To change modes it is necessary to:
stop the current scan
check the cccversion
To check the cccversion: caccc> cccversion. It should be “newzmsy” for all modes except VLBI and “VLBItwommlos” for VLBI observing.
in cacor, open the edit menu
select the edit button in the cacor gui,
select the 64MHz zoom config
edit> get cfb_64_32_2f_zm16
and program the correlator
edit> prog
Check that all blocks program correctly. If any blocks report an error, they need to be programmed again, e.g.:
edit> prog 8 or edit> prog 8 27
<CR> to return out of the edit menu
There is no extra flagging needed in this config
start observing a calibrator, e.g.
caobs> track 1
There are several “standard” schedule files with a set of zooms for setting - e.g. setup_1934_64 (and similar).
Check for rear transition module (rtm) problems
F1 and F2 Datalinks (separate tabs in the CACOR gui) should not have red error messages, and the 1ms values are in the range 2000 - 13000 (i.e. not 260 000). Fix with:
cacor> rtminit
Check (in spd) for “big humps”
F1 and F2 datalinks list the antennas, frequencies (1 or 2) and pols (A or B) and the block associated with that ant/freq/pol. Reprogram that block to fix big humps.
Check (in spd) zoom amplitudes and phases - continuum phases and amplitudes will probably be a mess, but zoom products should be clean.
Check every zoom that will be used in at least two consecutive cycles.
If there are a significant number of zooms, it is often quicker to set up differences:
spd> rsave Saves every cycle into a reference so that you display a running difference of the current cycle against previous cycle.
spd> d -100 100 To actually see the difference
Note:You need to wait several cycles after a reprogram until you are actually getting a relevant difference.
To work out which block needs programming, use the cacor “where” command:
cacor> where zoom zn chan
zn is the offending zoom (e.g. z21)
chan is a channel in the range of bad data - for single zooms this can be any channel in the band, but you will need to be more specific for stitched zooms
To reprogramming a block:
config> prog [offending block] calc or
config> prog [offending block] x 7,8 or
config> prog [offending block]
Note: calc resets the c bus;x 7,8 resets CABB board memory controllers; and no suffix will reprogram the whole block (delays may need to be recalculated after this).
Iterate until all is good
The reprogramming is finished......... but array calibration is still to be done - and is more complicated than continuum or 1MHz zoom calibration.
The issue is that, unless the delays are small, the phases will wrap by more than a turn within a single continuum channel - resulting in the phases just being noise.
To calibrate, first check (in spd) that the delays aren't already more-or-less calibrated..... Observe a calibrator with high flux density: If the phase across the first zoom in each frequency more-or-less flat (i.e. significantly less than one turn across the band), no extra calibration is needed and you can just proceed as normal.
Assuming that the delays are large, they need to be determined in two steps:
First across a zoom band (i.e. one continuum channel)
Then across the whole continuum band
The setup scan needs to be on a calibrator with a high flux density with the continuum frequency at the same centre frequency as the observations and the first zoom, in each frequency must be 1 zoom wide (you cannot do the zoom calibration with stitched zooms).
Set calbands to “z z”
cacor> calband z z
In spd, check the phases in the first zoom of each frequency (z1 and z[n+1] where n is the number of zooms in the first frequency)
If these are not clean, that problem needs to be sorted before going any further. (Stop/Start is often a good thing to try first.)
In vis check that the delays are flat in all pols
(if they are noisy but spd looks good, the first thing to check will be delavg: increasing delavg will give better signal to noise, but it will need to be decreased if the phases are wrapping quickly).
When they are clean, delay calibrate:
caobs> corr dcal
This applies the delay found in the single zoom (z1 and z[n+1]) to the whole continuum band in each frequency.
The delays should now be sufficiently small that the phases in the continuum bands (f1 and f2) should be fairly clean and possible to calculate a linear slope across the band. (Again, if the delays in vis are noisy, there is a problem and it needs to be sorted out before going any further.)
Set calband back to “f f” - i.e. using the whole continuum band to calculate delays (and Tsys etc.)
cacor> calband f f
Check delavg: If it had to be increased for the zoom calibration, it will probably be too large for continuum
Check (in spd) baseline amplitudes and phases
Check that tvch is set to appropriate values
Check that delavg is set to an appropriate value
Standard setup follows (i.e. dcal, acal, pcal etc.)
If you get to the end of the calibration, and appear to have a working system, but that the Tsyss are high, check that the noise diode phases are flat.
In spd, select the autocorrelation ab products (i.e. the noise diode products)
Note: The AB products are only generated when in the continuum calband (i.e. calband f )
spd> acs on
spd> sel ab
spd> p -200 200
Check that the phases are not wrapping.
(Note:Though the phases generally wrap over many channels in the cross correlation products, they can be wrapping quite quickly in the auto-correlation products and delavg may need to be a small number, which may mean that you need a calibrator with more flux density to sort this out.)