Touchpad-ControlledTouchpad-ControlledParametric EqualizerParametric Equalizer

ECE 445ECE 445Group #24Group #24

Anthony MangogniaAnthony MangogniaAlexander SpektorAlexander Spektor

Farsheed Hamidi-ToosiFarsheed Hamidi-Toosi

TA: Chad CarlsonTA: Chad Carlson

IntroductionIntroduction

Goal:Goal: To create a real-time audio To create a real-time audio filtering solution for musicians and filtering solution for musicians and sound engineerssound engineers

Goal:Goal: To provide independent To provide independent control over filter parameters: control over filter parameters: center frequency, bandwidth, and center frequency, bandwidth, and gaingain

Goal:Goal: To create a geometrically- To create a geometrically-intuitive input device for the filterintuitive input device for the filter

Existing AlternativesExisting Alternatives

X-Band EqualizersX-Band Equalizers Sliders or knobsSliders or knobs Limited controlLimited control Takes time to adjustTakes time to adjust Discrete frequency bandsDiscrete frequency bands

Digital Parametric EqualizersDigital Parametric Equalizers Pseudo-continuous frequency sweepPseudo-continuous frequency sweep Cumbersome software-based controlCumbersome software-based control

Design OverviewDesign Overview

Input Device:Input Device: Pressure-sensitive Pressure-sensitive touchpad.touchpad. Horizontal position: center frequencyHorizontal position: center frequency Vertical position: gainVertical position: gain Overall Pressure: bandwidthOverall Pressure: bandwidth

Input-Filter Interface:Input-Filter Interface: RS-232 serial RS-232 serial connectionconnectionFilter:Filter: DSP-implemented algorithm DSP-implemented algorithm Second-order IIR filter (based on Mitra-Second-order IIR filter (based on Mitra-

Regalia topology)Regalia topology) Filter coefficients update in real-timeFilter coefficients update in real-time

Design Block DiagramDesign Block Diagram

Audio In

Audio Out

DSP:Parametric Equalizer Filter

Microcontroller:Positioning Algorithm &

Touchpad DSP interface

Touchpad

1. Touchpad Process1. Touchpad Process

Bias cancellation/Tuning Circuit

830x Amplifier

PIC A/D

Pressure Sensor Voltage Differential

TouchpadTouchpad

Four corner-mounted pressure Four corner-mounted pressure sensors on height-adjustable shelvessensors on height-adjustable shelves

Pressure sensor output voltage Pressure sensor output voltage varies with finger positionvaries with finger position Positioning algorithmPositioning algorithm

Surface much larger than typical Surface much larger than typical commercially-available touchpadscommercially-available touchpads

Touchpad DesignTouchpad Design

Sensor Mount

Sensor Mount

PressureSensors

TouchpadSurface

Slide-in touchpad Bird’s Eye View

Height-adjustablesensor mount

Pressure SensorsPressure Sensors

Honeywell Honeywell 125PC30G1125PC30G1

Pressure range: 0-Pressure range: 0-30 psi30 psi

Sensitivity: Sensitivity: 8.33mV/psi8.33mV/psi

y = -28.116x + 6.814

-50

0

50

100

150

200

250

300

350

-12 -10 -8 -6 -4 -2 0 2

Voltage (V)

Pre

ssu

re (

mm

Hg

)

-12

-10

-8

-6

-4

-2

0

2

-25 25 75 125 175 225 275 325

Pressure (mm Hg)

Vo

ltag

e (V

)

V- GND V+ +10

Amplifier

Touchpad Signal Touchpad Signal AmplificationAmplification

Instrumentation Amplifier: AD622ANInstrumentation Amplifier: AD622AN Low-cost: $4.90Low-cost: $4.90 Gain: 2-1000x, external resistor controlGain: 2-1000x, external resistor control

5656ΩΩ 830x gain 830x gain Easy integration: wide power supply Easy integration: wide power supply

voltage gain (±2.6V-±15V)voltage gain (±2.6V-±15V)

Large gain Large gain Large bias voltage Large bias voltage Solved with 1MSolved with 1MΩΩ pull-down resistors at pull-down resistors at

inputs and 100k potentiometer inputs and 100k potentiometer (calibration)(calibration)

2. Microcontroller-Enabled2. Microcontroller-EnabledTouchpad Touchpad DSP Interface DSP Interface

Four sensor A/D

Positioning Algorithm

Serial Transmission

Analog to Digital Analog to Digital ConversionConversion

8-bit A/D conversion for quicker 8-bit A/D conversion for quicker calculationscalculations 10-bit possible for PIC16F877A, but doubles 10-bit possible for PIC16F877A, but doubles

number of bytes for mathematical operationsnumber of bytes for mathematical operations Decreases resolution to 256 points maximumDecreases resolution to 256 points maximum

More than enough to simulate continuous More than enough to simulate continuous operationoperation

Read 8 values for each sensor and Read 8 values for each sensor and averageaverage Functions as a digital LPFFunctions as a digital LPF

Positioning AlgorithmPositioning Algorithm

Sensor 2: V2

(Xmax, Ymax)

Sensor 3: V3

(Xmax, 0)

Sensor 1: V1

(0, Ymax)

Sensor 0: V0

(0, 0)

Touchpad Surface

(X,Y) – Finger Position

X2 = V2 Xmax / (V2 + V1)

Y2 = V2 Ymax

/ (V2 + V3)

X1 = V3 Xmax / (V3 + V0)

Y1 = V1 Ymax

/ (V1 + V0)

Positioning AlgorithmPositioning Algorithm

Sensor 2: V2

(Xmax, Ymax)

Sensor 3: V3

(Xmax, 0)

Sensor 1: V1

(0, Ymax)

Sensor 0: V0

(0, 0)

Touchpad Surface

X2

Y2

X1

Y1 (Xavg,Yavg) – Average

Positioning AlgorithmPositioning Algorithm

Sensor 2: V2

(Xmax, Ymax)

Sensor 3: V3

(Xmax, 0)

Sensor 1: V1

(0, Ymax)

Sensor 0: V0

(0, 0)

Touchpad Surface

X’ = X1 (Ymax-Yavg) + X2 (Yavg)

Y’ = Y1 (Xmax-Xavg) + Y2 (Xavg)

(X’,Y’) – Weighted Average

Data Sent to DSPData Sent to DSP

Three-byte start sequence: “230” x3Three-byte start sequence: “230” x3

Four sensor readings: S0, S1, S2, S3Four sensor readings: S0, S1, S2, S3

Two one-byte positioning words (x3)Two one-byte positioning words (x3)

One three-byte stop sequence: “232” x3One three-byte stop sequence: “232” x3

230 S0230 230

S1 S2 S3 X

X X Y Y

Y 232 232 232

Serial Data TransmissionSerial Data Transmission

Data transmitted at 38400kbpsData transmitted at 38400kbps Default rate for DSPDefault rate for DSP

Data FormatData Format

Sent over standard serial cableSent over standard serial cable DB-9 connectorDB-9 connector

1 START BIT 8 BIT WORD 1 STOP BIT

RS-232: Voltage Level RS-232: Voltage Level ConversionConversion

PIC output at TTL PIC output at TTL levelslevels ~ 0 - 5V~ 0 - 5V

DSP input at RS-232 DSP input at RS-232 levelslevels ~ ±12V swing~ ±12V swing

Conversion with Conversion with MAX232 line driverMAX232 line driver

MAX232Line Driver

38400kbps serial data atTTL from PIC

38400kpbs serial data atRS-232 to DSP

3. DSP: Audio Filtering3. DSP: Audio Filtering

Receive/Decode Data from Touchpad

Update Filter Coefficients

Apply Filter to Audio Input and Send to Speakers

Filter DesignFilter Design

Based on Mitra-Based on Mitra-Regalia second-order Regalia second-order IIRIIR

Design Equations:Design Equations: ββ = – cos( = – cos(ωωcc)) k = 10k = 10(GAIN/20 dB)(GAIN/20 dB)

αα = = (1 – tan (BW/2)(1 – tan (BW/2)

(1 + tan (BW/2)(1 + tan (BW/2)

Programmed in C for Programmed in C for TI-54x fixed-point TI-54x fixed-point DSPDSP

A(z) = All-Pass LatticeA(z) = All-Pass Lattice

Mitra-Regalia TopologyMitra-Regalia Topology

Src: Montana University Web site. http://www.coe.montana.edu/ee/rmaher/ECEN4002/lab4_020226.pdf

Design Challenges: Design Challenges: TouchpadTouchpad

Pressure sensor noisePressure sensor noise Problem: 30mV peak-to-peak noise levelProblem: 30mV peak-to-peak noise level Solution: 8-point averaging filter after PIC A/DSolution: 8-point averaging filter after PIC A/D

PIC A/D crosstalkPIC A/D crosstalk Problem: Changes in one pressure sensor affected Problem: Changes in one pressure sensor affected

values read for othervalues read for other Solution: Pull-down 0.1Solution: Pull-down 0.1μμF capacitors at A/D input F capacitors at A/D input

pinspins

Serial communication pinsSerial communication pins Problem: PICProblem: PICPC communication and PICPC communication and PICDSP DSP

communication use different DB-9 transmit pinscommunication use different DB-9 transmit pins Solution: Internal rewiring to accommodate bothSolution: Internal rewiring to accommodate both

Design Challenges: FilterDesign Challenges: Filter

Filter type changeFilter type change Problem: Original algorithm (Chamberlin) produced Problem: Original algorithm (Chamberlin) produced

undesirable resonance frequenciesundesirable resonance frequencies Solution: Switched to Mitra-Regalia topologySolution: Switched to Mitra-Regalia topology

IIR InstabilityIIR Instability Problem: Direct form two implementation caused Problem: Direct form two implementation caused

overflowoverflow Solution: Implemented lattice structure to reduce Solution: Implemented lattice structure to reduce

overflowoverflow

QuantizationQuantization Problem: Fixed-point quantization of coefficientsProblem: Fixed-point quantization of coefficients Solution: Lattice structure ensures pole-zero Solution: Lattice structure ensures pole-zero

cancellationcancellation

Internal Component TestInternal Component Test

Pressure Sensor + AmplifierPressure Sensor + Amplifier Unwanted signal oscillation: 60mV peak Unwanted signal oscillation: 60mV peak

to peakto peakDue to conflicting RC networksDue to conflicting RC networks

Too high for 1V sampling rangeToo high for 1V sampling range

Solution: Removed analog smoothing filterSolution: Removed analog smoothing filter

Bandwidth TestBandwidth Test

Center Frequency TestCenter Frequency Test

Gain TestGain Test

GWN Input Boost TestsGWN Input Boost Tests

GWN Input Cut TestsGWN Input Cut Tests

Finished Product TestFinished Product Test

Amplitude test performed Amplitude test performed with oscilloscopewith oscilloscope Input: 11 kHz Sine Wave, 200 Input: 11 kHz Sine Wave, 200

mV peak-to-peakmV peak-to-peak Tests: gain = 2 and gain = .5 at Tests: gain = 2 and gain = .5 at

11kHz by measuring peak-to-11kHz by measuring peak-to-peak voltage of output using the peak voltage of output using the scopescope

Amplitude resolution is .24 dB Amplitude resolution is .24 dB from -6dB to 6dBfrom -6dB to 6dB

Exceeded design requirement Exceeded design requirement which stated 2-3 dB amplitude which stated 2-3 dB amplitude resolutionresolution

GainGain Output Output P-PP-P

11 ~200m~200mVV

22 ~400m~400mVV

0.50.5 ~100m~100mVV

Finished Product TestFinished Product Test

Frequency tests performed with scopeFrequency tests performed with scope Fix bandwidth, set gain = 2Fix bandwidth, set gain = 2 Inputs: Sine at set frequencies, Gaussian White Inputs: Sine at set frequencies, Gaussian White

NoiseNoise Using FFT on scope, can see if frequencies Using FFT on scope, can see if frequencies

boosted by 2x at desired center frequencyboosted by 2x at desired center frequency Proposed frequency resolution: 1 Hz (not Proposed frequency resolution: 1 Hz (not

necessary)necessary) Total of 200 center frequencies possible, Total of 200 center frequencies possible,

distributed them logarithmically since hearing is distributed them logarithmically since hearing is logarithmiclogarithmic

High resolution for low frequencies, less High resolution for low frequencies, less resolution for higher frequenciesresolution for higher frequencies

Results:Results: Filter works for all frequencies within Filter works for all frequencies within hearing range (20Hz-20kHz)hearing range (20Hz-20kHz)

Finished Product TestFinished Product Test

Bandwidth TestsBandwidth Tests As total pressure increases, increase As total pressure increases, increase

bandwidthbandwidth Tested using FFT on scopeTested using FFT on scope Input signals: Gaussian white noise, Input signals: Gaussian white noise,

musicmusic Test to see if bandwidth varies from Test to see if bandwidth varies from

50Hz to 22050Hz as pressure increases50Hz to 22050Hz as pressure increases Passed tests, including aural testsPassed tests, including aural tests

Finished Product TestFinished Product Test

Latency TestsLatency Tests The target of less than 100ms system latency The target of less than 100ms system latency

was achieved was achieved Delay due to pressure sensors negligibleDelay due to pressure sensors negligible Delay due to PIC negligible (assembly code Delay due to PIC negligible (assembly code

minimized cycles)minimized cycles) DSP initially had some latency, but code was DSP initially had some latency, but code was

optimized by eliminating FOR loops (less than optimized by eliminating FOR loops (less than 30000 cycles at ~MHz)30000 cycles at ~MHz)

Usability tests confirm that system latency is Usability tests confirm that system latency is not an issue when using this system, negligible not an issue when using this system, negligible

Finished Product TestFinished Product Test

Usability Tests and ConclusionsUsability Tests and Conclusions Tested on music signals and white noise Tested on music signals and white noise

signalssignals Qualitative analysis: Qualitative analysis:

Was the filtering audible?Was the filtering audible?

Did the touchpad respond as desired?Did the touchpad respond as desired?

How intuitive was it to “find” a desired How intuitive was it to “find” a desired frequency?frequency?

Is this design marketable? If so, why?Is this design marketable? If so, why?

How much would this cost to manufacture?How much would this cost to manufacture?

Final ThoughtsFinal Thoughts

Improvement: Better pressure sensorsImprovement: Better pressure sensors Higher output voltage Higher output voltage Less amplification Less amplification

Improvement: Cascade featureImprovement: Cascade feature A button to keep current settings and use A button to keep current settings and use

new settings in cascadenew settings in cascade

Other applications: Large touchpad has Other applications: Large touchpad has many applicationsmany applications Computer input device for the disabled and Computer input device for the disabled and

kidskids

AcknowledgementsAcknowledgements

Chad Carlson Chad Carlson

Marty & other ECE 445 TAsMarty & other ECE 445 TAs

Profs. Haken & BeauchampProfs. Haken & Beauchamp

Machine Shop: Scott McDonaldMachine Shop: Scott McDonald

Parts ShopParts Shop

Questions?Questions?