5 Technical Information


Next: 6 VLBI Up: wsrtGuide Previous: 4 Observing Modes Contents

Subsections

5.1 Observing Bands

The WSRT array is equipped with Multi Frequency Front Ends (MFFE), which have cooled receivers at 3.6, 6, 13, and 18+21 cm, and uncooled receivers at 49 cm and 92 cm, in addition to the uncooled UHF high (700-1200 MHz) and UHF low (260-460 MHz) systems. In addition to this, the WSRT has now available one receiver for 5cm (6GHz) work (single dish). The accessible frequency range is given in Table 2. Multi-frequency observations are possible and described below.

92 dual, linear 310-390 125 0.250b,c 300
49g dual, linear 560-610 75 0.150c 50
21/18 dual, linear 1150-1750 27-31 0.012-0.013 5
13 dual, circular 2215-2375 60 0.021 1.3
6 dual, linear 4770-5020 65 0.021  
3.6 dual, lineard 8150-8650 110 0.042  
UHF-lowh dual, linear 250-460e 120-250f 0.500c  
UHF-high dual, linear 700-1200 120-180f 0.085  

Table 2

 Notes to Table 2. Table updated at Nov2004, please check the Time Exposure Calculator for the most updated values.

a) Theoretical continuum sensitivity for a standard 12h measurement with 91 equally weighted baselines, 2 polarizations and a bandwidth of 160 MHz (unless otherwise stated).

b) The noise in the 92 cm band has 3 components: thermal noise (about 0.25 mJy/beam in one 10 MHz band), sidelobe confusion noise ( 1 mJy/beam, depending on the uv-coverage which depends on the number of 12 h runs and number of bands) and, finally, the classic confusion noise. The latter is reported in the last column of this table.
c) The value is based on 10MHz bandwidth, but up to 8 x 10MHz is available.
d) Circular polarization to be implemented in the future.
e) The more sensitive 92 cm FE is used in region 310-390 MHz
f) Large values at the edges of the bands, strongly affected by the contribution of the galactic background,several regions not accessible due to external RFI.

g) Using the 49cm band is no longer possible due to strong RFI.

h) UHF-low is not longer available.

 

5.2 Correlator Setup

The powerful 8x20 MHz, quarter-million complex spectral channel IVC/DZB backend system  (the combination of the IVC - intermediate-to-video converter - and digital backend DZB) is used at the WSRT for all continuum and line observations. This implies that all observations are now done in ``spectral line mode'' (with a minimum of 64 channels per band). We discuss below the default configurations used for continuum observations and the possibilities for spectral line observations.

Please also consider that the new correlator is highly flexible and testing of new configurations (e.g. different channels and/or bandwidth per band) is still an ongoing effort. These new configurations will be offered to the general users as soon as possible.

Continuum observations. For continuum observations the new IVC+DZB backend is used. This has improved the sensitivity as it provides 8x20 MHz bandwidth (8x10 MHz below 1150 MHz). The default continuum configuration provides 8x64 spectral channels in 4 polarizations (for the other available configurations see also Table 3). The size of the datasets resulting from a 12h observations, with 60 sec integration time, is about 2 Gb. Averaging of the channels to reduce the size can be, of course, done in the off-line software in order to reduce the size of the datasets. However, it is recomended to performe the bandpass calibration first. Of the 20 MHz of each band, about 18 MHz are actually useful due to the shape of the bandpass at the edge of the band. Note that the bandwidth available is a significant fraction of the observing frequency, which allows, in a single observation, the study of Faraday Rotation or source spectral index.

For the L-band, we have defined a default set of frequencies for the 8x20MHz bands. They are 1450,1432,1410, 1392,1370,1350,1330,1311 MHz. They have been selected to avoid galactic emission and strong, known RFIs. It is worth noticing, that because of the spectral line mode in which the observations are done, search for redshifted HI in the observed field is also possible. The velocity resolution obtained in the default setup is about 60 km/s (and can go down to about 30 km/s if 128 channels per band are used, at the expens of the polarization, only two polarizations can be measured, se Table 3). Five of the bands in the default setup cover redshifted HI from almost zero to 28000 km/s redshift. However, if one is really interested in looking for HI at velocity higher than this, other frequency ranges can be used, e.g. from 1410 to 1270 MHz, although they are more likely to be affected by RFI.

At frequency different than the L-band, default frequency setups are used in order to minimize the effect of interferences. For 13cm, the eight bands are centered from 2210 MHz to 2338 MHz . At 18cm the bands are centered between 1655 and 1783 MHz and at 6 cm, the centers of the bands are between 4838 and 4966 MHz. Finally at 3.6 cm, the default centers of the bands are between 8400 and 8528 MHz. In all these cases, the adjacent bands overlap an integer number of channels in order to provide a continuous frequency coverage. Different setups are, of course, possible. If a non-default setup is needed for a project, please contant the staff to check the RFI situation in the requested frequency range.

Spectral line observing. One, two, four or eight of the IVC bands may be used simultaneously as outlined in Table 3. 160MHz total IF bandwidth (80 MHz below 1150MHz) can be covered. Restrictions on subdivision of bands and spectral channels are being lifted, and there is now a great deal of flexibility. Also, recirculation is implemented and provides double the number of channels for bands of 10 MHz or smaller. The broadest bandwidth that can be assigned to each band is 20 MHz. In addition to this, bandwidths of 10, 5, 2.5, ..., 0.156 MHz can be chosen. The spectral channels of the correlator are distributed uniformly over the required number of bands. Autocorrelation data are also recorded in addition to all cross-correlations. The chosen bands may be independently tuned to a different frequency, although the useful relative tuning range is fixed by the 160 MHz total IF bandwidth.

Line observations which do not need the full capacity of number of bands and channels of the backend could obtain a modest (5%-10%) increase in sensitivity by using the excess capacity to duplicate the setup on the same frequency span; this effectively amounts to digital oversampling (and comes at the expense of a larger data volume, of course).

Doppler tracking can only be specified for one band at this time, so multi-band spectral line users may wish to forego online Doppler tracking. Users should also bear in mind that the a fraction of each band will not be very useful due to the IF filter properties. This is typically about 2% at the video side of the band and about 8% at the other end. This implies that an 8x20 MHz observation, in which one desires continuous effective frequency coverage should employ a band spacing of about 18 MHz, rather than 20 MHz. A related consideration is that the required total bandwidth should be covered with a minimum number of different sub-bands. For example, if a 18 MHz total bandwidth is required, it would be more appropriate to employ a single 20 MHz band rather than 8 of 2.5 MHz. The same spectral resolution would be available in either case.

For observations that use Doppler tracking, the frequency at which the calibrators are observed is the middle of the time range over which the corresponding Doppler tracked target observation is scheduled to take place. The small shifts in frequency during the 12-hours observation of the target, relative to the calibrators, are relevant only in very special cases. Note that for short Doppler tracked observations (i.e. < 12 h) the frequency at which the calibrator is observed will not be the «average» frequency of the target, which may be relevant for very narrow band observations. An alternative, except for very narrow lines, is to observe the science target at fixed frequency and apply the Doppler corrections off-line. A special correlator mode is also available for observing exclusively auto-correlations. The characteristics of this mode are summarized in Table 3. 

1 1 2048 ...
1 2 1024 ...
1 4 512 ...
2 1 1024 ...
2 2 512 ...
2 4 256 ...
4 1 512 ...
4 2 256 ...
4 4 128 ...
8 1 256 ...
8 2 128 ...
8 4 64 ...
1 2 8196 autocorr. only
8 2 1024 autocorr. only
1 4 4098 autocorr. only
8 4 512 autocorr. only
Table 3.

a) 4 polarizations: XX, XY, YX, YY; 2 polarizations: XX, YY; 1 polarization: XX
b) maximum number of channels available may be doubled by recirculation, available for bandwidths of 10 MHz or less.

5.3 Sensitivity

The system temperature and the sensitivity achievable in the continuum using the 160 MHz band of the IVC+DZB backend, are given in Table 2. In the case the of low frequencies 92 and 49 cm, the noise is based on the 10 MHz band although at 92cm a slightly broader band can be used (but often is partly affected by interference). At these wavelengths a 12h observation will be confusion limited and will not reach the theoretical noise. Values for the classical confusion limit for the WSRT are also given in Table 2.

The r.m.s. noise per channel is given in Table 5 as function of the channel width (the channel width for a given correlator configuration can be derived from Table 4).

A web based form is also available to help the observer to plan their observations and in particular to estimate the exposure time needed.

64 312 156 78 39 20 10 5 2.5
128 156 78 39 20 10 5 2.5 1.2
256 78 39 20 10 5 2.5 1.2 0.6
512 39 20 10 5 2.5 1.2 0.6 0.3
1024 20 10 5 2.5 1.2 0.6 0.3 0.15

 

Table 4. Channel separation  b=B/NF in kHz as function of the bandwidth B and the number of channels NF

Here is some more information about the system noise in some of the bands.

At 6 cm, measurements have been made throughout the band from 4670 - 5150 MHz (though most were in the 4830 - 4910 MHz region). For some receivers there is a noticable increase in Ts below 4700 MHz and above 5100 MHz, of as much as 30 K. Moreover, about half of the receivers show a dramatic peak in Ts   (reaching typically 150 K) between 4680 and 4750 MHz. It seems that the best part of the band to use is 4800 - 5100 MHz.

The S-band (13 cm) has not been extensively used, and most observing has been done between 2215 and 2295 MHz. The average system temperature is <Ts> ~ 60 K.

The L band (18/21 cm band) has been the most extensively used, and thoroughly tested of all. The sensitivity has been determined from 1150 - 1850 MHz. For most frontends, the best sensitivity is in the range 1300 - 1450 MHz, where <Ts> ~ 27 K at elevation  ≥30o. A typical example is shown in Figure 3. The 1300 - 1460 MHz range used as default for continuum observations with 160 MHz bandwidth (see Sec. 5.2) seems to be a good choice.

In the 18 cm band (1660 MHz), we find a higher value of <Ts> ~ 33 K.

312 66 79 132 0.22 0.17
156 33 40 66 0.32 0.25
78 17 20 33 0.45 0.35
39 8 10 17 0.63 0.49
20 4 5 8 0.9 0.69
10 2.1 2.5 4.2 1.27 0.98
5 1.0 1.2 2.1 1.79 1.39
2.5 0.5 0.6 1.0 2.54 1.96
1.2 0.26 0.31 0.5 3.59 2.78
0.6 0.13 0.15 0.26 5.07 3.93
0.3 0.06 0.08 0.13 7.17 5.55
0.15 0.03 0.038 0.065 10.14 7.85

Table 5. Noise per channel for 21-cm observations with Uniform and Hanning taper. b is the channel width and can be found in Table 4 as function of total bandwidth B and number of used channels, NF . Calculations were done for 2 bit mode and two polarizations and assuming a 12-h synthesis observation with 91 interferometers and natural weighting. This table has been updated at the value available on Nov2004, please check the Time Exposure Calculator for the most updated values.

 

\resizebox*{0.7\textwidth}{!}{\includegraphics{panel3.ps}}
Fig.3 System temperature across the L-band (21/18 cm) for three representative MFFEs.


The 49 cm band has been seriously affected by television transmission (TV Drenthe and others). Attempts to alleviate the situation by inserting filters have unfortunately not given acceptable results for a broad bandwidth. However, a 10 MHz band around 610 MHz is often usable. Measurements of Ts have generally been affected by interference. The best values are around 70 - 75 K.

Most of the 92 cm bands are affected, at least part of the time, by intermittent interference from a variety of sources.

It is best to rely on a set of 8 bands selected by the WSRT staff. Contact them if you prefere a non-standard set of bands.

In the UHF bands, the addition of filters to block the strongest sources of television interference has largely suppressed the intermodulation products which proved a hindrance in the past, though RFI remains a concern throughout both bands. With this in mind, we find that the system temperature in the best part of the UHFlow band is Ts ~ 110 K, and Ts ~ 120 K for the UHFhigh. Near the band edges, these values can be worse by a factor 1.5 - 2.

5.4 Multi-frequency observations

The MFFE's can be tuned to another band in a short time. For technical reasons the switching period is restricted to a minimum of 5 minutes if a rotation of the frontend wheel is required, while it is minimally 20 seconds, if a frontend change without rotating the receiver wheel is aspired. Hence, by switching frequently between frequency bands, quasi-simultaneous observations in one observing run are possible, on the expense of UV-coverage in the single bands. It is not possible to observe in two frequency bands simultaneously using all telescopes, but a subset of the telescopes can be used to observe in one band, while another subset can be used to observe in a different band.

This implies three modes of conducting a multi-frequency observation:

In the default mode the user defines a set of observations (e.g. using alternating frequency bands) each of which will be conducted consecutively, treated as single measurement. This mode implies a correlator setup overhead of a minimum of 3 minutes and 10 seconds between the single measurements. Since within this time no data are taken, one might decide for this mode of operation if the switching frequency is not too high, and if the time loss is not significant. This would have the advantage that the data output is the standard WSRT output of one data set per measurement.

In the cluster mode the user defines 2 or more clusters of telescopes where each cluster observes with its own frequency and MFFE band, simultaneously with other clusters. The DZB correlator gives sensible results only for data measured between 2 telscopes which are in the same cluster. The advantage of this mode is that it provides real simultaneous measurements at different frequencies, but this is done on the expense of UV-coverage (defining 2 clusters of 7 antennae will result in 21 baselines per cluster instead of 91 for the whole array). The data are stored in one single dataset.

 

In the frequency switching (frequency mosaicing) mode the sky frequency for all telescopes is swiched to another frequency at regular times without defining a new measurement but by supplying a frequency mosaic ASCII file for a single measurement, thus circumventing the large setup times. In is possible to switch from any avaible MFFE frequency and Band to any other available MFFE frequency and Band. The wheel rotation- and tuning overhead implies a time loss of a few tens of seconds after switching frequencies, and a minimum switching period of 5 minutes is required if the frequency switch implies a rotation of the receiver wheel (but you do not loose 5 minutes between the two consecutive measurements). The data are stored in a single data set that needs to be split for data reduction.

Table 4 describes the minimum switching period and the time loss when switching between bands

Band
3 cm 6 cm
13 cm 
L-band
50 cm
92 cm
UFH-HI
UFH-LO
3 cm
20/10 300/30
20/10
300/30
300/30
20/10 300/30
20/10
6 cm
300/30 20/10
300/30
300/20
20/10
300/30
300/20
20/10
13 cm
20/10 300/30
20/10
300/30
300/30
20/10
300/30
20/10
L-band 300/30 300/20
300/30
20/10
300/20
300/30
300/20
20/10
49 cm
300/30 20/10
300/30
300/20
20/10
300/30
300/20
20/10
92 cm
20/10 300/30
20/10
300/30
300/30
20/10
300/30
20/10
UFH-HI 300/30 300/20
300/30
300/20
300/20
300/30
20/10
20/10
UFH-LO 20/10 20/10
20/10
20/10
20/10
20/10
20/10
20/10
Table 4. Minimum time between the start of two observations in frequency switching mode and time loss in seconds when switching from one band to another in frequency switching mode.

 

5.5 Spectral Dynamic Range

A spectral dynamic range of several hundreds:1 can be achieved as matter of routine. For a dynamic range larger than 1000:1, extra care should be put on the calibration. For experiments that required high spectral dynamic range, more frequent calibration should be planned (it is recommended to observe a bandpass calibrator  at least every hour) together with special attention to prior phase alignment between the antennas (for point sources). Online Hanning or Hamming smoothing is possible. This can mitigate the effects of Gibbs ringing s at the expense of spectral resolution.

5.6 Time and bandwidth smearing

The integration time of the observations can be as short as 10 sec, although the default used is 60 sec. In the standard continuum and line configuration, 12h observations with 60 sec integration produce a dataset of about 2Gb. Given the large volume of data produced, integration time shorter than the defaults may be used only for special situation in which time smearing may be a limitation to the quality of the final image (e.g mosaic observations).

5.7 Primary beam and Mosaicing

The effect of the primary beam attenuation is the main (although not the only) limiting factor of the field of view. The WSRT primary beam attenuation can be described by the function  cos6(c x n x r)  where r is the distance from the pointing center in degrees, n the observing frequency in GHz and the constant c= 68 is, to first order, wavelength independent at GHz frequencies (declining to  c=66 at 325 MHz and c = 63 at 4995 MHz). The values of the FWHM field of view at the different wavelengths are given in Table 1. A plot of the primary beam at 21 cm can be found here (where the peak near-in sidelobes have amplitude about 0.3% of the peak, while the cross-like pattern due to the feed-supported legs has an amplitude about 0.03%).

It is quite often the case that a single primary beam area is insufficient in size to image a field of interest. Regions of arbitrary size can be observed with the WSRT in a mosaic mode observation. In this mode, an ordered list of desired field positions is provided, together with the basic integration time (a multiple of the minimum of 10 seconds) as well as a dwell time (the length of integration per position, which must also be a multiple of 10 sec). Each of the field positions is observed in turn for a dwell time and the entire pattern is repeated for the duration of the observation. By returning to each field position many times during the course of a 12 h track, good uv-coverage can be built up for each field. To insure adequate azimuthal coverage, one should count on a minimum of 12-18 cycles through the pattern of fields, although this will depend on the complexity of the region being imaged and the observing frequency. Since the observing time of each track is being distributed over a number of pointing directions, the sensitivity achieved per pointing will of course be degraded relative to a single pointing observation.

During the last 10 sec of each dwell time, the telescopes of the array are sent to the next position in the list and this 10 sec data sample is flagged. Ten seconds is sufficient to arrive at the desired new position for moves of 2o or less. For very large moves, one must take into account the slew speed of 16o-18o per minute, in determining the likely overhead during each source change. It is not necessary to loose an entire integration time to source move overhead if the integration time is greater than 10 sec. For example, it is possible to specify both an integration time and a dwell time of 60 sec for a mosaic observation. The result would be (for moves of less than 2o) that 50 sec of good data would be averaged for each position and placed in the single 60 sec integration for that position in the current repetition of the pattern. In this way, the data rate can be kept to a minimum.

To achieve uniform sensitivity over a given region of sky requires sampling at a frequency of at least twice per FWHM in all directions. For example, pointings might be separated by about 15 arcmin for a mosaic observation in the 20 cm band, where the primary beam has a FHWM of 35 arcmin. An hexagonal packing of the pointings provides the most efficient two dimensional coverage with uniform sensitivity. On a rectangular grid, a somewhat smaller spacing would be required to obtain uniformity.

The mosaic mode can also be employed to obtain snap-shot coverage or auto-correlation data for extensive lists of individual sources or large regions of the sky. Source lists of hundreds of positions have been successfully observed in a single mosaic mode observation with only the minimal overhead of telescope move time.


Next: 6 VLBI Up: wsrtGuide Previous: 4 Observing Modes Contents
Design: Kuenst.    Development: Dripl.    © 2014 ASTRON