POLARIS CALIBRATION
Polaris was calibrated in the lab at Calgary, and was subsequently
cross-calibrated towards the Gillam ASI (Wilbur) using data gathered
during the 2000 Polaris installation trip to Gillam, during which
Polaris operated in parallel with Wilbur.
The purpose of the calibration exercise is to produce a conversion
factor to allow users of Polaris data to convert from raw image pixel
`data numbers' (dn) to physical units of auroral column emission rate
(rayleighs, R) for each wavelength employed. These are the R-Values.
Also, necessary to enable the use of R-Values at any point within the
field of view (not just the zenith) is the generation of a representative
flat-field frame, similar to the Q-Arrays used with Wilbur ASI data.
The first section below describes the result of the Calgary calibration,
using the Calgary 1999 calibration data set. The second section describes
the result of the Wilbur vs. Polaris cross-calibration. The final section
briefly describes how the two calibration methods compare.
1. Polaris calibration (Calgary, the 1999 calibration data set)
---------------------------------------------------------------
Polaris was calibrated in the lab at Calgary, resulting in R-Values for
all wavelengths used on Polaris as well as effective bandwidths for each
of the filters. A representative flat-field correction frame was also
generated, and is shown here (in a GIF version):
The calibration procedure employed for Polaris is identical to the one
used over the years with the old Gillam ASI, Wilbur. The absolute responsivity
of Polaris was measured using our standard brightness source, LBS-A3. Filter
bandwidth data was obtained by measuring a beam of collimated light from our
Acton monochromator. These data are then combined to obtain the final R-Value
for each filter.
Due to the enormous amount of data associated with all combinations of
filters, integration times, gain settings, and LBS aperture settings,
and the fact that 10 frames are exposed for EACH setting, analysis had
to be restricted to the `standard' policy of 1-second integration time (T=1)
and medium gain setting (G=1). The result of the analysis is as follows.
--- BEGIN ---
Calibration series: POLARIS_CGY1999_001000ms
Exposure time is 1.00000 seconds
Imager Gain is 1
A. FILTER BANDWIDTH DATA
Filter BW(eff)
--------------------
4278 2.80319 nm
4806 3.07599 nm
5577 2.69494 nm
6300 2.30366 nm
--------------------
B. FINAL 'R' VALUES [dn/R/s]
Filter LBSd06 LBSd07 LBSd08 LBSd09 LBSd10 LBSd11
-----------------------------------------------------------------
4278 0.000412 0.000434 0.000456 0.000499 0.000564 0.000612
4806 0.000739 0.000799 0.000861 0.000944 -- --
5577 0.000681 0.000726 0.000785 -- -- --
6300 0.000401 0.000433 0.000466 0.000512 -- --
-----------------------------------------------------------------
Computer run at Wed Mar 8 14:20:47 2000
--- END ---
The highest LBS aperture with a valid R-Value is normally used as the `official' value.
This is the value, of all values for that wavelength, with highest signal-to-noise
ratio. I.e., for 557.7 nm, the `best' R-Value is: 0.000785 dn/R/s. (In this case, LBS
apertures of D09 to D11 caused saturation in image pixels and had to be discarded).
2. Polaris/Wilbur ASI cross-calibration (Gillam, data from 2000 field trip)
---------------------------------------------------------------------------
Polaris and Wilbur data from March 31, 2000 were used during this calibration
exercise. Polaris data was dark-frame subtracted and flat-field corrected.
Wilbur ASI data was dark-frame corrected and calibrated using the 1999 P- and
R-Arrays. Without referring to the lab calibration (Part 1, above), an attempt
was thus made to infer rayleigh intensities in Polaris images by closely studying
the calibrated Wilbur ASI data and making comparisons.
Specifically, Wilbur ASI data was calibrated as follows:
a. SUBTRACT the current hour's dark frame from image frame
b. DIVIDE resulting frame by P-Array (has the expected flattening effect)
c. DIVIDE the resulting frame by the R-Value for this wavelength
d. DIVIDE the frame that then results by (104 x 16)/1000.0 (exposure time in s)
The following empirically derived expression gives the approximate rayleigh
emission rate as a function of Polaris pixel data number (dn) for a given
exposure time (T) in seconds and gain (G, in the range 0 to 3) at a wavelength
of 557.7 nm:
Emission Rate [kR] = .6 * 1.5^(3 - G) * dn / T
This, again, is based on a comparison of raw (uncalibrated, but field
flattened) Polaris data and calibrated Wilbur ASI data.
The dynamic range of the inexpensive detector currently employed by Polaris
(a Pulnix TM-745 CCD RS-170 video camera) is in the range 25 - 30. The
analog-to-digital converter used is 8-bits, the most sensible choice for the
kind of detector used. The following table gives approximate ranges of
detectable auroral emission rates (in units of kR) for exposure times of
one, two, three, and six seconds and gain settings of zero to three:
--- BEGIN ---
+--------+--------------------------------------------------------+
| 5577 | T = 6 s T = 3 s T = 2 s T = 1 s |
+--------+--------------------------------------------------------+
| G = 0 | 3.5 - 86 7.0 - 170 10 - 260 20 - 515 |
| G = 1 | 2.3 - 57 5.0 - 115 7.0 - 170 15 - 345 |
| G = 2 | 1.5 - 40 3.0 - 75 5.0 - 115 9.0 - 230 |
| G = 3 | 1.0 - 25 2.0 - 50 3.0 - 75 6.0 - 150 |
+--------+--------------------------------------------------------+
* Emission-rate ranges within table are in units of kilo-rayleighs [kR] at 557.7 nm.
This table Tue Jun 6 11:00:35 MDT 2000
--- END ---
3. A comparison of the two above calibration approaches
-------------------------------------------------------
As a test of how the above two procedure compare, we insert T=1 and
G=1 in the empirically derived formula of Part 2 above, and compare
directly to the result of Part 1.
For an exposure time of T = 1 second and a gain setting of G = 1 (medium gain)
we obtain a value of,
Responsivity = 1.35 kR/dn = 1350 R/dn (557.7 nm)
In Part 1 we obtained for 557.7 nm, with the same gain setting,
R = 0.000785 dn/R/s
For a 1-second exposure this corresponds to 1273 R/dn (557.7 nm).
The results thus compare as well as could be expected, allowing one to
estimate emission-ranges as shown in the table of Part 2 also for the
other wavelengths involved by using the calculated R-Values of Part 1
as a starting point. Note finally that the same LBS (and calibration
procedure) was used in calibrating both instruments. Note further that
resulting Wilbur ASI Rayleigh values have always traditionally seemed
somewhat too high, by about a factor of two [E.P. King, personal
communication]. This again brings up the ancient desire of achieving a
cross-calibration with the HIA LBS, which has not yet been done. This
whole calibration issue is thus not (yet) considered an exact science,
and DERIVED RAYLEIGH VALUES SHOULD THUS BE USED WITH CAUTION.
Trond S Trondsen - Tue Jun 6 16:51:24 MDT 2000