Imrm tricky things

[yihan Liu]

hello,

I recently found some tricky things about the rm and pa0 fitting.
I used L, S, C, and X band data to fit the rm and pa0 for a source. I used the qcut=0.001, while the fitting often looks as poor as the attached image.

That looks like some serious nπ ambiguity problems; while the final rm and pa0 images look pretty smooth, most pixels in the region I selected showed up, and no significant pa0 or RM value jumps seen as expected. More interestingly, I saw some negative slope in the rm pa0 linear fitting but never found a negative RM value in the final rm image. so, may I inquire what are the logic behind these tricky things?

btw, is it less proper to use C and X bands together for this fit? the imrm fitting algorithm may rely a lot on the first two frequency data (c and x bands). if the X band image looks noisy and as the λ^2 of c and x band are close to each other, would this introduce any troubles?

Many thanks for helping in advance!
Best

Yihan

[Mark.Wieringa]

Hi Yihan,

I haven't tried this program myself, but reading the docs it seems it can be tricky to get sensible results. It suggests to enter the images from high to low frequency (X, C, S, L), but if the highest frequency is very noisy you may need to try a combination with better S/N first.

I did not find an attached image with your post, so cannot see the ambiguity problem you noticed.

If your lower frequency images (L,S) have good S/N you could try splitting those up further into subbands (2 or more) and using those directly - the rotation between subbands may be less ambiguous and give a better estimate of the RM. Another option if you have good S/N is to use rmclean on the lower band.

Cheers,

Mark

[yihan Liu]

Hi Mark,

Many thanks I will try rmclean later to see if it is good!

Best
Yihan