Design of a Sensor Board for an Acoustic

Design of a Sensor Board for an Acoustic Traffic Monitoring System Gina Colangelo Tufts University Masters’ Project 12/02/2008

Project Goal To enable the study of a new technique for traffic monitoring using an acoustic array sensor network by: defining the system architecture of the sensor array board deriving the system specifications of the sensor array board designing a PCB for the sensor array board and the interface to the wireless link

Presentation Outline Introduction to Traffic Surveillance Networks Background on Passive Acoustic Sensors System Architecture System Specifications Part Selection Schematic Capture PCB Layout Future Goals

Technologies used for Rural Traffic Surveillance National Survey conducted in 2004 by ITS 1 found: – – 26 states use Inductive Loop Detectors (ILD) 21 states use Radar Detectors 11 states use Video Image Detectors (VIDS) 5 states use Acoustic Detectors Vehicle info collected by the sensor networks include: – – – Traffic Volume Vehicle Speed Vehicle Classification Travel Time Incident Occurrence ITS or Intelligent Transportation System is a division of RITA which is part of the U. S. Department of Transportation (DOT). 1

Inductive Loop Detectors (ILD) How it works: – 1 or more loops of wire are embedded under the road & connected to a control box. – When a vehicle passes over or rests on the loop, inductance is reduced showing a vehicle is present. Benefits – – Established Technology Not impacted by environmental conditions Accurate in detecting vehicle presence Performs well in both high and low volume traffic Disadvantages – – – High Cost (up to $10 k for initial costs) Invasive Installation Potential poor reliability due to improper installation – Not viable for certain locations and road conditions – Unable to directly measure speed or direction

Video Image Detection (VIDS) Employs machine vision technology to automatically analyze traffic data collected w/ Closed Circuit Television Systems (CCTV) Benefits: – Rapid Incident detection – Wide area detection capabilities (multilane, multi-direction) – Vehicle classification – Estimates traffic queues and speed – Installation does not require lane closures – Can be integrated with other sensor networks Disadvantages: – High Cost (initial cost > $50 k) – Higher Power than other sensor networks – Light conditions can affect surveillance

Radar-based Roadside Sensor How it works: – Transmits radar pulses – A portion of the energy is reflected or scattered from the vehicle and roadway back toward the sensor – This energy is received and interpreted. Benefits – Low Power – Most accurate technology for detecting speed – Traffic Count accuracy – Easy installation – Low cost Disadvantages – Accuracy can be affected by weather conditions (hail, snow, rain) – Directional Detection is poor – Interference could occur with other RF devices

Passive Acoustic Sensors Passive Road-side Sensor that receives sound waves from passing vehicles. Benefits: – – – Low Power Low Cost Easy to Install Directional and Multi-lane Detection Accurately measures traffic count Disadvantages: – Accuracy affected by environment factors – Speed measurements are not as accurate as other methods

Summary of Traffic Surveillance Technologies Employed Today: Technology Traffic Volume (Moving) Traffic Volume (Stopped) Vehicle Speed Vehicle Classification Incident Detection Power Cost ILDs A A C D C B C Radar A A A C C B A VIDs B B B A A C C Passive Acoustic B C B B C A A Code: A – Excellent; B – Fair; C- Poor; D – Nonexistent; U – Unknown Radar and Acoustic sensors are the least expensive to deploy and are the lowest power. VIDS collect the most information and the data processing possibilties are endless. Acoustic sensors are the least accurate, but the technology is relatively new. Acoustic sensors for traffic monitoring has room for improvement: – – – Improve accuracy Ability to detect idle traffic Intelligently process data (Vehicle Classification, Incident Detection)

Acoustic Array Sensor Research has shown that using arrays of acoustic sensors narrows the detection zone for improved SNR & better accuracy For this project the following configuration will be evaluated: – 4 microphone arrays per sensor: 2 arrays form a pair parallel to the road 2 arrays for a pair orthogonal to the road Each array contains 12 microphones – Sensors to be deployed road-side about 10 m above the road.

Sensor Network System Details Sound detected by each element in a single array will be summed together and amplified. 4 analog outputs will be digitized separately and processed. Processed signals will be transmitted to a gateway node within the wireless sensor network.

Signal Processing: Parallel Pair A vehicle moves across the detection zone (D 2 to D 1): – As a vehicle approaches, sound reaches D 2 earlier than D 1. The time delta will be negative (τ<0). – When a vehicle is at the center of the detection zone, τ=0. – As a vehicle exits the detection zone, τ will be positive. The rate of change or slope across the detection zone corresponds to the vehicle speed. To extract the time delay from the actual signals, the cross correlation of D 1 and D 2 is calculated:

Signal Processing: Orthogonal Pair A vehicle moves across the detection zone: – As the vehicle approaches, sound will reach D 1 earlier than D 2. The time difference will be positive and increasing (τ<0). – When the vehicle is at the center of the detection zone, τ will peak. – As the vehicle exits the detection zone, τ will be positive and decreasing. The sound map for vehicles in lanes closer to the sensor (smaller y) will have smaller peaks than those in lanes further from sensor.

System Architecture 2 Boards to form complete system: – Sensor Board Microphone Arrays Summing Stage Amplification Stage – System Board Analog-to-Digital Converter Processor RF Transceiver (ISM band of 2. 4 GHz) e. Z 430 -RF 2480 Demo Kit from TI chosen for System Board – USB-based wireless demo tool – MSP 430 F 2274 Mixed-Signal Microcontroller – CC 2480 2. 4 GHz Zig. Bee network processor – 2. 4 GHz Antenna

Sensor Board Architecture For initial prototype, 1 sensor board per array to allow for array spacing experimentation Majority of System Gain implemented on Sensor board to maximize SNR. Clipping Circuit and Anti-Aliasing Filter may be needed to condition signal for ADC.

System Board Architecture MSP 430 Mixed-Signal Microcontroller – AUX Op-Amps can provide extra gain if needed – On-chip 10 -bit ADC can multiplex in 4 analog channels for digitizing – CPU can be programmed to compute cross-correlation functions CC 2480 – Zig. Bee Processor to transmit data to gateway node to main control center over 2. 4 GHz ISM band

Determining Sensor Board Specs Power Supply Requirements Maximum Voltage Output Swing Element Spacing Summing Stage Configuration System Gain Anti-Aliasing Filter

Power Supply Requirements 2 Specifications need to be determined: – Supply Voltages – Maximum Current Sensor Board needs to be portable solution – Low power – Battery operated Reuse Battery Board included with e. Z 430 RF 2480 Demo Kit: – 2 AAA Batteries in series: 3 V supply – Capacity of AAAs: 900 -1155 m. A/h

Maximum Voltage Output Swing Needs to be limited to Input Range of ADC. – Depending of ADC front-end architecture, over voltage on inputs can cause conversion errors and in some cases damage the ADC. 10 -bit ADC integrated on the MSP 430 will be used. – ADC input range is programmable from Vcc to Vss. – For this project, ADC will be programmed to accept an input range of 3 V to 0 V. No Clipping Circuit needed on Sensor Board – Amplifiers on Sensor board will not produce a voltage higher than its supply voltage – ADC on the MSP 430 clips any signals greater than the programmable upper input range limit. – Benefits from using the same power source solution!

Element Spacing b/t array elements will be chosen to achieve the desired detection angle. Desired Detection Angle: – Sensor board mounted 10 m above road (z=10 m) – Desired detection zone is 2. 5 m in any direction at road level. – Detection Angle can be calculated:

Experiment to Determine Element Spacing Several 3 x 4 element arrays bread-boarded w/ summing stage: – Array 1: 1. 75 cm x 1. 75 cm – Array 2: 2 cm x 2 cm – Array 3: 2. 5 cm x 2. 5 cm Detection Angle measured by moving a 4 k. Hz sound source across the array from a fixed distance. Vpk-pk measured at summer output with Oscilloscope

Results of Element Spacing Experiments

Summing Stage Configuration Original Design used 1 op-amp to sum all 12 elements New Design sums the elements in 2 stages: – First stage sums the 3 rows of 4 elements separately – Second stage sums the 3 outputs of the 1 st stage New Design allows for more gain without large R – Improved Rise/Fall Times (τ = RC) – Increased Bandwidth – less Gain per stage (f 3 d. B = GBW/Gain)

Determining System Gain Experiments conducted to estimate the amount system gain required in the real traffic monitoring environment. Microphone array prototype board is not portable, and output cannot be stored. A portable digital sound recorder was used to collect field samples: – 2 Omni-Directional Electret Condenser Microphones – Mic positions can be configured from 45º to 135º (90º was chosen) – 7 Pre-Amp Levels – Recorded sound clips saved as *. WAV

System Gain Experiment Step 1 – Calibrate the digital recorder to the microphone array in a controlled environment with same sound source for each Pre-Amp Level. – Measured in Lab w/ 4 k. Hz sound source – Array and Recorder in same location for the measurement Step 2 – Collect Field Data for each applicable Pre-Amp level – 5 -10 s traffic samples were recorded from a height of 25 -30 ft from the road – Average of 5 vehicles passed the recorder during each sound clip including motorcycles, small cars, and SUVs. Step 3 – Download and process *. WAV files to determine peak sound levels – – MATLAB code was written to analyze these files The start and end of each file was removed to eliminate interference The 2 microphone outputs were summed The max peak and rms levels were calculated. Step 4 – Relate the peak sound levels from digital recorder back to the microphone array and determine the required system gain.

System Gain Results Array Output = 100 m. Vpk-pk Pre. Amp Level 3 optimal in field Digital Recorder Peak and RMS Levels from 4 k. Hz Signal in Lab vs. Pre-Amp Level To calculate Vpk-pk at the array output the following formulas were used: 1. Digital Recorder Peak and RMS Levels from Traffic Sound Clips 2. Calculating the Sound Source Ratio 3. Total System Gain Needed = 4 x 3. 5 = 14 Estimated Array Output Vpk-pk and Recommended Gain Settings

Anti-Aliasing Filter Anti-aliasing filter should be placed before ADC – Prevents harmonics, spurs and broadband noise outside of Nyquist from aliasing back inband – Improper filtering leads to a decrease in SNR, a reduced dynamic range and an increase in unwanted spurs. ADC input bandwidth/channel: – clock range: 450 k. Hz - 1. 5 MHz – Nyquist BW: 225 – 750 k. Hz – 4 chs muxed, BW per ch = 56. 25 – 187. 5 k. Hz Frequencies of Interest < 10 k. Hz Filter Design – – – Low-pass Butterworth – flat pass- & stop-band F 3 d. B = 20 k. Hz for flatter phase in pass-band 1 st order provides ~20 d. B attenuation at FADC/2. Since microphones have a frequency roll-off response, 1 st order should be adequate.

Summary of Board Specifications Parameter Specification Power Supply 3 V only (2 AAA batteries) Current Dissipation As low as possible Microphone Spacing 2. 25 cm x 2. 25 cm (detection angle ~ 16º) System Gain 14 or 11 d. B (10*log(14)) Maximum Output Swing 0 -3 V Low Pass Filter f 3 d. B = 20 k. Hz, Order = 1

Schematic - Circuit Design Architecture is finalized Active Components have been selected: – Analog Devices AD 8544 Quad Rail-to-Rail Op. Amp Single Supply Operation: 2. 7 V to 5. 5 V GBWP ~ 1 MHz Low supply current 45 u. A/amplifier – Emkay MD 9745 APZ-F Omni-Directional Microphone Operating Voltage: 2. 0 V to 10. 0 V Frequency Range: 100 Hz to 10 k. Hz S/N Ratio: > 55 d. B Altium Designer 6 software chosen for PCB Schematic and Layout

Schematic Page 1

Schematic Page 2

PCB Layout 4 Layer Board – 2 Signal Layers – GND and Power layer Board Size is 4’’ by 2. 5’’ All symbol footprints are either finalized or common footprints

PCB Layout – Top Layer

PCB Layout – Bottom Layer

PCB Layout – Layer 2 (VCC)

PCB Layout – Layer 3 (VSS)

Prototype Results (Breadboard) Current Dissipation – 2. 6 m. A from single 3 V supply (original design 15 m. A from 5 V, +9 V, and -9 V supplies) – Sensor board can operate for approx 400 hrs before recharging Frequency Response: – Op-Amp 1: f 3 d. B= GBWP/Gain = 1 MHz/3 = 333 k. Hz – Op-Amp 2: f 3 d. B = 1 MHz/5 = 200 k. Hz – Microphone: f 3 d. B = 10 k. Hz – LPF: f 3 d. B = 20 k. Hz Note: Using 1 stage summer, f 3 d. B=1 MHz/15 = 66 k. Hz

Future Work Fabricate, Assemble, and Test Sensor Board PCB Interface Sensor board with Main system board – Configure ADC on MSP 430 – Write code to perform cross-correlation and store results Perform experiments on Array Spacing Collect Field Data with this Portable prototype

References Chen, Shiping, Ziping Sun, and Bryan Bridge. “Traffic Monitoring Using Digital Sound Field Mapping. ” IEEE Transactions on Vehicular Technology 50. 6 (2001): 1582 -1589. Intelligent Transportation Systems. 2008. http: //www. itsdeployment. its. dot. gov/ ITS Decision, Project of the California Center for Innovative Transportation at UCal Berkeley. 2007. http: //www. calccit. org/itsdecision/

Acknowledgements Yuping Dong – Ph. D Student conducting research in this area Professor Chang – Project Advisor Tufts Wireless Lab