Minimum Integration Time - Continuum
For continuum observing we routinely support an integration interval
of 1-2/3 seconds. It is possible to specify integration intervals as
short as 5/12 seconds if needed, but this places a noticable strain on
the system and is discouraged.
Minimum Integration Time - Spectral Line
In the case of spectral line observations, the greater amount of data
and processing make a short integration interval impossible. The
table below shows the minimum achievable integration time (in seconds)
as a function of the total number of channels per baseline
Nch, and the number of IFs, for a single 27 antenna
subarray. To arrive at Nch do the following:
- start with the number of channels per IF from e.g.
http://www.vla.nrao.edu/astro/guides/vlas/current/node35.html
- multiply this by the number of IFs (1, 2, or 4). Use 4 for modes
PA and PB
- multiply this by another factor 2 in case online Hanning
smoothing is being used
Note that this table differentiates between whether, say, 32 channels
per baseline are divided over one IF of 32 channels, two IFs of 16
channels each, or 4 IFs of 8 channels each.
Minimum Integration Times: One Subarray of 27 Antennas
| Nch* |
1 IF |
2 IFs |
4 IFs |
| 16 |
1-2/3 sec |
1-2/3 sec |
3-1/3 sec |
| 32 |
1-2/3 |
1-2/3 |
3-1/3 |
| 64 |
3-1/3 |
3-1/3 |
3-1/3 |
| 128 |
3-1/3 |
3-1/3 |
3-1/3 |
| 256 |
3-1/3 |
3-1/3 |
3-1/3 |
| 512 |
5 |
5 |
5 |
|
* See definition above!!
|
In all cases baselines (antennas) may be given up to allow shorter
integration intervals. The processing time is, more or less, linear
to first order. In case not all IFs will produce the same number of
channels, straightforward averaging should produce a sensible limit.
Maximum Integration Time
Given current computer capabilities there does not appear to be a
compelling reason to use integration times exceeding 5 seconds,
except for scans to determine reference pointing. Longer
integration times may lead to unnecessary data loss, especially when
using short scans.
A word of caution
As we are finding out details of the constraints posed by the new
Modcomp independent system, the numbers given above are subject to
change. Though we will attempt to keep this table up-to-date, this
should be taken as a warning that in the future it may not always be
accurate. When in doubt, please consult Ken Sowinski or Gustaaf van Moorsel.
|
National Radio
