Zero fringe rate problem


Problem

VLBI recording system writes on magnetic tapes/disks two signals: signal from a quasar and signal from the phase calibration generator. Due to Earth rotation normally the fringe rate from the quasars is significant, so the fringing procedure picks up the fringes which corresponds the signal from the quasar. However, under some circumstances fringe rate for a moment may appear zero even for very large baselines. It turned out that the current version of Fourfit can be confused and pick up fringes from phase calibration signal. It really happens. Mike Titus tells that he used to see about a half a dozen of such cases per experiment. In some cases Fourfit detects that something was wrong with the observation and assigns a letter quality code, in some cases it considers such an observation good and assigns quality code 9.

It is not a new problem, but apparently very well forgotten by analysis community. I found 5 such observations in gravt experiment and 1 observation in recent r1033 experiment among the observations with good quality code. In all these cases Fourfit computed group delay for phase cal signal, not for the signal from the quasar. Karen Baver reported zero fringe rate observations in NEOS Intensive sessions.

Examples of cases with zero fringe rate:

  1. Experiment grav01t, scan 079-1837, baseline HARTRAO-NYALES20
    Fringe plot
    Fringe amplitude
    Amplitude in the 14-th channel
  2. Experiment r1033 ($02AUG19XZ), scane 231-2353, baseline TSUKUB32/WESTFORD
    Fringe plot
    Fringe amplitude
    Amplitude in the 3-th channel

These plots are rather distinctive. These are fringes from phase cal. Since fringe rate is changing with time, it affects not all accumulation periods (AP), but only part of them. There are cases when zero fringe rate plague affects the observation at both bands, there are cases when it affects observations at only one band. Normally such points should appear as outliers since they have nothing common with observations of quasars and supposed to be suppressed during data analysis.

Solution?

Although, certainly, it is not the main reason of data losses, it would be desirable to find a solution. Possible solutions:
  1. Upgrade SCHED: to add a new check. If delay rate during integration time becomes less by modulo than the specified limit, then sched does not allow to include such an observation in schedule.


  2. Upgrade Fourfit: to add a feature which would allow to recognize such a case and blank APs with the fringe rate less by module than the specified limit out. Or maybe invent more clever procedure.


  3. Develop a stand-alone program which would analyze fringe phase and amplitude and detect the points where Fourfit found fringes from phase cal instead of fringes from quasars. The list of such points will be printed and the correlator analyst will try to manipulate with Fourfit windows and instruct Fourfit how to pick up fringes from the quasar.


Feedback

Ideas, comments, proposed solutions are solicited.

  1. Letter from Brian Corey of 2002.10.30_18:03:03


Back to Leonid Petrov's discussion page.

Leonid Petrov
30-OCT-2002 18:52:17