Opened 9 years ago

Closed 9 years ago

#354 closed bug (fixed)

Hawk 2010 spectral calibration shifted

Reported by: mggr Owned by: mggr
Priority: immediate Milestone: 2010 data processing completion
Component: Specim Keywords:
Cc: mark1, benj Other processors:

Description

On examination, the Hawk spectral calibration appears to have shifted between 2009 and 2010. However, when processing using the new, shifted calibration, we see absorption features misplaced by the same amount as the shift.

This implies the spectral calibration is incorrect.

Attachments (1)

hawk_2010_dataset_vs_smarts.png (6.3 KB) - added by mggr 9 years ago.
2010 calibrated data vs model

Download all attachments as: .zip

Change History (3)

Changed 9 years ago by mggr

2010 calibrated data vs model

comment:1 Changed 9 years ago by mggr

Detailed description of problem in response to Specim's request for more information:

Hi Markku,

Thanks for getting back to us :)

On 12/08/10 10:58, support.aisa wrote:
[paraphrased: please describe what you did, so we're both sure you didn't make a mistake]

Yes, we were worried we were doing something wrong and checked pretty
carefully.  I'll explain exactly what we've done so you can verify.

We took the wlcalSWIR.wls files directly taken from the Hawk system and
from the 2010 calibration.  To convert, we just map from one to the
other (ie. line 1 -> line 1, line 2 -> line 2, etc), which is equivalent
to looking up the raw band number and relabelling the wavelength.
Here's a combined version to show you what I mean:

   (in use wls)              (2010 wls)
   wavel    fwhm           wavel    fwhm
  904.82    6.33          928.41    6.31
  911.14    6.33          934.72    6.31
  917.47    6.32          941.03    6.31
  923.79    6.32          947.35    6.31
  930.12    6.32          953.66    6.31
...
  999.65    6.32          1023.10    6.31
  1005.97    6.32         1029.41    6.31
  1012.29    6.32         1035.73    6.31
..
  2478.00    6.21         2512.91    6.31
  2484.21    6.21         2519.23    6.31
  2490.42    6.21         2525.54    6.31
  2496.63    6.21         2531.85    6.31
  2502.83    6.21         2538.17    6.31

You can clearly see the shift in the .wls files.  I've attached these
wls files again so you can compare yourself.  I've also attached the
combined version.

It's also interesting to note that the old Hawk fwhms are not constant.
 Checking our copy of the 2009 calibration, I see the bandset that Ops
have been using does not match 2009, nor 2006 (we don't have 2007 or
2008 calibrations).  In my previous email I said it was from 2009, but
it appears this is not true.  We don't know where this bandset / wls
came from - we just know it's the one they've used.  This shouldn't
matter though, as we only use it to compute the mapping above (ie. which
wavelength maps to band N, and how band N maps to the 2010 bandset).

We also verified that the 2010 wlcalSWIR.wls matches the 2010
calibration.  We did this by just checking the wavelengths in the .hdr
file (attached), and they match.  Here's the relevant snippet:

----------
Wavelength = {
 928.41,
 934.72,
...
2525.54,
2531.85,
2538.17
}
fwhm= {
   6.31,
   6.31,
----------

At this point, we believed we'd established:
 - that Ops have been using an incorrect wavelength labelling
 - that we can translate this to the 2010 one by mapping against the raw
band number (using all 256 bands via the .wls file)
 - that the calibration for 2010 was shifted

We then wanted to test to see which was correct, and decided to check
for absorption bands on a real flight line.

We converted the raw hdr file of a real flight line from March 2010 that
had been captured with the old calibration.  The conversion was done
using the wls mapping above, e.g. wavelength 999.65 -> 1023.10 - I've
attached an original and translated .hdr file.  We processed it and
compared the output spectrum to the model.  The old wavelengths appear
to line up correctly with absorption features, while the new ones do not.

I think our conversion process is correct (please tell us if we've made
a mistake!), which implies the calibration may be wrong.

Cheers,

Mike.

comment:2 Changed 9 years ago by mggr

  • Resolution set to fixed
  • Status changed from new to closed

Since no-one else has update this ticket.. The Specim calibration was confirmed to be incorrect by a quick test, by an internal ARSF calibration and by Specim. We have used our own calibration for processing 2010 data. No cause is known for the error, but presumably it was just a mistake rather than anything physical as the wavelength calibration was unchanged from 2009.

Note: See TracTickets for help on using tickets.