LABOCA Photometry with CRUSH

Updated for 2.03 release

Table of Contents

• Introduction
• The 'woodoo' observing mode
• Reducing photometry data with CRUSH
• What's new?
• Analysis
• How errors are calculated
• What is missing
• Estimating a Systematic Error Multiplier
• Conclusions

   

Introduction

The woodoo observing mode

After some less successful earlier attempts, the photometry mode was finally commissioned for LABOCA (and SABOCA) during the March 2010 (03/11–03/15) technical run (under T-084.F-0001-2009). The observing mode is the basic chopped photometry, with a symmetric nod. I.e., the wobbler is used for the fast switching (i.e. chopping) between the source of interest and an off position on a selected bolometer. After spending some time with the source in the left beam, the telescope nods to the other side where it continues chopping with the source now in the right beam. The cycle is repeated until the desired depth is reached.

This symmetry is required because the left and right beams illuminate the telescope primary differently (due to their off-axis positions), resulting in a large and arbitrary power inbalance between them. By combining the symmetric nod phases together the source flux can be separeated from the inbalance, as long as the atmosphere is sufficiently stable such that systematics can be ignored.

Reducing photometry data with CRUSH

As of version 2.02, CRUSH provides the capability to reduce photometric data obtained in this symmetric chop mode (a.k.a. woodoo). Simply specify the scans on the command-line, just like you would do for mapping. CRUSH will automatically recognize that these are photometry scans and will reduce data accordingly. E.g.,

Modifiers, such as faint and deep are encouraged when appropriate. Not using these on a weak source will result in sub-optimal discrimination of bad data, while using these options incorrectly on a bright source will result in over-flagging. The results from the commissioning run in March 2010 are summarized in Table 1 (tip: hover over the object names to see the scan numbers used). Fluxes are shown for default, faint and deep reductions, while fluxes obtained from imaging are shown in the last column for comparison (when available). The results were obtained using version 2.03-1, and should be representative of the 2.03 releases.

Table 1. Photometry Results

Object	Time	(default)	faint	deep	Imaging
Vesta	0.6 min	2.24 ± 0.040 Jy	2.26 ± 0.030 Jy	2.38 ± 0.031 Jy	2.25 ± 0.045 Jy
Metis	8.1 min	350 ± 14 mJy	339 ± 12 mJy	362 ± 13 mJy	345 ± 33 mJy
Angelina	49.4 min	32 ± 5.5 mJy	28 ± 4.1 mJy	29 ± 4.4 mJy
BR1202	155.2 min	37.5 ± 2.6 mJy	35.8 ± 1.8 mJy	37.7 ± 1.9 mJy
SMM14011	127.8 min	16 ± 2.9 mJy	12 ± 2.0 mJy	13 ± 2.2 mJy

As you can see, the photometry reductions recover the same flux as the imaging, within the indicated errors. There is good agreement between results obtained with the default pipeline and with the faint and deep options. This is good news, since latter the two options mainly differ in the way they treat large-scale (extended) emission, which is of no consequence for single-beam photometry. Below, you can find more details about the reproducibility and accuracy of the photometric reductions, if you are interested.

What's new?

The photometry reduction of CRUSH is improved from time to time, as the pipeline incorporates new ideas or fixes various bugs. Here is an overview of what has changed since the initial 2.02-1 release supporting photometric reductions for the first time.

Release 2.03-1

The 2.03-1 release brings a number of improvements in the photometry reductions. The main changes are:

• Better flux calibration. Decorrelation steps no longer reduce photometry flux, leading to more consistent fluxes across the brightness options.
• Check for pixel saturation increases reliability by discarding unresponsive pixels.
• Updated photometry configurations fine-tuned for precision photometry.

The net effect is fluxes that are about 10–15% higher than previously (i.e. these were about 10–15% too low before), and improved systematics mean that for >1Hz chops, the scatter of measurements is more in line with the estimated errors (around 8% excess vs. around 35% before). As such, the reported flux errors are confirmed to be reasonably precise.

Analysis

The close agreement between fluxes obtained in photometry and in imaging modes ought to convince you that the values are reliable. However, of equally great concern is whether the indicated uncertainties provide a fair estimate of the measurement errors. Ideally, repeated measurements of the same source should yield values that are scattered around the true flux of the source with a spread equal to the estimated uncertainty.

How errors are calculated

CRUSH estimates the error of the photometry directly from the data. The estimate has two components:

• (1) the residual scatter (white noise level) of the detector time-streams, after all correlated signals are removed, bad data is identified and flagged, and noise whitening was performed.
• (2) from the observed scatter of the measurements from each chop cycle inside a nod-phase (i.e. subscan).

The first component provides a very accurate measure of the detector sensitivities, providing a lower limit to the measurement uncertainty. The systematic sources of error in chopping increase this theoretical noise limit. This is where the second method of estimation provides some relief. Given a 1Hz chop (typical for LABOCA) and a minute long nod-phase (also typical, although shorter would probably be better, see further below), there are around 60 repeated measurements of the relevant fluxes (power inbalance plus/minus source) in the subscan. The actual scatter of these can be used to estimate the systematic errors coming from the chopping motion itself.

What is missing

Thus the CRUSH photometry values do include a reasonable measure of the systematic errors of the chopping. However, only the systematics of the chopping motion are measured in this way, whereas the systematics of the nodding are not accounted for. The reason for this omission is that a typical scan may have only a couple or a handful of nod phases. Thus, even moderately deep observations can have too few nods for a robust estimation of the systematics involved. Therefore, CRUSH provides no automatic way to estimate the excess uncertainty resulting from the nodding motion. You simply have to be aware of this, and come up with your own estimate, perhaps by using the information provided below...

Estimating a Systematic Error Multiplier

The commissioning data sets contains repeated observations of a few faint sources: the asteroid Angelina, the submillimeter galaxy SMM14011 and the distant quasar BR1202. The scatter of the individual measurements around their mean values can be used as a guiding measure of the level of systematic error not included in the CRUSH photometry results. Tables 2 (a) – (c) summarize all the photometry data for these three sources, tabulated against the chopping frequency used.

You may observe that all measurements scatter around their mean values, more or less consistent with the indicated (calculated) levels of error. A closer inspection of this scatter (see figure below) reveals, that the true error of the measurement is somewhat larger, and may be dependent on the chop frequency (apart the nod phase duration!). For chopping at 1Hz or faster, the unaccounted systematic error of the measurement is around 8% higher than the value estimated by CRUSH (dashed horizontal line). If there is a ~10% excess error, it is probably due to the slow nodding (1 minute per nod phase), a time-scale over which the sky may not be sufficiently stable to make a more accurate measurement.

The higher error multiplier for the 0.5Hz chop, may be indicative of a 1/f-type noise, which would be expected for variations of the sky and/or detector temperatures. The dashed curve indicates the best-fit scenario, in which the noise excess is due to unmodelled 1/f-type behaviour.

(As a side note, the original SHARC 350µm camera on the CSO used a 4Hz chop with a 15 second nod — anything more prolonged than these showed a clear excess in the residual noise values. The 870µm sky at APEX may be somewhat less unstable, but it is likely, nonetheless, that the systematic errors would go down with faster chopping and/or nodding.)

Conclusions

CRUSH provides a viable approach for reducing chopped (woodoo) photometry scans for LABOCA (and SABOCA), yielding reliable and reproducible results. The reported flux values include an estimate of some systematic errors, although not all. As a result, the true uncertainty of photometry may be perhaps 10% higher than indicated. The excess noise, if real, is probably due to the very slow nodding of the telescope during the observations used for commissioning. In all likelihood, the systematic errors will decrease with a faster nod. Observers are much encouraged to chop fast (at least at 1Hz) and nod more frequently (e.g. every 15 seconds) to obtain more precise photometry.

 

---

Copyright © 2011 Attila Kovács (attila[AT]submm.caltech.edu)