Figure 17: The standard calibrator signal (solid line) with an overlay of a shifted calibrator
signal (dashed). The A/D cards are indicated and a shift is clearly seen on card B.
Between early 1997 and 15th September 1997 there was an intermittent fault with the SCUBA data acquisition system (DAQ) that led to a particular form of data corruption. The problem is most serious in MAP/JIGGLE observations but can also affect PHOTOM. The DAQ fault was identified and fixed on 15th September 1997 and checking routines were added to the data-taking software to warn of any such data synch problems that occur in the future. Observers with data taken after that date need read no further.
Those with data taken in the problem period may remember that the SCUBA bolometer channels are collected into 9 groups of 16, with each 16 channel group being handled by a different A/D card in the DAQ. The name of each bolometer reflects the A/D card and channel number on that card which handles the signal; for example, B12 will be routed through channel 12 on card B. The 9 A/D cards have single letter identifiers running from A to I, the channel numbers on each run from 1 to 16.
Roughly once or twice a night the fault would insert a spurious number into the data stream from one or more of the A/D cards to the transputer system that does the digital demodulation. The system design is such that until the system was reloaded a spurious shift would be applied to the data. Thus if the fault occurred in the B card then after that point the system would see the data from that card as shifted up by one channel number; data for channel B5 would appear in channel B6, B7 in B8, etc. The end channels wrap around so that channel B16 would appear in channel B1 of the next dataset. The effect is cumulative, so that if 2 faults occurred on the B card then the data would then be shifted by 2 channels.
If the fault occurs for a card whose bolometers are measuring source signal then the image of the source will be reconstructed incorrectly by both the real-time display and the SURF package. In fact, the problem was first noticed when jiggle map images of Uranus showed an apparent double source.
Faulty data can be patched up using the scushift utility. The difficulty is in finding when and where in your data the problem has occurred since, without a high signal-to-noise signal to judge by, you cannot go on the appearance of the final image.
The situation is saved by SCUBA’s internal calibrator. During all jiggle-type observations a sinusoidal signal from a source inside the cryostat is superimposed on the astronomical data. The digital demodulation recovers the amplitudes of both the calibrator and astronomical signals. The pattern of the calibrator signal from the SCUBA bolometers forms a signature that is constant over long periods and can be used to detect shifts in the data.
To illustrate this point there are four data files distributed with this package.16 These files contain the calibrator signal for this period for the 450- and 850 m filters17. The files are:
Calibrator signal for the long-wave array with the 850 micron filter.
Calibrator signal for the short-wave array with the 450 micron filter.
Calibrator signal for the short and long-wave arrays at 450 and 850 microns.
Calibrator signal for the short and long-wave arrays at 450 and 850 microns but for a photom observation.
The ‘_map’ files are intended for comparison with JIGGLE/MAP data whereas the ‘_photom’ file is intended for use with PHOTOM data (there are two extra bolometers in this case). In addition, there is also a file containing shifted data, calsig_450_850_bp2.sdf, where the signal from the B card has been shifted by 2 channels. Fig. 17 shows this file overlaid on the correct, unshifted, calibrator signal.
Demodulated data can be checked as follows:
The dimensions of the data array in the demodulated file should be [5, n_bols, n_jiggles] (§E.3), where n_bols is the total number of bolometers measured (128 for both short and long arrays), and n_jiggles is the total number of jiggle positions measured in the observation.
In summary, manic should be given the following parameters: ONDIM=1, ELINE2=‘Y’,
YLIMITS=default, XRANGE=[3,3] and ZRANGE=default (where default is the default value suggested
for the parameter)
Alternatively, it should be possible to use ndfcopy to extract the NDF section, using the TRIM parameter to reduce the resultant section to 2 dimensions.
where ‘some colour’ is a different colour to that used to display the first calibrator signal.
scushift is very verbose! The ‘before’ and ’bolchan’ entries simply tell the user the form of the correction used. If you look carefully you will see that the 17th number in the list has changed from a 1 to a 15 indicating that the shift was successful.
This command works ‘inplace’ and, in fact, will not run on the raw data file; instead it should be run on the file produced by reduce_switch or flatfield.
There is one last wrinkle to the process of extracting calibrator data.. PHOTOM observations after 3rd June 1997 store signals from 2 channels in addition to the arrays, if the arrays were being used. Thus n_bols for a PHOTOM demodulated data array will be 130 rather then 128. In this case you should compare the calibrator signature with that in the file calsig_450_850_photom.sdf.
16The files can be found in $SURF_DIR/
17for the calibrator signal prior to April 1997 or for different filters please contact your SCUBA support scientist for more advice