A Portable Mid-Infrared Chemical Detection Sensor

Submitted To

Dr. Ray T. Chen, UT-AustinGary Hallock, EE 464H Professor

Heng-Lu Chang, Technical TA

Prepared By

Travis BrannenJohn Elson

Steven PrickettKaarthik Rajendran

Aaron TreptowParker Wray

EE 464H Honors Senior DesignElectrical and Computer Engineering Department

University of Texas at Austin

Spring 2015

CONTENTS

TABLES................................................................................................................................iv

FIGURES...............................................................................................................................v

EXECUTIVE SUMMARY.................................................................................................vi

1.0 INTRODUCTION........................................................................................................1

2.0 DESIGN PROBLEM STATEMENT.........................................................................1

3.0 DESIGN PROBLEM SOLUTION.............................................................................2

3.1 Principles of Detection........................................................................................3

3.2 Operating the Laser............................................................................................4

3.3 Operating the Detector.......................................................................................5

3.4 On-Board User Interface and Wireless Tablet Platform................................6

4.0 DESIGN IMPLEMENTATION.................................................................................8

5.0 TEST AND EVALUATION......................................................................................10

5.1 Subsystem Testing.............................................................................................10

5.1.1 Laser Controller to Lab-On-Chip Testing................................................11

5.1.2 Detection Testing.......................................................................................12

5.1.3 User Interface Testing..............................................................................13

5.1.4 Power Testing............................................................................................14

5.2 System Testing...................................................................................................15

6.0 TIME AND COST CONSIDERATIONS................................................................17

6.1 Time Considerations.........................................................................................18

6.2 Cost Considerations..........................................................................................19

7.0 SAFETY AND ETHICAL ASPECTS OF DESIGN...............................................20

8.0 RECOMMENDATIONS...........................................................................................21

9.0 CONCLUSIONS........................................................................................................23

REFERENCES....................................................................................................................25

APPENDIX A – LAB ON CHIP SENSOR ....................................................................A-1

APPENDIX B – ABSORPTION SPECTROSCOPY....................................................B-1

APPENDIX C – LASER DRIVERS EXPLAINED.......................................................C-1

ii

CONTENTS CONT’D

APPENDIX D – ANALOG FRONT END EXPLAINED.............................................D-1

APPENDIX E – DIGITAL LOCK-IN EXPLAINED....................................................E-1

APPENDIX F – OVERALL I/O SYSTEM SPECIFICATIONS..................................F-1

APPENDIX G – BILL OF MATERIALS......................................................................G-1

APPENDIX H – APPLICABLE STANDARDS............................................................H-1

iii

TABLES

1 ChemSense Cost Benefit Analysis.....................................................................................9

2 ChemSense’s Prototype Power Results..........................................................................17

2 Cumulative ChemSense Project Costs............................................................................19

iv

FIGURES

1 Block Diagram of ChemSense..........................................................................................3

2 Schematic of ChemSense’s Detection Process.................................................................5

3 User Defined Set Points (Android App)............................................................................7

4 Android App Data Export Function..................................................................................7

5 Final Prototype Dimensions and Layout........................................................................10

6 QCL Output with ChemSense Negative Pulsed Laser Driver........................................11

7 Lock-In Detection Test....................................................................................................12

8 ChemSense Laser Driver Powered by Power Subsystem...............................................15

9 Integration Test of Laser Driver, DSP MCU, and Detection Circuitry..........................16

10 Android App Real-Time System Test...............................................................................16

11 Android App Real-Time System Test with Threshold Alert............................................17

v

EXECUTIVE SUMMARY

Our team’s senior design project, ChemSense, is a portable, mid-infrared chemical sensor system that has applications in national security and industrial safety. Current infrared chemical detection technologies are bulky and costly. Utilizing Dr. Ray Chen’s lab-on-chip QCL/QCD, our challenge was to design, build, and test a portable, affordable mid-infrared chemical detection sensor.

Our design solution takes traditional, benchtop spectroscopy components (pulse generator, laser driver, and lock-in detector) and shrinks them to fit on a custom-designed printed circuit board (PCB). Thanks to our unique PCB design and use of digital signal processing (DSP) detection, our solution addresses key design requirements in portability (<2 lbs), real-time detection (within 100 ms), and power consumption (2 W).

ChemSense’s design evolved over the course of our project. With every roadblock we faced, our team seized the opportunity to make a smarter design and stronger product. These modifications include: the addition of negative voltage laser driver, increased end user digital filter flexibility, and a second user interface in a dedicated Android application using Bluetooth low-energy (BLE) wireless connectivity. Ultimately, these additions make ChemSense more adaptable to a wide range of sensors and end users while still keeping product costs under $200.

Testing and evaluating ChemSense first required that critical subsystem metrics be met before system integration and final prototype build. Our laser driver was designed to meet the voltage (15 V) and current demands (2.5 A) of the lab-on-chip sensor. The analog and digital detection circuits were evaluated to define their sensitivity (100 ns pulses), and our user interface components were tested to ensure error-free end user interaction. During system testing our device with the lab-on-chip, our laser driver was able to run the laser in a manner comparable to a benchtop laser driver, but we are still working on successfully processing the output of the laser sensor.

Design hurdles in the digital lock-in (DLA) development and PCB test circuitry affected our design timeline. The DLA was delayed three weeks due to difficulties in writing the Assembly biquad filters code. The sensitivity and precision of the laser driver and detection circuitry made it impossible to use mock circuits on breadboards. We were forced to rely on multiple PCB test designs that were time intensive to design, order, and build. Still, our costs for the project stayed under budget ($810) thanks to creative use of circuit component samples from multiple vendors and affordable PCB board house providers.

In the future, we could improve ChemSense in ways that enhance component safety and portability. To reduce the lab-on-chip’s exposure to transient voltages, the design could include a configurable sensor pad. To make ChemSense even more portable, we could integrate our DC/DC power convertors onto the main PCB, switch to smaller components, eliminate redundancies in the laser driver subsystem, and employ a flexible PCB. Given the complexity and challenges we faced, we are proud of the work we accomplished to help Dr. Chen realize his goal for a portable, affordable mid-infrared chemical detection device.

vi

1.0 INTRODUCTION

This document is describes our final design for ChemSense, a portable laser spectroscopy system

that detects hazardous airborne chemicals. By utilizing Dr. Ray Chen’s lab-on-chip laser sensor,

ChemSense offers chemical concentration detection on a mobile platform. We designed

ChemSense to save lives by alerting users of harmful airborne chemicals that cannot be detected

by human senses. Our design integrates tunable, mid-infrared laser spectroscopy in a portable,

low-power package. Previously this semester, we presented our initial design for ChemSense and

described our testing plan. This document presents the results of using those documents to build

and test a physical device. First, we describe how our design addresses the specific need for

portability and accurate detection required by our project sponsor, and the important metrics that

define success for our design. Next, we will describe the laser driver, analog filter, DLA, and UI

subsystems and how each is integrated into the overall design. We then discuss how challenges

such as learning that we needed to generate negative laser pulses affected our design solution.

Our system and subsystem-testing summary discusses, among other topics, our success at

generating laser pulses and our progress at detecting chemical concentration using the

QCL/QCD. Finally, we will discuss how time constraints limited our success at laser detection,

how we were able to stay under budget through multiple prototypes by using sample materials,

and provide recommendations for how to make ChemSense more flexible, accurate, and

portable.

2.0 DESIGN PROBLEM STATEMENT

Utilizing newly developed laser spectroscopy technology, our design of ChemSense solves a

need for a mobile chemical concentration measurement device. Current laser chemical sensing

technologies either require off-site analysis, or bulky onsite equipment. A portable chemical

identification device (PCID) could save lives by alerting the user when a dangerous airborne

chemical is present. Modern advancements in QCLs, QCDs, and optical waveguides in the mid-

infrared region (Mid-IR) have resulted in accurate and sensitive lab-on-chip chemical sensors,

approximately one centimeter in area [1, 2]. Despite these advances, an integrated mobile

controller does not currently exist. Dr. Ray Chen’s company, Omega Optics, has developed a

QCL/QCD lab-on-chip chemical sensor. Our device capitalizes on the small size of Dr. Chen’s

chip by integrating it into a portable controller.

Our design requirements come from the power needs of the lab-on-chip sensor and our design’s

goal to be a portable chemical sensor. For proper operation, the QCL requires a high fidelity

pulse shape at an optimized duty cycle, voltage, and current [3]. In order to lase, the QCL

requires a minimum of 10-20 V and 2.5-4 A, depending on fabrication quality (See Appendix A).

Above this threshold, additional power enhances the accuracy of chemical detection. On the

other hand, providing too much continuous power will destroy the lab-on-chip. To provide

enough power for the QCL to lase but avoid damaging the chip, we must provide short duration

(<100 ns) pulses at a maximum frequency of 50 kHz. As we are providing the QCL with these

short duration, high intensity pulses, we will be measuring the current output from the QCD.

The QCD output will be on the order of a few μV, so our detector circuit will need to be able to

convert a time dependent μV scale signal into a concentration measurement. We will evaluate

the system as a whole by repeating the tests conducted to characterize the chip by Dr. Chen’s

group. Since their test was conducted using a benchtop setup, if our device can obtain results that

are within 1 % of the results obtained by their test, we can conclude that our device is as good as

the benchtop system they are using in lab. To achieve our goal of portability, our system must be

low power (<2 W continuously) so that a lightweight battery can serve as the power source.

Additionally, the system needs to be lightweight enough to be easily carried (< 5 lbs). Finally,

given that ChemSense’s primary application is detecting hazardous chemicals, the device must

operate in real time, notifying the user within 100 ms if a dangerous substance is present.

3.0 DESIGN PROBLEM SOLUTION

ChemSense provides a fully customizable, lightweight, and low power solution to the need for

mobile broad spectrum chemical detection. ChemSense is composed of a dual mode laser driver,

quantum cascade laser (QCL) and detector (QCD), digital lock-in amplifier, user interface

(onboard touch screen and/or android app), and power electronics. These elements are integrated

on a custom printed circuit board (PCB) weighing less than 2 pounds and requiring less than 2

watts of power. From the user interface, the user will input operating parameters. Using this

information, ChemSense will operate the laser using the laser driver and read the output from the

detector using an analog filter and a digital lock-in amplifier. Once the sensor’s output has been

filtered, the data is sent to the user interface where the Beer-Lambert law is used to determine the

2

concentration of the chemical of interest. This data can either be output to ChemSense’s 7”

onboard touch screen, for real time data analysis, or sent via bluetooth to an android device.

ChemSense’s android application offers all the functions found on the attached touch screen as

well as the ability to export data to a computer. Both the laser driver and signal processor are

controlled by a microcontroller dedicated specifically for digital signal processing (DSP MCU).

Concentration calculations, the user interface, and bluetooth communications are handled by the

user interface microprocessor (UI MCU). A high-level block diagram for ChemSense is shown

in Figure 1, below.

Figure 1. Block Diagram of ChemSense.

ChemSense is a highly customizable chemical sensor. Because it can source and interpret

variable frequency, duty cycle, amplitude, and signed pulses, ChemSense could be used to run

nearly all solid-state lasers and detectors on the market. This means ChemSense has the potential

to sense nearly any chemical using the same control system by changing the sensor used.

3.1 Principles of Detection

ChemSense senses chemicals by absorption spectroscopy, a method of chemical detection where

light of a particular wavelength passes through the chemical of interest and is attenuated based

on the chemical’s absorption of light at that wavelength. A detector then collects the light and the

attenuation is used to infer chemical concentration. This process is governed by the Beer-

3

Lambert law and described in detail in Appendix B. One of the most important factors of

absorption spectroscopy is the choice of light wavelength. ChemSense uses QCL’s because these

lasers operate in the mid-infrared, the most selective wavelength region in the electromagnetic

spectrum for chemical detection. However, ChemSense was designed so that the user could

insert their own laser and detector and sense chemicals by detecting their absorption of light in

other wavelength regions, if desired.

For custom lasers and detectors, the user will need to specify the wavelength, pulse repetition

frequency and duty cycle for the laser. This defines how the laser will be modulated and how the

signal-processing unit will detect the signal. For proper chemical detection the user also needs to

input the molar absorption coefficient of the chemical they want to detect at the wavelength of

the laser they choose to operate. With this information, ChemSense can use the Beer-Lambert

law to determine concentration based on what it detects. Once this is done ChemSense will

repeatedly update the Beer-Lambert equation to provide a real time concentration of the

chemical of interest.

3.2 Operating the Laser

ChemSense’s laser driver can provide either positive or negative voltages in either continuous or

pulsed mode, depending on the application. In providing these options, we attempt to provide a

laser driver that can operate most solid state lasers on the market today. When the user sets a

laser modulation frequency and duty cycle, this information is sent to the DSP MCU. The DSP

MCU generates a sequence of pulse triggers, which are used as the control signal for the laser

driver. The laser driver then sends high voltage and current pulses to drive the laser.

Providing both positive and negative pulses on the same board requires proper bootstrapping.

Because of this, the negative pulse driver requires two negative power supplies, one to drive the

laser and the other to create the artificial reference to drive the MOSFET gate. Currently,

ChemSense’s portable power electronics do not provide negative voltages. Instead connections

are available for the user to input their own battery (or power supply) to provide the required

negative voltage. More information about this and how the pulse drivers operate can be found in

Appendix C.

4

3.3 Operating the Detector

Once the attenuated light signal is converted to a microamp scale current by the QCD (or the

users photodetector of choosing), it is sent through a transimpedance amplifier, low-pass filter,

gain, then into the ADC of the DSP MCU. From there a digital lock-in amplifier (DLA) extracts

the signal. This process is shown in Figure 2, below. We chose to use a transimpedance

amplifier because it provides the best gain interface for photodetector loads. The tunable anti-

alias filter employs a first order RC filter that both broadens the pulse and ensures the Nyquist

condition is met at the ADC. The final tunable gain allows the user to ensure the maximum

dynamic range of the ADC is used, so ChemSense can recognize small changes in concentration.

More information about the analog front end can be found in Appendix D. Once the signal is sent

to the ADC, the DLA extracts the signal from all out of band noise and give the UI MCU a

recording of the amplitude of the signal at that instant in time. The DLA works under the

principle of frequency locking to isolate the signal among potentially large amounts of noise.

Simulation has shown that ChemSense’s DLA is capable of finding the signal in an environment

where the noise can be up to six times stronger than the signal of interest. More information

about the DLA can be found in Appendix E.

Figure 2. Schematic of ChemSense’s Detection Process.

ChemSense’s detection process is designed to handle signal to noise ratios below .5 while still

offering a fully tunable setup for the user. The DLA has support for the user to implement their

5

own band-pass filter, lock-in modulator, and low-pass filter, if desired. They need only download

their filter and modulation coefficients into the system and ChemSense will do the rest. This

provides a near 100% customizable DLA, up to the limit of the DSP MCU’s processor speed.

Written entirely in Assembly, the DLA provides a more efficient signal translation than would be

possible using higher level languages. Since both the DLA and laser driver controller are both

implemented on the same DSP MCU, the DLA shares the same pulse trigger. This ensures

frequency locking because the modulation frequency in the DLA is the same as the one used in

the laser driver. This is important for signal recovery because if the DLA is not phase matched to

the input signal it will not function properly. Due to the limit that the ADC in the DSP MCU can

only collect voltages between 0 and 3.3 V, all negative pulses are rectified in the analog front

end. This means when ChemSense is operating in negative pulse mode, positive pulses are

detected. After the DLA generates a DC concentration measurement, it is sent to the UI MCU.

3.4 On-Board User Interface & Wireless Tablet Platform

ChemSense provides the option to interface with the sensor either through an onboard 7” touch

screen to interface with the user as well as a wireless Android tablet. The idea of the onboard

touch screen is to provide full control and immediate data analysis directly from ChemSense,

without out the need for external equipment or interfacing. The Android application provides

external control and analysis. This allows the user to attach ChemSense to a robot to analyze

dangerous chemicals in environments not safe for people. Additionally, if the user desired to

monitor an area for a prolonged period of time, they could leave ChemSense running and check

the Android application periodically. Using an external UI has the added benefit of reducing the

power needs of the device itself.

The user interface is run by ChemSense’s dedicated UI MCU. This processor handles all user

requests, ensures the requests are sent to the proper subsystem, and manages all communications

with ChemSense’s android application. The processor also performs all concentration

calculations and updates a concentration versus time graph for the user every 100 ms. The user

can also define a concentration set point at which an onboard alarm will sound. This feature is

also mimicked on the tablet. Written in Java, the app allows the user to perform all of the same

functions with a wireless tablet as with the on board user interface (Figure 3, on page 7) as well

6

as the ability to export recorded data (Figure 4, below). Utilizing the onboard memory and

processing power of the tablet, ChemSense’s Android app can record up to one hour of sensing

data (36,000 samples) and wirelessly email this data to an address of your choosing. This allows

for additional post processing, if desired.

Figure 3. User Defined Set Points (Android App).

Figure 4. Android App Data Export Function.

7

4.0 DESIGN IMPLEMENTATION

Throughout the process of designing, building, and testing ChemSense, unforeseen obstacles and

opportunities shaped the final product. The major changes to our initial design were the addition

of a negative voltage laser driver, the inclusion of an android application, and the removal of the

temperature control module. We were forced to develop a negative voltage laser driver when,

during testing, it became clear that the QCL required a negative voltage pulse to operate. We had

completed all of our preliminary design with the belief that the laser required a positive pulse,

and because of the fact that obtaining lab time with the lab-on-chip was difficult, we were

already fairly far along in the design process before we tested the chip. Since we had already

designed the system with a positive laser driver, we chose to leave both positive and negative

laser drivers on the chip in the final product. We chose to add an android application when our

project sponsor suggested it as a way to make our project more appealing at open house.

Implementing an android user interface ended up having the additional benefits of allowing us to

offload a significant portion of the power needs of our device (running the touch screen) to the

user’s cell phone, or allowing the removal of the LCD all together for increased portability. We

removed the temperature controller from the final design because we determined that it would

require eight watts of power which would not work with our design constraints. As a low power

alternative, we redesigned the sensor pad on the PCB to act as a heat sink.

Because of what we learned and the changes we made during the execution of our project,

several alternate designs for ChemSense are now possible. These designs center around positive

versus negative laser pulsing, the location of the user interface, and making the design wearable.

Our final prototype includes positive and negative laser driving, but the negative laser driver

requires an external power source. Alternate forms of ChemSense could provide only positive or

only negative laser driving or both, ideally with power supplied by the system’s batteries. The

simplest version would be having only positive laser driving, since this design has been

completed and tested. More design would be needed to work out how to power a negative laser

driver using the system’s batteries. A version of ChemSense offering both positive and negative

laser drivers would be able to operate a larger set of laser detectors, but would be slightly more

costly to produce.

8

After having completed both the android and onboard UI’s, it is clear that our onboard UI was

more difficult to implement than the android app and also draws power from the system’s

batteries that could be used to increase our operating time. Additionally, the onboard UI adds to

the size of the system. Because cell phones are so ubiquitous, we could easily offload our entire

user interface onto an android app, reducing our system’s power consumption, weight, and cost.

Finally, when we began implementing ChemSense we did not have a clear idea of how big the

final system would be; we were able to produce a system that is small enough to be easily carried

(Figure 5, on page 10), but there are a few optimizations that could make it even smaller. By

removing the onboard UI, putting the power supply module on the same board as the rest of the

system, reducing the battery size, and using only one microcontroller, we believe that

ChemSense could be as much as 60% smaller. Additionally, by using a flexible PCB, we could

make the device into an armband or a belt. Even without these improvements, ChemSense is

still far superior to a benchtop system, offering the same performance at one-fiftieth the cost (See

Table 1 below). A full bill of materials can be found in Appendix G.

Table 1. ChemSense’s Cost Benefit Analysis.

Item Lab Equipment ChemSense

Computer and UI $600 $100

Dual (+/-) Supply $670 $50

Laser Driver $340 $25

Digital Lock-In $4,000 $10

Pulse Generation $4,000 $10

TOTALS: $9,610 $200

9

Figure 5. Final Prototype Dimensions and Layout.

5.0 TEST AND EVALUATION

To determine how well we achieved our project goals we tested each subsystem and system as a

whole. We ensured our subsystems were ready for integration after we characterized each one’s

I/O and then combined all our hardware on our PCB to verify the software modules connected

with the hardware elements. Finally, we constructed our final ChemSense prototype and

evaluated whether it met our overall system specifications. (see Appendix F) In the following

section we will describe the testing process and results for each subsystem, including our

evaluation of integrating our laser driver with the QCL and tuning our analog front-end and DLA

detection circuitry. The following sections describe ChemSense’s system test process and results:

the conditions for collecting data, our testing results, and an evaluation on whether the test

outcomes were successful. In general our tests were successful, but we have not yet

demonstrated our detection circuit with a QCD input.

5.1 Subsystem Testing

Since our project contained different, interdependent systems, our design success depended on

being able to test each subsystem accurately in isolation. There were four major modules that had

to be independently tested: the laser controller, the detector, the user interface, and power. The

analog part of the system is comprised of the laser driver, the lab-on-chip sensor, and the

detector’s analog front end. The analog system feeds into the detector’s DLA (DSP MCU) that

produces a DC signal sent to the UI MCU. The UI MCU sends BLE data to the android app. The

10

following sections describe our critical subsystem tests. First, we discuss our success in

interfacing the QCL to our laser driver. Then, we describe what our detector circuit is capable of

detecting. Finally, we describe how we evaluated our user interface and power systems.

5.1.1 Laser Controller to Lab-On-Chip Testing

We tested the customized laser driver using different loads to simulate the QCL before

interfacing with the QCL. Because QCL's are expensive to fabricate, we used an 8 ohm resistive

load to confirm that both our positive and negative voltage laser drivers achieved voltages

greater than |15| and average currents equivalent to 10 mA. After verifying with an oscilloscope

that the high-voltage high-current output pulse was an approximately rectangular waveform with

a pulse width and a pulse repetition frequency matching the input pulse, we then used the QCL

as the load.

Testing our laser driver with the QCL, we were able to induce the QCL to lase using negative

voltage pulses (See Figure 6 on the following page). During our test procedure, we increased the

amplitude of our voltage pulses while monitoring QCL temperature to ensure that it was not in

danger of overheating. Once we observed measured power generated by the QCL greater than

200 mW, we concluded that our laser driver had successfully powered the QCL.

Figure 6. QCL Output with ChemSense Negative Pulsed Laser Driver.

11

5.1.2 Detection Testing

In order to certify the correctness of the results from our lab-on-chip, we had to prove that its

energy loss was detectable by the lock-in. We needed to spread the pulse out in time so that the

ADC of our microcontroller could detect it. In order to quantify the temporal spreading, we built

an analog filtering circuit on a breadboard with tunable resistors. We fed a square wave from the

lock-in microcontroller through the front end and through the digital filter. We then gradually

decreased the duty cycle in an attempt to test the minimum pulse width we were capable of

detecting. Figure 7 below shows the contrast between the lock-in detecting a signal and failing to

detect one. The lock-in can convert a square wave into a DC signal if the square wave has a large

enough duty cycle. These graphs show that, with the help of the analog filtering circuit, the lock-

in is capable of detecting pulses with 100 ns width but not 75 ns width.

Figure 7. Lock-In Detection Test

12

Because we implemented the digital lock-in amplifier in a microcontroller, we had to debug for

both code correctness and performance. Debugging for code correctness means ensuring the

program works as intended. This was done comparing the response to a given input of our DLA

with that of an identically designed set of filters in MatLab. Our filters produced values within

0.01 % of the MatLab simulated values, indicating that the only difference between our filter and

MatLab’s was in rounding.

Once we were sure our filter was correct, we had to make sure we could meet our performance

requirements. Specifically, the ADC samples at 100 kHz, so we had to be able to run the filter

100,000 times a second. Through empirical testing, we concluded that the bandpass and lowpass

filters could require a combined number of coefficients of at most 66, below the theoretical limit

of 100 coefficients that we had designed around. Empirical testing also revealed that each sample

took at most 8.2 microseconds to process, less than the 10 microseconds allocated to each sample

by the ACD’s sample rate. After proving that our filter algorithm worked as expected within the

required timeframe, we chose filter coefficients and performed functional testing of our filter by

trial and error. For further details on what software tools were used to accelerate this process,

please see Appendix E.

5.1.3 User Interface Testing

We tested the Touchscreen UI in two ways: unit tests and functionality testing. Our unit tests

were implemented using software modules that passed values into individual functions within

our code and tested whether the output of each function matched our hand-calculated expected

values. This was done until each function worked for a specific range of input values and

produced predictable outputs. Functionality testing was conducted by navigating the display

modes of UI touchscreen to ensure the interface responded as intended. Our functionality testing

did not ensure that the UI is bug-free, however, after we completed unit testing of all UI software

elements and functionality testing of all UI features we are confident that our UI is error free. To

interface with the signal processing subsystem (DLA), we wrote a program in Java to collect

serial data from the output of the lock-in. We used this both to test data to design our system and

to simulate the interface between the user interface microcontroller and the lock-in

microcontroller. Because the Java application used the same communication format as the two

13

microcontrollers, when we had to hook the lock-in microcontroller into the user interface

microcontroller, we were able to get the two communicating with minimal effort.

Testing the Android application required creating both a BLE client and server. Utilizing an open

source app (nRF UART 2.0 app from Nordic Semiconductors) that can send and receive BLE

data we were able to determine whether our UI MCU could send and receive bluetooth

transmissions. Once we verified that the UI MCU was handling data appropriately, we used our

UI MCU to troubleshoot the BLE functionality in our android app. After connecting the android

app to bluetooth, we were able to test the plotting and data handling abilities of the app. The BLE

connection allowed us to stream in data at 10 Hz. Testing 1000 packets, we determined that the

android app is successfully receiving >99.9% of packets transmitted by the MCU. We also

performed functional testing to determine that the android app is also able to send control signals

to the UI MCU.

5.1.4 Power Testing

Our power system testing’s primary goal was to ensure the laser driver, MCUs, and LCD

touchscreen could receive consistent power from the DC/DC power convertors, via the battery,

and not fail during operation. It was critical that our power system be able to provide the pulsed

current demanded by our laser driver in order for ChemSense to be portable. Figure 8 on page 15

shows the laser driver output as driven by our battery. The battery is able to provide the 15-25V,

>2A pulse required by the laser driver. To test the power system’s endurance, we interfaced the

power subsystem to directly power the laser driver, MCUs, and touchscreen to monitor battery

performance. During five, independent 8-hour tests we noted that the battery voltage deviated, on

average, by 0.4 V after the sixth hour. Our final system power performance evaluation can be

found in the next section.

14

Figure 8. ChemSense Laser Driver Powered by Power Subsystem.

(Yellow line is the pulse generation output, Blue line is the voltage across our laser diode

terminals with a resistive load.)

5.2 System Testing

After our subsystem tests and evaluations were complete we were ready to integrate the entire

system. At the time our design reached system testing the lab-on-chip sensor was not available

for testing, nor would it be available to showcase at open house due to the chip’s vunrability at

its current state. In addition, the exposed QCL mid-IR lasers posed a safety threat. In order to

progress in our testing, and have a fully realized prototype for open house, we chose to configure

a near-IR sensor on ChemSense. While our team has not yet been able to complete the lab-on-

chip mid-IR detection tests, we were successful with the positive pulsed system integration and

provide the results in the section below.

Before full system integration we tested the critical connection between our near-IR laser driver

and detection subsystems. We tested absorption detection by using three translucent films of

differing index and observed the results. Figure 9 on page 16 confirms that we were able to prove

our DLA, DSP MCU, and laser driver integration was a success.

15

Figure 9. Integration Test of Laser Driver, DSP MCU, and Detection Circuitry.

We were then ready to fully integrate the system, including our UI and power, and again use the

same test to confirm detection results. We had already used simulated data to display a real time

graph on both the onboard and android UI’s, and we had sent bluetooth data between the UI

MCU and the android app. So, after we interfaced the laser driver to the photodiode to the

detector subsystem, completing full system integration was straightforward. Figures 10 below

and 11 on page 17 are conclusive results of the full system performing chemical absorption

analysis. Figure 10 shows real-time (10Hz) data plotted in blue and the average for each minute

in green. Figure 11 shows the user-defined threshold being met by ChemSense and triggering

the alarm warning.

Figure 10. Android App Real-Time System Test.

16

Figure 11. Android App Real-Time System Test with Threshold Alert.

Once our final full system tests were proven we then used the benchtop power supply to feed our

DC/DC convertors to observe the power consumption of our prototype. Table 2 below shows the

results. When our full 7” LCD display is plugged in ChemSense draws slightly less than 2 W

(while the LCD screen is lit). We then disconnected the on-board LCD and then confirmed that

our power consumption was significantly less at 0.72 W while just utilizing the BlueTooth LE UI

functionality.

Table 2. ChemSense’s Prototype Power Results.

Prototype Setup Power Consumption

Full 7” Prototype 2 Watts

Prototype (without LCD) 0.72 Watts

6.0 TIME AND COST CONSIDERATIONS

While our team was able to meet our project timeline and goals, it was not without challenges

that altered or delayed our design schedule. The following section details our specific time

management concerns related to designing our custom PCB and troubleshooting our DLA

algorithm. We then highlight the choices our team made that allowed us to complete our project

in time and under budget.

17

6.1 Time Considerations

Time constraints were a significant issue for our project, largely due to the scope of our

design. Turn around times on circuit board manufacturing, DSP algorithm development,

and sensor operation issues caused some unexpected delays. The low-noise requirements of

our project made it necessary for us to have several circuit boards manufactured to get a

realistic idea of how our circuits would function in practice. The DLA we chose to

implement had it’s own challenges, and took much longer than expected to design and test.

Finally, the main sensor in our project was still in an experimental state for the majority of

our timeline, making it hard to nail down the exact operational characteristics we needed to

meet.

The main time challenge we faced was in the creation of the custom circuit boards needed

to maintain the precision necessary for QCD signal retrieval. We had difficulty using

breadboards to test circuits because they introduce a significant amount of parasitic

capacitance. These parasitics, along with unreliable breadboard contacts, affected the

operation of the sensitive analog components our project required. Designing a custom

circuit board was the only solution. Unfortunately, unpredictable manufacturing and

shipping times resulted in delays, during which we had to go about testing our proposed

circuitry by other means. Typical turnaround time for a single PCB was about ten to

fourteen days, and we ultimately ordered several sets of PCBs.

We chose to use a digital lock-in-amplifier algorithm largely because of the flexibility it

allows, but the implementation presented more issues than we initially expected. Precision

and speed requirements and constraints in the compiler we were using necessitated that the

DLA be written entirely in assembly language. Assembly language is much harder to debug

than code written in higher level languages such as C, and can end up requiring ten to one

hundred times the number of lines of code to accomplish the same goal. However,

assembly offers an increase in execution speed and, in this particular case, precision, that

we could not achieve with higher-level languages. We initially estimated that our DLA

would only take a couple weeks to develop, but in practice, it took nearly two months. An

upside to the development time invested in the DLA is that a significant amount of

18

modularity was built into it in order to help us test the design along the way, which allows

the end-user flexibility in customizing the DLA to their needs.

6.2 Cost Considerations

Our team was able to complete the project under the $1,000 limit (See Table 3, below) by

taking advantage of sample materials, modern low-cost PCB manufacturing services, and

use of existing equipment provided by Dr. Chen. As we were not provided with any

specific technology requirements beyond the use of the lab-on-chip sensor, our design was

completely custom, and required careful evaluation of many different parts. Throughout the

design process we evaluated over twenty different chips for our analog circuitry and

designed, built, and tested three different major revisions of our circuit board, along with

several revisions of smaller subsystem test boards.

Table 3. Cumulative ChemSense Project Costs.

Item Cost

PCB Orders $285

Near-IR Lasers and Detectors $150

All IC’s and Components $110

MCU Development Boards $100

7” LCD Screen and Driver $80

Li-ion Battery & Charger $75

BlueTooth LE Module $20

TOTAL: $810

Another cost factor we took advantage of was the use of new, affordable budget PCB board

houses for our design. In the past few years, many new-generation PCB manufacturing

services have sprung up which enable both professional and hobbyist access to high quality

printed circuit boards. Only a few years ago, one could easily spend over a hundred dollars

to have a single PCB made. These days, services such as OSH Park, Cheap Dirty PCBs,

and SeeedStudio offer extremely affordable PCB manufacturing services, with tolerances

19

and quality that rival and even exceed many larger, more established board houses. By

utilizing Cheap Dirty PCBs, our team was able to order three different revisions of our

board, receiving ten copies of each, all for under $300. The quantity of boards received

allowed us to create several copies of each version for testing.

Furthermore, since our design required that we take existing bulky, expensive equipment,

and distill the necessary functionality into our design, we had a significant amount of

proven (and unproven) hardware available to us. Some of the laser drivers we were

provided proved to be unusable for the lab-on-chip, which necessitated the creation of our

own design - a challenge we did not initially expect. Yet another unexpected challenge was

presented in the polarity requirements of the lab-on-chip itself. Our solution to these

problems resulted in tunable laser pulse drivers for both positive and negative polarities -

both costing significantly less than any commercial solution available to us.

7.0 SAFETY AND ETHICAL ASPECTS OF DESIGN

We designed our project to be safe to operate, reliable in detecting hazardous chemicals, and

proactive in alerting users to danger. Designing with safety in mind was important because

ChemSense’s operation is not inherently safe.The lasers required for mid-infrared spectrometry

can damage eyes. In addition, users will be trusting ChemSense to detect hazardous chemicals so

ChemSense must not fail to alert users to the presence of these chemicals. ChemSense avoids

failing to alert the user by testing the audible alarm on startup and failing safe if the laser fails.

The first and most serious safety issue the team faced in building and testing ChemSense was

dealing with the risks posed by infrared lasers. Because infrared lasers lase outside of the visible

wavelength the power of the laser isn’t obvious. If we had been dealing with similarly powerful

light sources in the visible spectrum (burning magnesium, for example) their brightness and the

resulting visual distortions would have kept us vigilant to the danger. Being unable to see the

laser in no way makes it less dangerous: infrared lasers can cause corneal burns and cataracts

[5]. (See Appendix H, Applicable Standards) Early on in the lab, we protected ourselves by

wearing safety glasses. Once we moved to the prototyping stage, we built a large foil covered

box and performed tests inside it. In the final product, we 3D printed a small cover to fit over the

20

laser / detector pair. We know that the material we used is effective because passed a small

sliver of it between the laser and detector and verified that it blocked 100% of the passed energy.

So long as this cover remains in place, user exposure to dangerous lasing will be minimal. If this

cover were to become dislodged, any operator would risk optical damage. Therefore, we

recommend a warning label advising against using ChemSense if the cover is displaced.

The second issue is avoiding false readings, either positive or negative. A false negative is when

our project is being used correctly but fails to alert the user to the presence of the chemical that

they’re attempting to detect. One way the system could create a false negative result is if the

alarm fails to function. It is not reasonable to expect the user to be constantly monitoring

ChemSense’s screen for hours at a time. The alarm is programmed to sound at startup as a

diagnostic to prevent this type of failure. At the same time, a notification informs the user that

the alarm should be sounding. If the user sees this notification but not the alarm, they will know

that their unit is malfunctioning. Thankfully, a malfunction of the laser / detector pair is unlikely

to cause a false negative. ChemSense fails safe because if the laser stops lasing or the detector

doesn’t detect anything, this is electrically equivalent to extremely high concentrations of a

chemical blocking out 100% of the laser’s power. Thus, in the most common failure case, the

system will alarm constantly rather than not at all.

8.0 RECOMMENDATIONS

While we have succeeded in constructing modules for the laser driver, pulse detection front-end,

digital lock-in amplifier, and a user interface, we have several recommendations for optimizing

the system design. Some important changes include replacing the QCL/QCD packaging from

pads to through-holes, placing the power electronics directly on the printed-circuit-board, and

using an alternative circuit design for detecting short duty cycle pulses. In this section, we will

identify alternatives for modules at the board level.

Optimizing the QCL/QCD packaging and PCB routing will significantly decrease overshoot and

ringing during high-to-low and low-to-high transitions (See Figure 9 on page 16 for an example

of ringing). Smoother rise and fall transitions on the input pulse prevents damage to the laser and

other electrical components. The lab-on-chip device currently sits on large area pads, which

21

effectively behave like capacitors because a dielectric insulator separates the gold contact from

the copper ground plate.

Additionally, we recommend packaging the laser in a metal can TO-56 or TO-9 package with

short leads. Because our system delivers low duty-cycle pulses to the laser, and likewise,

receives low duty-cycle pulses from the detector, it is critical to keep the wiring between

modules as short as possible. During our system testing, we recognized that longer connections

between modules significantly degraded the signal-to-noise ratio and introduced distortions.

While we succeeded in maintaining short trace lengths on the PCB we submitted, the next

revision would ideally incorporate the QCL/QCD with short leads directly soldered to the PCB.

To make ChemSense more portable, it needs a negative voltage source and for the power

electronics subsystem to be incorporated onto the same PCB as the rest of the system.

ChemSense's negative voltage driver requires two dc negative voltage supplies. We currently use

DC power supplies to provide those voltages, which is sufficient for testing in the lab. A future

revision needs a positive-to-negative voltage generator with dc-dc converters compatible with

negative voltages. Additionally, our power subsystem could be put on the same PCB as the rest

of the system. Currently, our system uses a single lithium-ion 7.4 V battery to power the

electronics. As part of the positive laser driver, a buck converter steps down the battery voltage

to the required voltage needed for the laser. Because we were concerned about noise

introduction, the buck converter is currently a separate module on its own PCB. The next

revision of the PCB should incorporate a buck converter.

The laser driver modules can be upgraded to save space on the PCB and produce pulses with

minimal noise. First, the isolated gate driver on the negative voltage pulse generator can be

replaced with a comparable surface mount component. A surface mount component saves space

and reduces parasitics because the leads are shorter. Currently, our negative pulse generator uses

two dc power supplies with a common ground. If we replace one of the dc power supplies with

an isolated power supply, the laser circuit and the digital circuits do not share a current path,

decoupling noisy circuits. After performing the recommended heat dissipation study mentioned

previously, the power MOSFET can be replaced with a compatible surface mount component

22

such as a small outline package to save space and reduce trace length. Additionally, the NMOS

transistor can be replaced with a different MOSFET family with improved switching

characteristics including faster rise and fall times and reduced parasitic capacitances. Because a

laser's operating point can change with temperature variations, we recommend adding a circuit

with feedback control. A current sense monitor together with a thermoelectric (TEC) module will

ensure constant current and voltage are delivered to the load despite external changes.

On the detection side, revising the analog front-end with a sample-and-hold (SH) circuit is

critical for detecting low-duty cycle pulses. Our current solution is inherently inaccurate because

it manipulates the pulse amplitude and the ADC sample rate is inadequate. We recommend

replacing the low-pass filter and gain stage with a sample-and-hold circuit. The sample-and-hold

circuit charges a low-leakage capacitor to the maximum pulse level for a time period comparable

to a 50% duty cycle. This eases the ADC requirement, and because the fundamental has more

energy at a 50% duty cycle, the lock-in will more accurately reflect the actual pulse amplitude

from the detector. The SH requires a low-leakage capacitor and several vendors provide

integrated circuit components.

9.0 CONCLUSION

At the culmination of this project our team is proud of our design, ChemSense. After careful

thought and planning in the Fall 2014, we spent approximately 2,000 man-hours designing,

building, and testing our modules in Spring 2015.  We laid out three printed circuit boards, each

with new analog circuits to adapt to our project's challenges, and wrote over 5,000 lines of

Assembly, C, and Java code. We are proud that our final design solution exceeds several of our

benchmarks including portability (< 2 lbs.), power (2 W), real-time detection (100 ms), and cost

($200). In our testing phases, we successfully proved that our negative pulse driver provides

sufficient voltage and current for the QCL to emit light and that our user interface relays data to

the user via a screen in real-time. We successfully showed that our system is capable of driving

commercial infrared lasers and detecting short duty cycle pulses with our custom lock-in-

amplifier. While we did not succeed to demonstrating pulse detection with the QCD, we believe

we have laid the groundwork for making this possible in a future revision. Team ChemSense was

23

honored to contribute to a project with important applications in national security and

environmental protection.

24

REFERENCES

[1]    B. Mizaikoff. “Waveguide-enhanced mid-infrared chem/bio sensors”. Chem Soc Rev, pp. 8683-8699, RSC Publishing. May, 2013.

[2]    R. Soref. (2010, May 1). Mid-infrared photonics in silicon and germanium. [online]. Available: http://www.nature.com.ezproxy.lib.utexas.edu/nphoton/journal/ v4/n8/full/nphoton.2010.171.html

[3]    S. Chakravarty, “Proposal 7-4721”, unpublished.

[4] Occupational Safety & Health Administration, OSHA Technical Manual (OTM) Section III: Chapter 6 (1999) [Online]. Available: https://www.osha.gov/dts/osta/otm/otm_iii/otm_iii_6.html

[5] Department Of Defense Test Method Standard Environmental Engineering Considerations And Laboratory Tests, MIL-STD-810G (2008) [Online]. Available: http://www.atec.army.mil/publications/Mil-Std-810G/Mil-Std-810G.pdf

[6] Department Of Defense Test Method Standard Electronic And Electrical Component Parts, MIL-STD-202G (2002) [Online]. Available: http://snebulos.mit.edu/projects/reference/MIL-STD/MIL-STD-202G.pdf

[7] T. Koch, “Infrared Spectroscopy: Theory”. University of Colorado, Boulder, Chemistry and Biochemistry Department, 2003.

[8] Y. Zou, H. Subbaraman, S. Chakravarty, X. Xu, A. Hosseini, W. Lai, R. Chen, "Integrated strip and slot waveguides in silicon-on-sapphire for Mid-Infrared VOC detection in Water" Proc. SPIE 8990, Silicon Photonics IX, 89900X (March 8, 2014).

25

APPENDIX A – LAB ON CHIP SENSOR

A-1

APPENDIX A – LAB ON CHIP SENSOR

ChemSense is designed around the operational characteristics of Dr. Chen’s lab-on-chip sensor.

This means the initial framework for a successful design lies in understanding how the sensor

emits light using its QCL, how it recollects the light using its QCD, and how changes in the

channel in between correlate to a sensed chemical. For us, this translates to finding the correct

models for the QCL’s drive characteristics, the QCD’s output characteristics, and for the channel

in between. We find that the electrical properties of the sensor can be modeled as a resistive

network due to reciprocity from the QCL and QCD. The channel can be modeled as an

attenuator, based on the Beer-Lambert law. Finally, the operational characteristics of both the

QCL and QCD can be found experimentally through benchmarking.

The lab-on-chip sensor is composed of a QCL, germanium waveguide, and QCD monolithically

integrated onto one chip. The sensor works by generating a potential across the QCL. This

potential causes the QCL to emit light. The generated light then travels to the waveguide, which

confines the light and forces it to travel in a localized region. When the light reaches the other

side of the waveguide it is collected by the QCD, which generates a voltage based on the light it

collected. For visibility, this process is shown without the waveguide in Figure A-1. The light

confinement characteristics of the waveguide are shown in Figure A-2.

A-2

Figure A-1. Lab-on-chip sensor without waveguide

Figure A-2. Graph of light confinement in slot waveguide

When thinking about how to operate the lab-on-chip sensor we need to model it electrically. One

of the most important facts about this sensor is that the QCL and QCD are both made from the

same bi-functional quantum cascade active region. Meaning when the region is given an external

potential, it acts as a laser. When the region is zero biased, it acts as a detector. This bi-

functionality means there is reciprocity between electron generating light and light generating

electrons. Therefore, the QCL and QCD have no frequency component in their impedance

characteristics. Given this knowledge we are able to model the QCL and QCD as a network of

resistances, based on the model shown in Figure A-3, where both the QCL and QCD share a

ground at the bottom of the substrate. The second model we need to make is the channel model,

for the area between the QCL and QCD. Since the chip is designed to perform absorption

spectroscopy, we create our channel model based on the Beer-Lambert law for absorption

spectroscopy. In optics, the Beer-Lambert law relates light absorption to the chemicals in the

lights path. This relationship is shown in the equation below,

Transmission= II o

=10−αγ l → Absorbance=log10( 1T )=−aγ l

Where T is the transmission percentage, a is the absorption coefficients of chemicals in the lights

path, l is the length of their interaction, and γ is the medium-specific absorption factor. This tells

us how the light will be attenuated as it travels from the QCL to the QCD. Given this knowledge

we model the channel as a variable attenuator where a is the variable based on the chemical

present between the QCL and QCD. This model is shown in Figure A-4 below.

A-3

Figure A-3. Electrical model for lab-on-chip sensor

Figure A-4. Channel model for lab-on-chip sensor

Given these two models, we can look at the benchmark characteristics of the lab-on-chip sensor

to determine our full operating range. The benchmark characteristics are shown in Figure A-5

below and were provided to us by Dr. Chen’s silicon photonics graduate group. The peak

operating performance of the laser occurs at a frequency of 50 kilohertz, pulse width of 50

nanoseconds, with 15 volts and 3.7 amps across the laser. Given these conditions, the peak

output power is approximately 150 milliwatts. The electrical and channel model for the chip, in

conjunction with the chips benchmarks, provides the starting point for the laser controller and

signal processor modules of ChemSense. The channel model sets the fundamental range for

concentrations we are able to detect, for a given intensity from the QCL. The chips benchmarks

tells us the maximum intensity that can be produced from the QCL as well as the maximum

QCD output value. The electrical model shows us how to interface with the chip to produce light

from the QCL and gather data from the QCD. It also tells us that the chip has no reactive

component so maximum power transfer to the QCL can be achieved by resistance matching.

A-4

Figure A-5. Benchmark characteristics of the lab-on-chip sensor

APPENDIX B – ABSORPTION SPECTROSCOPY

B-1

APPENDIX B – ABSORPTION SPECTROSCOPY

Absorption spectroscopy measures the amount of radiation absorbed by a chemical as a function

of radiation wavelength. When an electromagnetic wave impinges on a sample, the sample will

absorb energy according to the following formula:

E=hv=hcλ

,h=6.6 x 10−34 joules−sec (1)

A material's absorption spectrum is defined as the fraction of radiation absorbed by a material,

normalized by the incident radiation, over a range of frequencies. Energy will be absorbed at a

particular wavelength if the energy difference (from equation 1) matches the energy difference of

the chemical’s two quantum mechanical states [7].

In optics, the Beer-Lambert law relates light absorption to the properties of the material the light

is traveling through. The law states there is a logarithmic dependence between the transmission,

T, of light through a substance and the product of the absorption coefficient of the substance, α,

and the distance the light travels through the material, ℓ [8]. The absorption coefficient can be

rewritten as a product of the species’ molar absorptivity ε and molar concentration c [8]. The

Beer-Lambert law is summarized as follows:

T= II o

=10−α l=10− lℇ c (2)

I 0 and I are the intensity (power per unit area) of the incident light and transmitted light, respectively. Absorbance is measured from tranmittance using the following formula:

A=log10( 1T ) , T=% transmission (3)

Absorbance can be plotted as a function of frequency, wavelength, or wavenumber.

B-2

APPENDIX C – LASER DRIVERS EXPLAINED

B-3

APPENDIX C – LASER DRIVERS EXPLAINED

The laser driver's components include a microcontroller (TM4C123) to provide a trigger pulse, a

gate driver to provide the voltage and current to turn on the MOSFET, a MOSFET with fast

switching capabilities, and dc power supplies to power the circuit. ChemSense's two laser

drivers use a microcontroller to generate a pulse using timer functions. The pulse is digital logic

(0 to 3.3 V) and can be configured for pulse widths as short as 20 ns. The pulse is fed to the input

of a gate driver integrated circuit (IC). A gate driver is necessary to drive a capacitive load such

as a MOSFET with enough current to quickly turn it on and off. The MOSFET is operated as a

switch and if the gate-to- source VGS voltage is greater than the threshold voltage (typically 2 to

4 V), the MOSFET is no longer in cut-off region. Once the drain-to-source voltage VDS is

greater than or equal to the saturation voltage VSAT = VGS-VTH, the MOSFET is in saturation.

Both pulse generators are configured as a high-side circuit in which the MOSFET is closer to the

power supply compared to the load (i.e. laser). The positive pulse generator uses a LTC4440-5

gate driver and an IRF510 MOSFET while the negative pulse generator uses the FOD3182 and

an IRF510 MOSFET. Diagrams of both the positive and negative pulse laser driver can be seen

in Figures C-1 and C-2.

Figure C-1. Positive pulse driver.

B-4

C-1

C-2

Figure C-2. Negative pulse driver.

B-5C-3

APPENDIX D – ANALOG FRONT END EXPLAINED

B-6

APPENDIX D – ANALOG FRONT END EXPLAINED

The analog-front end consists of a transimpedance amplifier (TIA) to convert the generated

photocurrent to a voltage, a first-order low-pass filter, and a gain stage. All stages are single

supply (3.3 V). Along with a feedback resistor, the TIA stage is calibrated to produce a

maximum voltage at the supply rail corresponding to the peak current. A pad for capacitors is

included to provide some stability. The TIA is followed by a RC filter with tunable with a

potentiometer. The cutoff frequency is variable and can be adjusted according to the required

pulse repetition frequency. After the pulse is broadened by the RC filter and gained by the

amplifier, the signal is sent to the ADC to be digitized. A circuit diagram can bee seen in Figure

D-1, below.

Figure D-1. Analog front-end circuit.

B-7

D-1

APPENDIX E – DIGITAL LOCK-IN AMPLIFIER EXPLAINED

B-8

APPENDIX E – DIGITAL LOCK-IN AMPLIFIER EXPLAINED

In order to accelerate the filter design and testing process, we added a number of debugging

switches. Figure E-1 below are our filter debugging switches. These define statements can be

changed from a 0 to a 1 to enable or disable parts of the system. Because this is embedded code,

it was important to maintain consistent timing characteristics across different versions. Instead

of removing the bandpass filter when BP_FILTER is set to 0, we instead made the bandpass

filter transparent. This means that it takes the same amount of time, but produces no change in

the data.

Figure E-1. Filter Debugging Switches.

Figure E-2 are our debugging switches. These define statements let us test different parts of the

system independently. For example, if we wanted to test the system on simulated data instead of

real data from the ADC, all we had to do was set REAL_SAMPLES to 0, then supply a test

pattern. When testing the connections between the lock-in and the rest of the

system, we often set REAL_TXMT to 0, so that we could send a test pattern

instead of real output.

B-9

E-1

Figure E-2. Debugging Switches.

Figure E-3 below are our Load Statements. These statements initialize the

registers for the Filter subroutine.

Figure E-3. Load Statements.

B-10

E-2

E-3

APPENDIX F – OVERALL SYSTEM I/O SPECIFICATIONS

B-11

APPENDIX F – OVERALL SYSTEM I/O SPECIFICATIONS

ChemSense will require three types of inputs (sensor, user, and power) and two types of outputs

(sensor and data). Sensor inputs come from the local environment, user inputs are

variables/settings that define how the user wants to operate ChemSense, and the power input will

energize the device. Sensor outputs are used to operate the lab-on-chip sensor based on user

specified inputs and data outputs inform the user about collected data on a detected chemical.

Tables F1 and F2, below, detail the I/O operational and precision ranges for our design.

Table F1. System Input Specifications.

Input Description Input Type Operational Range Precision

Lab-on-Chip Temperature

Temperature of the lab-on-chip chemical sensor Sensor 20° to 150° C +/- 0.1° C

QCDFrequency

Pulse frequency the QCD is detecting in the lab-on-chip chemical

sensor

Sensor 5 kHz to 100 kHz +/- 1 Hz

QCD Voltage

Emission wavelength of the QCL in the lab-on-chip chemical sensor

Sensor 0 uV to 10 uV +/- .1 uV

Set QCL Frequency

User desired pulse frequency for the QCL

in the lab-on-chip chemical sensor

Userλ: 9.5 um

f: 5 kHz to 100 kHzλ: +/- 1 umf: +/- 1 kHz

Set QCL Duty Width

User desired duty width for the QCL in the lab-

on-chip chemical sensorUser 12.5 ns – 200 ns 12.5 ns

Power Electrical power to run the system.

Power -10 V to + 10V 10 mV

B-12

F-1

F-2

Table F2. System Output Specifications.

Name Description Type Range Precision

QCL Voltage Output voltage from ChemSense used to drive the QCL

Sensor 0 to 10 V +/- 0.1 V

QCL Current Output current from ChemSense used to drive the QCL

Sensor 0 A to 4 A +/- 0.1 A

Pulse Generator Frequency

Output pulse rate used to drive the QCL at the user set frequency. Sensor

5 kHz to 100 kHz +/- 1 Hz

Screen Visual interface for the user. Data N/A N/A

Graph Plots QCD Power / QCL Power versus time

Data N/A N/A

Concentration Quantifies concentration of a chemical based on user specified “set chemical” and sensor data

Data 0% to 100% concentration

5%

Threshold Alarm

Signal unsafe conditions. (Only for pre-defined chemicals.)

Data Yes / No N/A

B-13F-3

APPENDIX G – BILL OF MATERIALS

B-14

APPENDIX G – BILL OF MATERIALS

B-15

G-1

G-2

B-16G-3

APPENDIX H – APPLICABLE STANDARDS

B-17

APPENDIX H – APPLICABLE STANDARDS

Because ChemSense’s purpose is to detect hazardous chemicals using lasers, potentially in

military applications, we believe several standards will affect our project’s specifications. The

Occupational Safety and Health Administration (OSHA) and the American National Standards

Institute (ANSI) provide detailed standards for operation, classification, and labeling of laser

devices, as well as the handling of hazardous chemicals, while the Department of Defense

(DOD) provides MIL-SPEC, MIL-PRF, and MIL-STD equipment design standards.

It should be noted that, “An American National [ANSI] Standard implies a consensus of those

substantially concerned with its scope and provisions. These standards are intended as a guide to

aid the manufacturer, the consumer and the general public. There is, however, no inherent

requirement for anyone or any company to adhere to an ANSI standard. Compliance is voluntary

unless specifically required by some alliance” [6, Appendix III: 6-3]. Although ANSI standards

may not be compulsory, DOD (MIL-SPEC, MIL-PRF, and MIL-STD) standards will apply to

our device if our end user is a military entity.

The ANSI Z 136 Laser Safety Standards series is a DOD accepted standard, which covers the

safe use of lasers in a variety of settings, as well as the testing and labeling of laser devices. As

the ANSI standards are not freely available, we referred to the OSHA

standards covering categories relevant to ChemSense. Regarding classification of

laser devices, Section IV.B.1 of the OSHA Online Technical Manual (OTM)

states “Virtually all of the U.S. domestic as well as all international standards divide lasers into

four major hazard categories called the laser hazard classifications. The classes are based upon a

scheme of graded risk. They are based upon the ability of a beam to cause biological damage to

the eye or skin” [4]. Even though the laser we will use in our project will not be powerful enough

to pose any serious risk to users, appropriately classifying and labeling that risk will be part of

our responsibility. From a design standpoint, OSHA OTM Section VI.J.1 specifies: “A laser

shall have an enclosure around it that limits access to the laser beam or radiation at or below the

applicable MPE level. A protective housing is required for all classes of lasers except, of course,

at the beam aperture” [4]. As such, a protective enclosure should be a requirement for our device.

B-18

H-1

H-2

Lastly, OSHA provides guidelines on protective eyewear and clothing we will wear while

operating lasers.

DOD standards that may apply to our project cover the testing and operational durability of our

device. One DOD standard will need to consider is MIL-STD-810 [5], which specifies how to

design and test equipment for the conditions it may experience during use. Such considerations

include exposure to extreme temperatures, shock, vibration, sand and dust, rain, and a variety of

other environmental concerns. Another DOD standard we may need to consider is MIL-STD-

202, which “establishes uniform methods for testing electronic and electrical component parts,

including basic environmental tests to determine resistance to deleterious effects of natural

elements and conditions surrounding military operations, and physical and electrical tests” [6].

Our project will need to evaluate the applicability of the requirements in these standards.

B-19

H-3