Football Wristband for Measuring Throwing Speed Group 18 Kevin He & Darryl Ma ECE Senior Design November 30, 2006
Mar 30, 2015
Football Wristband for Measuring Throwing
Speed
Group 18Kevin He & Darryl Ma
ECE Senior DesignNovember 30, 2006
Motivation
Provides a cost effective solution for measuring speed
Does not require a dedicated operator
Can be applied to other sports: baseball, boxing, cricket, etc.
Objectives
Accuracy within 10% of measurements obtained from radar gun
Weighs less than 500 grams Lasts up to 5 hours per battery/charge Temperature is within 3C of ambient Able to measure a velocity range of 0mph to
70mph in increments of 1mph
Original Design Overview
Hardware Power Supply Module, Pressure
Sensor, Accelerometer, LCD, PIC Microcontroller
Software PIC programming for user interface
and speed calculation algorithm
Block Diagram
Block Diagram of the Device
System Schematic
Original Hardware Design
Power Supply Module Converts 3V from coin cell battery to stable
5VDC Pressure Sensor
Informs microcontroller when the ball is in the user’s hand and when the ball has been released
Accelerometer Measures acceleration of the wrist
LCD Displays velocity and user prompts
PIC Microcontroller Performs integration function to calculate speed
Power Supply Module
Battery Rating: 160mAh Converts 3VDC to a stable 5VDC Total Current Drain: 20-22mA
Schematic of Power Supply Module
Power Supply Output
Vmax = 5.445V Vaverage = 5.4305V Vmin = 5.416V Vripple = 14.69mV
Voltage output waveform from power supply module with 120kΩ load resistance
Pressure Sensor
Polyester variable resistor (32kΩ to 420kΩ)
Used voltage divider circuit to convert varying resistances into varying voltages
5VDC
Schematic of Pressure Sensor Circuit
Pressure Sensor
Pressure Sensor Resistance Vs Voltage
Vhigh => hand unoccupied Vlow => ball in hand
Output Voltage Vs Pressure Sensor Resistance
Resistance of Pressure Sensor (Ω)
Accelerometer
±5g ADXL320 Dual Axis Accelerometer
5V supply with 2mA current drain Capacitors attach across output
terminals for signal conditioning
Accelerometer
Capacitors for signal conditioning
Accelerometer Sensitivity
Verify datasheet sensitivity value of 312mV/g
Voltage output from accelerometer when dropped
Accelerometer Orientation
Voltage outputs change depending on orientation
LCD
Displays user prompts and final ball velocity
16X2 RT1602C character LCD 5V supply with 1mA current drain
PIC Microcontroller
PIC16F877A Performs integration of acceleration
input to calculate ball velocity 5V supply with 8mA current drain Operating frequency: 8.00MHz
Original Software Design
Written in PIC-C Purpose
Obtain Inputs from Pressure Sensor and Accelerometer
Calculate Velocity Output results to LCD
Original Flow Chart
Riemann Sum
Acceleration to speed conversion
a(t) = acceleration function v(ti) = velocity at time, ti
Positive and negative accelerations cancel to zero at apex of wind-up
Baseline Detection System
When the PIC senses a relatively stable signal, it resets the net velocity to zero
Integration sum is reset to zero here.
Voltage output from accelerometer when simulating a throw
Debugging
Major Problems we encountered:1) Poor battery life2) PCB/Protoboard Issues3) Inconsistent Results
Poor Battery Life Battery rating of
160mAh and current drain of 21mA suggests battery life of ~8 hours.
Actual battery life only about ~4 hours, with frequent fadeouts
Traced problem to battery brand
Energizer is the way to go!
PCB/Protoboard Issues
Main issues were related to durability Protoboard wires and components
often came loose Solution: Make PCB
PCB wire pads were often stripped of copper lining
Solution: Use 30 gauge wires to reinforce connections
Inconsistent Results
Causes Explored1) Accelerometer output noise2) Insufficient sample size3) Accelerometer Orientation4) Pressure Sensor Sensitivity
Accelerometer Output Noise
Measurement of the accelerometer’s axis show a large uncertainty in the output voltage. The uncertainty is on the order of ~1V, as seen on the graph.
Output voltage waveform from accelerometer when simulating a throw
Accelerometer Noise Solution
Stability Testing: Output from PIC stable within 9.8mV
Testing Accuracy of PIC Reading
from Accelerometer
Our solution was to add capacitors across the power terminals of the PIC, oscillator, and accelerometer. This further stabilized the input voltage waveform, which helped the accelerometer output consistent results.
Output voltage waveform after adding capacitors to PIC, oscillator, and accelerometer
Insufficient Sample Size
What is the processing time per sample?
Are there enough samples to perform an Riemann Sum integration?
Accelerometer Orientation
Changing the accelerometer orientation changes plane of acceleration
Pressure Sensor Sensitivity
Is the pressure sensor sensitive enough to detect the football? Lab tests show that even a gentle grip
causes a sizeable voltage drop However, while actually throwing a
ball, the user may loosen his grip even though the ball is still in his/her hand
Removal of Pressure Sensor
Radar Gun Speed (mph)
Device Speed (mph)
Trial #1 26 31.8Trial #2 28 26.2Trial #3 25 14
1) The trials below represent three out of ten trials that returned results. The other seven trials produced no measurement due to insufficient number of samples. The number of samples is directly controlled by the pressure sensor.
2) Inhibits throw
3) Reduces production cost significantly
Threshold Algorithm
Threshold
Starts integrating as soon as the 1.4g threshold is reached and continues to integrate until acceleration falls below this threshold
Eliminates need for pressure sensor and baseline detection
Will give relative velocity rather than exact velocity, so offset factor needed
Velocity Correction Factor
Why is it needed? Threshold algorithm only measures
acceleration beyond a certain limit, so not all acceleration is captured
The acceleration per sample is sqrt(x2 + y2), but taking the square root every sample reduces the number of total samples we can take, so we only used x2 + y2
Velocity Correction Factor
We know there is a correlation between the device speed and radar gun speed, so we need to apply an offset to make them equal
The velocity is:
25 1 9.8 / 1( ) 1
1024 0.312 1 0.44704 /
V g m s mphVelocity AccelerationVoltage ms
V g m s
The offset factor was determined solely based on experimental data. The speeds that we validated were the ones we could obtain with the radar gun (25mph – 40mph).
1( 0.0005 5) [27 ( 0.0005 5)]
3ActualVelocity Velocity Velocity
Our final equation including offset is:
Velocity Correction Factor
Measurements from Radar Gun (mph)
Measurements from our device (mph) – without
correction factor(Velocity * 0.0005)
Measurements from our device
(mph) – with correction factor
Trial #1
27 31.5 26.7
Trial #2
25 26 23
Trial #3
28 35.2 29.1
Trial #4
32 41.4 33.3
Trial #5
27 32 27
Trial #6
35 46 36.3
Trial #7
38 49.5 38.7
Trial #8
28 34.8 28.9
Average percentage difference
Without Correction Factor: 17.9%
With Correction Factor: 6.68%
Std. Dev. of percentage difference
Without Correction Factor: 3.25%
With Correction Factor: 2.61%
New Flowchart
Summary of Final Design
Original Power Supply Circuit Original LCD Added capacitors to clean up
accelerometer output Removed Pressure Sensor Modified Velocity Algorithm
Verification
Radar Gun Vs Wristband Correlation Check Tolerance Analysis Temperature Measurements
Radar Gun Vs Wristband
Radar Gun Measurement
Measurement from Device
Percentage Difference
27 30 10%
29 26.6 9.02%
34 33.9 0.29%
38 36.5 4.11%
36 35.8 0.56%
25 24 4.17%
27 27.5 1.82%
28 26 7.69%
33 34.2 3.51%
32 33.3 3.90%
Percentage Difference
Average: 4.507%
Std Dev.: 3.381%
Correlation Check
Since the radar gun only measured speeds greater than 25mph and we were not able to throw the football faster than 38mph, we performed a correlation check to make sure there was some correlation between the relative speed of the arm and the measured throw speed.
Correlation Check Results
Relative Speed
Trial #1
Trial #2
Trial #3
Trial #4 Trial #5
Slow 6mph 3mph 4.6mph 4.2mph 7.6mph
Medium 11.1mph
13.1mph
10.3mph
12.5mph 14.4mph
Fast 30.2mph
28mph 23.7mph
30.6mph 27.1mphResults show a definite correlation between the relative speed of the
arm and the measured throw speeds (throws were performed empty-handed)
Tolerance Analysis Concern: Accelerating over 5g could
damage the accelerometer or other components
∆Voltage = 1.891VSensitivity = 0.312V/gg = ∆Voltage/Sensitivityg = 5.66g
Tolerance Analysis on Y-Axis
Tolerance Analysis
Waveform shows no saturation at accelerations greater than 5g and when integrated back into device, there were no adverse effects.
∆Voltage = 2.031VSensitivity = 0.312V/gg = ∆Voltage/Sensitivityg = 6.5096g
Tolerance Analysis on X-Axis
Temperature Measurements
Room Temperature: 24.8°C Device-Wristband Surface: 26.8°C Wristband-Skin Surface: 26.4°C Normal Skin Temperature: 32.9°C Fulfills our performance requirement of
±3°C of ambient temperature
The Wristband
SWOT Analysis
Strengths: Easily removable and comfortable Powered by one 3V coin cell
battery Cost effective Large range of measurement
Weaknesses: Inconsistent results due to
human variation Slightly inhibits throw Measures acceleration in one
plane
Opportunities: Useful for other sports
applications
Threats: Low consumer demand
Comparison with Radar Gun
Cost Accuracy
Precision Ease of Use
Measurement Range
Battery Life
RadarGun
$85 - $400
Within +/-0.5mph
+/- 1mph Requires dedicated operator
Relative angle to ball should be less than 15 degrees for best accuracy
~20 hours using specialized batteries, or 6 AA batteries
Our Device
$80 for prototype, $45 if mass- produced
Within +/- 10% of Radar Gun
+/- 0.2mph
Does not require dedicated operator
Accuracy independent of ball path
~4 hours using a single 3V coin cell battery
Ethical Considerations
Safety Temperature Electrostatic Discharge (ESD)
Being honest/realistic about what our device is capable of
Future Steps
More accelerometers to achieve absolute acceleration, and enable more accurate measurements
Convert all parts to surface mount components to reduce device size
Improve durability Improve battery life Apply for a Patent
Credits
Ms. Hye Sun Park Mr. Mark Smart Professor Jonathan Makela Coach Dan Hartleb & Coach Eric
Snider
Questions?