Spectral Sensor Calibration Methods: Application Note

Application Note
AN000633
Spectral Sensor
Calibration Methods
AS7341 EVK Evaluation Kit
v2-00 • 2021-May-20
Document Feedback Spectral Sensor Calibration Methods
Content Guide
Content Guide
1 General Description ...................... 3 2.3 NIR Correction .............................................. 7

2.4 Normalization / Scale .................................. 11
1.1 Filter Correction ............................................ 3 2.5 Calibration with Matrix Based Algorithm ..... 14
1.2 Diffuser Compensation ................................. 4
1.3 Disturbances................................................. 5 3 Revision Information ................... 30
1.4 Measurement Setup and Parameters for
Measurement Accuracy................................ 5 4 Legal Information ......................... 31
2 General Correction ........................ 6
2.1 Calculations with Basic_Counts ................... 6
2.2 Offset ............................................................ 7
Application Note • PUBLIC

AN000633 • v2-00 • 2021-May-20 31 │ 2
General Description
1 General Description
This Application Note describes how to implement correction and calibration methods by considering
different effects based on the Spectral Sensor (and derivatives) Evaluation Kits (EVKs). It shows
various steps, procedures, and approaches used for alternative methods of spectral and multi-spectral
sensors. Everything is explained using the AS7341 example. Other spectral sensors are similar in
spectral technique, differing in the number of filters and peaks, but not in the algorithm and spectral
sensor correction functions shown here. Therefore, the details shown here for the AS7341 example
can also be used for all other similar spectral sensors with adaptation.
1.1 Filter Correction

The AS7341 is an 11-channel Sensor for spectral identification and color matching applications. The
spectral response has a wavelength from approximately 350 nm to 1000 nm. Eight optical channels
cover the visible spectrum. One channel can be used to measure near-infrared light. The ‘Clear’
channel is a photodiode without a filter (‘clear’) for monitoring tasks, and the ‘Flicker’ channel is
prepared for flicker measurements.
Figure 1 :
Typical Spectral Behavior of Each Channels in AS7341 (Sensitivity ‘1’-normalized)
AS7341 Typical Filter (normalized)

1
0.8
0.6
0.4
0.2
0
720
860
1000
300
328
356
384
412
440
468
496
524
552
580
608
636
664
692
748
776
804
832
888
916
944
972
1028
1056
1084
f1 f2 f3 f4 f5
f6 f7 f8 Clear Nir
The spectral filters, in combination with the diodes and electronics, provide details about the measured
spectrum of the light on the sensor. The sensor results are dependent on the sensor arrangement and
other direct effects such as series-related disturbances and deviations, as well as effects in the
measuring process itself. Therefore, in a final system setup, a correction of the raw sensor values is
necessary to eliminate the unavoidable disturbance effects and deviations. Furthermore, the
conversion of the sensor results as raw digital values into physical parameters (the application) is
necessary after the correction of the influences and can be part of the correction and calibration.

AN000633 • v2-00 • 2021-May-20 31 │ 3
General Description
Theoretically, a narrow band filter in spectral Sensors should only allow a unique light frequency that
matches the filter spectrum in transmission to pass through (like a spectrometer).
In practice and for the AS7341 sensor specifically, the transmission range and blocking outside of it is
not ideal, resulting in optical interference. Secondly, the filters and sensitivity are not in ideal form.
They are sometimes different in wavelength and vary slightly in the series in the specimens and lots.
Figure 2 shows the sensor-specific spectral response of filter 1 and 2, filter overlapping (doubled
active filter in transmission), and also the spectral values of the channels are not limited to the exact
band wavelength (transmission and filter function), instead it has an out of band spectral value in VIS
and NIR (rest transmission in blocked wavelength). This is due to the opening of the optical channel
band filters and affects the Sensor results.
For other sensors, it looks similar but different. Scaling factors, matrices, or special algorithms during
the calibration and correction process will reduce such effects.
Figure 2:
AS7341 Channel F1 + F2 Spectral Response with Filter Overlapping and Rest Transmission
1.2 Diffuser Compensation

The photodiodes inside the AS7341 or other spectral sensors have a near cosine response to
incoming light. Typical spectral filters used for channel separation are specifically developed
interference filter stacks on top of the photodiodes. Due to physical influences in the filter stacks, the
interference filter technology is limited for an incidence angle range (AOI) and expects a Lambertian
power distribution. The maximum angle of incidence to the photodiodes is limited to the design
requirements of the filter stack by the aperture/pinhole of the package. The rays with the most obtuse
angle hit the edges of the photodiodes form the opposite edge of the aperture. It is necessary to get a
diffused light on the Sensor to meet these requirements of power distribution. In the case of a non-
diffuse application, the use of an achromatic diffuser is required, which emits light with Lambertian
characteristics to the Sensor, regardless of the angle of incidence. If the diffuser is very close to or
directly on the sensor package, then its structure has to be very fine to get the same distribution to
each photodiode of the detector array. On the other side, the diffuser also changes the spectral

AN000633 • v2-00 • 2021-May-20 31 │ 4
General Description
response and transmission of the Sensor system because they always have their specific
transmission curve, which is greater than zero and not constant. Therefore, a correction of the diffuser
transmission may be required as one part within the calibration. The manuals list more details for
diffusers and EVKs.
1.3 Disturbances
Disturbances influence the Sensor results statically or dynamically. To obtain optimum results, a
verification and optimization process must correct or eliminate all these negative effects. The following
list includes some examples, which can affect the accuracy, more or less depending on the
application.
● Basic noise (e.g. dark current)
● Non-linearity Integration Time
● Gain Error
● Temperature and ageing effects from Sensor and luminary (e.g. LEDs)
● Ambient Light
● Reflections inside the Sensor System
Examples of such faults and their possible corrections can be found on the following pages.
1.4 Measurement Setup and Parameters for Measurement

Accuracy
For Sensor applications, the measurement setup and optical stability play a crucial role in accuracy
and calibration. For the purpose of calibration, the most stable and reliable measurement setup is
selected. Monochromatic test systems and/or spectrometers are required as reference devices to
specify the target pattern, check the behavior of the light source, and proof the test setup. Further,
they are important to verify the Sensor results and their corrections/calibrations. Reference devices
are depending on the sensor’s dynamic range and should be active for AS7341 from VIS (UV) to NIR,
with a spectral range of 350 nm to 1000 nm. The highest accuracy for a reference device is essential
when high accuracy for the Sensor is expected. The reference instrument should be at least ten times
more accurate or higher than the sensor requires.
The test setup should be stable and free of any disturbances and drifts. They must be checked
individually and systematically for each application, possibly after adding modules, to obtain
applications typical or device-specific correction values.
Use the Sensor EVK for feasibility projects. The hardware, software, and adapters have been
designed to be stable and ensure high accuracy for standard applications. However, the EVKs
consists of bare hardware and are not shielded against any environmental conditions. They also only
supply digits and raw values. Therefore, the customer has to adjust the EVKs + results and the
calibration + corrections for their specific application. For more details, please refer to the Sensor
ECGs manual.

AN000633 • v2-00 • 2021-May-20 31 │ 5
General Correction
2 General Correction
2.1 Calculations with Basic_Counts

Sensor results depend on Sensor specific setup - the selected parameters for Gain (AGAIN) and
Integration Time (TINT). However, changing these parameters under constant conditions in
measurement should not change the real Sensor response – RawSensorValues. On the other side,
ADC results are directly dependent on and more or less proportional to Gain and TINT1. The higher
the Gain and TINT, the better the ratio between signal and noise. Raw_Counts (= Raw Sensor Values)
from the ADC must be transformed into a result, which is not dependent on the parameter setup but
should achieve a maximum as possible. All Sensor calculations are on Basic_Counts. The definition
is:
Equation 1:
Raw_Counts
Basic_Counts =
(Gain × Integration Time)
Figure 3 shows an example from a protocol file with Setup, Raw_Counts, and Basic_Counts.
The Basic_Counts in this example for F1 is calculated by:
0.013264 = 1236 / (512 ∗ 182)
Figure 3 :
Example for Basic_Counts
1
TINT (Integration time) selection can affect the counter for the sensor results. It means TINT directly determines the Full Scale
Range and saturation.

AN000633 • v2-00 • 2021-May-20 31 │ 6
General Correction
For all corrections and calibrations, always use Basic_Counts or other calculated values without
dependence on the setup and parameters, especially for dynamic gain and the like.
2.2 Offset
Offset is defined here as a constant interference signal that continuously affects a Sensor via the
measuring process, for example, dark values, ambient lighting, or overcrossing. Each sensor channel
has its offset characteristics. Therefore, consider offsets individually per channel.
The first step of the correction is to measure the offset. This often requires a special device setup. It is
also recommended to check more than one Sensor to see their individual, lot, and series deviations.
Averaging can be a method to get an approximated and typical value for correction or to use the
individual offsets as part of a single device calibration.
The second step is to calculate or define correction values based on the offset measurements (and
averaging). These offset correction values will reduce the raw Sensor values in the sensing process
by a simple subtraction:
Equation 2:
SensorCorrectedValueOffset = Basic_Counts − Basic_CountOffset
The offset can be set in the GUI initialization and calibration files, as shown in the following example. It
is important to use for offset correction always the Basic_Counts.
Figure 4 :
Example Calibration Files with Specified Offset and Factors for Corrections in EVK
2.3 NIR Correction

Figure 2 shows the transmission and rest transmission for two AS7341 filters (F1 + F2). Spectral
reconstruction creates some issues for light spectra with huge NIR parts (e.g. filament lamp). In these

AN000633 • v2-00 • 2021-May-20 31 │ 7
General Correction
cases, the matrix algorithm will badly interpret the NIR crosstalk of the channels F1 … F8, and the
colorimetric calculations will get colder than warm-white. Therefore, the rest transmission affects the
sensor results, depending on the NIR size and series tolerances. NIR effects are typical but also
individual. In general, the higher the rest transmission, the greater the error if it cannot be corrected.
Offset or matrixing methods can correct rest transmission in case of it is constant over time, and the
sensor system was prepared to correct it. However, NIR radiation can affect an application and does
not necessarily have to be constant over time and within applications. Therefore, if NIR is an unknown
effect or is unstable for measurements, then special situational methods are necessary.
Figure 5:
Reconstruction Issues Based on NIR Crosstalk (Red = Reconstructed ; Blue = Mask)
The following pages show a method of using typical measured filter values to show typical corrected
results. Nevertheless, if such results can be replaced by an individual calibration (device calibration),
then the accuracy for such results of this NIR correction can increase. The method presented here is
based on the known proportions of the rest transmission of the F1...F8 filters in NIR 2 and relates these
to the measured values of CLEAR and NIR per measurement, to determine the NIR proportion for
correction in the individual channels F1...F8.
From the sensitivity spectra (𝑆λ for all filters of AS7341 F1…F8, see Figure 1), the ratios between the
single Fn sensitivity values and ‘Clear’ filter values at peak n (for F1 at 410nm etc.) can be calculated
from the design data3. For general calibration, typical data from a table (Figure 6) are used here as
values. For individual calibration, calculations from reconstructed sensitivity spectra will increase
accuracy.
2
From the design’s data filter specifications, and typical measurements - with acceptance of typical values of the spectra NIR
crosstalk of each sensor channel F1 … F8, it is possible to estimate the NIR part in the sensor signals, and from this an
estimated compensation (weighting of the signals) is possible.
3
peak = wavelength with maximum of sensitivity

AN000633 • v2-00 • 2021-May-20 31 │ 8
General Correction
Equation 3:
𝑆𝜆𝑘,𝜆 𝑘,𝑝𝑒𝑎𝑘
𝑆𝜆_𝑠𝑐𝑎𝑙𝑖𝑛𝑔𝑘 (1:8) =
𝑆𝜆𝐶𝑙𝑒𝑎𝑟,𝜆 𝑘,𝑝𝑒𝑎𝑘
Equation 4:
750
∫1100 𝑆𝜆𝑘 (𝜆)
𝑆𝜆_𝑟𝑎𝑡𝑖𝑜_𝑁𝐼𝑅𝑘 (1:8) = 750
∫1100 𝑆𝜆𝐶𝑙𝑒𝑎𝑟 (𝜆)
Figure 6:
General Values for Sλ_scaling and Sλ_ratio_NIR
F1 F2 F3 F4 F5 F6 F7 F8
Sλ_scaling 1.48 1.87 1.85 1.92 1.79 2.0 1.92 2.32

Sλ_ratio_NIR 4.47% 5.51% 5.31% 6.15% 3.91% 7.01% 5.31% 4.28%
In the NIR correction process for an actual measurement, the first step is to balance the filter results
F1…F8 (D_Basic – Basic-count Sensor) by using typical or individual Sλ_scaling factors (Equation 3).
Results are D_scaled_to_Clear as ‘Clear’ scaled/balanced Sensor results F1 ... F8.
Equation 5:
𝐷_𝐵𝑎𝑠𝑖𝑐 𝑛, 𝑘
𝐷_𝑠𝑐𝑎𝑙𝑒𝑑_𝑡𝑜_𝐶𝑙𝑒𝑎𝑟𝑛,𝑘 =
𝑆𝜆_𝑠𝑐𝑎𝑙𝑖𝑛𝑔𝑘
D_scaled_to_Clear are then the new scaled to clear sensor results for all channels for following
calculations.
Equation 6:
𝐷_𝑠𝑐𝑎𝑙𝑒_𝑡𝑜_𝐶𝑙𝑒𝑎𝑟𝑛𝐶𝑙𝑒𝑎𝑟 − ∑8𝑘=1 𝐷_𝑠𝑐𝑎𝑙𝑒_𝑡𝑜_𝐶𝑙𝑒𝑎𝑟𝑛𝑘

𝐷_𝐶𝑙𝑒𝑎𝑟_𝑁𝐼𝑅𝑛 =
𝐷_𝑠𝑐𝑎𝑙𝑒_𝑡𝑜_𝐶𝑙𝑒𝑎𝑟𝑛𝐶𝑙𝑒𝑎𝑟
D_Clear_NIR (Equation 6) is the ratio between the balanced integral of ‘Clear’ and the filter channels F1
… F8. It represents the signal ratio in ‘Clear’ and NIR.
Equation 7:
𝐷_𝐹_𝑁𝐼𝑅𝑛,𝑘 = 𝑆𝜆_𝑟𝑎𝑡𝑖𝑜_𝑁𝐼𝑅𝑘 ∗ 𝐷_𝐶𝑙𝑒𝑎𝑟_𝑁𝐼𝑅𝑛
The D_F_NIR (Equation 6) represents the sum of sensitivities of ‘Clear’ outside the wavelength of the
visible range. In this estimation, the sum of all scaled channels F1:F8 is set approximately equal to the
sum of the sensitivities of ‘Clear’. All inaccuracies in the estimation are therefore accepted.

AN000633 • v2-00 • 2021-May-20 31 │ 9
General Correction
NIR_weight (Equation 8) is the calculated ratio of NIR and ‘Clear’ results of the actual measurement.
The used factor of 0.1 represents a correction factor to compensate for different diode sizes of the NIR
and ‘Clear’ channels.
Equation 8:
𝐷_𝐵𝑎𝑠𝑖𝑐 𝑛,𝑁𝐼𝑅
𝑁𝐼𝑅_𝑤𝑒𝑖𝑔ℎ𝑡𝑛 = 0.1 ∗
𝐷_𝐵𝑎𝑠𝑖𝑐 𝑛,𝐶𝑙𝑒𝑎𝑟
Equation 9 corrects the measured value of the filters by the scaled portion of ‘Clear’ in NIR. Here, the
already calculated or typical ratios of actually Clear to NIR, typically F1 to F8 to NIR as well as the
ratio of ‘Clear’ in NIS and VIS are calculated together as a correction of channels F1 to F8.
Equation 9:
𝐷_𝑤𝑛,𝑘 = 𝐷_𝐵𝑎𝑠𝑖𝑐𝑛,𝑘 − 𝐷_𝐹_𝑁𝐼𝑅𝑛,𝑘 ∗ 𝑁𝐼𝑅𝑤𝑒𝑖𝑔ℎ𝑡 ∗ 𝑆𝜆_𝑟𝑎𝑡𝑖𝑜_𝑁𝐼𝑅𝑘

𝑛
The results for filters ‘Clear’ (𝐷_𝑤𝑛,𝐶𝑙𝑒𝑎𝑟 = 𝐷_𝑠𝑐𝑎𝑙𝑛,𝑐𝑙𝑒𝑎𝑟 ) and NIR ( 𝐷_𝑤𝑛,𝑁𝐼𝑅 = 𝐷_𝑠𝑐𝑎𝑙𝑛,𝑁𝐼𝑅 ) do not
change. It means they will not be corrected by the algorithm.
Figure 7:
Reconstructed Spectrum of Standard A Light Source with/without NIR Correction (1)
(x-wavelength, y-relative sensitivity in Basic_Counts)
(1) ams offers a published XLS sheet where all formulas of the NIR correction can be followed with an example like shown
in this diagram.
The result of the NIR correction with this algorithm is a modified corrected spectrum, with reduced
signals in channels 1-8 corresponding to the NIR portion of the current measurement. Here, the
correction of all sensors is carried out with standard values from a typical sensor. The result is always
good, depending on the current sensor from the typical sensor. Therefore, the NIR correction may also
worsen the result if the deviation from the typical sensor is large, and the NIR portion of the
measurement is very large. Figure 8 shows an example as a diagram, before and after the NIR
correction.

AN000633 • v2-00 • 2021-May-20 31 │ 10
General Correction
2.4 Normalization / Scale

Scale procedures are corrections when Sensor results show a percentage error compared to
reference values. Such values can be the results of targets measured by reference. In general, the
correction factor for each channel is the result of a balancing of reference values and Sensor readings.
The result of scaling is a correction vector, which includes correction factors for each Sensor channel.
All values are Basic_Counts.
Figure 8 :
Spectral Channels Before and After ‘1’-Scale (or Balance)
Scaling procedures are often used to adjust the behavior of Sensors for one defined reference point,
e.g. the minimum, maximum, or any other point from the series test (Figure 8).
According to the objectives, a one-point correction or two-point correction is applied. A typical formula
for such a scaling using one reference point is:
Equation 10:
ReferenceValue
SensorCorrectedValueScaled =
SensorCorrectedValueOffset
One-point correction means the Sensor response of all Sensors in an application will be calculated to
be scaled at One-point (e.g. Minimum min, Maximum max, or somewhere between min and max). The
title of this method is ‘White or Black Scale’.
In case the Sensor results are normalized between Two-points (e.g. ‘Dynamic Scale’ or ‘Black/White
Scale’), then these Two-points will be used in the formula:

AN000633 • v2-00 • 2021-May-20 31 │ 11
General Correction
Equation 11:
(X − Xmin)
SensorCorrectedValueScaled =
(Xmax − Xmin)
Here, X is SensorCorrectedValueOffset and Xmin/Xmax are the two Reference Values min/max.
Figure 9 and the following figures show an example for scaling in a light application.
The example describes the light detection via the spectral Sensor and inserts the spectrometer target
data (Figure 9), the Sensor results as raw Sensor values, and the corrected Sensor results after
scaling (Figure 11).
The correction uses scaled ‘Sensor values to spectrometer targets’ for one defined light source –
daylight used here. The result of scaling is the correction vector, which includes for each channel a
value representing the deviation Sensor RAW to spectrometer (Figure 10). These correction values
are useful to correct Sensor results also for other light sources. In the example, a Daylight source and
its results and target show the accuracy of the Scale procedure.
The results in the diagram(s) are good for such a primitive correction method but can be better using
matrices. The difference between scale and matric methods is the number of used targets. A higher
number of reference targets can increase accuracy for calibration dramatically.
Figure 9 :
Comparison Spectrometer Results and Sensor RAW Data for Daylight
Figure 10:
Correction Factors Based on Daylight Scale (Spectrometer Values / Sensor Raw-Data)
F1 F2 F3 F4 F5 F6 F7 F8
CorrFact 3.20 3.00 2.07 1.30 1.07 0.93 0.78 0.71

AN000633 • v2-00 • 2021-May-20 31 │ 12
General Correction
Figure 11 :
Comparison Spectrometer Results and Sensor Corrected Data for Daylight (1)
(1) Curves lie on top of each other.
Figure 12 :
Comparison Spectrometer, Sensor RAW and Sensor Corrected Data for LED Light
Figure 12 shows a comparison between the results of a spectrometer, Sensor RAW, and corrected
data for an LED Light under the condition of using daylight for calibration. The Sensor corrected
results (broken green) are much closer to the Spectrometer results than the unscaled results (green).
The accuracy after correction for LED lighting is lower than for daylight because, for both corrections,
a scaling based on daylight was used.

AN000633 • v2-00 • 2021-May-20 31 │ 13
General Correction
Figure 13:
Comparison Spectrometer Results, Sensor Corrected Results (based on daylight scaling) for
Daylight, LED, A, CWF
2.5 Calibration with Matrix Based Algorithm

Alternative calibration methods are used in case there are more than one or two reference values
(white and/or black balance), which affect each other. Such an algorithm considers a relationship
between n reference values = n targets T, measured with a reference device, and the Sensor results =
values S (individual or typical) of the identical target. This relationship between T and S is described in
a Calibration Matrix K - which represents a correction function that corrects and matches the Sensor
results into the application-specific values measured with the reference device. It means such a
method is corrected and matched in one-step. The target data determine the dynamics of the
measuring range. That is not only a three-dimensional color space but also a reconstructed spectrum
or others. Therefore, the matrix dimensions of the target and the algorithm must be matched.
If the target represents the entire dynamic range of the application, we are talking about a global target
(e.g. the entire color space of natural colors) and a ‘Global Correction’ matrix. If the target represents
only a small part of the application (e.g. only red colors as part of the entire color space), we speak of
a ‘Local Correction’ and local correction matrix. However, a localized correction also only produces
corrected results for this part of the target and is therefore only applicable to this smaller part.
However, it can lead to better results under certain conditions, since the correction function has to
consider fewer targets. Therefore, a two-step calibration can be advantageous if the global correction
only determines the subspace in which one is located, then the local corrections are applied to this
smaller part of the target. For example, a global color correction in CIE1931 RGB space is used to
define an approximate position in the color space. Then, for example, a local matrix defined as ‘red

AN000633 • v2-00 • 2021-May-20 31 │ 14
General Correction
oriented matrix’ can be used, which corrects ‘red’ more precisely (but does not take green as a color
location into account).
Figure 14 shows for a color x (Dark Skin of the Color Checker), the comparison of the targets and
results for the Global and Local Correction, if the target was reduced from 24 colors to 12, to the
colors which have the smallest color distance Delta E around color x. The error in Delta ab can thus
be reduced from 1.3 to 0.4 (or Delta E from 1.7 to 1.4) for the correction for this color x (Dark Skin).
Figure 14:
Comparison Results and Targets for ‘Dark Skin’ Global (Left) and Local Correction (Right)
Delta Global Correction Delta Local Correction

Delta ab 1,31426 Delta ab 0,37337
Delta E 1,71191 Delta E 1,42966
The error here is always the deviation of the sensor values corrected by the matrix from the reference
value (measured with a spectrometer).
It is important to make all measurements with the Sensor and reference device under identical
conditions closed to the application. Each deviation from calibration and application decreases the
accuracy.
The method of ‘Linear Regression’4 is often an algorithm, where the Calibration matrix values are
determined from S and T.
The following formulas of Linear Regression with transposed and inversed matric calculations define
Calibration Matrix K by using S and T:
4
Alternative to ‘Wiener Inverse’

AN000633 • v2-00 • 2021-May-20 31 │ 15
General Correction
Equation 12:
𝐾 = (𝑇 ∗ 𝑆 𝑇 ) ∗ (𝑆 ∗ 𝑆 𝑇 )−1
Here, S = Sensor Values as Matrix (including Offset correction and based on Basic_Counts),
T = Reference Values as Matrix and Target measurement from a Spectrometer, and K is the
Calibration Matrix CM that can be used to correct and match Sensor results.
Figure 15 shows the general flow in the generation of the calculation of the matrix. From the point of
view of calibration, the Sensor accuracy, as a result of calibration, is dependent on the calibration
procedure, used targets (number, quality, relationship), as well as from the generation and validness
of the relationship of Target (T) and Sensor results (S).
The algorithm and target must be adapted to the application and required accuracy and conditions of
the sensing process. The number and quality of targets must represent the application-specific
product. Targets can be optimized to get a minimum of deviation per target (as ‘typical, min or max
error’ or averaged over all targets), compared with the reference device.
A calibration matrix can refer to a device-specific calibration, a GD5-batch calibration, or a GD-type

calibration. It depends on which sensor data are used for calibration.
Device Calibration: This method is the most complex but has the highest accuracy. The targets must
be measured separately with all the Sensors. Then, the data of each Sensor is compared with the
reference data to get a device specific and individual calibration matrix. It is necessary if there are
deviations between the individual Sensor systems or devices.
Batch Calibration: Here, the targets are captured with one Sensor from a batch and calculated with
the reference data. The result is typical for all sensors in a batch. Therefore, this method is less
complex, but does not take into account the individual deviations of the Sensors. It is recommended if
there are only very small deviations between the individual Sensor systems in the batch.
5
GD = Golden Device

AN000633 • v2-00 • 2021-May-20 31 │ 16
General Correction
Figure 15 :
Parts and Effects During the Process Calculation of Calibration Matrix
Type Calibration: Here, the targets are captured with a Sensor as a prototype and calculated with the
reference data. The resulting calibration matrix is for all Sensors of this type, without consideration of
the individual deviations or changes over time and all the batches. This method has the smallest effort
but produces the worst results. Use this method in combination with a scaling where the individual
deviations are corrected. In this case, the advantages can be used – the low effort of a Device-to-type
calibration by a simple scaling to correct individual issues.
The GUI from the AS7341EVK was prepared to measure and correct the Sensor’s data from
luminaries6 or in reflection mode7.
For the luminary function, two alternative calibration matrices were prepared. The first matrix corrects
the Sensor spectral values directly into CIE1931 XYZ values (Figure 16), the second interpolates and
6
For Ambient Light Sensing like detection CCT, lx, Lu’v’ for D65, D50, A, CWF, TL84, LED, etc. by using a general calibration
matrixes (CIE1931 XYZ and spectral reconstruction) = results of a type calibration what is valid for all typical light sources
7
For using a general calibration matrix based on 24 Color of X-Rite Color Checker Large

AN000633 • v2-00 • 2021-May-20 31 │ 17
General Correction
generates a reconstructed spectrum with a step size of 1 nm (Figure 17). A spectrum evaluates itself,
allows a spectral fingerprint, or works with CIE1931 XYZ quantities after XYZ mapping.
The following figures show the AS7341 GUI for both modes, from ‘Basic_Counts to the corrected
values via Offset, scaling, and calibration by Matrices’.
Depending on the mode and the deposited calibration matrix (see marked black boxes in Figure 16
and Figure 17), Raw_Counts are transformed by the calibration into a spectrum (then spectrum into
XYZ) or direct into XYZ.
Figure 16 :
Basic_Counts to XYZ Values Based on XYZ Calibration (1)
(1) Filter names in GUI and in initialization files can vary based on different GUI versions.

AN000633 • v2-00 • 2021-May-20 31 │ 18
General Correction
Figure 17 :
Basic_Counts to XYZ Values Based on Spectral Calibration (1)
(1) Table for Calibration matrix interrupted. See the original spectral CM file after the EVK installation.
In the following subchapters, two Excel sheets show typical processes for calculating and using
calibration matrices. The algorithm and data are identical to the AS7341 GUI, whose matrix usage is
described above.
2.5.1 Example Reflection Mode
The first excel sheet8 is an example of reflection mode. It shows the calculation and usage of a
Spectral to XYZ based Global Calibration matrix9. The matrix calculation considers 24 measured
reference targets from the Color Checker10, measured with the Sensor (raw spectral values) and
reference device (in XYZ values).
Figure 18 shows a part of the excel sheet where the reference values (Target T) from a spectrometer
are listed as XYZ, and the Sensor values (S) and Offset are listed as corrected Basic_Counts per
channel in rows. In the table, 24 columns include the 24 color targets.
8
Ask the ams support team for the original XLS file “#Template Spectral to TCS Sensor calibration Refl.xlsx” to check the data
and formulas of the AS73XX.
9
Example for local calibration on request.
10
https://www.xrite.com/categories/calibration-profiling/colorchecker-targets.

AN000633 • v2-00 • 2021-May-20 31 │ 19
General Correction
Figure 18 :
Part of Excel Sheet (1) with Target, Sensor Values and Sensor Offset (2)
Action 1: Copy the Spectrometer XYZ values of the measured reference colors into cells [c6:z8] OR use the values h
TARGET DATA T Reference values 1 2 3 4 5 6 7 8 9 10
Dark Skin Light Skin Blue Sky Foliage Blue FlowerBluish GreenOrange YellowYellow Green PurpleModerate RedPurp
X 11,22 36,35 16,9 10,93 24,56 30,76 46,31 33,91 8,51 28,12
(Reference Values) Y 10,23 33,51 17,92 13,63 22,76 42,37 43,45 44,63 6,39 18,87
Z 7,15 24,55 33,54 7,13 43,25 45,05 7,83 11,31 14,41 13,56
Offset Sensor Offset Sensor Action 2: Copy the 'Basic Counts' of Offset into the Cells [C11:C18]. In our example, the values represent the direct r
415 0,0037
445 0,0139
(e.g. Optical 480 0,0064 Background of the calibration is a calculation of target values T and sensor values S with the
Overcrossing) 515 0,0285
measured with a sensor as sensor raw data. The sensor raw data are corrected (only offset is
555 0,0302
590 0,0301 used in the future under the same conditions as in the calibration matching of the sensor valu
630 0,0407 to input values to understand the calculations.
680 0,0121
Action 3: Copy the Sensor 'Basic Counts' from the protocol log file of all 24 targets into the Cells [C23:Z30].
Sensor Matrix S Basic Counts 1 2 3 4 5 6 7 8 9 10
Dark Skin Light Skin Blue Sky Foliage Blue FlowerBluish GreenOrange YellowYellow Green PurpleModerate RedPurp
415 0,029 0,088 0,077 0,027 0,107 0,108 0,075 0,066 0,041 0,059
445 0,087 0,252 0,327 0,076 0,427 0,400 0,121 0,123 0,148 0,159
480 0,050 0,182 0,131 0,054 0,166 0,249 0,117 0,141 0,048 0,088
515 0,190 0,634 0,421 0,303 0,512 1,066 0,633 1,007 0,135 0,240
555 0,314 0,859 0,464 0,396 0,491 1,116 1,380 1,397 0,150 0,378
590 0,405 1,305 0,417 0,320 0,566 0,745 1,695 1,063 0,215 1,031
630 0,309 1,048 0,292 0,227 0,633 0,390 1,171 0,611 0,264 1,017
680 0,107 0,407 0,115 0,082 0,269 0,153 0,417 0,239 0,108 0,347
(1) Tables were interrupted. See the full tables in the original MS Excel File.
(2) Filter names in spreadsheets and other sensor files can vary based actual development status.
Figure 19:
Used Target (Global) CIE1931 Color Space from Color Checker for Calibration in xy-Diagram

AN000633 • v2-00 • 2021-May-20 31 │ 20
General Correction
Figure 19 shows the used reference target 'Color Checker' (24 colors) in an xy diagram and
represents the dynamic range of the sensor target. This means that all colors at an imaginary outer
boundary line form in their sum, the color space, which is realized later by this correction. All colors in
the color space can thus be included in the measuring range of the sensor. Colors outside the
measuring range are not, or incorrectly, corrected. Fewer or more colors in the measuring range can
increase/decrease the dynamic range of the sensor, if the color location changes the boundary lines.
A change in the number of colors in the target usually leads to changed accuracies, both positive and
negative.
The target determines not only the accuracy and dynamics of a calibration. The targets will determine
the form of output values also. Using XYZ based targets will result in XYZ sensor values after
calibration. If the target reference values are in spectral form, e.g. 1nm step size, then the result is
also a spectrum with a step size of 1nm.
The following example in XLS shows the calculation of a correction matrix K, using a linear regression
algorithm based on a 24xtarget (dark skin, light skin, etc. - see the columns in the first lines of
Figure 20). The reference values of this target were measured with a spectrometer and the sensor
values with a Multi-channel spectral sensor in RAW values of S. The correction matrix is calculated
using matrix multiplications (inverse and transposed) of the sensor and reference matrix.
Customers can systematically verify the algorithms for calculating the calibration matrix in the XLS
sheet to insert the algorithm into their software.
Figure 20 :
Part of Calculation of Direct XYZ Matrix based on Reference and Sensor Data (1)
Sensor Matrix corr Basic-Offset Dark Skin Light Skin Blue Sky Foliage Blue Flower Bluish GreenOrange Yellow
Yellow Green
offset 415 0,026 0,084 0,074 0,024 0,103 0,104 0,071 0,063
445 0,073 0,238 0,313 0,062 0,413 0,386 0,107 0,109
480 0,043 0,175 0,124 0,047 0,160 0,242 0,111 0,134
S' = S - offset 515 0,161 0,605 0,393 0,275 0,484 1,037 0,605 0,978
555 0,284 0,829 0,434 0,366 0,461 1,086 1,350 1,367
590 0,375 1,275 0,386 0,290 0,536 0,715 1,665 1,033
630 0,268 1,007 0,251 0,187 0,593 0,349 1,131 0,570
680 0,095 0,395 0,103 0,070 0,257 0,141 0,405 0,227
Calculation:
67,197 212,134 117,689 487,974 716,428 785,182 554,435 206,290
Matrix for linear
A = T * S'_trans 69,061 219,038 123,399 524,119 755,390 790,918 542,426 202,059
transformation CM 68,752 249,548 121,031 450,237 584,552 599,090 423,157 162,273
9143,937 -1887,998 772,049 -177,796 -85,805 -88,149 -51,852 -283,612

-1887,998 434,899 -356,178 82,478 -11,049 41,494 14,796 16,562
772,049 -356,178 994,726 -248,824 143,994 -128,202 -2,768 131,171
-177,796 82,478 -248,824 71,265 -45,716 38,745 -2,846 -34,572
B = [S' * S'_trans]^-1
-85,805 -11,049 143,994 -45,716 37,298 -29,608 9,883 21,944
-88,149 41,494 -128,202 38,745 -29,608 29,496 -15,955 1,018
-51,852 14,796 -2,768 -2,846 9,883 -15,955 50,035 -98,000
-283,612 16,562 131,171 -34,572 21,944 1,018 -98,000 265,505
Action 4: See and use the calculated Correction Matrix.

Calculated Correction 93,354 3,256 -5,881 1,634 8,483 9,474 17,403 -20,620
CM = A * B 64,077 -7,336 13,801 16,200 13,837 2,913 11,605 -21,562
Matrix
-23,483 99,493 40,677 3,939 -4,126 -0,391 -0,487 -3,863
(1) Tables interrupted. See the original Excel sheet for the full table.
The result of this sheet is the Calibration Matrix K, marked in green in Figure 20, which can be used to
match the Sensor results after measuring the RAW data. This method of Linear Regression by using

AN000633 • v2-00 • 2021-May-20 31 │ 21
General Correction
these data is only one example to get a correction matrix. Other algorithms are known and must be
tested to get optimized sensor results.
Another method, which was shown earlier in this document, is to use a multi-stage matrix-based
calibration concept, by mixing the methods one after another. For example, use a complete (global)
color target (e.g. 24 colors) for calibration as a first step to get a local position in the CIE1931 color
space, and then a local calibration around this color position, with a reduced color (selected 12 colors)
target but higher accuracy.
Using a specific calibration matrix for each Sensor separately is a Device-specific calibration. If the
matrix is used for an alternative Sensor from the same batch/for this specific product, then it is a
batch-/type calibration. The calibration method, algorithm, and target must be harmonized with the
application and its requirements.
The offset from Figure 18 and the Calibration Matrix from Figure 20 should be used as input data for
the AS7341 GUI and Reflection Mode. For more details, see the manual for EVK.
Figure 22 shows the usage of the Calibration Matrix in comparison with the reference values. The
formula can be checked systematically. It considers the calculation of the XYZ Tristimulus result for
the Sensor, xy coordinates and the actual La*b* (D65 illumination, 2° observer). Deviations are given
by comparison with the reference values.
In Figure 21, all results are present in one diagram. It can be seen that each color has its accuracy
after the calibration. Therefore, the minimum, maximum, and average (typical) deviation of the color
target is always interesting. Accordingly, calibration can be optimized to minimize color variations or
min/max/typ. for a range of colors
Figure 21 :
Delta E after Correction, Using Direct XYZ Device Calibration

AN000633 • v2-00 • 2021-May-20 31 │ 22
General Correction
Figure 22 :
Using 24 Targets(1) to Correct Sensor Results and Comparison with References
Sensor Values Action 2: Copy the 'Basic Counts' of Offset into the Cells [C16:C23] or use the offs
Offset Dark Basic Counts 0,0600
415 0,0037
Dark Basic Counts = LED Spectrum Raw Counts
445 0,0139 0,0400
480 0,0064
515 0,0285 0,0200
555 0,0302
590 0,0301 0,0000
630 0,0407 400 450 500 550 600 650 700
680 0,0121
Action 3: Copy the Sensor 'Basic Counts' from the protocol log file or use them fr
Sensor Values 1 2 3 4 5 6 7
Sensor Values and Results
Basic counts Dark Skin Light Skin Blue Sky Foliage Blue Flower Bluish GreenOrange Yellow
415 0,02948 0,08753 0,07736 0,02741 0,10674 0,10792 0,07495
445 0,08678 0,25168 0,32701 0,07578 0,42676 0,39957 0,12075
480 0,04986 0,18179 0,13060 0,05357 0,16627 0,24859 0,11736
515 0,18989 0,63372 0,42114 0,30315 0,51243 1,06551 0,63308
555 0,31438 0,85895 0,46402 0,39605 0,49125 1,11608 1,37982
590 0,40503 1,30530 0,41656 0,32049 0,56569 0,74547 1,69494
630 0,30905 1,04787 0,29192 0,22727 0,63343 0,38996 1,17124
680 0,10734 0,40668 0,11507 0,08183 0,26913 0,15322 0,41703
Corrected counts Dark Skin Light Skin Blue Sky Foliage Blue Flower Bluish GreenOrange Yellow
415 0,02576 0,08381 0,07364 0,02369 0,10302 0,10420 0,07123
445 0,07286 0,23775 0,31308 0,06186 0,41283 0,38564 0,10683
480 0,04342 0,17535 0,12416 0,04713 0,15983 0,24215 0,11092
515 0,16137 0,60521 0,39263 0,27464 0,48392 1,03699 0,60456
555 0,28416 0,82874 0,43380 0,36584 0,46104 1,08586 1,34960
590 0,37494 1,27522 0,38647 0,29041 0,53561 0,71538 1,66486
630 0,26836 1,00718 0,25123 0,18658 0,59274 0,34927 1,13055
680 0,09520 0,39455 0,10294 0,06970 0,25700 0,14108 0,40490
Calibrated Results Dark Skin Light Skin Blue Sky Foliage Blue Flower Bluish Green Orange Yellow
X 11,32 37,06 17,40 10,25 24,81 30,41 45,88
Y 10,42 34,21 18,32 12,73 22,89 42,11 43,02
Z 7,23 25,27 33,56 6,61 43,67 44,38 7,52
x 0,39083 0,38386 0,25113 0,34628 0,27155 0,26015 0,47587
y 0,35959 0,35438 0,26447 0,43025 0,25055 0,36021 0,44619
z 0,24957 0,26176 0,48441 0,22347 0,47790 0,37964 0,07794
XSP-XS -1% -2% -3% 6% -1% 1% 1%

XYZ
Diff
YSP-XS -2% -2% -2% 7% -1% 1% 1%

ZSP-XS -1% -3% 0% 7% -1% 1% 4%
Diff xy
dxy 0,00241 0,00129 0,00455 0,00138 0,00084 0,00170 0,00164
Average dxy 0,00287

Min dxy 0,00021
Max dxy 0,01214
Spektrometer
L (D65/2°) 38,25210 64,57175 49,39893 43,69756 54,82426 71,12527 71,85934
a (D65/2°) 11,43062 15,64004 -0,73245 -14,17288 13,19162 -32,25669 14,73924
b (D65/2°) 12,84954 17,18788 -22,31491 22,31357 -24,90771 1,18557 68,31215
Lab results
Sensor
L (D65/2°) 38,57729 65,13169 49,88239 42,35960 54,96364 70,94629 71,56802
a (D65/2°) 10,75451 15,57349 -0,08978 -13,56698 13,68309 -32,78591 14,77548
b (D65/2°) 13,11603 16,97310 -21,50346 22,00004 -25,14127 1,61832 68,94041
Color Dark Skin Light Skin Blue Sky Foliage Blue Flower Bluish GreenOrange Yellow
Delta E (D65/2°) 0,79617 0,60340 1,14245 1,50185 0,56171 0,70667 0,69346
Max DeltaE 2,30337

Min Delta E 0,19500
Average DeltaE 0,98487 2,50000
(1) Tables were interrupted. See the full tables in the original MS Excel File.

AN000633 • v2-00 • 2021-May-20 31 │ 23
General Correction
2.5.2 Example Light Detection
In the second Excel sheet, all steps from the ADC Sensor values to the photometric results in light
detection, are shown based on two alternative calibration methods – ‘reconstructed spectrum’ and
‘direct XYZ matching’. Both methods are also part of the GUI. The following figure shows the used
processes of the alternative methods as a block diagram.
In spectral reconstruction, the calibration matrix CM for light detection is a spectral matrix with a step
size of 1nm, within the wavelength 400 nm to 1000 nm. The calibration then yields a reconstructed
spectrum in the given wavelength, which can be used for an XYZ calculation by multiplying the
reconstructed spectrum and the CIE1931 standard observer function XYZ.
If XYZ calibration is used, the target must be a reference between the sensor ADCs and the reference
values measured with a spectrometer. The results after calibration are direct XYZ values for the actual
measurement.
The calibration matrices for both variants were created with a ‘Golden Device GD’ sensor selected
with a typical sensitivity. The sensor was stimulated with typical light sources and referenced with the
spectrometer results to the calibration matrices representing device-to-type calibrations.
Since all the sensors have deviations from this GD, all the sensors must be adapted or scaled to the
GD. A scaling to daylight or another homogeneous light spectrum can do this.
The results of the calibration for light detection are XYZ photometric values that can be used to
calculate Lu'v', CCT, and Lux. The use of spectral calibration also allows typical calculations based on
spectral values such as CRI or advanced functions (e.g. light source detection by Spectral Mask
Compare).

AN000633 • v2-00 • 2021-May-20 31 │ 24
General Correction
Figure 23 :
Alternative Calibration Process for ‘Light Detection’
The following figures show a part of an Excel sheet with a calibration procedure based on the
algorithms ‘spectral calibration’ and ‘XYZ calibration’. The excel sheets explain all the steps for both
algorithms - from spectral ADC calculated Basic_Counts, corrections, calibration, and XYZ
calculations to get the photometric results. All formulas can be checked over the full algorithm
systematic or are helpful to benchmark own software. Figure 24 shows the Sensor data from ADC as
Basic_Counts, before/after corrections, and lists all correction values that can be used also direct in
EVK GUI’s ALS initialization and calibration files. For more details, see the manual for EVK. Figure 25
presents the process from ‘reconstructed spectrum’11 to ‘photometric results’. Figure 26 contains the
photometric result, but it is based on the direct XYZ calibration. The reason for the small differences
between both photometric results is the different algorithms and calibration matrices despite the
identical Sensor and reference targets in the calibration process.
Regardless of the small differences, both algorithms differ not only in the algorithms but also in the
effort required to correct them. Depending on which filters are included in the calibration, the XYZ
matrix for direct calibration has a dimension of [3x8…11]. The spectral reconstruction requires a matrix
dimension of [400x8…11]. Since the matrices have to be stored and used in the microcontroller, this
should be taken into account when selecting the method. On the other side, spectral reconstruction
also offers greater flexibility and contains more information. Spectral results, such as CRI and spectral
comparisons of light sources and perturbations, can be calculated via the spectrum, which is not
possible via color coordinates. Thus, each application has to be adjusted and optimized regarding its
necessary functions, accuracies, and effort in calibration.
11
Based on a calibration file, which is inserted in another sheet. The calibration file is based on a General Calibration, which is
described in chapter 2.5 of this document.

AN000633 • v2-00 • 2021-May-20 31 │ 25
General Correction
Figure 24 :
(1)
Sensor Data Plus Corrections are the Basics for Calibrations
Measured Sensor's Channel Data's switch on/off NIR correction ---v

Channel Basic Counts Basic Counts Corrected
AS7341 Wavelength from Gain Corr Sensor NIRNIR
Corr Sensor
Channel (WV) protocol file Corrected Factor Offset On/Off
Correction Data's
F1 415 0,011526 0,011057 1,00000 0,00000 0,011057
F2 445 0,020497 0,019664 1,00000 0,00000 0,019664
F3 480 0,029501 0,028302 1,00000 0,00000 0,028302
F4 515 0,033869 0,032492 1,00000 0,00000 0,032492
F5 555 0,036251 0,034778 1,00000 0,00000 0,034778
F6 590 0,035457 0,034016 1,00000 0,00000 0,034016
F7 630 0,041338 0,039658 1,00000 0,00000 0,039658
F8 680 0,044997 0,043168 1,00000 0,00000 0,043168
Clear 750 0,133147 0,127734 1,00000 0,00000 0,127734
NIR 900 0,054001 0,051806 1,00000 0,00000 0,051806
(1) Filter names in spreadsheets and other sensor files can vary based actual development status.
Figure 25 :
(1)
Reconstructed Spectrum and Photometric Results after Spectral Reconstruction
Spectral Reconstruction based on CIE1931 based on Golden Unit

Channel data's Matching Spectrum into XYZ Spectral Calibration Matrix
Reconstructed
Reconstructed Sensor Spec
WV Sensor Spec Normalized WV X Y Z
380 0,000849 0,482 380 0,000001 0,000000 0,000005 X 0,13598
381 0,000858 0,486 381 0,000001 0,000000 0,000006 Y 0,14678
382 0,000867 0,491 382 0,000001 0,000000 0,000007 Z 0,17377
383 0,000876 0,496 383 0,000002 0,000000 0,000007 x 0,29786
384 0,000884 0,501 384 0,000002 0,000000 0,000008 y 0,32151
385 0,000893 0,506 385 0,000002 0,000000 0,000009 z 0,38063
386 0,000902 0,511 386 0,000002 0,000000 0,000011
387 0,000911 0,517 387 0,000003 0,000000 0,000012 Lx 100 lx
388 0,000921 0,522 388 0,000003 0,000000 0,000014 L 1,33
389 0,000930 0,527 389 0,000003 0,000000 0,000016 u' 0,19026 ***
390 0,000939 0,532 390 0,000004 0,000000 0,000019 v' 0,46206 ***
391 0,000949 0,538 391 0,000005 0,000000 0,000021
392 0,000958 0,543 392 0,000005 0,000000 0,000024 CCT 7478 K
393 0,000968 0,549 393 0,000006 0,000000 0,000027 CCT outside range
394 0,000978 0,554 394 0,000007 0,000000 0,000031
395 0,000988 0,560 395 0,000008 0,000000 0,000036 *** for 2° observer and illuminant C
(1) Tables were interrupted, see the full tables in the original MS Excel File.

AN000633 • v2-00 • 2021-May-20 31 │ 26
General Correction
Figure 26 :
Photometric Results After XYZ Calibration
CIE1931 based on Golden Unit

XYZ Calibration Matrix
X 0,13498
Y 0,14594
Z 0,17119
x 0,29856
y 0,32279
z 0,37865
Lx 100 lx
L 1,32
u' 0,19027 ***
v' 0,46287 ***
CCT 7413 K
Calibration is therefore dependent on the method, the target, disturbances, sensor settings, and
results. If necessary, several methods may have to be applied, one after the other. The following
figure shows general sensor results after calibration of different light sources for matrix-based
methods.
● For (A), a general calibration matrix based on the design data of the filters,
● For (B), a general calibration matrix based on the measurement data from a typical sensor (golden
device), and
● For (C), a general calibration matrix based on the measurement data from a typical sensor (golden
device) with additional adjustment by scaling.
Figure 27:
Typical Results for Alternative Calibration Methods for ALS

AN000633 • v2-00 • 2021-May-20 31 │ 27
General Correction
2.5.3 Example Display Measurement
In the display measurement application, the ALS correction matrix (see chapter 2.5.2) should produce
logically correct results (see Figure 28, e.g. Delta xy = 0,01) as a reconstructed spectrum or in
CIE1931 color space.
In this cases and regarding GUI tests, no other correction matrix than the in standard installed matrix
should be used. In case of using a Golden Matrix (lot or type calibration), a correction vector to adapt
sensor raw counts to Golden Device or a White Balance can be helpful to increase accuracy.
Another way is using a direct spectral to XYZ correction matrix based on targets which are measured
on display with sensor and reference device. It is the same procedure like described in chapter 2.5.1
but by using another target. ams provides templates in which calibration and correction are shown as
examples.
Be attended to the number of linearly independent targets, which must be greater than or equal to the
number of filters used in the sensor to obtain a stable matrix.

AN000633 • v2-00 • 2021-May-20 31 │ 28
General Correction
Figure 28:
Display Test Results in Color Diagram ‘White’ (by Using Trace Mode For Measured RGBWCMY)
Reference Device
Sensor From Reconstructed Spectrum To XYZ Oriented Results

AN000633 • v2-00 • 2021-May-20 31 │ 29
Revision Information
3 Revision Information
Changes from previous version to current revision v2-00 Page
all chapters all
● Page and figure numbers for the previous version may differ from page and figure numbers in the current revision.
● Correction of typographical errors is not explicitly mentioned.

AN000633 • v2-00 • 2021-May-20 31 │ 30
Legal Information
4 Legal Information
Copyrights & Disclaimer
Copyright ams AG, Tobelbader Strasse 30, 8141 Premstaetten, Austria-Europe. Trademarks Registered. All rights reserved.
The material herein may not be reproduced, adapted, merged, translated, stored, or used without the prior written consent of the
copyright owner.
Information in this document is believed to be accurate and reliable. However, ams AG does not give any representations or
warranties, expressed or implied, as to the accuracy or completeness of such information and shall have no liability for the
consequences of use of such information.
Applications that are described herein are for illustrative purposes only. ams AG makes no representation or warranty that such
applications will be appropriate for the specified use without further testing or modification. ams AG takes no responsibility for
the design, operation and testing of the applications and end-products as well as assistance with the applications or end-product
designs when using ams AG products. ams AG is not liable for the suitability and fit of ams AG products in applications and
end-products planned.
ams AG shall not be liable to recipient or any third party for any damages, including but not limited to personal injury, property
damage, loss of profits, loss of use, interruption of business or indirect, special, incidental or consequential damages, of any
kind, in connection with or arising out of the furnishing, performance or use of the technical data or applications described
herein. No obligation or liability to recipient or any third party shall arise or flow out of ams AG rendering of technical or other
services.
ams AG reserves the right to change information in this document at any time and without notice.
RoHS Compliant & ams Green Statement

RoHS Compliant: The term RoHS compliant means that ams AG products fully comply with current RoHS directives. Our
semiconductor products do not contain any chemicals for all 6 substance categories plus additional 4 substance categories (per
amendment EU 2015/863), including the requirement that lead not exceed 0.1% by weight in homogeneous materials. Where
designed to be soldered at high temperatures, RoHS compliant products are suitable for use in specified lead-free processes.
ams Green (RoHS compliant and no Sb/Br/Cl): ams Green defines that in addition to RoHS compliance, our products are free
of Bromine (Br) and Antimony (Sb) based flame retardants (Br or Sb do not exceed 0.1% by weight in homogeneous material)
and do not contain Chlorine (Cl not exceed 0.1% by weight in homogeneous material).
Important Information: The information provided in this statement represents ams AG knowledge and belief as of the date that
it is provided. ams AG bases its knowledge and belief on information provided by third parties, and makes no representation or
warranty as to the accuracy of such information. Efforts are underway to better integrate information from third parties. ams AG
has taken and continues to take reasonable steps to provide representative and accurate information but may not have
conducted destructive testing or chemical analysis on incoming materials and chemicals. ams AG and ams AG suppliers
consider certain information to be proprietary, and thus CAS numbers and other limited information may not be available for
release.
Headquarters Please visit our website at www.ams.com

ams AG Buy our products or get free samples online at www.ams.com/Products
Tobelbader Strasse 30 Technical Support is available at www.ams.com/Technical-Support
8141 Premstaetten Provide feedback about this document at www.ams.com/Document-Feedback
Austria, Europe For sales offices, distributors and representatives go to www.ams.com/Contact
Tel: +43 (0) 3136 500 0 For further information and requests, e-mail us at [email protected]

AN000633 • v2-00 • 2021-May-20 31 │ 31

Spectral Sensor Calibration Methods: Application Note

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Spectral Sensor Calibration Methods: Application Note

Uploaded by

Copyright:

Available Formats

Application Note

1 General Description ...................... 3 2.3 NIR Correction .............................................. 7

Application Note • PUBLIC

1.1 Filter Correction

AS7341 Typical Filter (normalized)

Application Note • PUBLIC

1.2 Diffuser Compensation

Application Note • PUBLIC

1.4 Measurement Setup and Parameters for Measurement

Application Note • PUBLIC

2.1 Calculations with Basic_Counts

The Basic_Counts in this example for F1 is calculated by:

0.013264 = 1236 / (512 ∗ 182)

Application Note • PUBLIC

SensorCorrectedValueOffset = Basic_Counts − Basic_CountOffset

2.3 NIR Correction

Application Note • PUBLIC

Application Note • PUBLIC

Sλ_scaling 1.48 1.87 1.85 1.92 1.79 2.0 1.92 2.32

𝐷_𝑠𝑐𝑎𝑙𝑒_𝑡𝑜_𝐶𝑙𝑒𝑎𝑟𝑛𝐶𝑙𝑒𝑎𝑟 − ∑8𝑘=1 𝐷_𝑠𝑐𝑎𝑙𝑒_𝑡𝑜_𝐶𝑙𝑒𝑎𝑟𝑛𝑘

𝐷_𝐹_𝑁𝐼𝑅𝑛,𝑘 = 𝑆𝜆_𝑟𝑎𝑡𝑖𝑜_𝑁𝐼𝑅𝑘 ∗ 𝐷_𝐶𝑙𝑒𝑎𝑟_𝑁𝐼𝑅𝑛

Application Note • PUBLIC

𝐷_𝑤𝑛,𝑘 = 𝐷_𝐵𝑎𝑠𝑖𝑐𝑛,𝑘 − 𝐷_𝐹_𝑁𝐼𝑅𝑛,𝑘 ∗ 𝑁𝐼𝑅𝑤𝑒𝑖𝑔ℎ𝑡 ∗ 𝑆𝜆_𝑟𝑎𝑡𝑖𝑜_𝑁𝐼𝑅𝑘

(x-wavelength, y-relative sensitivity in Basic_Counts)

Application Note • PUBLIC

2.4 Normalization / Scale

Application Note • PUBLIC

CorrFact 3.20 3.00 2.07 1.30 1.07 0.93 0.78 0.71

Application Note • PUBLIC

(1) Curves lie on top of each other.

Application Note • PUBLIC

2.5 Calibration with Matrix Based Algorithm

Application Note • PUBLIC

Delta Global Correction Delta Local Correction

Application Note • PUBLIC

A calibration matrix can refer to a device-specific calibration, a GD5-batch calibration, or a GD-type

Application Note • PUBLIC

Application Note • PUBLIC

Application Note • PUBLIC

2.5.1 Example Reflection Mode

Application Note • PUBLIC

Application Note • PUBLIC

9143,937 -1887,998 772,049 -177,796 -85,805 -88,149 -51,852 -283,612

Action 4: See and use the calculated Correction Matrix.

Application Note • PUBLIC

Application Note • PUBLIC

XSP-XS -1% -2% -3% 6% -1% 1% 1%

YSP-XS -2% -2% -2% 7% -1% 1% 1%

dxy 0,00241 0,00129 0,00455 0,00138 0,00084 0,00170 0,00164

Average dxy 0,00287

Max DeltaE 2,30337

Application Note • PUBLIC

2.5.2 Example Light Detection

Application Note • PUBLIC

Application Note • PUBLIC

Measured Sensor's Channel Data's switch on/off NIR correction ---v

Spectral Reconstruction based on CIE1931 based on Golden Unit

Application Note • PUBLIC

CIE1931 based on Golden Unit

Application Note • PUBLIC

2.5.3 Example Display Measurement

Application Note • PUBLIC

Sensor From Reconstructed Spectrum To XYZ Oriented Results