next up previous contents
Next: Setting up Pulsar Gating Up: Fringe Searching and Pulsar Previous: Clock Synchronization   Contents

Finding Fringes and Calculating the Clock Model

On elleba, if you have only just started lbaboss, type:

lbaboss$>$ s2init

which will initialise the correlator

s2init should only be run once. If you are unsure whether it has already been done, then exit from lbaboss, restart it and then do an s2init.

Now let the correlator know which schedule file it should be reading;

lbaboss$>$ s2file v088d-3

to select the file v088d-3.sch

Now, you need to configure the correlator mode, which sets up the number of channels correlated, the bandpass, the number of pulsar bins (if required) etc. Go to the grey lbacor window (running from lbaccc on the lbacop-mm screen) and in the top right hand corner you will see a small one-line display window which will be showing the last correlator configuration used by the correlator. Below this display window is another small window where you can scroll and select the configuration you need. So, highlight the appropriate configuration with your mouse and hit the CONFIG button. A small window will pop up on screen with a kind of scrolling display indicating that the correlator is programming itself into the configuration you've requested. Occasionally there is a slight pause for a few seconds, which is perfectly normal. However, if the pause is for a significant length of time or if it hangs, then grab one of the correlator personnel - Warwick, Paul, Tasso or John.

The configuration filenames in general are of the form;

${\small <}$number of baselines${\small ><}$b $\vline$ ac $\vline$ bac${\small >}$_${\small <}$bandwidth MHz${\small >}$_${\small <}$#frequency channels per product${\small >}$_${\small <}$nothing $\vline$ 2p $\vline$ 4p $\vline$ 2f${\small >}$

nothing 1 frequency, 1 polarisation
2p 1 frequency, two polarisations (LL RR)
4p 1 frequency, four polarisations (LL RR LR RL)
2f 2 frequencies, 1 polarisation each
b Standard cross-correlation configuration, useful for most of the correlating that we want to do
ac Auto-correlations only (the number preceding is number of stations), useful for spectral line observations with lots of channels
bac cross- and auto-correlations only - useful for looking for fringes

Sometimes there is psr$\small <$number of pulsar bins$\small >$ at the front of the filename. The number of pulsar bins is: 8 16 32. If you have pulsar observations, you will set up your number of bins and gating a little further on.

If you are using less than 7 sets of transports (eg 2 sets only - at11, at06) type:

lbaboss$>$ s2off 3 4 5 6 7 to turn off transports 3, 4, 5, 6 and 7
lbaboss$>$ s2rewind to rewind tapes to beginning
lbaboss$>$ s2play to start tapes playing
lbaboss$>$ s2align 13:10:00 to align tapes on calibrator data (see below)
lbaboss$>$ cycle 2 .0625 2 second correlations, blank out 62.5 ms of data because of blank pulse at Mopra
lbaboss$>$ go to start correlating

After a period of time, the tapes will align. First each set of 8 tapes in each S2 playback unit aligns with each other to a round number of seconds, then the different S2 playback units align to the same time. This can take a minute or two and a lot of beeping will come out of lbaboss in the meantime. When everything is lined up and going the message:
Underway at hh:mm:ss
will appear in the lbaboss message window. If this doesn't happen after a few minutes, then the tapes may have a problem. Commands to the S2 playback units can be given in their windows, eg.
lbaboss$>$ align selfalign
lbaboss$>$ align 13:10:30 (at the time the other S2 units show this time)
If a tape within a set of 8 is having problems being read in that drive, it is often possible to swap two tapes around.

If it does get underway, check that the error rates on each S2 console is low ($< 10^{-4}$) by typing:

s2console$>$ transport display esterr

If errors on a particular transport are too high, wait and see if they reduce. If they don't, get help. Also swapping tapes around within a set of 8 can often help this with recalcitrant tapes or drives.

To redisplay the time information on the tapes type:

s2console$>$ transport display time

Now we want to examine the lag spectrum using SPD, being run on lbaccc:

SPD: l

we now have a plot of correlation coefficient - lag.

The sign of success here in finding fringes is the spike in the lag spectrum. This depends to a large extent, on having good parameters in the schedule file, the most important of which are the delay and the rate at which delay is changing. You should look at correlator schedule files from experiments around the same date (most VLBI runs occur over a given week with other experiments) to find good first guesses for each antenna used. The window over which the lag is calculated depends on the correlator configuration, so while you may eventually want to correlate say with a particular correlator configuration, to find the fringes on a rogue antenna with rather uncertain lag, a search with more channels (eg. 1b_16_1024_2f, 1 baseline, 16 MHz, 1024 channels) may be needed to give a wide lag search window.

For this particular experiment, there is a fringe at channel -22
We now work out the error in delay:
-22 channels $\rightarrow$ delay $=$ -(-22/32)$\times$1000 $=$ 688 ns
(the channel number is divided by 2$\times$bandwidth in MHz)

We can see this offset in TVD by making sure we have the delay scaled appropriately;

TVD: scale d 668 708 (to scale delay between 668 and 708 nanoseconds)

And we can also use TVD to work out the exact error in delay:

TVD: cal
which starts an automated delay calibrator program
TVD: ref antenna 1
TVD: select polarisation aa or bb
TVD:  

A small program within TVD will calculate the offset from zero for each baseline from the reference antenna and then give you an average. The number given should be added to the delay in the schedule file for that antenna. And in this example we see that the exact shift in delay is 683 ns.

So, in the v088d-3.sch file, we have for the second station (CAT), that the offset of CAT from Mopra was given as -57.783 ms. Thus we have to subtract 683 ns from this delay.

Thus new offset is -57.783 - 0.683 $=$ -58.466 ms. This new figure replaces the delay figure for CAT in the v088d-3.sch file. i.e. we replace -57.783 with -58.466

We've taken a while to do this, so we want to look at the data again. To change the relative alignment of the tape on lbaboss we type:

lbaboss$\small\bf >$ s2align -3m (to realign the tapes back by 3 minutes)

And we now check on SPD that the lag has moved to channel 0 and on TVD we should see the delay also come back to zero;

SPD: l (display letter L for lag spectrum)
TVD: scale d -50 50

It has worked !

We also check that the phase is flat, since a non-zero delay gives a phase gradient with frequency across the band, and again we can use SPD and/or TVD;

SPD: p (display letter p for phase)
TVD: tapd (to make sure we are displaying all 3 of amplitude, phase and delay)

Now we need to determine the rate at which the delay was changing. So, we work out the delay correction at another time (this may be some time later, from a different set of tapes). Thus we have;

Delay was -58.466 at 1997 212/1315 UT
Delay was -58.176 at 1997 213/1309 UT

And we can calculate the rate;

rate $=$ (change in delay)/(change in time)
  $=$ (-58.176 - -58.466 $\mu s$)/(86040 s)
  $=$

$3.3705 \times 10^{-6}\mu s/s$

This rate is substitued for the current rate (0) that is on the CAT21 line of the v088d-3.sch file.

That is we replace 0 $\rightarrow$ 3.3705 E-6

Once all the delays and rates have been checked and entered into the correlator schedule file, you are satisfied the phases are relatively flat and all is well - you are ready to start correlating and writing the data out to an RPFITS file. If you have pulsar observations though, you have one more step beforehand - setting up the pulsar gating.

next up previous contents
Next: Setting up Pulsar Gating Up: Fringe Searching and Pulsar Previous: Clock Synchronization   Contents

Paul Jones 2003-06-13
ASKAP
Public