The tied-array uses the Digital Continyum Backend (DCB) for fringe stopping and as a check on the amplitudes and phases. The tied-array has the same setup procedure as that used for a new wavelength, see the WSRT Inregel Procedure, so check
Total power levels to the DCB are important because the total power readout system is badly affected by incorrect levels. If a telescope is not in use its inputs to the equalizer should be connected to the noise source. The total power levels should be carefully checked to see that the level is correct, stable and that the noise source step is seen. The tolerance is within 3dB of the nominal value. If very different frequencies are used within one wavelength band it may be necessary to change the attenuator settings.
It is very important for the phasing of the array to have good phase-zero values for all frequency and bandwidths will be used. To get accurate polarization characteristics it is also also necessary to accurately calibrate the phase difference between dipoles, the X-Y phase difference. This can be done using an observation with crossed dipoles. Often there is only one frequency and bandwidth setup to be used; the major exception is with OH maser VLBI experiments at 18cm. Any strong point source with an accuratly known position can be used to check phases, so there are often many in a VLBI session. A 15degree agreement should be considered normal, and up to 30 degrees acceptable, although the `X-Y phase difference' for any telescope should be stable to within 1 degree. As in normal observations new phase-zero calibrations should not be determined on observations less than 30minutes, and the observations should have good consistencies. Shorter observations can be used to check the phase behaviour.
It is necessary to have accurate delay offsets to get a good addition of the signals in the adder box. Delay offsets determinations for the DCB need at least one fixed and one movable telescope at present. As in normal observing the telescope operator should check for good consistencies (<5%) and small errors in the delay determination (<10ns).
System temperatures from system temperature observations on very strong sources with values derived by decomposition from standard (strong) calibration sources. A 2% agreement should be considered normal, although an agreement of 5% is more typical at 92cm. At short wavelengths (6cm) the amplitudes can be badly affected by tropospheric extinction, especially if there are rain clouds present.
Note that the 9A spacing used for the tied-array is not usually critical, because most of the sources we observe with the tied-array are very compact. The last setting is normally adequate. It is useful to avoid the 36m spacing to avoid too much shadowing for low declination sources.
The sensitivity is effectively a constant per telescope at any given wavelength:
92cm
49cm
21cm
18cm
6cm
For n telescopes this would be exactly n times larger, but forthe WSRT tied-array there are some digital loss factors which should be determined in advance. This `sensitivity' is also known as the `antenna gain' (or simply `gain')
System temperatures have to be determined in advance, as happens anyway in the normal setup at any wavelength. Typical values can be found in WSRT Users Documentation.
The phase offsets for each telescope must be calibrated and removed so that they add in phase, thereby making the signal strong enough to find correlation.
The Total Power of added signals is basically described in Note 315 (analogue), and Note 289 (digital). There are digital losses occurred in the sampling (at DCB) and resampling (at Mk3) but their quantitative amounts are not yet well known. The best summary is in Note 496.
One can estimate the added total power step as follows: System equivalent flux density (SEFD) of the tied array off source is:
SEFD = Tsys/( n * G )
Where
The program MOSYS (MOnitor SYStem)
calculates this System Equivalent Flux Density,
with an allowance for the losses from phase errors at the telescopes.
The step seen when putting the tied array on a calibration source
(if the array is correctly phased up) is then:
step = S * (n-1)/n *1/D
Where
When the noise source is on, the added total power appears to drop, because the correlated fraction of the signal drops. This effect is usually masked by using a noise tube blanking circuit in the signal path to the total power recorders.
There are losses from the digitization at the DCB being redigitized at the VLBI formatter. This was theoretically studied for the single-bit case, see NFRA Note 289 ,but not for the real DCB sampling with three levels and oversampling. Bregman quotes results for adding single-bit (2 level) and 2-bit (4 level) digitized signals, but does not tackle the case used in the Westerbork digital tied-array (3 level oversampled)
In particular
DCB LO3 synthesizers are all locked to the maser, but their phase (and maybe also delay) differences are not easily measured. A phasecal signal (also known as a `1MHz rail' because of the way it has been implemented in VLBI ) should be injected at a point before the different bands split off, so that its relative phase between bands can be detected later and used to correct the astronomical signal. That is why a phasecal system is important for VLBI use.
The hardware has been made such that all the delay differences should be less than 1ns, although phase offsets can be arbitrarily large, and even change once the DCB LO3 synthesizers have been turn off and on again (or equivalently if the maser signal to which they lock has been lost).
Tests up to the present suggest that there may be delay differences up to 50ns, although there is no clear place where such large differences could be introduced.
The phase of signals in the RFDS steps in increments of 22.5 degrees. How much extra decorrelation do we get from this in practice with different numbers of telescopes? There is a worst-case estimate by van Ardenne of less than 2%, but only general arguments about this effect.
This is seen in VLBI polarization observations but is not understood. It may be happening in the RFDS or adder box, but since both sidebands of any DCB band are treated the same until they unfolded (at the RFDS) it is not possible that differences could be introduced before then.
Why is there a 2dB difference between X and Y total powers appearing at the Mk3 rack from the ASTI? (X is higher for the same attenuator settings). There must be about a 2dB difference in the output X and Y amplifiers although they are nominally the same.
Are the ouputs from the ASTI properly terminated (with or without FFB, or WADDS connectors). Yes
The adder sometimes gives large numbers of sync errors (where there is a timing discrepancy between the external and internal 10 second pulses) within a short time and then the problem disappears spontaneously. Why? Attention during the VLBI session in May '95 focussed on temperature effects, but these were not proven. There is clearly some drift in the timings in the electronics, and the most plausible reason is the temperature, but it is not at all clear which of the two timings is drifting. During all tests of this the effect has not occurred.
The phase centres of telescopes do move with hour angle and declination, so shouldn't the tied array centre move also? Yes. See the document The movement of the tied-array.
The DCB delay works with steps of 3.125ns. These must have some (low level?) effect in the VLBI observations. What are they? See the document The movement of the tied-array.
There is a nominal 12.8microsec offset in the WSRT system. Is that accurate? More input is needed from JIVE about clock offsets from local GPS and VLBI fringes