greg davis chris johnson scott hambleton jon holton chris monfredo

Greg DavisChris JohnsonScott HambletonJon HoltonChris Monfredo

10/29/13 Rochester Institute of Technology 1

P14251Underwater Acoustic Communication

Underwater Acoustic Communication


AgendaBrief Review of ProjectSubsystems Analysis:

CE Software Subsystems CE Hardware Subsystems EE Subsystems

Communications Power Systems

ME Subsystems Box Subsystem Thermal Subsystem



Customer Requirements

Customer Rqmt. #

Importance Description

CR1 9 Send signalCR2 9 Send signal at a rate in kb/sCR3 9 System must function underwaterCR4 9 Reliable communication schemeCR5 9 Communications must resist frequency contaminationCR6 9 Must have 2-way communication capabilitiesCR7 9 Must be able to operate in frequency contaminated environments

Two-way communication at 15 kb/s of data15 Watts of powerOperating depth of 10mMax operating temperature of 85 deg F


Rochester Institute of Technology 410/29/13

Software Subsystems-Communication Protocol-Control Unit-Receivers and Transmitters-Compression/Decompression-Encryption/Decryption-Data Framing-Error Checking and Correction



Communication Protocol: CSMA/CA(Carrier Sense Multiple Access/Collision Avoidance)

• Little Noise• High Throughput via CA• Functions with Swarm Expansions• Little Overhead (“11” for RTS, wait for “00” for CTS)

StartAssemble a

Frame of DataChannel

Idle?

Back off for random amount

of time

Transmit Request to Send (RTS)

Clear to Send (CTS) Received?

Transmit Data End

NO

NO

YES

YES



Control Unit•The main program running on the microcontroller:It initializes all of the other software modules and manages them•Receives data, decides what to do with it, and sends it off to its destination



Receivers and Transmitters•Four total software modules:

• Incoming Message Receiver (Rx Hardware to MC)• Outgoing Message Transmitter (MC to Tx Hardware)• PC Receiver (PC to MC)• PC Transmitter (MC to PC)• Rx modules send incoming data to the control unit• Control unit sends outgoing data to the Tx modules



Compression/Decompression Module•One method for compression, one for decompression•Each method takes in and returns a bit array•From engineering analysis, some level of compression is needed to achieve a 15kbit data rate (timing analysis done later)•Most lossless compression algorithms can achieve a 2:1 compression ratio



Encryption/Decryption•One method for encryption, one for decryption•Each method takes in and returns a bit array•Encryption will be implemented as time allows, therefore:

• Assume a publically shared key • No need to worry about timing analysis for now



Data Framing•A frame contains data being transmitted, and provides some additional information about the data as well•More frames = Easier to correct errors (Less data to check over at a time)•Start with a 7.5kbit frame, this can be lowered as needed•Frame header will start with a “11” to signify the beginning, followed by the amount of bits contained (a 14-bit number)•Frame footer will simply have a “00” to signify the end of the frame.



Encoder

k data bits

n encoded bits



User Interface•Data is sent and received using Rx and Tx Modules•A Text-based Interface (i.e. unix terminal, cmd prompt) is sufficient for sending and receiving messages•If time allows, a more user-friendly GUI may be implemented



Software ArchitecturePC

(User Interface)

RX (PC)

TX (PC)

Control Unit (Microcontroller)

RX (Incoming Message)

TX (Outgoing Message)

DSP

Error Handler (Encoding, Decoding, Correcting)

Encryptor/Decryptor

Compressor/Decompressor



Microcontroller: Raspberry Pi•Widely-used, cost-effective microprocessor•Price: About $50 after tax + shipping•700MHz ARM11-based processor (CPI is just over 1)•512 MiB SDRAM•Easy to interface with (multiple serial, i2c ports)•Numerous written resources and strong developer community

Phase Shift Keying

Digital modulation scheme that stores data by modulating the phase of the carrier frequencyThe modulation will allow each phase to represent a unique pattern of bits, with each phase containing the same number of bitsThere are two main ways of demodulating a PSK signal

1. By viewing the phase itself as conveying info2. By viewing a change of phase as conveying info





Quadrature Phase Shift Keying (QPSK)

Each point in the constellation represents a 2 bit binary number based on the in phase and quadrature components the signal

00 = A*cos(2πfct)01 = A*sin(2πfct)10 = -A*cos(2πfct)11 = A*cos(2πfct)

Initializing the constellation in this manner is known as Gray Coding. This allows for a lower bit error rate due to only one bit changing per 90 degree shift in phase.



Spectral Efficiency

Specifies the information rate that can be sent over a given bandwidthBeing that PSK is a double-sideband modulation scheme, the symbol rate W cannot exceed the N (bit/s)/HzSince we are considering QPSK to be our modulation scheme, we have an alphabet of M = 4 symbolsFrom this we know:

N = log2(M) = 2, and thus we cannot exceed 2 (bit/s)/Hz

From our Engineering Specs. our data rate must be 15 k(bit/s) 15k(bit/s) <= 2x(bit/s)/Hz Therefore our Bandwidth, x, must be at least 7.5kHz

To account for coding overhead and non-perfect signals, we have decided to set our bandwidth to 10kHz



• Being that the frequency range of our speaker is 2kHz to 15 kHz, we will center our bandwidth around 8 kHz

• This will give us an overall bandwidth ranging from 3kHz to 13kHz



Demodulation: 2 Schemes

AMPBPF

sin(2πfct)

cos(2πfct)

ADC

ADC

To MC

To MC

si(t)



Mixing:



This modulation scheme will take in the signal, si(t), bandpass filter the signal around the frequencies contained in our bandwidth, and then pass that signal to the in phase and quadrature phase branches

In each branch, the signal will either get mixed with a cosine or a sine to leave

only the part of the signal which is in phase with the mixing signal

Each modified signal will now be passed to an analog to digital converter and fed to pins on the micro controller

Pros: 1) we only need to differentiate between two signals on each pin

• Cons: 1) requires more components and therefore space 2) slightly more complicated than other demodulation schemes



This scheme is similar to the first scheme only without the phase discriminant part of the circuit.

Pros: 1) requires fewer components

Cons: 1) more error prone due to comparing between 4 signals vs. 2 signals

AMP BPF si(t) ADC To MC



Transmitter Amplifier StageVoltage Gain: Non-inverting Op-Amp Circuit

Current Gain: Class B or AB Amplifier

(5V )(AV )=12V

AV =2.4V /V

-Class B uses less power-Class AB has lower distortion



Common Mode Choke:

These are very useful for removing electromagnetic interference and radio frequency interference from the power supply linesA CMC is composed of either 2 windings around a magnetic core or a ferrite bead.A CMC is essentially 2 inductors in series, and just like any otherinductor, resist changes to currentTherefore, alternating currents athigher frequencies are resisted muchmore than current changes at lowfrequenciesThis is to say that the chokes impedance increases with freq.



Specifying a CMC:

The cutoff frequency of the CMC can be derived to be fc = 1/(2π*(2L/R))

where L is the value of the inductor in Henneries and R is the value of the load resistor, determined by the speaker, in OhmsSince we are working with very low frequencies we are able to the allow the CMC to filter any frequency above 50 kHz without much worry

fc = 50 kHz (L) = R/(4π*50k)



Signal Power TransformationsElectrical Power to Acoustic Power

P ac=P elec Eff speaker

P ac=(10W )(0.01)

P ac=0.1W



Signal Power TransformationsAcoustic Power to Transmitting Sound Pressure Level (SPL)

SPL=√((P ac Z )

A)

SPL= 1216 Pa

Z =1.48 MNs /m 3

P ac=0.1W

A=1ft2=0.1m2

LT=20log(SPL1uPa

)

LT=181 dB



Signal Power TransformationsTransmission Losses1. Spreading

2. Absorption

TLS=20 log R

TL A=1dB / km

TL( A@ 30m )=0.03dB

TLS=20 log30mTLS=29.54 dB



Signal Power TransformationsTransmitting to Receiving SPL

LR= LT −TL S−TL A

LR=181dB−29.54dB− .03 dB

LR=151.43 dB

SPL10W=35.5Pa

SPL2.5W=18.7Pa



Signal Power TransformationsReceiving SPL to Hydrophone Voltage

@0m @30m

Hydrophone Sensitivity=20 log( (? uV /Pa )(1V /uPa )

)=−190dB

Hydrophone Sensitivity= 316uV / Pa

V R=(316uV /Pa )(1216Pa )

V R=.381V

V R=(316uV /Pa )(35.5Pa )

V R=.0112V



Bandpass Filtering:

Simplest implementation, is an RC high pass filter followed by an RC low pass filter in series

The cutoff frequency is defined as fc=1/(2π*RC)

fcLP = 13kHz= 1/(2π*RC)RC = 1/(2π*13k)

fcHP = 2kHz = 1/(2π*RC)RC = 1/(2π*2k)



Bandpass Filter Simulation:

Parameters:RL = 5kΩCL = 2.3nF

RH = 10kΩCH = 8nF

Results:wcL = 8.696e+4 rad/secwcH = 1.25e+4 rad/sec



Receiver Amplifier GainPSKAutomatic Gain Amplifier-amplifies the input signal such that the RMS voltage matches a reference voltage-allows for amplification to same voltage level independent of distance-can be used because information is not stored in amplitude

AMMaximum Gain

Resulting Gain of lowest voltage @30m

ADC Level Division

AV=5.381

=13

(18.7Pa )(316uV / Pa )(13 )= .074 V

5

210=.005V



Noise Squelch



Signal to Noise

NL10k=NL1k−17log(f

1000)

NL10k=70 dB−17log( 100001000

)dB

NL10k=53dB

SPLnoise=450uPa

SNR=SPLSig

SPLnoise

SNR= (35.5Pa )(450 uPa)

=98dB



Symbol Error Rate

Ps = 2Q[(Es/No)(1/2)]-Q[(Es/No)(1/2)]2

Q[x] = Q-function or the tail probability: Gives the probability that a normal, random Gaussian variable will be larger than x

Es/No = signal to noise ration of each symbol (in dB)

Ps = 2Q[(98)(1/2)]-Q[(98)(1/2)]2 = 4.183825607779467e-23(That’s pretty low…)



Level ShiftingOnly necessary is hydrophone voltage varies between a positive and negative voltage



Interfacing with Modulation/Demodulation SchemesModulation:•AD9835 Direct Digital Synthesizer, Waveform Generator•Takes 16-bit commands, can store phases and frequencies•Outputs an analog signal based on selected phase and frequencyDemodulation:•Still looking at potential chips•If no chip can be found, an ADC can just pass the input wave to the microcontroller and DSP can be performed



Communication Hardware Diagram

Microcontroller

Demodulation Chip/Circuit

Modulation Chip

PC

Voltage/Current Amplifier

Common Mode Choke

Speaker

HydrophoneBandpass

FilterAutomatic

Gain Amplifier



System Power



Buck Converter Buck-Boost Converter

D=V 0

V S

D=V 0

(V S+ V 0)

L=(1−D )2 R2f

L=(1−D) R2f

C= D

(R (ΔV 0

V 0

) f )C= (1−D)

(8L(ΔV 0

V 0

) f 2)



Design Specifications

Buck Specifications Buck-Boost Specifications



Battery SelectionBattery Voltage:12 VoltsSystem Power: 15 WattsSystem Current: 1.25 AmpsBattery Lifetime > 1 HourBattery Energy > 1.25 Ah

Data Rate Analysis





Data Rate Analysis (Compression/Decompression)•FLZP compressor was chosen for reference analysis (selected a compression algorithm with a low rating)•Can compress at 171968kbps and decompress at 674608kbps on a 2.9 GHz processor (2:1 ratio)•RPi is roughly 24% as fast. Slowdown translates to 41272 kbps compression and 161905kbps decompression.•15kbits can be compressed in roughly 360µs, 7.5kbits can be decompressed in roughly 46µs•The time needed for compression and decompression is negligible, even for a poorly rated compressor



Data Rate Analysis (Encoding/Decoding)•Using Reed-Solomon error correcting codes,a 206MHz processor can correct 10Mbps if 10% error rate•RPi Speedup ≈ 340%•The RPi can theoretically correct 34Mbps if 10% error rate•At most, we are looking to correct 750 bits. Using the above rate, this can be accomplished in 22µs•The time needed to correct errors is negligible for a widely used EEC scheme. •It is assumed that correction is more complex than both encoding and detection (more complex operations)



Thermal Analysis



Rapid Prototype Design•Uses Corrosive resistant plastic and rubber O-rings

•Rapid Prototyping Machines

•Quick build time

•Easy implementation

•Watertight connectors on back

•Interchangeable top and front panels for future integrations

•Higher cost



Sheet Metal Design•Uses Corrosive resistant sheet metal and rubber gaskets

•Metal bending and welding techniques

•Low cost of material

•Quick local fabrication vendor

•Interchangeable front and back panels for future integrations

•Watertight connectors on back



Structural Analysis

Using stainless steel, ¼” thickSalt water density is 1035Pressure requirement is increased



Final Tests Overview•Mounting in pool

• Base structures with cables for manipulating•Test data transfer rate•Test error correction

• Test with/without•Test range

• Test at 30 m distance•Test operating temperature range

• Hot tub for hot range, ice bath for cold•Test pressure resistance

• Lower box to 10 m depth



Cost Analysis

Estimate is under budgetLeaves room for any surprises later on in the project

Item Qty Cost/item Total CostRaspberry Pi 2 50 100Speakers 2 100 200Hydrophone 2 200 400Electronics 2 100 200Box 2 300 600Material & equipment tests 1 80 80

Total System 1580



Material Testing (B117-11)

Samples are sprayed continuously with salt solution for 1 weekSamples are weighed before and after test5-6 Samples of 316 Stainless, 464 Naval Brass, ASC Plastic, and 6061 Aluminum



Questions?

greg davis chris johnson scott hambleton jon holton chris monfredo

Documents

incoming data

outgoing data

themreceives data

data framinga frame

kbit data rate timing

level of compression

encryptiondecryptionone

kbit frame