2.4.3 Level 1b products and algorithms
2.4.3.1 Algorithms
The goal of the level 1b
processing is to transform the
interferograms generated at the end of the
level 1a processing into calibrated and
corrected spectral radiance spectra. The
overall processing , divided in high level
functions, will be processed in the
following order:
The following flowchart shows how
the Level 1b ground processing is organised:

Figure 2.5 Level 1B processing flowchart 
During the Level 1b processing there are also
some auxiliary functions, applied to both
the scene and the calibration data, that are
part of several of the above high level
functions. These auxiliary functions are:
All these algorithms are
described in more details in the "Algorithm Technical
Baseline Document for MIPAS Level 1b
Processing
Ref. [1.6 ]
".
2.4.3.1.1 Calculate Offset Calibration function
Level: 1b
Main objectives:
The main objective of the
Calculate Offset Calibration function is
to deliver offset calibration
measurement data in a form suitable for
radiometric
calibration of the spectra by the
Calculate Radiance function.
Specific objectives:
Specific objectives of
the function are:
 Perform spikes detection
 Sort offset data according to
the direction of interferometer
sweep.

Coadd six
interferograms in each band.
 Detect and correct fringe count
errors in spectral bands C and D.

 Gain spectral interpolation
 Calculate coarse spectra
 Calculate calibrated spectra
 Responsivity scaling
 Correct for detector nonlinearity.
 Equalize and combine
interferograms in band A.
 Assess NESR performance.

 Accumulate statistics
from deep space readings
to obtain the NESR
of the instrument
 Check the validity of
incoming readings
Organigram:

Figure 2.6 
Input:
Output:
Detailed description:
The radiometric
offset is an estimate of the instrument
contribution, due to its elfemission,
to the total measurement. This estimate
of the instrument contribution is
made by simply pointing the instrument
to deep space and performing a
measurement cycle as in the nominal
case. For practical reasons, the deep
space measurement is taken at a
tangential height of around 150 km.
Also, due to the potential
difference in phase between different sweep directions of
the instrument, a measurement is
taken in each of the forward and reverse
directions of the interferometer. In the
ground segment, the closest in time
offset measurement (in the correct sweep direction) is
simply subtracted from each
interferogram during processing. Click here for details
on the offset measurement 1.1.3.4.1. .
The signals detected during
offset measurements arise mainly
from noise sources in detectors/
amplifiers and from thermal emission
of the optical components
within the interferometer. Even if
the spectrum will be weak, it is
believed that fringe count errors
can be effectively determined.
The scheme applied to scene and
calibration measurements will most
probably detect the occurrence of
fringe errors, and the use of
all interferograms (including
offsets) maximizes the chance of
detecting and correcting the errors
as soon as possible after
their occurrence.
The zero offset measurements will be
subtracted from the relevant
individual interferograms.
Logically, these measurements should
be made at the same spectral
resolution as the scene measurements
themselves, in order that the
vectors are directly comparable.
However, it is not expected
that any high resolution features
will be present in the offset
spectra, which means that the
measurements may be made at
low resolution, with an
interpolation on the ground segment.
This has the advantage of reducing
the duration of the offset measurement.
Offset calibration is performed such
that the closest in time available
valid offset measurement is used
until a new valid offset
measurement becomes available. If no
offset data are found at the
beginning of the Level 0 product
input data set, then the first
available offset found leading to
valid measurement shall be used for
all initial scenes. If one or more
invalid offset measurements are
detected in the middle of the input
stream, then a “closest in
time strategy shall be applied,
which means that complete scans
shall be calibrated with the closest
valid offset. If no valid offset at
all is found in the input data, then
the offset calibration data
contained in the offset validation
file shall be used.
2.4.3.1.2 Calculate Gain Calibration function
Level: 1b
Main objectives:
The main objective of the
Calculate Gain Calibration function is
to deliver a file representing the
radiometric gain of the instrument,
computed using gain calibration
measurements, in a form suitable for radiometric
calibration of the spectra by the
Calculate Radiance function.
Specific objectives:
Specific objectives of
the function are:
 Perform spikes detection
 Sort the gain calibration
measurements according to types
of measurement and sweep direction.

Coadd
interferograms to increase SNR.
 Detect and correct fringe count
errors in spectral bands C and D.

 Gain spectral interpolation
 Calculate coarse spectra
 Gain shift correction
 Calculate calibrated spectra
 Responsivity scaling
 Correct DS and CBB
measurements for nonlinearity
of each affected detector.
 Subtract offset due to
contribution of the instrument.
 Equalize and combine
interferograms in band A.
 Compute coarse spectra using a
FFT algorithm
applied on the zeropadded interferograms.
 Interpolate gain spectral
vectors to provide the gain on a
predefined spectral axis.
 Calculate expected blackbody
radiance from temperature
readings corresponding to
blackbody measurements.
 Calculate the complex ratio of
theoretical to calculated
spectrum (gain computation).
 Gain coaddition
 Check for radiometric accuracy
of the incoming data.
Organigram:

Figure 2.7 
Input:
Output:

Calibration
gains [MIP_NL__1P: GAIN
CALIBRATION ADS #1 ] 6.4.2.

Spectral
accuracy data for validation
[MIP_NL__1P: GAIN CALIBRATION
ADS #2 ] 6.4.2.
Detailed description:
The radiometric gain
calibration requires all deep space and
blackbody measurements of the gain
calibration sequence. Since in this case
the instrument is again contributing to
the observed signal, it is also
necessary to perform deep space
measurements before the blackbody
measurements in order to subtract the
appropriate instrument offset. (In this
instance, the term "Deep Space
Radiometric Calibration" is
used to distinguish the measurements
from the regular Offset Calibration made
with the scan sequences. The Deep
Space (DS) Radiometric Calibrations are
used only to correct the Calibration
Blackbody (CBB) measurements and
must be explicitly commanded. In fact,
several measurements of each kind will
be needed. This is because the
signal to noise ratio of a single,
offsetcorrected, blackbody measurement
is not high enough, particularly in band D, to achieve the
required radiometric accuracy. Therefore
a single gain calibration implies
several successive measurements.
It is expected that there will be no
high frequency features in either the CBB spectrum or
in the instrument contribution (as
assumed also for the offset
calibration). These assumptions will be
verified on the ground during
instrument Assembly and Integration Test
(AIT), but the assumption is reasonable.
Therefore, each CBB or Deep Space
sweep of the instrument will be made at
lowspectral resolution, i.e. with
a duration of 0.4 seconds. The baseline
scenario uses 300 sweeps at low
resolution in both forward and reverse
directions for both CBB and DS measurements.
The gain data is processed by the Level
1b processor at the beginning before
scene data. During processing, the gain
file is not be modified by the
processor.
2.4.3.1.3 Calculate Spectral Calibration function
Level: 1b
Main objectives:
The Calculate Spectral
Calibration function performs the
processing of some selected
(radiometrically) calibrated scene
measurements and generates the
spectral calibration data.
Specific objectives:
Specific objectives of
the function are:
 Compute a corrected spectral axis
Organigram:

Figure 2.8 
Input:
 Radiometrically and spectrally and
locally calibrated (RSL) atmospheric spectra
 List of reference spectral lines
Output:

Corrected spectral
axis (spectral calibration data)
Detailed description:
Spectral calibration is
performed in MIPAS using standard
limb measurements from the
atmosphere already corrected for the Doppler effect by
the Calculate Radiance function.
Specific reference spectral lines will
be retrieved in the observed spectra
according to the extremities of specific
microwindows listed in a reference lines
database. From these microwindows will
be performed the line position
identification, with respect to a
database containing the exact known
theoretical position of the reference
lines. In order to reduce noise,
equivalent scenes are coadded,
i.e., scenes with altitude included in
the range of the processing parameter
file. The computed known values of the
reference lines positions will be used
to establish the assignment of the
calibrated wavenumber to the index of
spectral data points. Following
this operation, spectral calibration
will be used for the wavenumber
assignment of all subsequent
measurements until a new spectral
calibration is performed.
The Calculate Spectral Calibration will
be performed when it is appropriate to
update the spectral calibration, with a
current baseline of twice per
day. Because it is related
to the same parameters, the spectral
shift can be considered as a part of the
instrument line shape. The
disadvantage is that it is then
necessary to perform a deconvolution of
the ILS from an
observed spectrum to get the proper
wavenumber assignment. Here we will
assume that the spectral shift is
included in the spectral
calibration, i.e. it is calibrated out
by the spectral calibration procedure
without any ILS deconvolution.
It is also assumed that the spectral
calibration will be the same throughout
the spectral range. It is assumed that
the definition of the optical axis
is common to all four detectors on the
output ports, for both output ports. It
is also assumed that the residual
misalignment between the two output
ports is low enough so that the
difference in wavenumber is
negligible. Two algorithms
have been proposed to perform spectral
calibration: the Peak Finding Method
(PFM) and the CrossCorrelation Method
(CCM). The feasibility of both
these methods have been demonstrated,
and both algorithms have demonstrated
strengths and weaknesses. The PFM has
shown to be a little simpler to
implement and faster to execute, but the
CCM presents the
advantage of giving information related
to the precision of a given fit. A
switch between in the level 1b processor
setup allows the selection of one
of these two methods. The
PFM method uses an
analytical model to describe
the target line minimising the
squared difference between the
modelled and observed spectral lines
within preselected spectral
windows. The optimisation involves
the simultaneous fit of four
independent parameters using a
simplex algorithm. The fitted
parameters correspond to an additive
offset, the line width, a line
amplitude scaling factor and the
line centre wavenumber.
For the
CCM method, the
crosscorrelation function of the
measured spectral line and a
modelled spectrum (within predefined
spectral windows) is computed. The
frequency shift in the observational
data is obtained by computing the
position of the peak in the
crosscorrelation function.
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 1 and 5
(to attain stability and a
precision equal or less than 0.001
cm^{“1}), and will be
defined in auxiliary data.
Spectral calibration is performed
such that the latest available valid
spectral measurement is used until a
new valid spectral measurement
becomes available. If in the middle
of the input stream invalid spectral
calibration are calculated, then a
“previous closest in
time strategy is applied, which
means that complete scans shall be
calibrated with the previous valid
spectral calibration. If no
valid spectral calibration at all is
available, then the spectral
calibration data contained in the
current ILS and spectral
calibration file is used. Spectral
calibration data is written to
auxiliary file simultaneously with
ILS retrieved
data. Otherwise the file is not be
modified by the processor.
2.4.3.1.4 Calculate Spectral Radiance function
Level: 1b
Main objectives:
The Calculate Radiance
function performs the processing of the
scene measurements and generates a
radiometrically calibrated spectrum.
This function assumes that gain, offset
and spectral calibrations are available
as soon as they are produced, so that
they can be used for the
processing of all scene measurements
following these calibrations. If this is
not the case, then processing will
proceed with the latest available
calibration data.
Specific objectives:
Specific objectives of
the function are:
 Perform spikes detection
 Detect fringe count errors in
spectral bands C and D,
and in the case of
misalignment adjust the phase of
the gain and offset according to
the current fringe count.

 Gain spectral interpolation
 Calculate coarse spectra
 Calculate calibrated spectra
 Responsivity scaling
 Correct scene measurements for
nonlinearity of each affected detector.
 Equalize and combine
interferograms in band A.
 Subtract offset due to
contribution of the instrument.
 Compute spectra using a FFT algorithm
applied on the zeropadded interferograms.
 Correct spectral axis for Doppler
shift and perform spectral
interpolation onto a predefined
uniform spectral axis.
 Interpolates spectrum over a
predetermined user's grid
 Radiometric calibration by a
complex multiplication of the
actual scene spectrum with the
actual gain.
 Perform scene measurement
quality verification.
 Report of NESR

Organigram:

Figure 2.9 
Input:
Output:
 Radiometrically and spectrally
and locally calibrated spectral
radiance of the scene
Detailed description:
