High Resolution AMR Compass

1
High Resolution AMR Compass
Dr. Andy Peczalski Professor Beth Stadler Pat
Albersman Jeff Aymond Dan Beckvall Marcus
Ellson Patrick Hermans
Honeywell
2
Agenda

• Introduction/Abstract Marcus E
• MATLAB Simulations Marcus E
• Software Pat H
• Hardware Jeff A
• Testing Pat A
• Results Dan B

Honeywell
3
Abstract
This projects purpose is to improve the accuracy
of a digital compass by using multiple compass
ICs. These will work together to collectively
improve the accuracy of the overall system.
Honeywell
4
Abstract
One benchmark is to try to increase the accuracy
of the system by the number of sensors
used. Increased precision and repeatability is
also desired.
Honeywell
5
Abstract
Customized hardware is necessary to implement the
multiple sensor system. Customized software to
manage the implementation is also necessary.
Honeywell
6
MATLAB

• Used to simulate single and multiple sensors
  before our hardware was complete
• Provided a vehicle to test the performance of our
  heading calculation algorithms
• 1702 lines of MATLAB simulations

Honeywell
7
Sensor Placement

• The placement of the sensors must create a system
  accurate across 360 degrees
• Each individual bridge of each sensor can be
  simulated independently in MATLAB
• Multiple arrangements can be simulated to
  determine the best implementation

Honeywell
8
Orientation Simulations

• Single IC Senor Output Wave Form

• Data Appears Evenly Spaced
• ICs at 0, 36, 72, 108, 144, 180, 216, 252, 288,
  324 Degrees

Honeywell
9
Orientation Simulations

• Single IC Senor Output Wave Form

• Data Evenly Spaced
• ICs at 0, 9, 18, 27, 36, 45, 54, 63, 72, 81
  Degrees

Honeywell
10
Software

• Three software realms involved with this project
• MATLAB
• C
• VB

Honeywell
11
C

• Written in MPLab
• Version 8.0
• CCS complier
• Version 4
• Run on PIC 18f4550
• 1326 Lines of C
• 2532 Lines of Assembly

Honeywell
12
Sensor Communication

• Sensor Commands
• Heading
• Adjusted voltages
• Raw voltages
• Calibrate
• Re-address
• Number of Summed measurements

Honeywell
13
Serial Communication

• Allows Compass to display results
• Very helpful in debugging
• Allows for VB to control sensor
• Easy to implement in CCS
• 115200 Baud allowable from the 20Mhz crystal

Honeywell
14
Honeywell
15
Honeywell
16
Weighted Averaging
Honeywell
17
Honeywell
18
VB

• Provides an end-user interface
• Synchronizes the compass and the rotation table
• Allows for automated data acquisition
• Provides a repeatable test benching system
• Requires a third board to handle adjusted ground
  on PMC
• 4733 Lines

Honeywell
19
Honeywell
20
Honeywell
21
Serial
Serial
Personal Computer (VB)
PMC Controller
PIC18F4520 (C)
Rot. Table
Parallel
Sensors
Honeywell
I2C
22
Final Hardware

• Abstract
• Initial Design
• Problems with Initial Design
• Changes Made
• Proposed Final Design

Honeywell
23
Abstract

• One compass, two boards
• Main Board
• Microcontroller
• Daughter Board
• Sensors

Honeywell
24
Initial Design
Main Board
Honeywell
25
Main Board

• Essentially a controller board
• Microcontroller
• RS-232 Communication
• I2C Communication
• Interfacing
• Daughter Board
• Front Panel

Honeywell
26
Initial Design
Daughter Board
Honeywell
27
Daughter Board

• Three functional systems
• Sensor array
• Power MUX
• Laser

Honeywell
28
Daughter Board
Dimensions

• Constraint One of the dimensions must be less
  than 3.5
• Opening of zero-gauss chamber is 3.5 in diameter

3.132
Honeywell
3.492
29
Daughter Board
Dimensions

• Constraint One of the dimensions must be less
  than 3.5
• Opening of gauss-free chamber is 3.5 in diameter

0.73
3.132
The Daughter Board meets size requirements
Honeywell
30
Daughter Board
HMC6352
Feedback Networks
Power
LED
Clock
Ground
Data
Decoupling Capacitor
Honeywell
31
Daughter Board
I2C Bus
Clock
Data
Honeywell
32
Daughter Board
Power MUX

• Design challenge
• Need to assign unique address to each sensor
• Each sensor is factory installed with address
  0x42
• In order to change addresses, a command must be
  sent to a sensor on the bus
• This command message contains
• How to change address of individual sensor if
  every sensor is receiving the command?

Start Address Ack Command Ack Stop
Honeywell
33
Daughter Board
Power MUX

• Solution Need to isolate communication to
  individual sensor
• How?
• Burn-in Socket
• Use a network of jumpers
• Multiplex I2C to each sensor
• Multiplex power to each sensor

Honeywell
Photo taken from http//www.locknest.com/newsite/p
roducts/qfn/index.htm
34
Daughter Board
Power MUX

• We chose to multiplex power
• Advantages
• Saves power
• Simplifies troubleshooting
• Disadvantages
• Signal loss through MUX
• Other unknowns

Honeywell
35
Problems with Initial Design

• Problems
• Main Board
• None
• Daughter Board
• I2C bus
• When powered off, the sensors interfere with I2C
  bus
• 5V data signal is pulled down to 2.5V
• Therefore communication will not work
• Problems not related to design
• Sensor 3 will not communicate
• Will not hinder project algorithm will still
  work
• Slight loss of sensitivity at sensor 3s axes of
  sensitivity (27 and 117 )

Honeywell
36
Changes to Initial Design

• I2C bus fix
• Remove MUX and feed power to all sensors
• Cut I2C traces
• Add jumpers to I2C vias and address them one by
  one
• Connect all jumpers to I2C bus

Honeywell
37
Changes to Initial Design

• Other changes
• No laser mount
• Laser mounted directly to plexi-glass case
• Saves cost (25)

Honeywell
38
Changes to Initial Design

• Other changes
• Main Board Layout

Before
After
Honeywell
39
Proposed Final Design

• Due to I2C bus issues, our current design does
  not work
• Two options
• Power all sensors and use burn-in or jumpers
  socket to isolate sensors
• Multiplex I2C bus

Honeywell
40
Proposed Final Design

• Option 1 Power all sensors and use
  socket/jumpers
• Advantages
• No MUX needed
• Reduces surface area of board
• Reduces signal loss of MUX
• Sleep mode on sensors
• Save power
• I2C bus has not been tested in this mode

Honeywell
41
Proposed Final Design

• Option 1 Power all sensors and use
  socket/jumpers
• Disadvantages
• Sockets can be expensive
• Footprint of HMC6352 is not common
• Hard to find socket
• No disadvantages if we add jumpers

Honeywell
42
Proposed Final Design

• Option 2 Multiplex I2C bus
• Advantages
• No need for a socket
• Sleep mode to save power (not tested)
• Disadvantages
• Side effects of multiplexing I2C unknown

Honeywell
43
Testing

• Prototype Final

Honeywell
44
Test Setup
Honeywell
45
Accuracy
Precision
Repeatability
Compare
Compare
ß field
Compare
Honeywell
46
Prototype Testing

• Given one sensor

• CCS compiler

Honeywell
47
Final Testing

• Elements of Final testing
• Pretesting (zero gauss values)
• Pretesting (offsets)
• Testing (accuracy, precision, repeatability)

Honeywell
48
Pre-testing (zero gauss)

• Place sensors in the zero gauss chamber
• Rotate 360 deg. while taking readings
• Analyze data and get zero gauss values

Honeywell
49
Pre-testing (offsets)

• Place sensors in artificial magnetic field
• Run VB script that finds sensor locations
• Finds zero gauss value of each chip
• Works using relativity
• Bang bang control
• Analyze data and find chip placements
• Hardcode this to software

Honeywell
50

Raw voltage readings with offsets
Honeywell
51

Raw voltage readings with offsets
Honeywell
52
Accuracy

• Test Procedure
• Determine the B field
• Find the zero crossing on each axis
• B field should be 90 degrees from zero crossing
• Average the 20 axes results
• Take measurement
• Compare result to actual
• Rotate to different position
• Repeat steps 2-5

113 deg
23 deg
Honeywell
53
Results

• Results Comprise of
• Determining Specs
• Comparison of Specs to Controls
• Ways to improve
• Future for Nanowires?

Honeywell
54
Results Specs - Repeatability

• Comprised of 5 readings taken at 0, 90, 180,270
• Our Product Min - 0.015 Max -0.089
• Control Min - 0.033 Max -0.051
• Honeywell - 0.030 Max - 0.120

Honeywell
55
Results Specs - Precision
Honeywell
56
Results Specs - Accuracy
Honeywell
57
How Can We Improve

• Currently using arcTan(x/y) to compute heading
• This assumes we have X and Y which need to be 90
  degrees apart
• In practice this is not true, we found this is
  actually only within -8 degrees
• Use different algorithms, better weighting
• More Sensors

Honeywell
58
Future For Nanowires?

• Nanowires are inherently less accurate
• Means greater room for improvement
• Small enough to use more than 10 bridges
• Weighting should have more of an effect
• Will have completely different obstacles
• All in all, from the results of this feasibility
  test they look very promising

Honeywell
59
Conclusion

• Questions/ Comments?
• Demo Upstairs?

Honeywell