next up previous contents index
Next: How do I average Up: AVERAGING DATA IN FREQUENCY Previous: AVERAGING DATA IN FREQUENCY

Should I average my continuum data in frequency ?

 

Now you should make a fairly fundamental decision. Do you want to retain the spectral information across each band for each frequency or would you prefer to average all the channels in each band to one channel. The answer to this question is that you will always need to average the data in frequency. Why bother asking the question then ? Well, you must always average in frequency for the initial determination of the antenna gains with time from your calibrators (§ 8), but you may want to retain the spectral information on your program sources for multi-frequency synthesis. I will address this latter issue here.

If you are only interested in continuum information, then you may want to do this averaging. However, you must estimate the advantage in terms of the (u,v) coverage that you gain by retaining the spectral information and gridding all channels together in the imaging task (this is multi-frequency synthesis) against the reduction in the amount of data you need to handle and degradation of the image that occurs when you average channels together. The latter effect is known as bandwidth smearing and is essentially a chromatic aberration. The effect is a radial smearing of the source with a reduced amplitude and increases with distance from the phase centre. It is discussed in detail in Appendix D. Note that calibrators are not affected by bandwidth smearing because they are at the phase centre.

The improvement in the radial (u,v) coverage by using the full bandwidth is easily estimated from an expression such as


displaymath5660

where tex2html_wrap_inline5662 is the baseline in wavelengths at the band centre (frequency tex2html_wrap_inline5664), and tex2html_wrap_inline5666 is the bandwidth. Note that for a given observation, the most additional coverage is gained by the longest baselines. With the compact array, it is possible to fill in some of the inter-spacing gaps in this fashion (and also by rapid frequency switching) and this is a significant advantage for complicated sources. At 13 and 20 cm, there is often significant benefit to be gained by retaining all the channels, especially in the 6 km array. At 3 cm it is very unlikely you would retain all channels, and at 6 cm the benefit is marginal.

If you conclude that there is no benefit to be gained with multi-frequency synthesis for your data, there is an additional point to consider before you average some channels together. This concerns interference, which is often troublesome in the 13 and 20 cm bands. You may want to edit the data first (see § 7) and remove channel-specific interference from the data, and then average the good channels. If you need to consider this option, read and apply the section on editing first, and then come back to averaging of the data later.


next up previous contents index
Next: How do I average Up: AVERAGING DATA IN FREQUENCY Previous: AVERAGING DATA IN FREQUENCY

nkilleen@atnf.csiro.au