Improvements on an ATRIR Spectrometer

1
Improvements on an ATR-IR Spectrometer

• By Dylan Wilks
• Union College
• March, 2005

2
Acknowledgments

• Paul Wilks, President, Wilks Enterprise, Inc.
• Don Lavery, Senior Engineer, Wilks Enterprise,
  Inc.
• Palma Catravas, Senior Project Advisor,
  Professor of Engineering Union College
• Michael Rudko, Professor of Engineering Union
  College
• Seyfollah Maleki, Professor of Physics, Union
  College

3
Introduction

• Infrared spectrometry has been used for over
  fifty years to identify substances
• Useful since it requires only a small sample and
  does not destroy the sample
• New devices are becoming much smaller, allowing
  them to be used in the field or factory

4
Attenuated Total Reflection (ATR) Explained

• A sample is placed on the ZnSe crystal
• An infrared source emits light into the crystal
• The light reflects on the inside of the crystal,
  eventually reaching the variable filter detector

5
Attenuated Total Reflection (ATR) Explained

• Each time the IR light reflects off the sample
  side, energy is absorbed by the sample
• The amount of energy absorbed by the sample is
  used to generate a waveform

6
Project Definition

• Improve upon the Wilks Enterprise VFA
  spectrometer through both hardware and software
  modifications, and conduct experiments that will
  lead to future improvements.

7
Project Definition, Cont

• The project was divided into five sections
• Design a spectral search algorithm.
• Reduce thermal cross-talk through software design
  to improve output spectral quality.
• Conduct IR source experiments to categorize
  impact of source setup on spectral quality.
• Comparison of sample grinding techniques
• a. output spectral quality
• b. particle size categorization

8
1. Spectral Search Algorithm

• An algorithm that compares an unknown spectrum to
  a library of known spectra was desired.
• 2 designs were implemented Mean Square Average,
  and Factor Analysis
• Both algorithms were implemented in MATLAB

9
1. Spectral Search Algorithm Mean Square Average

• A simple approach that takes the squared
  difference between each pixel from a library
  spectrum and the unknown spectrum.
• The squared differences are then averaged, and
  the library spectra with the lowest mean square
  average is returned as the most likely match.

10
1. Spectral Search Algorithm Factor Analysis

• Factor Analysis is a method that allows for
  spectral search improvement by considering only
  the important parts of the spectra
• The data is represented as vectors, which are
  placed in matrix form. Matrices are then reduced
  to include only important information through
  singular value decomposition
• At this point the algorithm was less accurate
  than the mean square average approach.

11
2. Thermal Cross-talk Reduction

• The detector consists of 64 pixels
• When one pixel is heated by infrared light, the
  pixels adjacent to the heated pixel will heat up
  as well. This is thermal cross-talk.
• An algorithm was implemented to compensate for
  this

12
3. IR Source and Detector Experiments

• Experimented with different source setups, and
  different detector arrangements

Tested effects of lens vs. reflector light
collimation on pixel cross-talk and energy
Experiments led to a modification of the source
arrangement from 5 sources to 3 with
hemispherical lenses
13
4a. Sample GrindingOutput Spectral Quality

• Spectra were taken From samples ground with both
  a ceramic and agate mortar and pestle

• Both techniques generated similar spectral
  quality spectra

14
4b. Sample GrindingParticle Size Categorization
Agate Ground Asprin, magnified 100x, scale
1um

• Particle size distributions were found for pills
  ground in an agate mortar and pestle and a
  ceramic mortar and pestle

• The particle size distribution was similar for
  both grinding techniques, 10-50 um clumps
  of particles

Ceramic Ground Asprin, magnified 100x,
scale 1um
15
Conclusions

• Design a spectral search algorithm
• successful, 2 algorithms implemented
• 2. Reduce thermal cross-talk through software
  design to improve output spectral quality
• successful, algorithm increased spectral quality
• 3. Conduct IR source experiments to categorize
  impact of source setup on spectral quality
• found ideal source setup of 3 sources with
  lenses
• 4. Comparison of sample grinding techniques
• a. output spectral quality
• both techniques generated similar spectra
• b. particle size categorization
• 10-50 micron particle size distribution for
  both techniques

16
Summary

• The experiments and improvements completed in
  this project have increased the resolution, as
  well as the marketability of the Wilks Enterprise
  VFA spectrometer. This spectrometer provides an
  approach to infrared spectroscopy that is
  reliable, accurate and small enough to use in the
  field, rather than in a lab.

17
Low Carb Guinness?
18
Questions?