02'(//,1* &$3$&,7,9( &283/('
(/(&7520$*1(7,& ,17(5)(5(1&(
)25 (/(&7520<2*5$3+< 6(16256
21 833(5 /,0% $0387((6
Deepam Yasvant Ambelal
A dissertation submitted to the Faculty of Engineering and the Built Environment,
8QLYHUVLW\ RI WKH :LWZDWHUVUDQG -RKDQQHVEXUJ LQ IXO¿OPHQW RI WKH UHTXLUHPHQWV
for the degree of Master of Science in Engineering.
Johannesburg, 2021
i
'(&/$5$7,21
I declare that this dissertation is my own unaided work. It is being submitted for a
degree of Master of Science in Engineering at the University of the Witwatersrand,
Johannesburg. It has not previously been submitted for any degree or examination
to any other university.
The financial assistance of the National Research Foundation (NRF) towards this
research is hereby acknowledged. Opinions expressed and conclusions arrived at,
are those of the author and are not necessarily to be attributed to the NRF
Deepam Yasvant Ambelal Date
13 November 2021
ii
$%675$&7
Prosthetics are used to substitute for the missing limb sections to aid the amputee's
mobility and quality of life. The shift towards electromyography (EMG) -controlled
hand prosthetics from mechanical ones is to offer multiple movements or gestures.
This transition in environments with electrically-powered devices results in the
challenge of noise interference. Capacitively-coupled electromagnetic interference
(CC-EMI) is the predominant noise source and it is found almost everywhere.
An example scenario is an upper limb amputee person JUDVSLQJ D KRXVHKROG ODPS¶V
cable with their non-amputated hand. EMI is expected to galvanically circulate
through the body from the dipole source. Therefore, the suitable selection of the
LQVWUXPHQWDWLRQ DPSOLILHU¶V common mode rejection ratio (CMRR) is important.
Limited literature shows it is possible to determine the EMI in the body using MRI
scans-based FEM modelling, however, it is costly and time-consuming to process.
Human body electric circuit models are used for high electrocution scenarios but
are not applicable for the scenario. This research addresses the problem by
modelling the hand-to-hand (H2H) region of the human body as an electric circuit,
to determine the CC-EMI magnitude at the site of an EMG sensor.
The model developed has three sub-models: The capacitive-coupling (CC) circuit
model, the human body circuit model, and the EMI measurement circuit model. The
circuit model for the CC between the power cable and the grasping hand was
developed using COMSOL simulations. The measurement system circuit model is
considered to be a pair of voltage followers with hydrogel electrode inputs. The
human body model is the main model and the development was based on modifying
the Freiberger human body model. The Freiberger model was modified into a ladder
Page iii of xix
network where the upper limb circuit path was split into two transverse paths using
parallel resistor theory. Along with genetic algorithm optimization, impedances for
the multiple loops were obtained between the two hands in order to mimic galvanic
circulation and its paths.
The model was applied for the scenario using 18 healthy, un-amputated participants
grasping an insulated, flat, two-core power cable for a lamp in a controlled
environment. The experiment evaluated four cases which covered the combinations
of using each hand to grasp the cable and two rotational orientations of the cable in
the grasp. The average error across all participants for modelling the four cases was
27.4 %. The lowest average error achieved for a single participant was 9.6 %. The
error magnitudes were considered acceptable because it was achieved using more
accessible equipment than using MRI scans-based FEM modelling.
Validation was conducted with a lightly modified scenario and confirms that
alternate scenarios can be simulated for the same participant by modifications to
the sensor position and/or the CC sub-model. Thus the CMRR can be determined
from the EMI evaluations of multiple scenarios.
iv
$&.12:/('*(0(176
The author would like to acknowledge with gratitude the support rendered by both
Dr M.V. Shuma-Iwisi and Mr A.K. Mohamed. The author would also like to
express gratitude to the National Research Foundation (NRF) for support funding
through the Grantholder-Linked Student Support Programme.
v
&217(176
Declaration ............................................................................................................... i
Abstract ................................................................................................................... ii
Acknowledgements ................................................................................................ iv
List of Figures ......................................................................................................... x
List of Tables......................................................................................................... xv
Nomenclature ..................................................................................................... xviii
Chapter 1 Introduction ....................................................................................... 1
Context to Research Carried Out .............................................................. 2
Electromyography .................................................................................... 2
Noise in EMG Sensors ............................................................................. 3
Electromagnetic Interference (EMI) noise ............................................... 4
1.4.1. Capacitively-Coupled Electromagnetic Interference (CC-EMI) ...... 4
Problem Identification .............................................................................. 6
1.5.1. Model Requirements ......................................................................... 7
1.5.2. Research Question ............................................................................. 9
1.5.3. Research Contributions ................................................................... 10
Chapter Summary ................................................................................... 11
Page vi of xix
Chapter 2 Human Body Electric Circuit Models & Human Tissue Electrical
Characteristics ....................................................................................................... 13
Galvanic Circulation ............................................................................... 13
2.1.1. Discussion ....................................................................................... 15
Human Body Impedance ........................................................................ 16
2.2.1. Freiberger Human Body Model [35] .............................................. 17
2.2.2. Discussion ....................................................................................... 19
Electrical Properties of Human Tissue ................................................... 20
2.3.1. Assumptions .................................................................................... 21
Chapter Summary ................................................................................... 21
Chapter 3 Proposed Model .............................................................................. 22
Capacitive-Coupling Source Model ....................................................... 22
3.1.1. Conceptualising the Electric Field on the grasped Type P3 cable .. 23
3.1.2. Modelling Capacitance on a grasped, flat power cable................... 24
3.1.3. Capacitance on a Grasped Flat Power Cable: COMSOL modelling
25
3.1.4. Discussion ....................................................................................... 28
Development of the Human Body Circuit Model .................................. 29
3.2.1. Modification 1: Modifying Percentages of the Hand-To-Hand
Impedance ..................................................................................................... 29
3.2.2. Modification 2: Reducing original Model Scope to Hand-to-hand
Model 30
3.2.3. Modification 3: Introducing the 2nd Pathway using Parallel-Resistor
Theory 32
Page vii of xix
3.2.4. Modification 4: Introducing Inter-Pathway Resistances for Multiple
Galvanic Flow Paths and H2H Capacitance ................................................. 32
3.2.5. Modification 5: Dividing Bicep area of Upper Arm for placement of
Bipolar electrode nodes ................................................................................. 35
3.2.6. Discussion ....................................................................................... 37
Chapter Summary ................................................................................... 38
Chapter 4 Experimentation & Results ............................................................. 39
Experimental Environment and Component Design .............................. 39
4.1.1. Environment Setup .......................................................................... 40
4.1.2. Measurement System ...................................................................... 43
Experiment Procedure ............................................................................ 48
4.2.1. Preparation ...................................................................................... 48
4.2.2. Participant Parameter measurements .............................................. 48
4.2.3. Experiment ...................................................................................... 51
4.2.4. Discussion ....................................................................................... 55
Results .................................................................................................... 56
4.3.1. Participant Parameter Measurements .............................................. 56
4.3.2. Experiment Measurements .............................................................. 57
4.3.3. Experiment Data Analysis............................................................... 65
4.3.4. Experiment Discussion.................................................................... 69
Chapter Summary ................................................................................... 70
Chapter 5 Model Assembly and Optimization ................................................ 72
Unified Model of Experiment ................................................................ 72
Page viii of xix
Model Optimisation Results and Error Calculations .............................. 75
5.2.1. µ1DWXUDOO\ 6HOHFWHG¶ 3DUDPHWHUV Z, Y, X, W and V Combinations ... 76
5.2.2. Averaged Output Voltages with Relative Error .............................. 78
5.2.3. V(Out_A) ± V(Out_B) Magnitude Comparison ............................... 80
5.2.4. Discussion ....................................................................................... 81
Chapter Summary ................................................................................... 81
Chapter 6 Model Validation and Critical Analysis .......................................... 83
Model Validation .................................................................................... 83
6.1.1. Validation Experiment .................................................................... 84
6.1.2. Validation Experiment Results ....................................................... 86
6.1.3. Validation Experiment Data Analysis ............................................. 87
6.1.4. Validation Model ............................................................................ 88
6.1.5. Simulated Results of Validation Model .......................................... 91
6.1.6. Discussion ....................................................................................... 91
Critical Analysis ..................................................................................... 93
6.2.1. Successes ......................................................................................... 93
6.2.2. Failures ............................................................................................ 93
6.2.3. Satisfying Problem Statement ......................................................... 94
6.2.4. Research Contributions ................................................................... 95
Chapter Summary ................................................................................... 96
Chapter 7 Future Works and Conclusion ......................................................... 97
Possible Improvements and Future Works ............................................. 98
Page ix of xix
7.1.1. Experiment ...................................................................................... 99
7.1.2. Model Development ...................................................................... 100
Chapter Conclusion .............................................................................. 100
References ........................................................................................................... 101
Appendix A ± Circuit Figures of Unified Model ................................................ 106
Appendix B ± Application Sample of Genetic Algorithm ± Participant A ......... 111
Appendix C - Determining Inter-Pathway Resistances using Genetic Algorithm
............................................................................................................................. 119
7.2.1. Model Amendments ...................................................................... 120
7.2.2. Genetic Algorithm Pseudo-Code .................................................. 121
7.2.3. Applying the Genetic Algorithm ................................................... 125
x
/,67 2) ),*85(6
Figure 1 ± Electrostatic Coupling between the Overhead Cables and EMG system. A)
Describes the inherent electric field coupling between the EMG electrode wires and
overhead power cables. B) Describes the parasitic capacitances that exist between the
subject, electrode wires and power cables. ......................................................................... 5
Figure 2 ± Electrostatic Coupling between Powered Cables and the subject describing the
inherent electric field coupling between the subject and power cables. Electric field
direction indicates the positive half cycle. The negative half cycle will reverse the field
direction. ............................................................................................................................. 6
Figure 3 ± Describing the parasitic capacitances that exist between the subject and
appliance or electrical component chassis. ......................................................................... 6
Figure 4 ± Electrostatic Coupling between Powered Cables and the Amputee subject. The
VXEMHFW LV JUDVSLQJ D ODPS¶V LQVXODWHG SRZHU FDEOH ZLWK WKHLU XQ DPSXWDWHG KDQG 7KH IORZ
of EMI induced current conducting through the body is indicated with orange dashed lines
DUURZV¶ SRLQWLQJ GLUHFWLRQ LV DUELWUDU\ 7KH QHJDWLYH KDOI F\FOH ZLOO UHYHUse the field
direction. ............................................................................................................................. 7
Figure 5 ± Intra-body signal propagation through the fore-arm. A simplified view of the
fore-arm with three tissue-layered skin (i.e. skin, muscle and bone). ............................... 14
Figure 6 ± Intra-body signal propagation through the fore-arm - Current flow paths for
galvanic coupling, where primary current flow (shown in green arrows) do not affect
receiver signal and the secondary current flow is the current that induces a potential
difference across the receiver electrodes(shown in blue arrows). ZT1 represents the
LPSHGDQFH RI WKH SULPDU\ ÀRZ SDWK =L1 and ZL2 refer to the impedance of the longitudinal
IRUZDUG DQG UHWXUQ SDWK RI WKH VHFRQGDU\ ÀRZ UHVSHFWLYHO\ =T2 refers to the transverse
Page xi of xix
SDWK RI WKH VHFRQGDU\ ÀRZ DQG l is the distance between transmitter and receiver electrode
pairs. .................................................................................................................................. 15
Figure 7 ± Total body impedance as a function of applied voltage for various contact areas;
dry hand-to-hand electrodes measured by Beigelmeier [44]. The image was extracted from
Reilly [47] and the various curves illustrated are identified by the contact area on each hand.
.......................................................................................................................................... 16
Figure 8 ± Internal Partial Impedances of the Human body. The numbers indicate the
percentage of the internal impedance of the human body for the part of the body concerned,
LQ SURSRUWLRQ WR WKH µKDQG-to-IRRW¶ path. (Red arrows indicate skin impedances from the top
layer to the nodes in the body). ........................................................................................ 18
Figure 9 ± Hand grasping a flat power cable, with cable in the cross-sectional view. Blue
and brown circles indicates insulation around the neutral and live wire, respectively.
Human fist outline adapted from Wikimedia Commons [52]. .......................................... 23
Figure 10 ± Simulated Electric Potential of a flat power cable with a floating potential
boundary surrounding sheath. The legend indicates the arrow potentials in V/m. Live and
neutral conductors are on the right and the left, respectively. .......................................... 24
Figure 11 ± Segmenting Cross-section of type flat power cable into three capacitance
regions ............................................................................................................................... 24
Figure 12 ± Simplified current paths in a cross-section of a flat power cable .................. 25
Figure 13 ± Simulated Capacitance C1 and Calculated Capacitances C2 and C3 for Various
Hand Widths Between 50 mm And 120 mm .................................................................... 27
Figure 14 ± Capacitive-Coupling Model, where magnitudes of C2 and C3 are determined
using Equation (4) ............................................................................................................. 28
Figure 15 ± Original Freiberger Model with human body outline. ................................... 31
Figure 16 ± Hand-to-hand Section of Freiberger Model ................................................... 31
Figure 17 ± Modification 2 of Hand-to-hand Human Body Impedance Model ................ 32
Figure 18 ± Modification 3 of Hand-to-hand Human Body Impedance Model ................ 33
Page xii of xix
Figure 19 ± Network of two series capacitors in parallel to another two series capacitors
/DEHOOLQJ LV XQUHODWHG WR WKH +XPDQ %RG\ FLUFXLW PRGHO¶V HOHPHQW QXPEHULQJ ........... 35
Figure 20 ± Modification 5 of Hand-to-Hand Human Body Circuit Model ..................... 36
Figure 21 ± View of Anechoic Chamber from the entrance ............................................. 40
Figure 22 ± Sample measurement of the 230 Vrms power supply found in the Anechoic
Chamber ............................................................................................................................ 41
Figure 23 ± FFT of Sample measurement of the 230Vrms power supply found in the
Anechoic Chamber ............................................................................................................ 41
Figure 24 - Orientation of electrical connections to the lamp. .......................................... 42
Figure 25 - High-level System Diagram of sensor employed ........................................... 43
Figure 26 ± EMI Measurement System Circuit Model. VCC and VEE are powered by the
lithium batteries ................................................................................................................ 46
Figure 27 ± Virtual Op-Amp Settings ............................................................................... 46
Figure 28 ± PCB design for the Measurement sensor ....................................................... 47
Figure 29 ± Pre-Experiment measurements of Body Parameter A) Measurement of
placement of the electrodes (marked with blue dots) and arms span. Human body outline
adapted from Clker-Free-Vector-Images [29]. B) Measurement region of hand width. Palm
image was available through Creative Commons Licence [61] ........................................ 49
Figure 30 ± Hand-to-hand Impedance measurement setup. A Type P3 (2 × 0.75mm) cable
was wrapped with copper tape on its end. The brown wire was connected to the copper
plate on the left to the red node of the LCR meter. The black node was connected to the
right copper plate with a jumper wire. .............................................................................. 50
Figure 31 ± Grasping the Hand-to-hand Impedance measurement setup ......................... 51
Figure 32 ± Simplified setup for the experiment in the Anechoic Chamber. Participant does
not hold any surfaces during baseline measurements. Figure of person was available
through Creative Commons Licence [63] ......................................................................... 52
Page xiii of xix
Figure 33 ± Two cases of the main experiment where the cable has are in two different
rotational orientation in the hand. Note that the experiment had the thumb was placed on
the cable for a more natural grip. Figure of fist was adapted through Creative Commons
Licence [64] ...................................................................................................................... 53
Figure 34 ± Simplified setup for the Case 1 and Case 2 experiment in the Anechoic
Chamber. Figure of person was available through Creative Commons Licence [63] ...... 54
Figure 35 ± Simplified setup for the Case 3 and Case 4 experiment in the Anechoic
Chamber. Figure of person was available through Creative Commons Licence [63] ...... 55
Figure 36 ± Two cases of the validation experiment where the cable has are in two different
rotational orientation in the hand. Note that the experiment had the thumb was placed on
the cable for a more natural grip vii. .................................................................................. 55
Figure 37 ± Baseline Measurements of Participant C. Measurements taken from the
Voltage Follower Outputs for Electrode A and B; and Plug point source with 20:1 Step
Down Transformer. ........................................................................................................... 62
Figure 38 ± Case 1 Measurement of Participant C. Measurements taken from the Voltage
Follower Outputs for Electrode A and B; and Plug point source with 20:1 Step Down
Transformer....................................................................................................................... 62
Figure 39 ± Case 2 Measurements of Participant C. Measurements taken from the Voltage
Follower Outputs for Electrode A and B; and Plug point source with 20:1 Step Down
Transformer....................................................................................................................... 63
Figure 40 ± Case 3 Measurements of Participant C. Measurements taken from the Voltage
Follower Outputs for Electrode A and B; and Plug point source with 20:1 Step Down
Transformer....................................................................................................................... 64
Figure 41 ± Case 4 Measurements of Participant C. Measurements taken from the Voltage
Follower Outputs for Electrode A and B; and Plug point source with 20:1 Step Down
Transformer....................................................................................................................... 64
Figure 42 ± Box and Whisker Diagram of Averaged Voltage Follower Output Voltages,
Baseline Offset Removed, for all Four Cases. .................................................................. 68
Page xiv of xix
Figure 43 ± Unified Model of Experiment, with the Noise Source connection shown in
Figure 43 ........................................................................................................................... 73
Figure 44 ± Noise Source Connection Diagram for the Four Different Cases. A) Case 1,
B) Case 2, C) Case3, and D) Case 4 ................................................................................. 74
Figure 46 ± Box-and-Whisker diagram describing the clustering of relative error data
shown in Table 21 for the four different cases. ................................................................. 80
Figure 47 ± Noise Source Connection Diagram for the Four Different Cases. A) Case 1,
B) Case 2, C) Case3, and D) Case 4 ................................................................................. 89
Figure 48 ± Unified Model of Validation Experiment, with the Noise Source connection
shown in Figure 45 ............................................................................................................ 90
Figure 49 ± Comparison of V(Out_A) ± V(Out_B) against Z Values used in Parameter
Optimization for Case 1 and Case 2, for Participant A ± Generation 1 .......................... 113
Figure 50 ± Comparison of V(Out_A) ± V(Out_B) against Z Values used in Parameter
Optimization for Case 3 and Case 4, for Participant B ± Generation 1 .......................... 113
Figure 45 ± Flowchart of the Pseudo-code for the desired Genetic Algorithm [66] ....... 122
xv
/,67 2) 7$%/(6
Table 1 ± List of Internal Partial Impedances Percentages of the Human body, in Proportion
WR WKH µ+DQG-to-)RRW¶ 3DWK ................................................................................................ 19
Table 2 ± Permittivity and Conductivity Values for Human Body Tissue at 50 Hz and 200
Hz [50], ............................................................................................................................. 21
Table 3 ± Simulated Capacitances for Various Hand widths between 50 mm and 120 mm
.......................................................................................................................................... 26
Table 4 ± Error on Capacitance values obtained with Linear Relationship equations (3) and
(4) ...................................................................................................................................... 28
Table 5 ± Factor adjusted percentages of Body Areas mentioned in Freiberger Model to
result in a 100% Hand-to-Hand Impedance ...................................................................... 30
Table 6 ± Defining the Regions for UArmA, UArmAB and UArmB .............................. 37
Table 7 ±Percentages of Body Segment Impedances after Modification 4 ...................... 38
Table 8 ± Specific Requirements for flat, two-core cable ................................................. 42
Table 9 ± Comparison of various Op-Amps Characteristics. Green fill indicates Best
Performance and Orange fill indicates Poorest Performance ........................................... 45
Table 10 ± Ranges of Non-Amputee Participant sample set ............................................. 56
Table 11 ± Participant Parameter Measurements recorded off Non-Amputee Participants
.......................................................................................................................................... 57
Page xvi of xix
Table 12 ± Condensed Table of Non-Amputee Baseline Measurements for Main
Experiment ........................................................................................................................ 59
Table 13 ± Condensed Table of Non-Amputee Case 1 and Case 2 Main Measurements of
the Experiment. ................................................................................................................. 59
Table 14 ± Condensed Table of Non-Amputee Case 3 and Case 4 Validation Measurements
of the Experiment. ............................................................................................................. 60
Table 15 ± Participant &¶V Parameter Measurements recorded from (sourced from Table
11). .................................................................................................................................... 60
Table 16 ± Summary of Measurements Recorded from Participant C ............................. 61
Table 17 ± Analysis of Output Voltages Difference between Electrodes for all Four Cases
.......................................................................................................................................... 66
Table 18 ± Averaged Output Voltages between Electrodes for all Four Cases ............... 67
Table 19 ± Analysis of Averaged Voltage Follower Output Voltages, Baseline Offset
Removed, for all Four Cases. Shaded Values indicate a less than ±5% difference between
the median and average measurements ............................................................................. 67
Table 20 ± 3DUWLFLSDQW 3DUDPHWHUV DQG WKH¶ 1DWXUDOO\ 6HOHFWHG¶ 3DUDPHWHUV Z, Y, X, W and
V ........................................................................................................................................ 77
Table 21 ± Comparison and Error Calculation of Averaged Output Voltages between
Experiment and Simulation ............................................................................................... 79
Table 22 ± Comparison of ܸሺݐݑ݊ܫ̴ܣܫሻ Ȃ ܸሺݐݑ݊ܫ̴ܣܫെሻ between from the Experiment
and the Simulation measurements..................................................................................... 81
Table 23 ± Clustering of Parameters Z, Y, X, W and V in Participants ............................. 84
Table 24 ± Participant Parameter Measurements recorded off Non-Amputee Participants
.......................................................................................................................................... 86
Table 25 ± Condensed Table of Non-Amputee Baseline Measurements for Validation
Experiment ........................................................................................................................ 86
Page xvii of xix
Table 26 ± Condensed Table of Non-Amputee Case 1 and Case 2 Measurements for
Validation Experiment ...................................................................................................... 87
Table 27 ± Condensed Table of Non-Amputee Case 3 and Case 4 Measurements for
Validation Experiment ...................................................................................................... 87
Table 28 ± Validation Experiment: Averaged Output Voltages between Electrodes for all
Four Cases ......................................................................................................................... 87
Table 29 ± Validation Experiment: Analysis of Averaged Voltage Follower Output
Voltages, Baseline Offset Removed, for all Four Cases. .................................................. 88
Table 30 ± Comparison and Error Calculation of Averaged Output Voltages between
Validation Experiment and Simulation ............................................................................. 92
Table 31 ± Extract of Table 21: Comparison and Error Calculation of Averaged Output
Voltages between Experiment and Simulation ................................................................. 92
Table 30 ± Desired Average Output voltages that the Model needs to Optimize to (values
obtained from Table 19) .................................................................................................. 112
Table 31 ± *HQHUDWLRQ í 7RS &RPELQDWLRQV RI 9DULDEOHV Z, Y and X that produced
Lowest Relative Error for Main Experiment and meet Natural selection requirements. 114
Table 32 ± *HQHUDWLRQ í 7RS Combinations of Variables Z, Y and X that produced
Lowest Relative Error for Validation Experiment and meet Natural selection requirements
........................................................................................................................................ 115
Table 33 ± *HQHUDWLRQ í 7RS &RPELQDWLRQV RI 9DULDEOHV Z, Y and X that produced
Lowest Relative Error for Main Experiment and meet Natural selection requirements. 116
Table 34 ± *HQHUDWLRQ í 7RS &RPELQDWLRQV RI 9DULDEOHV Z, Y and X that produced
Lowest Relative Error for Validation Experiment and meet Natural selection requirements
........................................................................................................................................ 117
xviii
120(1&/$785(
AC Alternating Current
CAD Computer-Aided Design
CC Capacitively-Coupled
CC-EMI Capacitively-Coupled Electromagnetic Interference
CMR Common Mode Rejection
CMRR Common Mode Rejection Ratio
DB board Distribution Box Board
DC Direct Current
DSP Digital Signal Processing
EMG Electromyography
EMI Electromagnetic Interference
FEM Finite Element Modelling
H2H Hand-to-Hand
HREC Human Research Ethics Council
HV High Voltage
IEC International Electrotechnical Commission
LCR Inductor, Capacitor, Resistor
MRI Magnetic Resonance Imaging
MU Motor Units
Page xix of xix
MUAP Motor Unit Action Potential
Op-amp Operational Amplifier
PCB Printed Circuit Board
PVC Poly-Vinyl Chloride
SOIC Small outline integrated circuit
SNR Signal-to-Noise Ratio
SPICE Simulation Program with Integrated Circuit Emphasis
Page 1 of 113
Chapter 1
,1752'8&7,21
Surface Electromyography (EMG) is a common, non-invasive medical procedure
used to record muscle signals for prosthetic devices. The sensor used in this
procedure can easily amplify millivolt signals but have the disadvantage of being
corrupted by a variety of external electrical signals [1]. Electromagnetic
Interference (EMI) at power-line frequency is known to be particularly
troublesome. In literature, there are several techniques used to deal with EMI, such
as Common Mode Rejection (CMR) [2], active and passive shielding [3]±[5],
analogue filters [6] [7], and Digital Signal Processing (DSP) techniques [8]±[16].
Most EMG sensors use Instrumentation Amplifiers (IA) to provide CMR of EMI.
They are recommended to have a Common Mode Rejection Ratio (CMRR) greater
WKDQ G% DQG UHVXOW LQ D QRLVH IORRU RI ȝ9RMS in the 20 Hz and 500 Hz frequency
band [17]. However, one can only confidently specify a CMRR if the actual
magnitude of the EMI is known. To determine this EMI at the input positions of the
EMG sensors, empirical data and/or model projection of the current induced by
EMI flowing through the body, from the source to the sensor inputs, is required.
Finite Element Modelling (FEM) methods can provide this information, but at the
cost of accessibility and time, which is unfeasible for evaluating multiple people
and/or scenarios [18] [19]. More accessible methods that can model EMI
Page 2 of 113
magnitudes as it propagates through the body and the resulting interference at the
site of EMG measurement are unknown.
Context to Research Carried Out
A missing limb or a loss of a limb can occur congenitally or from the amputation
of a severely diseased, injured or dysfunctional limb [20], [21]. This limb loss can
affect the amputee's mobility, psychological well-being, quality of life,
socialization, and employability [20]. Prosthetics are used to substitute for the
missing limb section. Most mechanical prosthetics have one degree of freedom
which results in limited movement capabilities. Over the past decade, there has been
a shift towards EMG-controlled hand prosthetics from mechanical ones. This shift
is because EMG-controlled prosthetic hand offer multiple movements or gesture
control through the use of multiple EMG inputs and modern control systems.
Electromyography
EMG, defined by De Luca [17] ³is the discipline that deals with the detection,
analysis and use of the electrical signals that emanate from contracting muscles´
However, this term refers to the act of achieving a graphical representation of the
signal, which was more prevalent before the computer age. With the ability to store
data digitally, the more appropriate term to define electrical signals emanating from
PXVFOHV LV µm\RHOHFWULF¶ VLJQDOV WKRXJK WKH WHUP (0* LV VWLOO FRPPRnly used [22].
The current flow generated by the ionic chemical flow across the muscle fibres¶
membranes is the source of EMG signals. The chemical changes result in the
contractions of muscle fibres and depends on the degree of stimulation. Muscle
fibres are grouped into Motor Units (MU) and share a common motor neuron. The
pulse voltages produced by the MUs are called Motor Unit Action Potentials
(MUAP). These are the most basic signal component into which an EMG signal
can be separated into.
Page 3 of 113
The frequency of EMG signals ranges between 2 Hz and 500 Hz [23] and the
dominant frequencies are between 20 Hz and 150 Hz, depending on the 08¶V
density, the stimulation intensity and/or the stimulation duration [1], [24]. Smaller
muscle like the Flexor Carpi Radialis (the muscle that controls the wrist) mainly
emits signals at lower frequencies at 55 Hz ± 95 Hz due to smaller set of MU. Larger
muscle like the Biceps Femoris mainly emits higher frequencies at 155 Hz ± 195 Hz
because of a larger set of MU [25]. The myoelectric voltages are known to range
between 10 ȝ9 DQG P9, however, it is expected that maximum myoelectric
voltage for amputated muscles to not peak above 1 mV because a portion of the
MU has been lost or shortened [24], [26]. Additionally, intensive muscle
stimulation over long durations UHVXOWV LQ D UHGXFWLRQ LQ WKH PXVFOH¶V FKDUDFWHULVWLF
frequencies and amplitude because the muscle fibres become fatigued. This is
called muscle fatigue [1].
The overlap in the characteristic muscle frequency and the frequency of various
electrical noise sources in the EMG measurement of amputated and/fatigued
muscles may lead to the poor performance of an EMG-controlled prosthetic.
Noise in EMG Sensors
Due to the small amplitude nature of EMG signals and a characteristic frequency
range of a few hundred Hz, the EMG sensors are required to be sensitive to changes
in signal voltages across the required frequency range.
The use of EMG-controlled prosthetics in environments populated by electrically
powered devices such as domestic homes and offices results in the challenge of
noise interference. Several noise sources are present in the EMG characteristic
frequency range of 10 Hz to 1 000 Hz, which deform or drown the EMG signal
measurements. These noise sources are:
x Electrode Noise,
x Movement Artefact Noise,
x Electromagnetic Interference (EMI) Noise,
Page 4 of 113
x Cross-Talk Noise, and
x Internal Noise.
Clancy et. al. [27], extensively reviews the above-mentioned noises. EMI is the
most common source of noise in EMG measurements, and it is found almost
everywhere.
Electromagnetic Interference (EMI) noise
Electrified spaces, such as homes and offices, contain wired electrical equipment.
Current flowing through cables generates electromagnetic waves which are
regarded as EMI. Continuous EMI noise sources in the EMG frequency bandwidth
below 1 kHz are from the powerlines (at 50/60 Hz), power supplies and their
associated harmonics. Impulse EMI noise sources can be from switching of heavy
load appliances, electrostatic discharges and lightning. The human body is largely
a conductive composite, and the EMI noise can either be capacitively or inductively
coupled to the human body [27]. Inductively coupled EMI is taken care of by using
shielding techniques but increases the exposure to Capacitively-Coupled (CC) EMI
[3]±[5].
1.4.1. Capacitively-Coupled Electromagnetic Interference (CC-
EMI)
CC involves charge movement from higher electrostatic potential surfaces to lower
potential surfaces. Figure 1 A show a diagram of an individual, attached to EMG
sensor probes, standing between an overhead powerline and a ground plane. The
gaps between the two different levels of potentials are parasitic capacitances and
result in stray charge flows from the powered cables to objects as shown in Figure
1 B [27] [28]. These potentials make the external environment influence the
outcome of the EMG measurement.
Page 5 of 113
Figure 1 ± Electrostatic Coupling between the Overhead Cables and EMG system. A) Describes the inherent
electric field coupling between the EMG electrode wires and overhead power cables. B) Describes the
parasitic capacitances that exist between the subject, electrode wires and power cables.
In Figure 1 B the individual is in a uniform electric field. In environments that may
be experienced by an amputee equipped with an EMG-controlled prosthetic, the
EMI is influenced by powered appliances and equipment (like electric stoves and
fridges), active electrical components (like power cords, extensions and wall
switches), and non-grounded metallic surfaces (like metal frame tables) [31]. This
results in a non-uniform electric field unlike in the uniform electric fields shown in
Figure 1 A and B.
The location of the largest concentration of stray electric fields is expected to radiate
close to the source. As seen in Figure 2, where the subject¶s hand is in proximity to
the appliance chassis or electrical component surface, the hand will provide a better
conduction medium, than air. A return path for the high potential electric field from
the live wire back to the neutral wire is established. Instead of parasitic capacitance
occurring across the air as shown in Figure 1 B, the capacitance is across the cable
insulation as shown in Figure 3. The contacts made by the hand results in a dipole
interaction and a potential difference develops between the areas of the hand that
are nearest to the live and neutral wires which would attribute to the EMI induced
current circulating through the body.
Page 6 of 113
Figure 2 ± Electrostatic Coupling between Powered Cables and the subject describing the inherent electric
field coupling between the subject and power cables. Electric field direction indicates the positive half cycle.
The negative half cycle will reverse the field direction.
Figure 3 ± Describing the parasitic capacitances that exist between the subject and appliance or electrical
component chassis.
Problem Identification
Modifying the interaction found in Figure 2 and Figure 3 to the potentially common
scenario with an amputee person. As shown in Figure 4, the subject has an EMG
Page 7 of 113
sensor on the bicep of the amputated arm. The amputee would use their non-
DPSXWDWHG KDQG IRU JUDVSLQJ D KRXVHKROG ODPS¶V SRZHU FDEOH It is expected that
the current induced by EMI will be conducted through the body, from the un-
amputated hand to the EMG sensor on the amputated arm. Therefore, a suitable
selection of an IA with an appropriate CMRR is important. However, there is
limited literature on the human body interacting with a dipole EMI source.
The conduction from the un-amputated hand to the EMG sensor is similar to the
galvanic circulation experiment conducted by Wegmüller [32]. Galvanic circulation
is expected to occur in the body, where the flow will begin from the grasping hand,
through the torso, to the bicep-placed sensor on the amputated arm, and back to the
grasping hand. This is shown by the orange arrows in Figure 4.
Figure 4 ± Electrostatic Coupling between Powered Cables and the Amputee subject. The subject is grasping
D ODPS¶V LQVXODWHG SRZHU FDEOH ZLWK WKHLU XQ DPSXWDWHG KDQG The flow of EMI induced current conducting
through the body is indicated with orange GDVKHG OLQHV DUURZV¶ SRLQWLQJ GLUHFWLRQ LV DUELWUDU\ 7KH QHJDWLYH
half cycle will reverse the field direction.
1.5.1. Model Requirements
The EMI measurements at various EMG electrode positions on the human body are
simple voltage measurements, often using buffer amplifiers to separate the subject
Page 8 of 113
from bench-powered oscilloscopes [31]. There are many different possibilities
involving a subject interacting with EMI sources in a household or office
environment and many more unique human body compositions. Therefore, a base
model that can be adjusted for the various possibilities DQG WKH VXEMHFW¶V XQLTXH
parameters can be developed to determine the potential range of EMI magnitudes
that an IA must account for. The ability to model copious quantities of possibilities
with reasonable accuracy would reduce the accumulated experimentation time and
resources.
Modelling the EMI due to grasping of the power cable as shown in Figure 4 requires
two types of sub-models:
1. sub-model of the capacitive coupling between the source and the surface of
the hand, and
2. sub-model of the electrical current flow pathways in the human body from
the grasping hand to the EMG sensor
The EMI source, being a two-core power cable, is a dipole because the electrostatic
potential will propagate from the live wire to the neutral wire, which is tethered to
the ground in the DB board. Modelling CC can be done using a physics-based
simulator, like ³COMSOL 0HWDSK\VLFV´.
The human body model will need sufficient parameters to uniquely describe the
subject. Enough nodes are also needed to distinguish between the positions of the
(0* VHQVRU¶V inputs and the CC interception points. Since the body interacts with
a dipole, the EMI will galvanically circulate through the body, similarly to the
experiment conducted by Wegmüller [32]. The circulation will begin from the
grasping hand, through the torso, to the sensor on the amputated arm where the
sensor is placed on the bicep, and back to the grasping hand.
Modelling the human body electrical pathways often occurs in two ways. The first
uses FEM to model the subject and the second uses electric circuit models.
Page 9 of 113
Finite Element Modelling (FEM)
FEM GLYLGHV WKH KXPDQ ERG\¶V WKUHH-dimensional volume into small blocks with a
certain characteristic or property [18]. FEM models can be constructed from CAD
models or MRI scans. MRI scans have the benefit of being true representations of
the subject however the model generated is not deformable. CAD models, being
inherently deformable, allows the model to perform a motion or a gesture.
Though FEM models may produce high-resolution data, the use of MRI as a source
of data has several shortcomings. MRIs are expensive to run and take a long time
to acquire complete data [19]. Highly detailed or high-resolution models also
require high-performance computing hardware to process the circuit analysis which
adds to the processing cost. Open-Source models are available online from several
sources; however, most models require procurement [18].
Electric Circuit Human Body modelling
The second method is to model the human body as an electric circuit which is
simpler and easier to compute. These models are often used to simulate the HV
electrocution of subjects due to lightning or coupling to HV power lines, but they
have not yet been implemented for EMG measurements. The circuit model can also
be adjusted for different scenarios as well as specific human body parameters.
1.5.2. Research Question
As explained before, electrically modelling multiple human bodies using circuit
models is more viable to practice than using FEM due to accessibility reasons. The
unavailability of a human body electric circuit model to model dipole-sourced EMI
in EMG-sensors, lead to the following research question:
³:KDW is the human body electric circuit model for the prediction of EMI due to a
dipole capacitively-FRXSOHG VRXUFH JLYHQ WKH IROORZLQJ FRQGLWLRQV"´
Page 10 of 113
x The subjects are non-amputees that have right-handed dominancy,
x The EMI sensor is placed RQ WKH ULJKW DUP¶V ELFHS RI WKH VXEMHFW
x The subjects are grasping a flat, two-core power cable used for lamp
applications, to induce EMI currents in the body
x The lamp will be powered by 230 VRMS (330 VPeak), 50 Hz, and
x The subjects are measured in a controlled environment.
Constraints
The limited accessibility to research participants resulted in the following
constraints:
x No access to amputee subjects. Therefore, modelling of non-amputee
subjects was carried out, and
x No dominant left-handed subjects were evaluated due to their low
prevalence. Therefore, modelling was conducted only on dominant right-
handed participants.
1.5.3. Research Contributions
The objective of this research is to develop a model that can determine the noise
voltages induced by EMI, at the inputs of the EMG sensor that are placed on the
upper limb, using empirical data. This model would be used to predict the voltages
induced by the EMI for an alternate but similar scenario. By achieving this
objective, some contributions made to this field of research are:
x Possible utilization or modification of the existing electric circuit model of
the human body.
x Providing an experiment method which can be used to examine the flow of
EMI interference that is due to near-field or capacitive coupling with
sources. This method was demonstrated for the events where WKH VXEMHFW¶V
measurements are not ground or common-node referenced to a non-foot
region of the human body.
Page 11 of 113
x Results of the experiment may provide useful empirical data to determine
the required CMRR specification for the sensor. Usually this is
approximaWHG (0, QRLVH ³URRI´ YDOXH.
x Likewise, to the empirical EMI measurements, CMRR can be determined
using the model projections of scenarios similar to the base scenario.
Chapter Summary
This chapter has introduced the possibility of problems due to the impact of CC-
EMI on EMG-controlled prosthetics. Upper limb amputees seek these EMG-
controlled prosthetics to regain mobility and independence. But the EMG sensors
used to operate the prosthetic have CMRR that are not specified according to
empirical data or model projections. Therefore, when the prosthetic user is
performing a daily task that introduces higher than normal EMI noise conditions
for the sensor, the user may have difficulty in operating the prosthetic using
amputated or weaker arm muscles. This circumstance LV EHFDXVH WKH (0* VLJQDO¶V
FKDUDFWHULVWLF IUHTXHQF\ EDQG RYHUODSV ZLWK WKDW RI WKH (0,¶V IUHTXHQF\ EDQG
In particular, the purpose of this research is directed towards modelling the
magnitude of the CC-EMI, at 50 Hz, that conducts through the human body and
could interfere with the Bicep EMG measurements. To be able to model the
magnitude of the EMI, it is required that a model of the human body, as well as the
model of the capacitive coupling, be developed.
Achieving a functioning electrical model of the human body that accounts for the
current flow from one hand to the other will potentially reduce the time and
resources required to capture EMI magnitudes found when an EMG-user is
interacting with the various EMI sources found in homes or offices, and thus a
CMRR can be prescribed.
The research requires the understanding and modelling of electric fields and
parasitic capacitances in homes and office space, the electrical properties of human
tissue and the human body as an electric circuit. Chapter 2 provides some
Page 12 of 113
background knowledge on the human body electrical models and electrical nature
of human tissue.
Chapter 3 documents the theoretical development of the capacitive-coupling and
the human body circuit model.
In Chapter 4 the experimentation and the measurement results are discussed.
In Chapter 5, the final model derived from the sub-models discussed in Chapter 3,
and the parameters recorded in Chapter 4 is presented.
In Chapter 6, model validation is conducted and research work is critically
analysed. Potential future works are also discussed.
The research work is concluded in Chapter 7.
Page 13 of 113
Chapter 2
+80$1 %2'< (/(&75,& &,5&8,7
02'(/6 +80$1 7,668(
(/(&75,&$/ &+$5$&7(5,67,&6
This chapter examines the literature on galvanic circulation, human body electric
circuit models and human tissue electrical characteristics. The magnitudes of body
impedance depend on several factors such as current path, touch voltage, duration
of current flow, frequency, degree of moisture of the skin, surface area of contact,
the pressure exerted and temperature. [33], [34]. The Freiberger Human Body
model [35] which does account for several of the factors is discussed in this chapter.
Galvanic Circulation
Wegmüller [32] attempted to describe the portions of the human body as a signal
propagation mediums for Intra-body Communication. Considering the areas of the
body with simple anatomical structures like the arms and legs. The modelling of
the geometrical and physical properties was approximated by three types of tissue:
Page 14 of 113
skin, muscle, and bone. This is shown in Figure 5. experimental setup was described
as being similar to electrical impedance tomography [36]. Electrodes that were
attached to a low-amplitude current source were placed on the lower part of the
VXEMHFW¶V IRUHDUP DQG GHWHFWLRQ HOHFWURGHV ZHUH DWWDFKHG WR WKH XSSHU SDUW RI WKH
forearm.
Figure 5 ± Intra-body signal propagation through the fore-arm. A simplified view of the fore-arm with three
tissue-layered skin (i.e. skin, muscle and bone).
However, to describe the galvanic current flow between the transmission electrodes
and the receiver electrodes. Kibret et al. describes the processes better in a graphical
form and this is shown in Figure 6. In this figure, there are two groups of currents
flowing between the electrodes. The primary flow occurs between the transmission
nodes and it can be considered the transverse path. The secondary flow moves
laterally across the limb and is intercepted by the receiver electrodes which
measures across second transverse path and it would be a fraction of the input
voltage [37]. However, this diagram is for a single, homogenous medium and will
differ when a multi-medium model is developed.
Page 15 of 113
Figure 6 ± Intra-body signal propagation through the fore-arm - Current flow paths for galvanic coupling,
where primary current flow (shown in green arrows) do not affect receiver signal and the secondary current
flow is the current that induces a potential difference across the receiver electrodes(shown in blue arrows).
ZT1 UHSUHVHQWV WKH LPSHGDQFH RI WKH SULPDU\ ÀRZ SDWK =L1 and ZL2 refer to the impedance of the longitudinal
IRUZDUG DQG UHWXUQ SDWK RI WKH VHFRQGDU\ ÀRZ UHVSHFWLYHO\ =T2 refers to the transverse path of the secondary
ÀRZ DQG l is the distance between transmitter and receiver electrode pairs.
2.1.1. Discussion
Current research into modelling CC-EMI on the human body is limited to scenarios
with overhead powerlines and its effect in electrocardiography measurements [38]±
[40]. Haberman [31] may have touched on proximity CC-EMI with the experiment
involving contact with a light switch. These works all reduced the human body
component of the model, between the electrodes, to arbitrary star or delta triple
resistor network or a potential divider pair.
Apart from the Freiberger model, to be discussed, human body circuit models have
been used to simulate the lightning or electrostatic discharge electrocution of
persons. 7KHVH PRGHOV DUH ³VWLFN-ILJXUH´ KXPDQ ERG\ LPSHGDQFH PRGHOV DQG DUH
described in Andrews [41] and Sutherland [42]. However, these do not support or
cannot be modified to support galvanic circulation.
Galvanic circulation is considered an important concept in the developed model.
Like the current propagation in Figure 6, the current propagation in the CC-EMI
H2H Human Body model will also circulate, however, across a larger transverse
path, i.e., from one hand to the other and with more nodes.
,ƵŵĂŶ
ŽĚLJ
TX
ZL1
V RX
A1
A2
B1
B2
ZT2
I
l
ZL2
Page 16 of 113
Adapting Figure 4 to describe the EMI coupling and propagation like the processes
occurring in Figure 6, the developed model will break up into three sub-models.
The transmission source containing the EMI source and the CC, the transmission
medium consisting of the human body model, and the receiver part consisting of
the measurement system. The developed human body model will essentially
perform as a two-port network, where two terminal inputs will interact with the CC
model and a three terminal output will connect to the EMI measurement sensor.
Human Body Impedance
Biegelmeier [43]±[45] conducted experiments showing the relationship between
dry Hand-to-+DQG DEEUHYLDWHG WR ³+ +´ impedance and applied AC voltage for
various hand contact areas between 1 mm2 and 8200 mm2. It was found that at 25 V,
the H2H impedance was inversely proportional to the contact area. The experiment
results are shown in Figure 7 [46].
Figure 7 ± Total body impedance as a function of applied voltage for various contact areas; dry hand-to-hand
electrodes measured by Beigelmeier [44]. The image was extracted from Reilly [47] and the various curves
illustrated are identified by the contact area on each hand.
Page 17 of 113
2.2.1. Freiberger Human Body Model [35]
Over 80 years ago, Freiberger [35] determined the impedance contribution of
different segments of the body and the impedance contributions of the skin in
different areas of the body. Figure 8 depicts the distribution of body impedance for
various current paths. The numbers indicate internal impedance for various
electrode placements as a percentage of the total hand-to-foot impedance body. This
model was also adopted and is easily accessible in the International Electrotechnical
Commission (IEC) technical standard TS 60479-1 [33], [34]
Page 18 of 113
Figure 8 ± Internal Partial Impedances of the Human body. The numbers indicate the percentage of the
internal impedance of the human ERG\ IRU WKH SDUW RI WKH ERG\ FRQFHUQHG LQ SURSRUWLRQ WR WKH µKDQG-to-IRRW¶
path. (Red arrows indicate skin impedances from the top layer to the nodes in the body). i
Page 19 of 113
Table 1 below, derived from Figure 8 lists the impedance contribution for each body
section and different skin contact areas. The total of 100 % µKDQG-to-IRRW¶ SDWK body
impedance does not include the outer skin impedance. Reilly [46] described that
50% of the internal impedance for hand-to-hand or hand-to-foot contacts resides in
the extremities, i.e., wrists or ankles, and is due to the prevalence of poorer
conducting bone and ligament. Reilly suggests that the impedance distribution in
Figure 8 is consistent with the investigations of Taylor [48], who used high-voltage
capacitive discharges on living people.
TABLE 1 ± LIST OF INTERNAL PARTIAL IMPEDANCES PERCENTAGES OF THE HUMAN BODY, IN PROPORTION TO THE
µ+AND-TO-FOOT¶ PATH
Body Part Areas Percentage
Head to heart 10
Heart to navel 1.3
Navel to Upper leg 5.1
Upper Leg 14.1
Lower Leg 32.3
Lower Arm 26.4
Upper Arm 10.9
Deltoid 6.9
Shoulder to Shoulder 6.1
Upper Arm to heart 9.9
Waist 8.7
Skin Areas Percentage
Lower Arm 1.8
Upper Arm 3.3
Shoulder 3.9
Heart 5.2
Navel 8
Upper Leg 3.6
Lower Leg 3.3
2.2.2. Discussion
The Freiberger model is an appropriate foundation for developing the human body
electrical circuit model based on dipole inputs because the body is separated into
similar characteristic segments, like the upper arm, lower arm, upper leg, etc. The
benefit of segmentation by percentages is that each iteration of the model is unique
WR WKH VXEMHFW¶V KDQG-to-hand or hand-to-foot impedances. The model also offers
Page 20 of 113
skin impedance percentages at numerous locations which allows more accuracy for
the meaVXUHPHQW VLWH¶V SRVLWLRQ
If the H2H impedance is measured with skin impedances are included, it will be
98.1 % of the hand-to-foot impedance measured below the skin surface. Therefore,
this model will require a scale adjustment to make it appropriate for H2H
impedance measurement matching up to 100%.
Electrical Properties of Human Tissue
To model a basic electric circuit model of the human body, one would need to know
the electrical properties, i.e., permittivity İ and conductivity ı, of affected tissue
such as skin, muscles, blood, and bone. Gabriel [49] mentions that there are few
reliable datasets on the human body tissue for frequencies below 100 kHz because
the tissue degrades due to decomposition and electric current. Reilly [46] provides
an in-depth work on the dielectric properties of tissue, effects of electricity on the
human body for a wide range of voltages, current and frequencies.
Gabriel et. al. [50], [51] is often referenced in literature and is currently assumed to
be a reliable source of data for conductance and permittivity of human body tissue
for frequencies between 10 Hz and 10 GHz. Table 2 lists the permittivity and
conductivity of the human body tissue of interest [50], [51]. As shown in Table 2,
tissue with high electrolyte content, like blood and muscle, have higher
conductivities.
It should be noted that skin conductance can change due to sources of hydration and
salinity, like perspiration. Also, subcutaneous adipose tissue or fat is less
conductive than muscle. Therefore, one would expect the attenuation of signal
voltage to be proportional to the quantity of fat that lies between the skin and
muscle.
Page 21 of 113
TABLE 2 ± PERMITTIVITY AND CONDUCTIVITY VALUES FOR HUMAN BODY TISSUE AT 50 HZ AND 200 HZ [50],
Tissue 50 @ ߝ Hz 200 @ ߝ Hz 50 @ ߪ Hz
(m/S)
Hz 200 @ ߪ
(m/S)
Blood 5.26 × 103 5.26 × 103 0.700 0.700
Muscle 1.77 × 107 3.70 × 106 0.233 0.291
Subcutaneous Adipose Tissue 4.58 × 105 6.56× 105 4.04 × 10-2 4.08× 10-2
Cortical Bone 8.87 × 103 4.79 × 103 2.01 × 10-2 2.01 × 10-2
Dry Skin 1.14 × 103 1.14 × 103 2.00 × 10-3 2.00 × 10-3
2.3.1. Assumptions
Based on Table 2, an assumption that wet tissues have negligible capacitances can
be made. Therefore, if capacitance is measured off the subject, it will be attributed
to the skin surface of the developed model.
Chapter Summary
In this chapter, human body electrical impedance and human body tissues
characteristics are discussed. The Freiberger human body model was presented and
discussed is expected to serve as a usable base for building the desired model.
Human body tissues are also discussed to identify the different degrees of
resistances and capacitances in the body. Making use of existing verified human
body models and data will benefit the development of the desired model by
reducing the difficulties and complexities of µstarting from zero ¶ This base model
developed in the next chapter should serve to emulate the galvanic circulation as
well as to provide the inputs and output interfaces for the CC model and the EMI
measurement sensor.
i NOTE : In order to calculate the total body impedance ZT for a given current path, the internal
partial impedances Zip for all parts of the body of the current path have to be added as well as the
impedances of the skin of the surface areas of contact. The numbers outside the body show internal
portions of the impedance to be added to the total, when the current enters at that point
Page 22 of 113
Chapter 3
352326(' 02'(/
In this chapter, the sub-models that make up the CC-EMI model for EMG sensors
are developed. These sub-models are the capacitive-coupling (CC) source model
and the human body circuit model.
Capacitive-Coupling Source Model
The deliberate selection of grasping a flat, two-core, power cable (2 × 0.75 mm2)
as the source of EMI was done to reduce the complexity of the CC component in
the final model. These cables have Poly Vinyl Chloride (PVC) insulation for each
of their twin-cores and further surrounded by a PVC insulation sheath, forming an
elliptical cross-section profile. The cables are identified by the code
³+ 99+ -)´
Power cables come in various geometries and several core numbers. Therefore, the
CC model used for this cable is not expected to be directly usable for other cable
variations. For example, there are other two-core wire geometries used in common
household and office environments. Some have circular cross-sections, and some
have thinner wires. These properties affect the electric fields between the wires and
therefore affect the CC induced due to grasp.
Page 23 of 113
3.1.1. Conceptualising the Electric Field on the grasped Type P3
cable
As mentioned in the last chapter, the subject can become capacitive coupled to the
power grid when the cable is grasped, as shown in Figure 9. This results in the tissue
that is in proximity and surrounds the cable becoming a floating potential medium.
Figure 9 ± Hand grasping a flat power cable, with cable in the cross-sectional view. Blue and brown circles
indicates insulation around the neutral and live wire, respectively. Human fist outline adapted from
Wikimedia Commons [52].
The electric potential of the flat power cable is modelled in COMSOL V5.3 and the
cross-section is shown in Figure 10. The copper cylinders are 2 mm apart (from
their centres), the insulation is 0.5 mm thick, and the sheath thickness is 0.6 mm
[53]. The material domains which were PVC were given a relative permittivity of
4.5 [54]. The source potential is set to 330 Vpeak and is based on the rounded-up
average source voltage measured during the experimentation of 329.9 Vpeak. The
result shows that the floating potential surface was at 165 Vpeak.
Page 24 of 113
Figure 10 ± Simulated Electric Potential of a flat power cable with a floating potential boundary surrounding
sheath. The legend indicates the arrow potentials in V/m. Live and neutral conductors are on the right and the
left, respectively.
3.1.2. Modelling Capacitance on a grasped, flat power cable
The analysis of the electric field lines in Figure 10, indicates two different travel
paths. The first path propagates directly between the Live and Neutral and is
concentrated in the region between the two conductors. The second path moves
from the live conductor to the floating potential, conducts along the floating
potential, and finally returns to the neutral conductor.
To convert the electrostatic model into an electric circuit model with minimal
elements, it can be assumed that the first path can be represented as a single
capacitor, C1. And the second path can be represented as two capacitors, C2 and C3,
each for the span across the dielectric. The regions transformed into capacitors are
shown in a graphical representation in Figure 11.
Figure 11 ± Segmenting Cross-section of type flat power cable into three capacitance regions
Page 25 of 113
Essentially, Figure 12 is the result of the electric circuit transformation, where two
current paths are between the live and neutral conductors. The current path that
connects C2 and C3, can follow any path along with the floating potential along the
sheath and in the figure. In theory, the capacitances C2 and C3 have the same
capacitance because the geometry of the cable is symmetrical along the tangent
where the two insulators meet.
Figure 12 ± Simplified current paths in a cross-section of a flat power cable
3.1.3. Capacitance on a Grasped Flat Power Cable: COMSOL
modelling
COMSOL V5.3 was used to model the capacitance between Live and Neutral for
the region that is grasped. COMSOL Electrostatics solver has the capability of
measuring capacitances that lie between terminal voltages and ground. To use a
superposition type calculation to determine the three capacitor values, two
simulation geometries were conducted, where one had the sheath as a floating
potential surface and the other did not.
Capacitance C1 was determined by simulating the electrostatic field of the flat
power cable without a floating potential where the charge moves directly from the
live wire to the neutral wire. When simulating the electrostatic field of the flat
power cable with the floating potential, the resulting capacitance measured is a
combination of series connected C2 and C3; which is parallel connected to capacitor
C1. In this case, the charge has two paths to flow, the first is directly like the last
case and the second is through the floating potential. Using the capacitor arithmetic,
C1C2 C3
Page 26 of 113
Equations (1) and (2), one can deduce capacitances for C2 and C3 from the total
capacitance when floating potential is present.
௨௩ܥ ൌ ଵܥ ଶܥ ڮ ܥ
(1)
ͳ
௦௦ ௨௩ܥ
ൌ
ͳ
ଵܥ
ͳ
ଶܥ
ڮ
ͳ
ܥ
(2)
Capacitances simulated and calculated were for the hand widths between 50 mm
and 100 mm at increments of 5 mm and are displayed in Table 3. The minimum
width is based on the average for 6-year-old child [55], and the maximum is slightly
larger than the 95th percentile for a 40-year-old male [56].
TABLE 3 ± SIMULATED CAPACITANCES FOR VARIOUS HAND WIDTHS BETWEEN 50 MM AND 120 MM
Capacitance
(pF)
Simulated Measurements Calculated
Values No Floating Potential With Floating Potential
Hand width
(mm)
CL-N = C1 CL-N = C1 + 2×C2 C2= C3
50 3.65 6.15 5.00
55 4.02 6.77 5.50
60 4.38 7.38 6.00
65 4.75 8.00 6.50
70 5.11 8.61 7.00
75 5.48 9.23 7.50
80 5.84 9.84 8.00
85 6.21 10.50 8.58
90 6.58 11.10 9.04
95 6.94 11.70 9.52
100 7.30 12.30 10.00
105 7.67 12.90 10.46
110 7.93 13.30 10.74
115 8.38 14.10 11.44
120 8.76 14.80 12.08
Figure 13 is a plot of the values found in Table 3 which also contains the trend-lines
describing the linear relationship between the hand width and the Capacitances C1,
C2 and C3. The linear relationships are described in Equations (3) and (4) . The
Page 27 of 113
relative errors between the linear Equations (3) and the simulated values are
described in Table 4 , along with the corresponding impedances at 50 Hz. Most
deviation errors were below 1 % except for the event with 110 mm hand width.
ଵܥ ൌ ʹǤͻ ܨȀ݉
(3)
ଶܥ ൌ ଷܥ ൌ ͻͻǤͺ ܨȀ݉
(4)
Figure 13 ± Simulated Capacitance C1 and Calculated Capacitances C2 and C3 for Various Hand Widths
Between 50 mm And 120 mm
Page 28 of 113
TABLE 4 ± ERROR ON CAPACITANCE VALUES OBTAINED WITH LINEAR RELATIONSHIP EQUATIONS (3) AND (4)
CL-N = C1 C2 = C3
Hand width
(mm)
Capacitance
(pF)
Error %
Impedance @
50 Hz (Mȍ
Capacitance
(pF)
Error %
Impedance
(Mȍ
50 3.65 0.14% 873 4.99 0.20% 638
55 4.01 0.26% 794 5.49 0.20% 580
60 4.37 0.14% 728 5.99 0.20% 532
65 4.74 0.24% 672 6.49 0.20% 491
70 5.10 0.14% 624 6.99 0.20% 456
75 5.47 0.23% 582 7.49 0.20% 425
80 5.83 0.14% 546 7.98 0.20% 399
85 6.20 0.22% 514 8.48 1.13% 375
90 6.56 0.29% 485 8.98 0.64% 354
95 6.93 0.21% 460 9.48 0.41% 336
100 7.29 0.14% 437 9.98 0.20% 319
105 7.65 0.20% 416 10.48 0.18% 304
110 8.02 1.12% 397 10.98 2.22% 290
115 8.38 0.04% 380 11.48 0.32% 277
120 8.75 0.14% 364 11.98 0.86% 266
3.1.4. Discussion
In this sub-section, the aim was to obtain an equation that estimates the coupling
capacitance values (C2 and C3) that describe a hand grasping a flat power cable. The
coupling capacitance values were determined by comparing simulation values of
the use case that has a floating potential with a set of simulations that do not contain
floating potentials. It can be deduced that the coupling capacitances C2 and C3,
between the cable conductors and the subject grasping the cable, was between
4.99 pF and 11.98 pF (638 0ȍ and 266 0ȍ UHVSHFWLYHO\ DW +] , for hand widths
between 50 mm and 120 mm. The CC circuit model, to be used in the final model,
is shown in Figure 14.
Figure 14 ± Capacitive-Coupling Model, where magnitudes of C2 and C3 are determined using Equation (4)
Page 29 of 113
Development of the Human Body Circuit Model
Using the Freiberger model as a base, several modifications are required to morph
it into the desired model with the following characteristics:
x Multiple circuit loops to mimic galvanic current flow, and
x Ability to divide the length of the bicep (or any other measured muscle) to
place the EMG electrodes,
To achieve the above characteristics, the below modifications to the Freiberger
model were made.
3.2.1. Modification 1: Modifying Percentages of the Hand-To-Hand
Impedance
The focus of this research is on upper limb measurements. Using the hand-to-leg
impedance percentages in Table 1 yield 88.12 % for the H2H impedance using
Equation (5). This equation was derived by calculating the impedance between the
two hands of Figure 8, where the paths from the torso to the head and the torso to
the legs are considered open circuits.
The H2H impedance percentage values were recalculated to make Equation (5)
amount to 100% H2H impedance, including the skin impedance. This resulted in a
scale-up factor of 1.135 applied to each of the impedance values and the new
proportioned values are listed in Table 5 .
ܴுଶு̴ ൌ ʹ ൈ ൫ܴ௦ಽಲೝ ܴ ܴ൯ ൫ʹ ൈ ܴಹೌೝ
൯ȁȁሺʹ ൈ ܴ௧ௗ
ܴௌ௨ௗሻ
(5)
Page 30 of 113
TABLE 5 ± FACTOR ADJUSTED PERCENTAGES OF BODY AREAS MENTIONED IN FREIBERGER MODEL TO RESULT IN A
100% HAND-TO-HAND IMPEDANCE
Body Part Areas Abbreviated name Original % New %
Head to heart Head_heart 10 11.35
Heart to navel Heart_navel 1.3 1.48
Navel to Upper leg Navel_Uleg 5.1 5.79
Upper Leg ULeg 14.1 16.00
Lower Leg LLeg 32.3 36.65
Lower Arm LArm 26.4 29.96
Upper Arm UArm 10.9 12.37
Deltoid Deltoid 6.9 7.83
Shoulder to Shoulder Shoulders 6.1 6.92
Upper Arm to heart UArm_heart 9.9 11.23
Waist Waist 8.7 9.87
Skin Areas Abbreviated name Original % New %
Lower Arm skin_LArm 1.8 2.04
Upper Arm skin_UArm 3.3 3.74
3.2.2. Modification 2: Reducing original Model Scope to Hand-to-
hand Model
The Freiberger model was SRWHQWLDOO\ VXLWDEOH IRU ³IDU-field capacitive coupling´
where the person is some distance away from an EMI source and the EMI induced
currents will dissipate to the ground plane YLD WKH VXEMHFW¶V IHHW. In other words, the
current path flows from one extremity to another. The model was plotted in LTspice
and is shown in Figure 15. Terms Left_Hand1 and Right_Hand1 on the extremity
nodes refer to the right hand and left hand of the participant.
Page 31 of 113
Figure 15 ± Original Freiberger Model with human body outline.
Based on the values in Table 5, it is assumed that the path of least resistance
between the hands does not pass through the head and legs. Therefore, the first
model modification was to remove parts of the model that would have a negligible
effect, which are the head and legs. The result is shown in Figure 16.
Figure 16 ± Hand-to-hand Section of Freiberger Model
Page 32 of 113
3.2.3. Modification 3: Introducing the 2nd Pathway using Parallel-
Resistor Theory
Modification 2 simplified the original model for a H2H transmission case.
Modification 3 was done to introduce the anterior and posterior electrical pathways.
This modification was the next step in creating a closed loop between the noise
source and the measurement site. Another reason for having separated anterior and
posterior paths was because the bicep muscle is adjacent to the inner elbow and not
to the elbow bone. It should be noted that the EMG sensor reference is placed above
the elbow bone which implies that this separation is important.
The upper limb sections of the model were duplicated and placed in parallel to
connect to the UArm_heart resistors, as shown in Figure 17. Since the upper limbs
have two paths each, the resistance values of skin_LArm, LArm and UArm, need to
be doubled to maintain the 100 % hand-to-hand impedance value. The resulting
H2H impedance is still 100 %.
Figure 17 ± Modification 2 of Hand-to-hand Human Body Impedance Model
3.2.4. Modification 4: Introducing Inter-Pathway Resistances for
Multiple Galvanic Flow Paths and H2H Capacitance
This modification is to introduce resistors between the anterior and posterior paths,
thus creating multiple meshes for current to circulate. Based on the literature
presented in Section 2.3, it is assumed that current flow path preference is through
the body tissue with the least impedance, which in this modelling case is muscle
tissue. Current will enter the muscles at attachment points on the skeleton.
Page 33 of 113
As shown in Figure 18 , eight resistors crossing between the anterior and posterior
paths have been introduced. Ideally, these inter-pathway resistors placed between
the anterior and posterior paths represent higher impedance body tissue like bone
or tendon that connects opposite muscles at joints for the wrists, elbows, and
shoulders. The inter-pathway resistors are positioned to join at the existing eight
nodes on both arms of the Freiberger model. It is assumed that the current leakage
to higher impedance tissues like bone and skin are negligible, therefore, no inter-
pathway resistors partway between the upper arm or lower arm segments.
The sub-sections below describe the modifications to the inter-pathway resistances,
H2H capacitance and the extremity nodes.
Figure 18 ± Modification 3 of Hand-to-hand Human Body Impedance Model
Inter-Pathway Resistors
Assuming both arms have the same impedance composition, the four resistors,
ResSkin, ResWrist, ResElbow and ResShoulder, were added in each arm circuit to
represent the Stratum Corneum and the joints for the wrists, elbows, and shoulders,
respectively. It should be noted that the resistor skin_LArm represents the resistance
experienced by the current flow through the Stratum Corneum layer and into the
sub-Stratum Corneum layers of the skin. And ResSkin represents the resistance
across the skin surface. Based on the literature that states cortical bone is more
conductive than dry skin, it is assumed that the magnitude of ResSkin must be
greater than ResWrist.
It is known that a sizeable portion of the internal impedance for hand-to-hand
contacts resides in the extremities due to the prevalence of poorer conducting bone
Page 34 of 113
and ligament [46]. Subsequently, regions surrounded by a greater proportion of
muscles to joint tissue will have lower resistance. It is assumed that the magnitude
of ResWrist would be larger than ResElbow.
To maintain proportionality of the magnitudes of ResSkin, ResWrist, ResElbow and
ResShoulder between participants, the values were derived as a factor of the Hand-
to-hand resistance, RH2H, magnitude resulting in Equations (6) to (10) .
݊݅݇ܵݏܴ݁ ൌ ܻ ൈ ܴுଶு ሺߗሻ
(6)
ݐݏ݅ݎܹݏܴ݁ ൌ ܺ ൈ ܴுଶு ሺߗሻ
(7)
ݓܾ݈ܧݏܴ݁ ൌ ܹ ൈ ܴுଶு ሺߗሻ
(8)
ݎ݈݁݀ݑ݄ܵݏܴ݁ ൌ ܸ ൈ ܴுଶு ሺߗሻ
(9)
݊݅݇ܵݏܴ݁ ݐݏ݅ݎܹݏܴ݁ ݓܾ݈ܧݏܴ݁ ݎ݈݁݀ݑ݄ܵݏܴ݁
(10)
The above mentioned four resistance values are unknown and are determined in
Chapter 5, Section 7.2.3.
Hand-to-Hand Capacitance (CH2H_0)
Capacitors with a µCH2H¶ capacitance magnitude were placed at the ends of
skin_LArm to represent the capacitance due to the Stratum Corneum. Assuming the
VXEMHFW¶V + + VHULHV capacitance was measured and stored at the value µCH2H_0¶.
The CH2H value could be derived by converting the single capacitor magnitude to a
capacitor network of two series capacitors in parallel to another two series
capacitors, as shown in Figure 19. The result is that each capacitors magnitude is
equal to the total capacitance of the system, CH2H_0.
ுଶுܥ ൌ ͲǤͷ ൈ ுଶு̴ ȁȁ ͲǤͷܥ ൈ ுଶுబܥ ൌ ுଶு̴ܥ
Page 35 of 113
(11)
Figure 19 ± Network of two series capacitors in parallel to another two series capacitors (Labelling is
XQUHODWHG WR WKH +XPDQ %RG\ FLUFXLW PRGHO¶V HOHPHQW QXPEHULQJ
Extremity Node Modification
The nodes Left_Hand1 and Right_Hand1 in Figure 17 were split into two nodes due
to the addition of ResSkin in Figure 18. The naming was revised to L_Hand1 and
L_Hand2 for the nodes on the left hand, and R_Hand1 and R_Hand2 for the nodes
on the right hand. These nodes will be connected as the left-hand or right-hand pairs
to the Noise model developed in the CC Source Model in Section 3.1.
3.2.5. Modification 5: Dividing Bicep area of Upper Arm for
placement of Bipolar electrode nodes
This modification was to subdivide the right-hand VLGH¶V XSSHU DUP UHVLVWRU R15 in
Figure 18) into regions separated by the bipolar electrode positions. The result
shown in Figure 20 has the three resistors (R23, R24 and R25) with the values of
UArmA, UArmAB and UArmB, respectively. The nodes between them are labelled
Pad_A and Pad_B and represent the electrode position. It should be noted that the
resistance percentage values of UArmA, UArmAB and UArmB are proportional to
the length
ϭ Ϯ
ϯ ϰ
Page 36 of 113
Figure 20 ± Modification 5 of Hand-to-Hand Human Body Circuit Model
Page 37 of 113
each region occupies of UArm. The regions are described in Table 6.
TABLE 6 ± DEFINING THE REGIONS FOR UARMA, UARMAB AND UARMB
Parameter Equation or Magnitude
A Length between Pad_A and Elbow electrode (m)
B Length between Pad_B and Shoulder bone (m)
AB Length between the Pad_A and Pad_B centres
The percentage resistance values of UArmA, UArmAB and UArmB are calculated
using the mathematical equations (12) to (15). Using the assumption by Wegmüller
[32] that the internal resistance across the length of the upper arm was uniform.
This allowed the simple subdivision described in the equations below.
݄ݐ݃݊݁ܮ̴݉ݎܣܷ ൌ ܣ ܤ ሺ݉ሻ ܤܣ
(12)
ܣ݉ݎܣܷ ൌ ܣ ൈ ሺΨሻ ݄ݐ݃݊݁ܮ̴݉ݎܣȀܷ݉ݎܣܷ
(13)
ܤ݉ݎܣܷ ൌ ܤ ൈ ሺΨሻ ݄ݐ݃݊݁ܮ̴݉ݎܣȀܷ݉ݎܣܷ
(14)
ܤܣ݉ݎܣܷ ൌ ܤܣ ൈ ሺΨሻ ݄ݐ݃݊݁ܮ̴݉ݎܣȀܷ݉ݎܣܷ
(15)
3.2.6. Discussion
In this sub-section, the aim was to obtain a human body electrical circuit model that
accounts for galvanic circulation using the knowledge found in the literature. The
developed model made use of the Freiberger model as a base. However, the model
is incomplete as there are undefined resistances, that account for current circulation
between anterior and posterior paths. The human body model requires experimental
data to derive unknown resistor values. The resulting body segment percentages
after all the modifications are shown in Table 7.
Page 38 of 113
TABLE 7 ±PERCENTAGES OF BODY SEGMENT IMPEDANCES AFTER MODIFICATION 4
Body Part Areas
Abbreviated
name
New
Percentage
Lower Arm LArm 59.91
Upper Arm UArm 24.74
Deltoid Deltoid 7.83
Shoulder to Shoulder Shoulders 6.92
Upper Arm to heart UArm_heart 11.23
Skin Areas
Abbreviated
name
New
Percentage
Lower Arm skin_Larm 4.09
Upper Arm skin_Uarm 7.49
Chapter Summary
In this chapter, a basic CC-circuit model was developed for the flat, two-core power
cable using COMSOL electrostatic simulations, and the Human Body circuit model
was developed from the literature presented in Chapter 2. The CC-circuit model
ZDV GHVLJQHG WR SURYLGH FDSDFLWDQFH DFURVV WKH LQVXODWLRQ EDVHG RQ WKH VXEMHFW¶V
hand width. The relationship between the hand width and capacitance is linear and
is represented by Equation (4).
The Human Body circuit model was developed from the Freiberger model and
accounts for galvanic circulation of EMI due to a dipole source. However, the
circuit is incomplete as there are abstract resistors that represent the resistance
across the skin, and joints of the wrist, elbow, and shoulder. These abstract resistors
currently do not have specified magnitudes and will be determined by use of
experimental data.
Page 39 of 113
Chapter 4
(;3(5,0(17$7,21 5(68/76
In this chapter, the experimentation procedure and the measurement results are
processed and discussed. The data obtained will be used for populating the input
parameters into the combined model and optimising to proximate the output
parameters in Chapter 5. The sections in this chapter will present the decisions taken
on and setup of the noise source, the component selections for the measurement
sensor, and the data collection procedure used during the experiment. The
measurement results are processed and analysed, and a sample data set is presented.
Experimental Environment and Component Design
The designed experiment will need to provide the input data, required to
characterise the participant based on their externally measureable parameters. As
well as output data required to calibrate the model for improved accuracy. To re-
iterate, the conditions are:
x The subjects are non-amputees that have right-handed dominancy,
x The EMI sensor is placed RQ WKH ULJKW DUP¶V ELFHS RI WKH VXEMHFW
x The subjects are grasping a flat, two-core power cable used for lamp
applications, to induce EMI currents in the body.
x The lamp will be powered by 230 VRMS (330 VPeak), 50 Hz, and
Page 40 of 113
x The measurements are done in a controlled environment.
The environment and lamp used during the experiment are discussed in Section
4.1.1 - Environment Setup. The component selection and design of the developed
EMI sensor are discussed in Section 4.1.2 - Measurement System.
4.1.1. Environment Setup
The requirement for a controlled environment leads to the use of the School of
(OHFWULFDO DQG ,QIRUPDWLRQ (QJLQHHULQJ¶V $QHFKRLF &KDPEHU IDFLOLW\ LQ WKH
Electromagnetic Laboratory. The chamber is virtually isolated from noise sources
such as power cables and telecom signals. 7KH FKDPEHU¶V RQH Vrms, 50 Hz plug
point source will be used for powering the lamp, as well as the oscilloscope that
will be recording the measurements obtained from the EMI sensor.
As shown in Figure 21 , a table was positioned in the chamber, and the lamp and
the oscilloscope were placed on the table. The 230 Vrms, 50 Hz power source was
measured via a generic, 20:1 step down transformer with a 300mA current rating.
Figure 21 ± View of Anechoic Chamber from the entrance
Page 41 of 113
Power Supply Characteristics
The power supply was measured using the step-down transformer to isolate the
oscilloscope¶V JURXQGLQJ DQG WKH PHDVXUHment ground reference. The
measurements were multiplied by 20 to compensate for the stepped-down voltage.
The measurement sample is shown in Figure 22 , and the RMS and peak voltages
were 230.7 V, and 329.84 V, respectively.
The 230 Vrms, 50 Hz power supply sample¶V frequency spectrum, with a 1 kHz
filter, is shown in Figure 23, where one can see that harmonics at 150 Hz, 250 Hz,
350 Hz, 450 Hz, 550 Hz and 650 Hz are above 50 mVp.
Figure 22 ± Sample measurement of the 230 Vrms power supply found in the Anechoic Chamber
Figure 23 ± FFT of Sample measurement of the 230Vrms power supply found in the Anechoic Chamber
Page 42 of 113
Noise Source ± Cable and Lamp
7KH FDEOH ZDV D P ³$EHUGDUH 9 + 99+ -)´, used for electrical
devices up to 3 A. The Type P3 cables dimensions are shown in Table 8 .
TABLE 8 ± SPECIFIC REQUIREMENTS FOR FLAT, TWO-CORE CABLE
Number of
cores and
conductor
size
Thickness of
insulation
(mm)
Thickness of Sheath
(mm)
Overall
breadth limits
(mm)
Overall width
limits
(mm)
(mm2) minimum nominal minimum nominal lower upper lower upper
2 × 0.75 0.35 0.5 0.41 0.6 3.2 3.8 5.2 6.3
$Q LPSRUWDQW DVSHFW RI WKH ODPS¶V FDEOH WR note was the orientation of the Live and
Neutral wires. With the flat sides of the cable facing as shown in Figure 24, the Live
and Neutral wires were on the left and right side, respectively. The light bulb used
LQ WKH H[SHULPHQW ZDV D ³(XUROX[ * %&´ : :DUP :KLWH /(' JOREH as it is
the most common source of illumination in homes and offices due to its energy
efficiency. It is known that LED lighting contain power regulation circuitry and are
known to possess switching frequency noise, it is expected that the switching
frequency will not occupy the frequencies adjacent to 50 Hz.
Figure 24 - Orientation of electrical connections to the lamp.
Page 43 of 113
4.1.2. Measurement System
As shown in Figure 25, the EMI measurement system comprises of the Electrodes,
the PCB-mounted Voltage Followers circuit, the Oscilloscope and the DC (Direct
Current) Power Supply. The dual-rail DC power was provided by a pair of 7.2 V
batteries. 7KH UHFRUGHG GDWD ZDV FROOHFWHG XVLQJ WKH ³5,*2/ 062 =´
oscilloscope. In this subsection, the following are discussed.
x Hydrogel electrodes used and their characteristics, and
x The component selection and PCB design of the EMI sensor.
The oscilloscope will have some effect on the experiment as it is AC-powered. To
reduce the effects of the any residual uncontrolled EMI within the anechoic
chamber, a baseline measurement will required to acknowledge the ambient EMI
on the body before interacting with the lamp cable.
Figure 25 - High-level System Diagram of sensor employed
Hydrogel Electrodes used and Characteristics.
7KH K\GURJHO HOHFWURGHV REWDLQHG IRU WKH (0, PHDVXUHPHQWV ZHUH ³6.,17$&7
FS-5* (&* (OHFWURGHV´ 7KHVH HOHFWURGHV KDYH GLPHQVLRQV RI PP î
40 mm.
The series impedance of the hydrogel electrode pads was measured by placing two
electrodes face to face and measuring the impedance between the two male press-
studs using the LCR meter 7KH PHWHU UHDG ȍ DQG ȝ) IRU WKH UHVLVWDQFH
and capacitance, respectively. For simplicity, the resistance and capacitance were
Voltage
Follower
Oscilloscope
DC Power
Supply
Electrodes
Page 44 of 113
rounded RII WR ȍ DQG ȝ) UHVSHFWLYHO\ &DOFXODWLQJ D VLQJOH HOHFWURGH¶V
impedance yielded the values of ȍ DQG ȝ)
Voltage Follower Circuit
The op-amps TL061 ACDR, TL071BCD and TL081BCD were screened from the
general-purpose op-amps section of the Texas Instruments free sample website
[57]. The op-amps chosen to have the following specifications.
x a single channel, SOIC, 8-pin surface-mount packaging,
x can operate with dual-power supply, and
x has Typical Input Voltage Noise Density < 50 Q9 ¥+]
The comparison between the three op-amps is displayed in Table 9. The
TL071BCD and the TL081BCD specifications are similar in most categories.
Though the TL071BCD is more purpose-built for low noise operation, the
TL081CDR was chosen due to the shorter rise time overshooting period which
should produce a more responsive circuit. It was expected that experiments would
yield measureable harmonics of 50 Hz thus for requirement of the faster op-amp.
However, this was not the case and therefore the TL071BCD would have been the
better op-amp. It should be noted that the use of either of the listed Op-amps will
SURYLGH D 7ȍ LQSXW LPSHGDQFH ZKLFK LV ODUJHU WKDQ DQ\ NQRZQ KXPDQ ERG\
impedance.
Page 45 of 113
TABLE 9 ± COMPARISON OF VARIOUS OP-AMPS CHARACTERISTICS. GREEN FILL INDICATES BEST
PERFORMANCE AND ORANGE FILL INDICATES POOREST PERFORMANCE
Parameter TL061
ACDR [58]
TL071 BCD
[59]
TL081 BCD
[60]
Optimization Low Power Low Noise General
Input Voltage Offset Range (mV) 3 to 7.5 2 to 5 3 to 5
Input Offset Current Range (nA) 0.005 to 3 0.005 to 2 0.005 to 10
Input Bias Current Range (nA) 0.03 to 7 0.05 to 7 0.03 to 7
Typ. Peak Output Voltage Swing (V) ± (VCC - 1.5)
Large Signal Differential Voltage
Amplification (V/mV)
6 200 200
Unity Gain Bandwidth (MHz) 1 3 3
,QSXW UHVLVWDQFH ȍ 1012
Max CMRR @ 1 kHz (dB) 86 100 86
Slew Rate (V/uS) 3.5 13 13
,QSXW 9ROWDJH 1RLVH # N+] Q9 ¥+] 42 18 18
Rise-time overshoot factor ȝV 0.2 0.1 0.05
Dual Power Supply Range (V) ± 5 to ± 15
Max Supply Current per channel (mA) 0.25 2.5 2.8
Temperature Range (°C ) 0 to +70
Figure 26 shows the circuit diagram of the electrical components described in
Figure 25. The measurement circuit made use of a virtual op-amp and the settings
used in the virtual Op-amp is shown in Figure 27 and are based on the parameters
obtained for the TL081BCD datasheet. The hydrogel pads impedances were also
added in series to Voltage Follower inputs.
7R FUHDWH D IORDWLQJ UHIHUHQFH SRWHQWLDO IRU WKH PHDVXUHPHQW VHQVRU D 0ȍ
resistor was placed between the DC ground node and ground and was used to
represent the air gap between the participant and the ground plane. The value of
100 0ȍ ZDV FKRVHQ IURP D UDQJH EHWZHHQ 0ȍ DQG *ȍ ZKLFK are
respectively ground separation and power line separation impedances of a person
standing in a middle of a room obtained from Serrano et. al. [39].
Page 46 of 113
Figure 26 ± EMI Measurement System Circuit Model. VCC and VEE are powered by the lithium batteries
Figure 27 ± Virtual Op-Amp Settings
Page 47 of 113
PCB design for Voltage Followers
The PCB for the measurement sensor was designed in EAGLE (version 9.5.2) and
is shown in Figure 28. The footprint size for the resistor and capacitors were 1206.
7ZR ȝ) FHUDPLF FDSDFLWRUV ZHUH XVHG DW WKH WHUPLQDOV DV D QRQ-DC bypass path.
Figure 28 ± PCB design for the Measurement sensor
7KH WZR PP GLDPHWHU KROHV ZLWK SDGV PDUNHG µ$¶ DQG µ%¶ ZHUH FRQQHFWHG WR
the inputs of the two voltage followers. The holes mounted female press-studs and
therefore able to fasten onto the male press-studs of hydrogel electrodes. The
centres of the 5.5 mm holes were 32 mm apart so that two hydrogel pads can be
positioned side by side without overlapping.
The terminals of the batteries were connected to wires that are soldered on the pads
on the top-left. The isolated VCC and VEE for the op-amps require jumper wires to
connect them to the battery bus. The yellow lines in Figure 28 indicate the jumper
wire relationship.
The pad below E$23 was used to attach the wire that connects to the elbow electrode
to the DC ground bus.
To minimise inductive coupling noise in the sensor, a shielded seven-core cable
was used to connect the PCB to the batteries and the oscilloscope. The two
unmarked holes on the left were used to fasten down the cable to prevent strain on
WKH 3&%¶V SDGV, and thus prevent jitter from loose wires or contacts.
Page 48 of 113
Experiment Procedure
The experiment had three parts: The participant preparation, participant parameter
measurements, and the experimental measurements.
4.2.1. Preparation
7KH DUHDV RQ WKH SDUWLFLSDQW¶V ULJKW DUP ZKHUH the hydrogel electrode pads were to
be placed were cleaned using alcohol cleansing pads to provide the best adhesion
between the hydrogel electrode pads and the skin. As mentioned in the literature
review, the gentle cleaning of the skin surface reduces skin impedance.
The two hydrogel electrodes were placed on the bicep of the ULJKW DUP¶V such that
the span between the proximal electrode on the bicep and the shoulder would be
equidistant to the span between the distal electrode on the bicep and the elbow. The
reference hydrogel electrode was placed on the right elbow¶V bone.
4.2.2. Participant Parameter measurements
The participant parameter measurements comprised of taking dimensional
measurements and the H2H impedance measurements of the participants.
Dimensional Measurements
Dimensional measurements were recorded to enable calculations to be made using
Equations (12) to (15), from Section 3.2.5, for each participant. The dimensional
measurements recorded are listed below and shown in Figure 29.
x Hand-to-hand arm span,
x Hand widths (taken in the middle of the palm where the width is the widest),
x The span between Elbow electrode to Electrode A (Span A), and
x The span between Shoulder to Electrode B (Span B).
Page 49 of 113
The distance between the two electrode centres was always 32 mm due to the
distance between the press-stud connectors on the sensor PCB
Figure 29 ± Pre-Experiment measurements of Body Parameter A) Measurement of placement of the
electrodes (marked with blue dots) and arms span. Human body outline adapted from Clker-Free-Vector-
Images [29]. B) Measurement region of hand width. Palm image was available through Creative Commons
Licence [61]
Hand to Hand Impedance Measurement Setup
An H2H impedance measurement experiment was devised to estimate the
impedance that would occur when grasping the flat power cable. Like the H2H
impedance measurement system described in Chinen et. al. [62], the devised
experiment used copper tape wrapped around a 2 × 0.75 mm2, flat power cable to
make grasping surfaces for each hand. These surfaces are intended to mimic the
JUDVSLQJ VXUIDFH DUHD RI WKH SDUWLFLSDQW¶V KDQG
Each of the copper pieces joined lead wires that connected to each of the LCR meter
terminals. When the participant grasps both the copper pieces with each hand, it
would close the path between the plates and impedance can be measured by the
inductance, capacitance and resistance (LCR) meter.
Page 50 of 113
The pieces of copper tape had the dimensions of 20 mm × 150 mm. When these
pieces were wrapped, there would be a 5 mm overlap which allows soldering a
seam. Figure 30 is the hand-to-KDQG LPSHGDQFH PHDVXUHPHQW VHWXS $ ³7(&3(/
/&5 ´ /&5 PHWHU ZDV XVHG WR PHDVXUH the series resistance and series
capacitance of the participants.
It was expected that the participants will grasp the copper surfaces with a
comfortable, medium force grasp as shown in Figure 31. Similar to the literature on
hydrogel pads, closing the skin surfaces from open-air increases skin moisture
retention and thus reduces the resistance, and increases the capacitan