Failed to determine covariance matrix in uvfit

Is MIRIAD being a pain? Let us know your experience.

Moderator: Mark.Wieringa

Post Reply
bnorfolk
Posts: 45
Joined: Fri May 17, 2019 3:56 pm

Failed to determine covariance matrix in uvfit

Post by bnorfolk »

Hi,

From my previous post I'm attempting to measure the flux of my target with uvfit.

Upon the previous suggestion to implement: uvfit vis=myvis line=chan,200,1,10,10 object=point

I receive the following errors:

uvfit: Revision 1.10, 2013/08/30 01:49:21 UTC

Reading the data ...
Applying bandpass corrections to hd100453.33000
Applying gain corrections to hd100453.33000
### Warning: Performing linetype averaging before applying bandpass!!
### Warning: ... this may be very unwise
Total number of correlations: 16472030
Performing the fitting process ...
### Warning: Failed to determine covariance matrix
------------------------------------------------
RMS residual is 6.648E-02

Source 1, Object type: point
Flux: 0.000
Offset Position (arcsec): 0.0000 0.0000
------------------------------------------------

I'm unsure why I'm receiving the bandpass error even though the uvfit task prints the bandpass solution is being applied and more importantly, I'm unsure why determining the covariance matrix fails and I receive a zero flux reading.

Any suggestions would be greatly appreciated.

Thanks,
Brodie
Mark.Wieringa
ATCA Expert
Posts: 297
Joined: Mon Feb 08, 2010 1:37 pm

Re: Failed to determine covariance matrix in uvfit

Post by Mark.Wieringa »

Hi Brodie,

The 'line type averaging' error appears in a few cases and happens because of the order of operations when reading/averaging/calibrating data on the fly.
The easiest way to avoid it is to use uvaver on the data, creating a fully calibrated dataset, and then use that as the input for uvfit.

The 'Failed to determine covariance matrix' might mean there is no source near the centre of the field? It probably has to be within a few beam widths for the fit to converge to it.
If the source is somewhere else, you can provide a first guess of the parameters using the 'spar' parameter.
If it still doesn't find the source, it may be too faint, or there may be too many other sources nearby that confuse the fit. Doing the fit in the image plane is then a better option (imfit).

Hope that helps, if not, let us know.

Cheers,

Mark
Post Reply