So they made the antenna surface rougher, and the final power arriving at the focal plane (Rx cabin) is less than 200kW/m2.


Designed to be removable without doing any work to the rest of the telescope, or other receivers. You just need to warm up the cabin and slide the Rx you want out.

3 plates at 4K, 20K and 80K, where the connections are made with nilon rings and copper springs. When the Rx is cooled, the nylon shrinks compressing the copper rings.

There is a window and an IR filter to prevent IR heat leakage.

For Band 1-6:

1 Feed horn followed by a ortho-mode transducer (OMT). Coming in there is a circular waveguide supporting both linear polarizations, and coming out there are 2 rectangular waveguides supporting linear polarizations.

Bands 7> : there is a polarizing grid that splits the 2 polarizations (1 is reflected, the other one transmitted).

The change on the way polarization is recovered is due to space: a grid a low frequencies would be big, and then not rigid. On the other hand, an OMT is complicated piece of hardware, with lost of small sections which would have to be even smaller at higher frequencies.

Band 3 4 6 7 8 and 9 are being built.

3 : canada 4 and 8 Japan 6 US 7 France 9 Netherlands

5: Sweden

2SB (Two side band) Rx: for each polarization you get 2 outputs, one called the Upper side band (USB) and Lover side band (LSB).

DSB (Double side band): Both USB and LSB come on the same IF range (they overlap).

In the

YIG Oscillator where changing the current that runs in the coil will change the tuning frequency, using the reference frequency that comes from the LO.

- Change phase of LO apply a non-frequency dependent delay correction.

- Apply time delay to coarse delay done in the BackEnd :

Correlator bulk time delay (at DGCK,

ALMA Block Diagram

- Orto transducer to split 2 polarization - mixers, DSB, 2sb, ssb, generate an output of 4-12 GHZ (in upper and lower, or DSB)

- at the Warm Cartridge Assembly arrive the master signal form the laser synthesizer and is used by to phase lock the YIG and generate a signal that is then multiplied and mixed with the sky signal. At the WCA also arrive the FLOOG signal, which is used to introduce phase offsets to the LO signal.

The the FLOOG signal is by using the timing events and phase references that come from the LO Reference Receiver (LORR). The LORR distributes to both the FLOOG and the LO2 synthesizer.

The LORR produces the timing events (48ms), a reference signal at 125MHz and something else called the SF.

The each IF (USB and LSB) is then divided in 2 pairs of baseband by the 2x2Matrix Switch. Each baseband is filtered by two Lower Passband Filter (LPF) in order to reduce the 8 GHz bandwith to 4 GHz. It is not clear why they use a 12 GHz LPF, as it would let all the signal through.

After the LPF, each baseband is mixed with each BB LO2 signal, which were created by using the LO2RR references. Each BB as its own LO2 synthetizer, so that they can be tuned to downcovert the respective portion of the 4-8GHz or 4-12GHz IF.

Each BB will be downconverted to a 1.5-5 GHz band, and subsequent filters will be used to remove any contamination arising from the mixing.

After more filters, amplifiers and equalizing, each BB is split, half goes to a Total Power detector used for monitoring, and the other half goes to the Digital Transmitter System, (DTS or DTX) where the signals from each BB get digitized. The digitized signal will be sent through optical fiber to the correlator.

One can actually decide to use 4 basebands in the USB or in LSB,


Most fringe rate in LO1, rotate as if we were observing at LO1 frequency

Residual due to BB is put in LO2

LO offsets- DGCK small offsets in the TFB (divides the BB in 32 sub-bands) in the correlator since each sub-BB is at different sky frequencies

FLOOG: First local Oscilator Offset Generator

FLOOG is the FTS module that is associated with tuning the first LO.

There is one FTS that is shared among the ten local oscillators (LO) of the ten front end bands (LO1). LO1 is typically synthesized at some lower frequencies and multiplied, except for bands 1 ⤓ 3, up to their final frequencies using cold multipliers. There are five FTS devices in each antenna: one FTS module for tuning the first LO and one FTS in each of four second LO modules. The FTS communicates with the CAN bus using an AMBSI2 card that is mounted in the FTS module. The CAN node shall operate at node address 0x32 (hex) for the FTS used in the first LO. The node addresses of FTS in the second LO can be found in ICD for the second LO

Photonics, LLC, Digitizer.

More info at AIV Back End wiki page.


TFB: Divides to make faster cross-correlation

Bulk time delay done by shifting sample by delays in time (multiply sample A1(t)A2(t-m) were m is the delay between Antenna 1 and Antenna 2 in units of quarter of nanoseconds which is how often the data is sampled ( or 1/sampling frequency (1/4GHz)) .

4GHz is the sampling frequency at the digitizers (it has to be twice the bandwith of the sampled band).


Specks to allow 4 simultaneous sub-arrays (could be TP, ACA and 2 sub-sets of the main array).



Fringe Rotation and Delay Tracking


ALMA (Thijs de Graauw, ALMA Board)

- Construction: Integrated Project Teams : (
  • Antenna (Jeff Zivick, MAsao Saito, Steffano Stanghellini)
  • FE (J. Webber Gie Han Tan, )
  • BE (Chris Langley, Fabio, )
  • Computing (B. Glendenning, K. Tatematsu, G. Raffi)
  • Correlator (J. Webber, Alain Baudrey,
  • Site (E. Donoso, ??, ??)
  • Management (Dick Kurz, Tetsuo Hasegawa (d), W. Wild, A. Russell, Satoru Iguchi)
  • Science (R. Hills, A. Peck, A. Wootten, Morita-san, L. Testi).

  • Systems: Nick Whyborn, Christoph Haupt (P. Napier, Dick Schramek)

  • AIV: Rick Murwinscky (Project Ingeneer, J.M. McMullin)

- Observatory (JAO):

  • Science (Lars-Ake Nyman)
  • Technical Services (Richard Prestage)

From NRAO ALMA webpage

Cruising EDM

Documentation -> Released Documentation -> Project-level documents -> Released Project -level Documents

-- AntonioHales - 19 Oct 2007
Topic revision: r19 - 2008-10-16, AntonioHales
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding NRAO Public Wiki? Send feedback