Touchpad-Controlled Parametric Equalizer

1
Touchpad-ControlledParametric Equalizer

• ECE 445
• Group 24
• Anthony Mangognia
• Alexander Spektor
• Farsheed Hamidi-Toosi
• TA Chad Carlson

2
Introduction

• Goal To create a real-time audio filtering
  solution for musicians and sound engineers
• Goal To provide independent control over filter
  parameters center frequency, bandwidth, and gain
• Goal To create a geometrically-intuitive input
  device for the filter

3
Existing Alternatives

• X-Band Equalizers
• Sliders or knobs
• Limited control
• Takes time to adjust
• Discrete frequency bands
• Digital Parametric Equalizers
• Pseudo-continuous frequency sweep
• Cumbersome software-based control

4
Design Overview

• Input Device Pressure-sensitive touchpad.
• Horizontal position center frequency
• Vertical position gain
• Overall Pressure bandwidth
• Input-Filter Interface RS-232 serial connection
• Filter DSP-implemented algorithm
• Second-order IIR filter (based on Mitra-Regalia
  topology)
• Filter coefficients update in real-time

5
Design Block Diagram
Audio In
DSP Parametric Equalizer Filter
Microcontroller Positioning Algorithm Touchpad
?DSP interface
Audio Out
Touchpad
6
1. Touchpad Process
Pressure Sensor Voltage Differential
Bias cancellation/Tuning Circuit
830x Amplifier
PIC A/D
7
Touchpad

• Four corner-mounted pressure sensors on
  height-adjustable shelves
• Pressure sensor output voltage varies with finger
  position
• Positioning algorithm
• Surface much larger than typical
  commercially-available touchpads

8
Touchpad Design
Touchpad Surface
Sensor Mount
Pressure Sensors
Slide-in touchpad
Birds Eye View
Height-adjustable sensor mount
Sensor Mount
9
Pressure Sensors

• Honeywell 125PC30G1
• Pressure range 0-30 psi
• Sensitivity 8.33mV/psi

V-
GND
V
10
Amplifier
10
Touchpad Signal Amplification

• Instrumentation Amplifier AD622AN
• Low-cost 4.90
• Gain 2-1000x, external resistor control
• 56O ? 830x gain
• Easy integration wide power supply voltage gain
  (2.6V-15V)
• Large gain ? Large bias voltage
• Solved with 1MO pull-down resistors at inputs and
  100k potentiometer (calibration)

11
2. Microcontroller-EnabledTouchpad ? DSP
Interface
Four sensor A/D
Positioning Algorithm
Serial Transmission
12
Analog to Digital Conversion

• 8-bit A/D conversion for quicker calculations
• 10-bit possible for PIC16F877A, but doubles
  number of bytes for mathematical operations
• Decreases resolution to 256 points maximum
• More than enough to simulate continuous operation
• Read 8 values for each sensor and average
• Functions as a digital LPF

13
Positioning Algorithm
Sensor 1 V1 (0, Ymax)
X2 V2 Xmax / (V2 V1)
Sensor 2 V2 (Xmax, Ymax)
Touchpad Surface
Y1 V1 Ymax / (V1 V0)
Y2 V2 Ymax / (V2 V3)
(X,Y) Finger Position
Sensor 3 V3 (Xmax, 0)
Sensor 0 V0 (0, 0)
X1 V3 Xmax / (V3 V0)
14
Positioning Algorithm
Sensor 1 V1 (0, Ymax)
X2
Sensor 2 V2 (Xmax, Ymax)
Touchpad Surface
Y1
(Xavg,Yavg) Average
Y2
Sensor 3 V3 (Xmax, 0)
Sensor 0 V0 (0, 0)
X1
15
Positioning Algorithm
Sensor 1 V1 (0, Ymax)
Sensor 2 V2 (Xmax, Ymax)
X X1 (Ymax-Yavg) X2 (Yavg)
Touchpad Surface
(X,Y) Weighted Average
Y Y1 (Xmax-Xavg) Y2 (Xavg)
Sensor 3 V3 (Xmax, 0)
Sensor 0 V0 (0, 0)
16
Data Sent to DSP

• Three-byte start sequence 230 x3
• Four sensor readings S0, S1, S2, S3
• Two one-byte positioning words (x3)
• One three-byte stop sequence 232 x3

230
S0
230
230
S1
S2
S3
X
X
X
Y
Y
Y
232
232
232
17
Serial Data Transmission

• Data transmitted at 38400kbps
• Default rate for DSP
• Data Format
• Sent over standard serial cable
• DB-9 connector

1 START BIT
8 BIT WORD
1 STOP BIT
18
RS-232 Voltage Level Conversion

• PIC output at TTL levels
• 0 - 5V
• DSP input at RS-232 levels
• 12V swing
• Conversion with MAX232 line driver

38400kbps serial data at TTL from PIC
MAX232 Line Driver
38400kpbs serial data at RS-232 to DSP
19
3. DSP Audio Filtering
Receive/Decode Data from Touchpad
Update Filter Coefficients
Apply Filter to Audio Input and Send to Speakers
20
Filter Design

• Based on Mitra-Regalia second-order IIR
• Design Equations
• ß cos(?c)
• k 10(GAIN/20 dB)
• a (1 tan (BW/2)
• (1 tan (BW/2)
• Programmed in C for TI-54x fixed-point DSP

A(z) All-Pass Lattice
Mitra-Regalia Topology
21
Src Montana University Web site.
http//www.coe.montana.edu/ee/rmaher/ECEN4002/lab4
_020226.pdf
22
Design Challenges Touchpad

• Pressure sensor noise
• Problem 30mV peak-to-peak noise level
• Solution 8-point averaging filter after PIC A/D
• PIC A/D crosstalk
• Problem Changes in one pressure sensor affected
  values read for other
• Solution Pull-down 0.1µF capacitors at A/D input
  pins
• Serial communication pins
• Problem PIC?PC communication and PIC?DSP
  communication use different DB-9 transmit pins
• Solution Internal rewiring to accommodate both

23
Design Challenges Filter

• Filter type change
• Problem Original algorithm (Chamberlin) produced
  undesirable resonance frequencies
• Solution Switched to Mitra-Regalia topology
• IIR Instability
• Problem Direct form two implementation caused
  overflow
• Solution Implemented lattice structure to reduce
  overflow
• Quantization
• Problem Fixed-point quantization of coefficients
• Solution Lattice structure ensures pole-zero
  cancellation

24
Internal Component Test

• Pressure Sensor Amplifier
• Unwanted signal oscillation 60mV peak to peak
• Due to conflicting RC networks
• Too high for 1V sampling range
• Solution Removed analog smoothing filter

25
Bandwidth Test
26
Center Frequency Test
27
Gain Test
28
GWN Input Boost Tests
29
GWN Input Cut Tests
30
Finished Product Test

• Amplitude test performed with oscilloscope
• Input 11 kHz Sine Wave, 200 mV peak-to-peak
• Tests gain 2 and gain .5 at 11kHz by
  measuring peak-to-peak voltage of output using
  the scope
• Amplitude resolution is .24 dB from -6dB to 6dB
• Exceeded design requirement which stated 2-3 dB
  amplitude resolution

Gain Output P-P
1 200mV
2 400mV
0.5 100mV
31
Finished Product Test

• Frequency tests performed with scope
• Fix bandwidth, set gain 2
• Inputs Sine at set frequencies, Gaussian White
  Noise
• Using FFT on scope, can see if frequencies
  boosted by 2x at desired center frequency
• Proposed frequency resolution 1 Hz (not
  necessary)
• Total of 200 center frequencies possible,
  distributed them logarithmically since hearing is
  logarithmic
• High resolution for low frequencies, less
  resolution for higher frequencies
• Results Filter works for all frequencies within
  hearing range (20Hz-20kHz)

32
Finished Product Test

• Bandwidth Tests
• As total pressure increases, increase bandwidth
• Tested using FFT on scope
• Input signals Gaussian white noise, music
• Test to see if bandwidth varies from 50Hz to
  22050Hz as pressure increases
• Passed tests, including aural tests

33
Finished Product Test

• Latency Tests
• The target of less than 100ms system latency was
  achieved
• Delay due to pressure sensors negligible
• Delay due to PIC negligible (assembly code
  minimized cycles)
• DSP initially had some latency, but code was
  optimized by eliminating FOR loops (less than
  30000 cycles at MHz)
• Usability tests confirm that system latency is
  not an issue when using this system, negligible

34
Finished Product Test

• Usability Tests and Conclusions
• Tested on music signals and white noise signals
• Qualitative analysis
• Was the filtering audible?
• Did the touchpad respond as desired?
• How intuitive was it to find a desired
  frequency?
• Is this design marketable? If so, why?
• How much would this cost to manufacture?

35
Final Thoughts

• Improvement Better pressure sensors
• Higher output voltage ? Less amplification
• Improvement Cascade feature
• A button to keep current settings and use new
  settings in cascade
• Other applications Large touchpad has many
  applications
• Computer input device for the disabled and kids

36
Acknowledgements

• Chad Carlson
• Marty other ECE 445 TAs
• Profs. Haken Beauchamp
• Machine Shop Scott McDonald
• Parts Shop

37
Questions?