sedfitter

sedfitter copied to clipboard

I am wondering if you, or anyone else reading this, has considered making filter definitions for Herschel PACS 70 and 160 microns, and ATLASGAL 870 microns. Or is it safe to consider these as identical to the existing MIPS 70 and 160 filters and SCUBA 863 microns? I was about to start trying the sedfitter on a sample of objects observed by those instruments. Thanks, Todd

@trhunter - this package includes a function to convolve the models with arbitrary transmission curves:

http://sedfitter.readthedocs.org/en/stable/convolution.html

I would recommend doing this rather than using the existing MIPS and SCUBA filters. I will email you the transmission curves in case you are interested.

Sure, if you send me the curves, I will try the convolution.

Just sent you an email!

Thanks. Do you also have the longest wavelength PACS band (160 um) shown here? http://herschel.esac.esa.int/Docs/PACS/html/ch03s02.html#fig:photo_filter_chain_transmission

I think I just found it here: http://svo2.cab.inta-csic.es/svo/theory/fps3/index.php?id=Herschel/Pacs.red&&mode=browse&gname=Herschel&gname2=Pacs

It seems that utils.integrate (and hence filter.normalize) assumes that the data are in increasing frequency order (decreasing wavelength), like all the curves that you sent me. When a file is in reverse order (like the PACS3 file above), filter.normalize produces a negative response curve. I suppose it would be friendlier to check the order, or just take the absolute value.

@trhunter - I think I sent you the PACS3 file? (or is it missing from the email?). You are right that I should have a better error message if the order is wrong - I'll leave this issue open as a reminder! :)

Indeed you did send it! It looks like my mailer failed to "save all" attachments. The curve I retrieved from the website is very similar, but not quite identical to yours.

While working on a SOFIA FORCAST proposal and building convolutions with their filter curves (which are also in increasing wavelength order), I discovered that taking the absolute value of the response is not sufficient. You must reverse the order of the curve so that it is in the expected order (increasing frequency), otherwise the SED fits that you get are non-sensical but no obvious error message.

@trhunter - I ran into the same error recently, but I think it should now be fixed (both the order and absolute value) in the latest version - can you confirm?

Was convolution with Herschel filters ever solved? I have attempted it myself, unsuccessfully. I followed what I thought was the correct prescription from Robitaille et al 2007 Appendix A, but test SED fit results are wonky. Attached is a zip file that will allow you to reproduce my attempt (see README). Perhaps you can tell me what I am doing wrong? Thanks!

filterattempt.zip

Hi all,

I also seem to have run into a similar issue. Using the same method and internal function (which was successful for defining UKIDSS JHK) seems to return odd SED fit results for any of the Herschel/PACS filters. This seems to have been resolved for the Hyperion models (I see their files in the convolved filter group) but of course that's a different set. Has anybody been successful thus far with the 2007 model set?

Hi, can someone tell me how to define UKIDSS JHK filter for sedfitter ?