Advance FT-NIR User Training

GNP – Good NIR Practice

Sirinnapa (Mui) Saranwong
Bruker Optik GmbH, Germany

Innovation with Integrity

Bruker FT-NIR products

TANGO-R Matrix-F
TANGO-T

MPA II

-Bruker Confidential- 2
Optimize all Settings to get the best
Spectra

Good Results

Good Spectra

Signal/Noise Ratio
Preamp Gain

32 16 8 128 64 32
cm-1 cm-1 cm-1 Scans Scans Scans

Resolution Measurement Time

-Bruker Confidential- 3
Good NIR Methods are based on good
Spectra and Reference Values

• Acquisition of good spectra is most important. Everything afterwards is

based and related on them.

• In the moment you realize that the spectra are not good, the samples are
probably gone.

• Make sure that an instrument is running with optimal settings. Otherwise

many samples are scanned and than it is too late to make a change.

-Bruker Confidential- 4
Good NIR Methods are based on good
Spectra and Reference Values

• For good NIR methods the following points must be optimized:

• Repeatable sample preparation (ground, heated, mixing, etc)
• Good sample presentation
• Suitable measurement parameters
• Resolution
• Preamp gain
• Signal/noise
• Right background setup
• Comparable amount of sample measured on NIR and analyzed with
reference method(s)
• Good reference values

-Bruker Confidential- 5
What is the Goal for the Spectrum?

• The spectrum should represent the sample. More scans will be needed if the
sample is heterogeneous.
• For good Chemometric evaluation methods, the signal to noise (S/N) ratio
will directly affect the repeatability and accuracy of model, i.e.
• Thresholds in Ident methods
• RMSECV/RMSEP of Quant methods
(noisy spectra -> noise in PLS factors -> noise in regression coefficients)

NOTE: Using derivatives as data preprocessing can improve results but noise
will increase

-Bruker Confidential- 6
Background Measurements

• Background should be measured in a way that it can be reproduced

easily.
• The best way is to measure in an open beam (air as reference).
• For reflectance measurement, Spectralon should be used for probe, and the
internal reference (gold plate) should be used for integrating sphere.
• Do NOT use for background:
• empty or filled vial or cuvette in sample compartment
• external gold mirrors on integrating sphere
• empty vial with transflection mirror on integrating sphere
• empty Petri dish with external gold mirror on integrating sphere
• empty Petri dish with external transmission
PLS can compensate the effect of sample holder (Quartz, glass,
plastic). The use of empty container as Background create risk of bias if
another empty container is used for Background measurement.

-Bruker Confidential- 7
What is the Goal for the Spectrum?

Optimize the
Signal/Noise Ratio

Sample Measuring
heterogeneity speed

Sample Spectral
absorptivity characteristics
(peak width)

-Bruker Confidential- 8
Amplitude and Signal/Noise Ratio

• Even if amplitude of interferogram is small,

the signal/noise ratio could be quite OK sometime.

• Tablets and some transparent polymers are often limited in amplitude

(typically several hundred counts) but spectra could still be OK.

• In the case of low peak amplitude, change the setting to lower resolution
(for example, from 16cm-1 to 32cm-1) could help improve Signal/Noise
ratio

-Bruker Confidential- 9
High Amplitude, good Signal/Noise

• Amplitude limit: 32000

• Preferable not to
exceed: 25000

• Focus only on the

value, not the
algebraic sign (+ or -)

-Bruker Confidential- 10
How to check Signal/Noise Ratio
visually

• Check
smoothness of
single channel
spectrum on flat
area, e.g. around
10,000 to
8,000cm-1.

• NOTE: in check
signal mode
always 13cm-1
resolution is
used!
Improvement of
S/N by lowering
resolution is not
visible here, only
in measured
spectra!
-Bruker Confidential- 11
 How to check Signal/Noise Ratio
visually

• Checking the
0,0002

noise level by
taking 1st
derivative of a
sample
spectrum.
Absorbance Units
-0,0004
-0,0008

10500 10000 9500 9000 8500 8000

-Bruker Confidential- 12
 How to check Signal/Noise Ratio
visually
0,0002

• Make sure the

appropriate
smoothing
point selection
Absorbance Units

is done
depending on
-0,0004

the used
resolution.

• Common
settings are
-0,0008

8cm-1 = 17pt
16cm-1 = 13pt
32cm-1 = 9pt
64cm-1 = 5pt
10500 10000 9500 9000 8500 8000

-Bruker Confidential- 13
Important Measurement Parameters

• The following parameters are most important for the quality of spectra
(i.e. Signal/Noise Ratio) and are fundamental for the robustness of
calibrations and repeatability of NIR predictions.
• Resolution
• Number of scans (measurement time)
• Preamp gain settings for sample and background
• Scanner velocity (only in process)

-Bruker Confidential- 14
Resolution

• Different resolutions can be used depending on sample’s optical property.

• In general, the measurement condition, especially resolution, should be kept

the same all through the spectral collection period. However, if required,
changing of resolution is also possible. Spectra of sample measured with the
resolution to be used in routine analysis shall be set as the first spectra file
in IDENT/QUANT set up to allow data interpolation.

-Bruker Confidential- 15
Resolution

• Use 8cm-1 for

• transmission measurements of clear liquids in sample compartment or
with transmission probes/ flowcells.
• highly reflecting samples on integrating sphere, such as white bright
powder.

• The resolution should be changed to 16cm-1 or lower (32 or 64cm-1), if

• low signal at detector (e.g. thick tablets),
• bad signal/noise (S/N), e.g. due to low reflectivity of the sample such as
meat, transparent pellets, etc.
• faster measurements are required (e.g. in process).

8 cm-1 16 cm-1 32 cm-1 64 cm-1

High resolution Low resolution

Slow measurement Fast measurement
-Bruker Confidential- 16
Resolution, S/N and Measurement Time

• The FT technology with mirror allows resolution changing which has direct
effects and consequences on model performance.

• If you reduce the resolution by a factor of 2,

e.g. from 8 to 16 cm-1 the signal/noise ratio will doubled (better)

• Alternatively, doubling the number of scans would increase the

signal/noise ratio just by a factor of square root of 2, approx. 1.41

• Lower resolution requires shorter acquisition time for the same number of
scans; therefore, for the same measurement time (for example 30 sec),
double number of scans can be used for low resolution set up.
• 8cm-1, 32 scans requires 22 seconds for MPA II/Matrix-F
• 16cm-1, 64 scans also requires 22 seconds for MPA II/Matrix-F

-Bruker Confidential- 17
Spectra measured at different resolution
Log (1/R)

8cm-1, 64 scans
16cm-1, 64scans

-Bruker Confidential- 18
Spectra measured at different resolution
First derivative

8cm-1, 64 scans
16cm-1, 64scans

-Bruker Confidential- 19
Benefits of high resolution
(White bright samples!)

Fructose measured at 8cm-1

Glucose measured at 8cm-1

-Bruker Confidential- 20
Information loss in resolution

Fructose measured at 64cm-1

Glucose measured at 64cm-1

-Bruker Confidential- 21
 Talc OH-Bands with different
Resolutions
0.40

2 cm-1 (max. FT-NIR)

8 cm-1 (std. FT-NIR)
0.35

> 25 cm-1 (Dispersive)

0.30
Absorbance Units
0.20 0.25
0.15
0.10
0.05

1380 1385 1390 1395 1400 1405

Nanometers

-Bruker Confidential-
Recommended Resolution/Scan Settings

-Bruker Confidential- 23
Recommended Resolution/Scan Settings

-Bruker Confidential- 24
Sample Gain Settings

• Use always
sample gain x1
instead of
‘automatic’.

-Bruker Confidential- 25
IMPORTANT: Preamp Gain Settings

• Separate settings
for preamp gain for
Sample and BGR
measurements.
Different preamp
gains can be used
for Sample and
BGR.

Ref: 1x (1x)
A: 3x (2.1x)
B: 30x(8.2x)
C: 300x (15x)
The number in () is
for MPA II Int sphere
For integrating
sphere of MPA
1/Matrix-I, changing
of preamp gain is not
possible
-Bruker Confidential- 26
 Preamp Gain setting for Background
(not for Int sphere of MPA 1/Matrix-I)

• Standard background gain settings are:

• Ref Sample compartment, external transmission
• A Liquid probe (IN236)
• B Powder probe (IN261)
• C Q412/A

• Different settings might be used for other probes, long fiber, and High
Intensity MPA.

• By the preamp gain the signal is amplified directly in the detector electronic,
without amplifying the noise (noise is manly introduced by the electronic
after the detector).

• The optimal preamp gain settings for background and sample can only be
found by trial and error. For inhomogeneous and/or moving samples this
must be done during movement of the sample.
-Bruker Confidential- 27
 Dark Sample, Same Resolution and
Measurement time, different Gains

Gain Ref
4

Gain A
Gain B
Gain C
3
Absorbance Units
2 1
0

12000 11000 10000 9000 8000 7000 6000 5000 4000

Wavenumber cm-1

-Bruker Confidential- 28
Preamp Gain
Moving Samples with proper Gain

-Bruker Confidential- 29
Preamp Gain
Moving Samples with too high Gain

-Bruker Confidential- 30
Preamp Gain
Moving Samples Spectra changes

-Bruker Confidential- 31
Saturation of Detectors

• An overload or saturation of a detector starts in theory with an amplitude of

32,000.

• In real live the detector can be overloaded even when the amplitude is
below 32,000. If the signal is exceeding the digital limit it can be converted
to a negative contribution leading to a lower amplitude.

• E.g. an amplitude of 28,000 can be that the real value is 36,000, but the
4,000 counts above 32,000 are flashed back and reduce the shown
amplitude.

That’s why you should always check the Single Channel Spectrum in the
Check Signal mode!

-Bruker Confidential- 32
Check signal (align mode)

• Display limits
set down to
0 cm-1!

-Bruker Confidential- 33
Check signal (align mode)

• Normal shape
of detector
signal with
sample in
place

-Bruker Confidential- 34
Saturation of detector (overloaded)

Signal overload
can be detected
by
• Strange shape
around 10,000
cm-1
• Signal goes up
from zero
beyond detector
limit (here
4,000 cm-1)

-Bruker Confidential- 35
Saturation of detector (overloaded)

Signal goes up
again below the
detector limit
after reaching
zero line (here
4,000 cm-1)

-Bruker Confidential- 36
Optical Slit & Total Absorbance

-Bruker Confidential- 37
Which Optical Path Length should I use?

• What are the components you are interested in?

• What are the concentrations of the components?
• E.g. if you are interested in small amounts of organics in water, water bands
could be total absorbed, CH bands are at different locations in comparison to
OH.
• E.g. if you are interested in small amount of OH bands in huge amount of
CH, the CH bands could be total absorbed!

• Optimized your pathlength regarding the precision required and range of

concentration

-Bruker Confidential- 38
Total Absorbance at Transmission
Measurements

• Total Absorbance = no light at the detector,

all light is absorbed by the sample.

• Spectral region of total absorbance is not allowed to use for evaluation

 these regions have to be removed (even in the optimization set up)

• How to recognize spectral regions of Total Absorbance?

 look at SSC data block!
 If signal is close to zero, it is a Total Absorbance region

-Bruker Confidential- 39
 Pentaerythritol and Dipentaerythritol in
Water

Signal Saturation
4.0

OBVIOUS
3.5
3.0

Can we use
this region?
Absorbance Units
2.0 2.5
1.5
1.0
0.5

10000 9000 8000 7000 6000 5000 4000

Wavenumber cm-1

-Bruker Confidential- 40
 Move to SSC
0.10
0.08
Single channel
0.06
0.04
0.02
0.00

10000 9000 8000 7000 6000 5000 4000

Wavenumber cm-1

-Bruker Confidential- 41
 Comparison SSC and AB Spectrum

Signal Saturation!!
4
3
Absorbance Units
2

SSC
1

AB
0

10000 9000 8000 7000 6000 5000 4000

Wavenumber cm-1

-Bruker Confidential- 42
 Comparison SSC and AB Spectrum
Detail

• Peak
3.0

around
Signal Saturation!!
7.000
cm-1
2.5

cannot
be used!
2.0
Absorbance Units
1.5 1.0

SSC
AB
0.5

8000 7500 7000 6500 6000

Wavenumber cm-1

-Bruker Confidential- 43
Fiber Optical Cables
Cut-Off

• Be aware of an increase of the cut-off frequency with

an increasing fiber length.
• Consider this for Method Setup & definition of Optical Slit

-Bruker Confidential- 44
 Long Wavelength Cutoff due to Fiber
Length
0.5

5m
10m
15m
0.4

20m
30m
50m
0.3
Single channel

100m
0.2
0.1
0.0

12000 11000 10000 9000 8000 7000 6000 5000 4000

Wavenumber cm-1

-Bruker Confidential- 45
 Fiber Optical Cable
Cut-Off Wavenumbers
0.16
0.14
0.12
0.10

500m 150m 2m
Single channel
0.08 0.06
0.04
0.02
0.00

5500 5000 4500 4000 3500

Wavenumber cm-1

-Bruker Confidential- 46
Fiber Optical Cable
Cut-Off Wavenumbers

Fiber Length [m] Cut-Off Wavenumber [cm-1]

2 3.900
5 3.990
10 4.150
20 4.290
30 4.360
40 4.400
50 4.440
60 4.460
70 4.500
80 4.520
90 4.530
100 4.560
120 4.580
Fiber cable measured
in short circuit 150 4.630
connection w/o probe
500 5.040

-Bruker Confidential- 47
Influence of Fiber Optical Cable
Length/Coupling

Single fiber: 145 m

Couplings: 20+20+50+50 m

-Bruker Confidential-
40 kHz Measurements

-Bruker Confidential- 49
 In-line Fermentation Spectra with
Scanner Velocities 10 and 40 kHz

10 kHz
1min
4

scan time

40 kHz
3
Absorbance Units

1min
scan time
2 1
0

11000 10000 9000 8000 7000 6000 5000 4000

Wavenumber cm-1

-Bruker Confidential- 50
 In-line Spectra of Fermentation with
different Scanner Velocities: O-H-Band
2.6

10 kHz
1min
2.5

scan time

40 kHz
2.4
Absorbance Units

1min
scan time
2.3 2.2
2.1
2.0

7100 7050 7000 6950 6900 6850 6800 6750

Wavenumber cm-1

-Bruker Confidential- 51
Aspects of NIR Analysis of Liquids

-Bruker Confidential- 52
Aspects of NIR Analysis of Liquids

• Pathlength
• 1, 2, 5 or 10mm?
• Sample presentation and handling
• Vial, flow cell or probe?
• Sample temperature
• Temperature control required?
• Water content of sample

• Viscosity
• Influences the selection of
pathlength and sample
presentation/handling
• Opaque or cloudy liquids
• Transmission measurements
limited; Transflection as
alternative

-Bruker Confidential- 53
Aspects of NIR Analysis of Liquids

Finally the selection of the pathlength is based on

handling and practical aspects.

• 8mm vials are

• Easy to fill without air bubbles
• Cheap consumable parts
• Not useful for aqueous solutions

• Important for transmission probes:

• At least 2 mm pathlength to avoid immobile
gas bubbles or particles in slit
• From 2mm on the slits are easy to clean
manually and allow visual inspection
• For high viscous liquids 5mm or more are
recommended and required

-Bruker Confidential- 54
NIR Spectrum of CHCl3

-Bruker Confidential- 55
Overtone of combination band and 2nd
Overtone of C-H vibration in CHCl3

-Bruker Confidential- 56
 Ethanol NIR Spectra
AB Spectra

optical slit
1 mm
2 mm
4

5 mm
10 mm
3
Absorbance Units
2
1
0

8000 7000 6000 5000 4000

Wavenumber cm-1

-Bruker Confidential- 57
 Ethanol NIR Spectra
Single Channel Sample Spectra

optical slit
0.14

1 mm
2 mm
5 mm
0.12

10 mm
0.10
Single channel
0.06 0.08
0.04
0.02
0.00

6000 5500 5000 4500 4000

Wavenumber cm-1

-Bruker Confidential- 58
 Cyclohexane NIR Spectra
AB Spectra

optical slit
5

1 mm
2 mm
5 mm
10 mm
4
Absorbance Units
2
1
0 3

9000 8000 7000 6000 5000

Wavenumber cm-1

-Bruker Confidential- 59
 Cyclohexane NIR Spectra
Single Channel Sample Spectra

optical slit
1 mm
0.25

2 mm
5 mm
10 mm
0.20
Single channel
0.15
0.10
0.05
0.00

9000 8000 7000 6000 5000 4000

Wavenumber cm-1

-Bruker Confidential- 60
 NIR AB Spectra of Water
measured with Transmission Probes
5

optical slit
1 mm
2 mm
5 mm
4

10 mm
3
Absorbance Units
2
1
0

11000 10000 9000 8000 7000 6000 5000 4000

Wavenumber cm-1

-Bruker Confidential- 61
 NIR Spectra of Water
Single Channel Sample Spectra

optical slit
1 mm
2 mm
0.20

5 mm
10 mm
0.15
Single channel
0.10
0.05
0.00

11000 10000 9000 8000 7000 6000 5000 4000

Wavenumber cm-1

-Bruker Confidential- 62
 NIR Spectra of Water
Single Channel Sample Spectra

optical slit
1 mm
2 mm
0.05

5 mm
10 mm
0.04
Single channel
0.03
0.02
0.01
0.00

7500 7000 6500 6000 5500 5000 4500

Wavenumber cm-1

-Bruker Confidential- 63
 Small Amount of C-H in O-H
3.5

Optical Slit
2 mm
5 mm
3.0
2.5
Absorbance Units
2.0

O-H
1.5
1.0

C-H
C-H
0.5

O-H
10000 9000 8000 7000 6000 5000 4000
Wavenumber cm-1

-Bruker Confidential- 64
 NIR Spectra of Water in Range of
0 to 95oC
1.4
1.2
Absorbance Units
1.0
0.8
0.6
0.4
0.2
0.0

9000 8000 7000 6000

Wavenumber cm-1

-Bruker Confidential- 65
 Toluene NIR Spectra in Range of
0 to 100oC
0.06
0.05
0.04
Absorbance Units
0.03
0.02
0.01
-0.01 0.0

9000 8000 7000 6000 5000

Wavenumber cm-1

-Bruker Confidential- 66
 Changes of Water Spectra with
Temperature

0 °C
5 °C
10 °C
1.45

15 °C
20 °C
25 °C
30 °C
Absorbance Units
1.40

35 °C
40 °C
45 °C
50 °C
1.35

55 °C
60 °C
65 °C
70 °C
1.30

75 °C
80 °C
85 °C
90 °C
1.25

95 °C
7100 7000 6900 6800 6700
100 °C
Wavenumber cm-1

-Bruker Confidential- 67
 Changes of Toluene Spectra with
Temperature1.165

25 °C
30 °C
1.160

35 °C
40 °C
Absorbance Units

45 °C
1.155

50 °C
55 °C
1.150

60 °C
65 °C
70 °C
1.145

75 °C
80 °C
85 °C
1.140

90 °C
95 °C
1.135

5958 5956 5954 5952 5950 5948

Wavenumber cm-1

-Bruker Confidential- 68
OPUS Practical Session

• Datablock
• Color assignment
• Peak picking
• Zoom in/out
• Show parameter including checking gain, amp, time for BG
• Compare parameter
• Change background
• Check signal and gain set up
• Velocity
• Correlation
• OVP  Set up, PQ Ref, LWN
• User setting and GLP mode
• Stray corr

-Bruker Confidential- 69
 Innovation with Integrity

©Copyright
Copyright Bruker
© 2011 Bruker Corporation.
Corporation. All rights
All rights reserved. reserved.-Bruker Confidential-
www.bruker.com