Chapter 2
SCUBA-2 Overview

 2.1 The instrument
 2.2 Observing modes
 2.3 The raw data

2.1 The instrument

The Submillimetre Common User Bolometer Array-2 (SCUBA-2) is a 10,000-pixel bolometer camera. It has two arrays operating simultaneously to map the sky in the atmospheric windows of 450 and 850μm. Each array is made up of four sub-arrays as shown in Figure 2.1.

How it works

The SCUBA-2 bolometers are integrated arrays of superconducting transition edge sensors (TESs) with a characteristic transition temperature, Tc. In addition, each TES is ringed with a resistive heater which can compensate for changes in sky power. The SCUBA-2 focal plane is kept at a base temperature slightly below Tc, however a voltage is applied across each TES resistance to position the bolometer at the transition temperature. From this point, any increase of temperature on the bolometers (e.g. from an astronomical signal) will increase the TES resistance and heat it up. This causes a drop in current and therefore a drop in temperature making the system self-regulating.

For properly performing bolometers, the change in current through the TES is proportional to the change in resistance, with the response calibrated using flat-field observations (described below). This changing current generates a magnetic field which is amplified by a chain of superconducting quantum interference devices (SQUIDs). This induces a feedback current which is proportional to the current flowing through the TES, and it is this feedback current that is recorded during data acquisition.


Before science data can be taken the system must be optimised. These ‘setups’ are performed after slewing to the azimuth of the source, where the SQUID, TES and heater biases are set to pre-determined nominal values, in order to position the bolometers in the middle of the transition range.


The shutter then opens onto the sky, and as it does so the gradual increase in sky power hitting the array is compensated for by a decrease in the resistive heater power via a servo loop designed to keep the TES output constant. This acts to keep the bolometers positioned at the centre of the transition range and is known as heater tracking.

The responsivity of the bolometers will change slightly between the dark and the sky; therefore, once the shutter is fully open a fast flat-field observation is carried out to recalibrate them. A flat-field measures the responsivity of each bolometer to changing sky power. It does this by utilising the resistance heaters which are ramped up and down around the nominal value. The change in current through the TES is then recorded for each bolometer giving a measure of its responsivity. The flat field solution is then the inverse linear gradient of the current as a function of heater power.

At this point bolometers with responsivities above or below a threshold limit are rejected, along with bolometers that display a non-linear response or have a poor S/N. A second flat-field is performed at the end of an observation so bolometers whose responsivity has changed over the course of the observation can be flagged.

For full details of the array setup and operation see Holland et al. (2013) [12].

Figure 2.1: The layout of the arrays at 850μm (left) and 450μm (right). The labels denote the name assigned to each sub-array. Raw data files are generated separately for each sub-array and must be co-added. This figure was made by running wcsmosaic on a raw sub-scan from each sub-array.

2.2 Observing modes

Two observing modes are offered for SCUBA-2—daisy and pong. As the bulk of SCUBA-2 observing involves large area mapping, both observing modes are scan patterns. Your choice depends on the size of the area you wish to map, where you would like your integration time concentrated and the degree of extended emission you wish to recover.

PONG A pong map is the scan strategy for covering a large area. The default options allow for three sizes—900 arcsec, 1800 arcsec and 3600 arcsec. A single pong map is a square of these dimensions and the telescope fills in the square by bouncing off the edge of the area. To ensure an even sky background it is recommended a minimum of three, but preferably more than five, pong maps are included in a single observation with a rotation introduced between each one. In this way a circular pattern is built up, (see the lower right-hand panel of Figure 2.2), with a diameter equal to your requested map size.

To recover large-scale extended structure you are advised to use larger pong maps which scan at a higher rate. This option is preferable to tiling multiple smaller maps. Ultimately it is the size of the SCUBA-2 field-of-view that determines the sensitivity to large-scale structure.

DAISY daisy maps are the option for point-like or compact sources ( <3 arcmin) by maximising the exposure time on the centre of the image. The telescope moves at a constant velocity in a ‘spirograph’ pattern that has the advantage of keeping the source on the array throughout the observation. This is shown in the top panel of Figure 2.2.
Why these patterns?

SCUBA-2 removes atmospheric noise in the data-processing stage (Holland et al. 2013) [12]. The power spectrum of data taken by SCUBA-2 has a 1/f noise curve at lower frequencies. To ensure astronomical signals are far away from this 1/f noise, fast scanning speeds are required.

In order to disentangle source structure from other slowly varying signals (e.g. extinction, sky noise, 1/f noise), the scan pattern must pass across each region of the map from different directions and at different times. The scan patterns themselves, along with the associated parameters (velocity and scan-spacing), have been designed and optimised to meet both these criteria.


Figure 2.2: The top row shows a daisy and the bottom row shows a pong. The left column shows the telescope track over a single rotation of the pattern. The right column shows the telescope track after multiple rotations of the pattern. The scan pattern for an observation can be visualised in this way with Topcat using the output from jcmtstate2cat. See section 9.3 for more details. Figure modified from Holland et al. (2013).

2.3 The raw data

A normal science observation will follow the following sequence.

Science scans

The SEQ_TYPE keyword in the FITS header may be used to identify the nature of each scan (see Section 9.2). When you access raw from the Science Archive you will get all of the files listed above. Later when you reduce your data using the map-maker you must include all the science files and the first flat-field. The final flat-field is not currently used.

Shown below is a list of the raw files for a single sub-array (in this case s8a) for a short calibration observation. The first and last scans are the flat-field observations,which occur after the shutter opens to the sky at the start of the observation and closes at the end (note the identical file size); all of the scans in between are science.

  % ls -lh /jcmtdata/raw/scuba2/s8a/20131227/00034
  -rw-r--r-- 1 jcmtarch jcmt 8.0M Dec 27 03:00 s8a20131227_00034_0001.sdf
  -rw-r--r-- 1 jcmtarch jcmt  22M Dec 27 03:00 s8a20131227_00034_0002.sdf
  -rw-r--r-- 1 jcmtarch jcmt  22M Dec 27 03:01 s8a20131227_00034_0003.sdf
  -rw-r--r-- 1 jcmtarch jcmt  22M Dec 27 03:02 s8a20131227_00034_0004.sdf
  -rw-r--r-- 1 jcmtarch jcmt  22M Dec 27 03:02 s8a20131227_00034_0005.sdf
  -rw-r--r-- 1 jcmtarch jcmt 6.8M Dec 27 03:02 s8a20131227_00034_0006.sdf
  -rw-r--r-- 1 jcmtarch jcmt 8.0M Dec 27 03:03 s8a20131227_00034_0007.sdf

The SCUBA-2 data acquisition (DA) system writes out a data file every 30 seconds; each of which contains 22 MB of data. The only exception is the final science scan which will usually be smaller (6.8 MB in the example above), typically requiring less than 30 seconds of data to complete the observation.

Note: All of these files are written out eight times, once for each of the eight sub-arrays.

The main data arrays of each file are cubes, with the first two dimensions enumerating bolometer columns and rows within a sub-array, and the third time slices (sampled at roughly 200 Hz).

A standardised file naming scheme is used in which each file name starts with the sub-array name, followed by the UT date of the observation in the format yyyymmdd, followed by a five-digit observation number, followed by the scub-scan number. The name ends with the standard suffix .sdf used by all Starlink data files. For instance, the files listed above hold data from the s8a sub-array for observation 34 taken on 27th December 2013.


Raw SCUBA-2 data come in uncalibrated units. The first calibration step is to scale the raw data to units of picowatts (pW) by applying the flat-field solution. This step is performed internally by the map-maker but can be done manually when examining the raw data—see Section 9.1.

The second step is to scale the resulting map by the flux conversion factor (FCF) to give units of janskys. When running the Orac-dr pipeline this is done automatically. However it is good to check that the FCF value applied to your data is sensible. Checking FCF’s must be done manually, instructions for this is given in Section 8.1.3.