





2.4.3.1.5 Calculate ILS Retrieval function
Level: 1b
Main objectives:
The Calculate ILS Retrieval function
performs the instrument line shape (ILS)
retrieval from radiometrically and
spectrally calibrated spectra. The
result of this operation is made
available to the output data
products.
Specific objectives:
Specific objectives of
the function are:
 Select specific microwindows
containing precisely one
reference peak of wellknown wavenumbers.
 Obtain or generate the reference
theoretical spectral line
corresponding to this microwindow.
 Fit an ILS to the
incoming spectrum by minimizing
residuals between the
reference line and the
parametric ILS.
 Store the iterated ILS parameter
set and the specific wavenumbers as
a Level 1b product.

Organigram:

Figure 2.10 
Input:
Output:
Detailed description:
ILS retrieval has been
studied extensively in the technical
note (see
Ref. [1.8 ]
). A
deconvolution approach has shown to be
inadequate, but a second approach
has shown to give good enough results.
The chosen ILS retrieval method
is called the “ Parametric
ILS Fitting Method
(PIFM). This method proceeds with a
theoretical ILS, obtained by a
modelization with a limited number of
parameters, convolved with the
theoretical line and iteratively fits
the results onto the experimental
data. Appropriate peaks for
spectral calibration that represent
known features of standard scene
measurements have been identified and
studied in the document (see
Ref. [1.8 ]
). The
precision of the peak identification
algorithm is proportional to the
number of equivalent scenes that are
coadded, as the noise affecting the
signal decreases when multiple readings
are superposed. This number will
probably vary between 2 and 10, and will
be defined in auxiliary data.
The operation of ILS retrieval is more
computer intensive than others tasks
presented up until now, but this
operation will be requested only from
time to time, not on a regular basis as
the computation of spectral
calibration for example. Topics of the
exact frequency at which the ILS
retrieval shall be done is addressed
here. It has been chosen to
extract the ILSin each detector
band of the instrument on an
appropriate spectral line located
anywhere inside the band. The list of
reference spectral lines will be stored
in a table kept as auxiliary
data. The auxiliary data
file containing retrieved ILS parameter data and
spectral calibration data shall be
produced by the Level 1B processor
according to the processing parameter
file. An initial ILS and spectral
calibration auxiliary file will be given
as an input to the processor at
all processing stations and shall be
used until the next file will be made
available. ILS and spectral
calibration data will be written to the
auxiliary file simultaneously (i.e.,
only ca. once per week). Otherwise
the file shall not be modified by the
processor.
2.4.3.1.6 Calculate Pointing function
Level: 1b
Main objectives:
The Calculate Pointing
function performs the line of sight
(LOS) pointing calibration in order to
generate corrected LOS pointing angles.
This includes:
 Compute correction of elevation
pointing angle,
 Compute corrected pointing
angles of actual scene (sweep).
Specific objectives:
Specific objectives of
the function are:
 Compute the actual pointing
error at time of ZPD crossing
 Compute actual azimuth pointing angle
 Compute correction of elevation angle
 Compute actual elevation
pointing angle
Organigram:

Figure 2.11 
Input:
 Measured LOS angles

LOS calibration data
Output:
Detailed description:
The Calculate Pointing
function is based on the following
assumptions. It is assumed that
commanded elevation angles are only
partially corrected with respect
to known pointing errors according to
the best knowledge based onground
characterisation and LOS calibration
measurements. The remaining elevation
error, obtained from LOS calibration
measurements, shall be computed in the
ground segment (PDS) and be used to
correct in measurement mode the measured
elevation angles. The corrected
elevation angles and the measured
azimuth angles are used to compute the
geolocation (height/longitude/latitude)
of the actual scene (target).
2.4.3.1.7 Calculate Geolocation function
Level: 1b
Main objectives:
The main objectives of
the Calculate Geolocation function are:
Specific objectives:
Specific objectives of
the function are:
 Compute orbital position of
spacecraft at ZPD time
 Compute tangent height,
longitude and latitude
 Estimate error on computed tangent height
Organigram:

Figure 2.12 
Input:
Output:
Detailed description:
The Calculate
Geolocation function calculates the tangent point
geolocation and related information. The
function has as input the orbit state
vector and corrected pointing angles.
2.4.3.1.8 Detection and correction of spikes
Level: 1b
Main objectives:
Detect and correct
spurious spikes in an interferogram
Specific objectives:
Specific objectives of
the function are:
 Inspect the interferogram around
the ZPD to detect
the presence of spikes
 Reject the interferogram if it
is in a calibration measurement,
otherwise replace detected
spikes by the mean of the
neighbor points.

Organigram:

Figure 2.13 
Input:
 Interferogram to inspect and correct
Output:
 Corrected interferogram
 Spikes position
Detailed description:
This function has the
purpose of detecting spurious spikes in
an interferogram. The
presence of spikes in an interferogram
can be caused by cosmic radiation or
transmission errors. The affected
points in a scene interferogram are
corrected by taking the mean between
immediate nonaffected points. This
scene will be flagged of having
corrected for one or more spikes. If a
spike is detected in a gain or in an
offset measurement, this measurement
will be discarded in order to
avoid corrupting all of the subsequent
calibrated spectra. The
algorithm performing spike detection
scans groups of points in the
interferogram (odd number with central
block corresponding to ZPD block) in search
of spikes. In each block, except for the
central ZPD region of the
middle block, the standard deviation of
the interferogram values is
computed, and a spike is identified if a
given point amplitude exceeds a
predefined threshold for values in the
real or the imaginary parts. To
improve the accuracy of the algorithm, A
second pass is done excluding the data
points identified as spikes to calculate
the final standard deviation of
the group. For each detected
spike, the value at the specific
wavenumber is replaced by a mean of the
two immediate points in the
interferogram vector. The real
part and the imaginary part are
corrected independently. The
spike correction will always cause some
distortions with respect to the original
spectrum, but it has been shown that
this distortion is within the
radiometric accuracy requirement.
Interferograms that have been corrected
for spikes are flagged as such.
2.4.3.1.9 Detection and correction of
fringe count errors
Level: 1b
Main objectives:
Detect and correct fringe
count errors (FCE) in the interferograms
Specific objectives:
Specific objectives of
the function are:
 For FCE detection:


FFT
the ZPD
region of
the interferogram
 Multiply the resulting
spectrum by the latest
available gain (interpolated)
 Calculate the spectral
phase of the roughly
calibrated spectrum
 Perform a linear
regression of the phase
vs. wavenumber
 Calculate the OPD shift
 For FCE correction:

 Perform the FFT of
the shifted interferogram
 Calculate the phase
function necessary to
correct the calculated shift
 Multiply the spectrum
with the calculated
phase function
 perform an inverse FFT on
the spectrum
Organigram:

Figure 2.14 
Input:
Output:
 Phase corrected interferogram
Detailed description:
The basic ground
processing for MIPAS contains no
explicit phase correction or
compensation. For a given interferometer
sweep direction, it is assumed that the
gain and offset calibrations and also
the scene measurements have the
same phase relationship, i.e. they are
sampled at precisely the same intervals.
This sampling is determined by a
metrology fringe counting system using a
reference laser source within the
interferometer subsystem, with the
fringe counts forming a “
clock signal to the ADC in the onboard
signal processor electronics (SPE). The fringes
trigger the sampling of the
interferogram. If, for any reason, a
fringe is lost, then the phase of
subsequent measurements will be affected
and if these are calibrated using a gain
or offset measurement taken before
the occurrence of the fringe loss, then
errors will be introduced into the final
spectrum. The ground processing scheme
includes a method for detecting and
correcting fringe losses by analyzing
the residual phase of calibrated
spectra, computed from the central
ZPD region of each
interferograms. Hence there is no
specific measurement required as
part of calibration for this aspect.
The proposed approach assumes
that fringe count errors occur at
turnaround, i.e. between two
measurements. Under this assumption,
the effect of a fringe count
error is to shift all measurements
following the error by N
points. The problem manifests itself
at calibration because all the
measurements involved may not have
the same sampling positions, i.e.
they do not have the same phase
relationship. Fringe
count errors occurrence within a
measurement is believed much less
probable, and its effect is the same
as if the error would have
been at the turnaround. Thus it
will be covered by the above assumption.
Fringe count errors can occur in all
types of measurements done by the MIPAS
instrument, except of course the LOS calibration
measurements during which the
sweeping mechanism is stopped.
Depending on the type of
measurement, the effect is not the
same and therefore, the detection
and correction approach will
be different. Because the phase is
not strictly the same for forward
and reverse sweeps, the fringe count
error detection and correction
will be done independently for the
two sweep directions. For all
measurements, the fringe count
reference interferogram of a
given sweep direction will be the
last gain interferogram of that
sweep direction. The last gain
interferogram can be either a
deep space or a blackbody
interferogram, depending on the
acquisition scenario requested.

Figure 2.15 Fringe Count Error handling 
Fringe count errors detection
The approach selected
for fringe count error detection
consists in a coarse radiometric
calibration of the actual
measurement at very low
resolution, followed by an analysis
of the residual phase. The
radiometric calibration is done
using the last available gain
measurement. When the optical path
difference (OPD) axis
definition of the actual
measurement is the same as the gain
used for radiometric calibration,
then the residual phase should be
zero. A shift will produce a
phase error increasing linearly with wavenumber.
The algorithm simply performs a
linear regression on the residual
phase of the calibrated spectrum to
reveal an integer shift due to a
fringe count error on the observed
interferogram. The spectral phase is
expressed as tan arctangente of the
ratio of the imaginary part
over the real part of the
spectrum.
Fringe count errors correction
Once the OPD shift is
known, the decimated interferogram
must be shifted by a
fractional number of points
corresponding to this shift divided
by the current DF. This requires
some sort of interpolation. The
current approach is to perform
a multiplication of the Fourier
transformed of the shifted IGM by
the phase function obtained in the
detection procedure.
With this method, no manipulation is
done on the OPD axis of the
interferogram, but each data
point is corrected to represent the
value of its desired current OPD
position. It should be
mentioned that fringe count errors
will affect interferograms of all
bands. For the MIPAS instrument,
detection is done only for bands C
and D. The approach for
fringe count error detection and
correction will be the same for all
types of measurements. However, the
implementation will be
somewhat different for the different
types. This is discussed below. The
fringe count error detection will be
performed systematically on
all incoming interferograms.
However, the correction procedure
will be applied only if a nonzero
shift is detected.
FCE handling in offset measurements
Detection and
correction are done with respect to
the last available gain calibration.
All the offsets corresponding to one
orbit are aligned to the
fringe count phase of this last
gain. If one or more fringe count
errors occur during the computation
of one orbit, the ground
processing will detect the same
shift for all subsequent offset
interferograms and will apply the
same (always recalculated)
correction on these offsets until
the end of the processing of the orbit.
FCE handling in gain measurements
At the beginning of a
gain measurement sequence, there is
no reference against which one can
check for fringe count errors. Thus,
there is no relation between
the actual measurement and the
previous fringe counting reference.
This is the main reason why we start
with a new gain
measurement. Fringe
count errors during gain calibration
are checked by comparison with the
first measurement of the sequence,
typically a blackbody
measurement (either forward or
reverse). The first step is to
determine the OPD shift
between that measurement and the
previous gain. The same procedure as
for normal error detection and
correction is then followed.
This corrected gain will then be
used for detection of fringe count
errors on all subsequent
interferograms. In principle, the
calibrated spectra obtained with
this corrected gain should show no
additional phase until a fringe
count error occurs. Then, all
errorfree measurements will be
coadded normally. Each time a fringe
count error will be detected, a new
coaddition group will be formed.
When the complete calibration
sequence is over, then all the
coadded measurements are corrected
with respect to the last measurement
and the remaining processing
of the radiometric calibration is
performed normally. Correcting the
gain with respect to the last
measurement presents the
advantage that all subsequent
errorfree measurements need no
correction. After
processing the data corresponding to
one orbit, if one or more FCE are
detected, the current gain is
shifted according to the last fringe
count error measured. This is done
in order to avoid correcting all
the offsets and scenes in subsequent orbits.
FCE handling in scene measurements
When a scene is
measured, its fringe count is
checked against the last available
gain calibration. All the scenes
corresponding to one orbit are
aligned to the fringe count phase of
this last gain. If one or more
fringe count errors occur during the
computation of one orbit, the
ground processing will apply the
same correction on these scenes
until the end of the processing of
the orbit. After that,
the gain is shifted according to the
last FCE to match the
offsets and scenes of
subsequent orbits. This way, the
worst that could happen is that all
the scenes of only one orbit would
need to be shifted. All the
subsequent processing of the orbits
to follow would not suffer
needlessly of a single previous FCE event.
This approach also minimizes the
accumulating of numerical error on
gains, that can be modified only
after successive orbits. In
practice, FCE are expected
to occur very infrequently during
processing of one orbit; but
even if this would be the case, the
fact of aligning offsets and scenes
to the last available gain
calibration would limit the
error accumulation on the gain
calibration vector. This
procedure will slightly increase the
throughput for the reference gains
used for the ground segment
computation. There will be one
each time at least one fringe count
is detected during one orbit. But,
as fringe count errors are expected
to occur infrequently, there
would usually still only be one gain
vector per week and, should an error
occur, only the gain would be
modified after the processing of
the corresponding orbit. The fact of
realigning gain calibration vectors
between orbits should save a lot of
operations, as one would
otherwise be correcting every
interferogram until the next gain
calibration (the saving occurs
independently of whether there are
frequent fringe errors or not).
2.4.3.1.10 Correct nonlinearity function
Level: 1b
Main objectives:
Correct the nonlinearity
of the response of the detectors of MIPAS.
Specific objectives:
Specific objectives of
the function are:
 apply the nonlinearity
polynomial on a detector per
detector basis for each interferogram

Organigram:

Figure 2.16 
Input:
 Interferogram to correct
 Set of nonlinearity coefficient for
each band/detector
Output:
Detailed description:
The detectors from the
first three MIPAS
bands (detectors A and
B) are photoconductive detectors,
subject to nonlinearity depending
on the total photon flux falling on
them. Here, the nonlinearity means that
the response of the detector differs
from a linear behavior as a function of
the incoming flux. This phenomenon
occurs at high fluxes. The
nonlinearity can be a source of
significant radiometric errors if it
is not properly handled (as much as
40% in band A). As explained
in (
Ref. [1.9 ]
), the nonlinearity
produces a change in the effective
responsivity as well as the
apparition of spectral artifacts.
The present method corrects for the
decrease of responsivity with DC
photon flux in the radiometric
calibration, within the required
radiometric accuracy. The approach
is the following: A
characterization must first be
performed on ground, and then in
space at specific intervals, at
instrument level, of the total
height of the unfiltered and
undecimated interferogram with the
onboard calibration blackbody at
different preselected temperatures.
These values will be used
during the characterization phase
for a computation of the nonlinear
responsivity coefficients. These
values will be used to correct
for the nonlinearity of the
detectors by means of a specific
algorithm called the Adaptive
Scaling Correction Method
(ASCM). Although they
are intended to be combined in a
single band, the optical ranges of
the detectors A1 and A2 are not the
same. They will then exhibit a
different behavior with respect to
photon flux. As a result, they will
require different nonlinearity
corrections. Because of this,
the signals from these detectors are
not equalized and combined on board
the instrument in the SPE. This
operation is instead performed by
the ground processor following
nonlinearity correction. The
other two PC detectors, B1
and B2, are not combined in any case
as they produce the bands AB
and B. Other than the need to keep
A1 and A2 separate in the baseline
output set up at the SPE, the
nonlinearity measurements and
correction has no impact upon the
calibration scenario.
The important effect of detector
nonlinearity is on the radiometric
accuracy performance. The present
radiometric error budget
allocated to the nonlinearity in
the 685 “ 1500
cm^{“1} (where the
detectors are the most nonlinear)
shall be better than the sum
of 2 x NESR and 5% of the
source spectral radiance, using a
blackbody with a maximum temperature
of 230K as source. A
polynomial correction is then
applied on each incoming
interferogram, at the very beginning
in the processing chain, with the
purpose of compensating for
the global effects of
responsivity. For the
measured responsivity curves of the
MIPAS engineering
and demonstration model (EDM),
the correction of the nonlinearity
error due to the change of effective
responsivity and from the cubic
artifacts have shown to lead
to an accuracy within the allocated budget.
The polynomial used to correct the
nonlinearity is of the form:
k = 1 +
d_{o} F +
d_{1} F
^{2} +
d_{2} F
^{3} + d_{3} F
^{4}
  eq 2.2 
where F is the
total flux on the detector (estimated as
the difference between the maximum and
minimum values of the digitized
interferogram by the ADC before filtering
and decimation and multiplied by the
amplification gains) and the d_{i}
are coefficients determined by the
nonlinearity characterisation for each
detector. In absence of
nonlinearity effect, these coefficients
are 0. The interferograms are corrected
by dividing them by this
polynomial.
2.4.3.1.11 Responsivity scaling
Level: 1b
Main objectives:
Scale interferograms
acquired with different gains to a
common baseline.
Specific objectives:
Specific objectives of
the function are:
 Scale interferograms acquired
with different gains to a common baseline.

Organigram:

Figure 2.17 
Input:
 Interferograms (gain, offset and
scene measurements)
 gains scaling factors
Output:
Detailed description:
The gains of MIPAS are adjusted
depending on the relative intensity of
the target so as to maximize the
dynamic range of the instrument. For
instance the gains are not the same for
deep space measurement and for CBB measurements.
Since interferograms acquired with
different gains will be combined
during the radiometric calibration
processing, it is necessary to scale
these interferograms to a common
baseline. In practice, three
scaling items need to be considered,
as a result of the preamplifier
warm (PAW) system:
1) A scaling to account for a
commanded gain change
The gains are predefined and are
commanded by an 8bit word sent via
the instrument control unit. Since
different gains may be
commanded, a data scaling in the
ground segment to equalize
performance must be foreseen. The
commanded gain is available in the
auxiliary data stream and so this is
a simple scaling effect based on the
extracted word.
2) A temperature dependent scaling to
account for changes in responsivity
of the detectors. The
detector units are specified to
provide a stable response based upon
assumed knowledge of their
temperature (i.e. the responsivity
may vary but it must be well
characterized). For this reason, a
correction of performance with
time/temperature must be foreseen.
This is made based on the
measured detector temperature
(available via thermistor values in
the auxiliary data) and using
characterization curves
generated during characterization
tests on ground.
3) A temperature dependent scaling
(gain & possibly phase) to
account for the variations in the
performance of the electronics of
the PAW and the SPE
around the orbit. At
present, it is not thought necessary
to correct for these effects around
the orbit as predictions show the
variations will not cause the
units to drift out of specification.
The InFlight
Calibration Plan
Ref. [1.1 ]
foresees to make around orbit
measurements during Commissioning
Phase to check whether there are any
such variations.



 
Keywords: ESA European
Space Agency  Agence spatiale europeenne,
observation de la terre, earth observation,
satellite remote sensing,
teledetection, geophysique, altimetrie, radar,
chimique atmospherique, geophysics, altimetry, radar,
atmospheric chemistry
