Experimental analysis and modelling of textile transmission line for wearable applications
The Authors
Michel Chedid, Saab Training Systems AB, Jönköping, Sweden
Ilja Belov, School of Engineering, Jönköping, University, Jönköping, Sweden
Peter Leisner, School of Engineering, Jönköping, University, Jönköping, Sweden
Acknowledgements
The work was partly funded by the Swedish Knowledge Foundation through the Industrial Research School for Electronics Design, and Vinnova through the Summit Programme.
Abstract
Purpose – The paper seeks, by means of measurement and modelling, to evaluate frequency dependent per-unit-length parameters of conductive textile transmission line (CTTL) for wearable applications and to study deterioration of these parameters when CTTL is subjected to washing.
Design/methodology/approach – The studied transmission line is made of Nickel/Copper (Ni/Cu) plated polyester ripstop fabric and is subjected to standard 60°C cycle in a commercial off-the-shelf washing machine. The per-unit-length parameters (resistance and inductance) and characteristic impedance of the line are extracted from measurements before and after washing. Using the measurement data an equivalent circuit is created to model the degradation of the line. The circuit is then integrated in a three-dimensional transmission line matrix (TLM) model of the transmission line.
Findings – Both an electrical equivalent circuit and a TLM model are developed describing the degradation of the conductive textile when washed. A severe deterioration of the electrical parameters of the line is noticed. Experimental and modelling results are in good agreement in the addressed frequency band.
Research limitations/implications – Analysis is performed for frequencies up to 10 MHz. The developed TLM model can be used to conduct parametric studies of the CTTL. To counteract the degradation of the line, protective coating is to be considered in further studies.
Originality/value – This paper extends knowledge of the subject by experimental and simulation-based characterization of the CTTL when subjected to washing cycles.
Article Type:
Research paper
Keyword(s):
Textiles; Electronics industry; Modelling; Impedance voltage.
Journal:
International Journal of Clothing Science and Technology
Volume:
19
Number:
1
Year:
2007
pp:
59-71
Copyright ©
Emerald Group Publishing Limited
ISSN:
0955-6222
Introduction
Both civilian and military markets of wearable applications are growing rapidly. Wearable electronics is designed to be body-worn while in use. It has to satisfy several design requirements: portability during operation, operation constancy, controllability, robustness and ergonomics (Chedid and Leisner, 2005). These requirements are crucial to guarantee good fitness of electronics to the human body. A wearable system has normally a modular architecture where different types of modules, e.g. sensors, radio and computer modules, are distributed on the body communicating to each other. A body area network based on conductive textile transmission lines (CTTL) that carries both direct current (DC) power and alternating current (AC) data signals, eliminates the need for traditional cables between the modules within the wearable system. This requires a low DC resistance to deliver power and known characteristic impedance in order to build matched transceivers.
In order to get a wearable system to comply with the requirements on robustness and ergonomics, the conductive textile should exhibit the following characteristics: durability, washability and flexibility. Therefore, the CTTL for networking requires an analysis of the mechanical and electrical characteristics of such material and how it reacts to environmental stresses.
Several studies have been done about conductive textile characteristics. In Cottet et al. (2003) conductive textile characteristics were studied and modelled for high frequency signal transmission. In Gorlick (1999) a separate power and data bus was developed where different modules could be connected to the bus through stingers. In Wade and Asada (2006) DC behaviour of conductive fabric networks was studied. In Jayoung et al. (2005) the effects of mechanical stresses on conductive textile were addressed at low frequencies. Robustness to washing induced stresses in combination with complex frequency dependent behaviour of the studied transmission line (TL) has not been investigated earlier.
In this research the deterioration of the electrical properties of the CTTL is investigated by means of measurement of the per-unit-length parameters when CTTL is subjected to environmental stresses in form of washing cycles. Based on experiment, an equivalent circuit model is derived to predict the deterioration of the electrical properties. The circuit is then integrated in a three-dimensional transmission line matrix (TLM) model and the results are compared to experiment.
Experimental setup
In Figure 1, the flowchart of experimental and modelling steps is presented.
The studied CTTL has a length of 1 m. It consists of two stripes of conductive textile with a low surface resistivity, sewed onto a non-conductive textile. Hardware modules are connected to the CTTL through conventional snap fasteners, see Figure 2.
The conductive textile is a Nickel/Copper (Ni/Cu) polyester ripstop from Laird Technologies. Ripstop is made of very fine polyester filaments plainly woven with coarse fibres ribbed at intervals to stop tears, see Figure 3.
To make it conductive, every filament is plated first with a thin layer of Copper followed by another layer of Nickel. The conductive layer has an approximate thickness of 1 μm. Electrical characteristics of the textile, taken from manufacturer's datasheet, are shown in Table I.
The snap fasteners are normally used in grounding products, e.g. electrostatic discharge (ESD) wrist straps. Before attaching the snaps to the textile, a highly conductive silicone based paste (volume resistivity: 0.01 Ω cm) is used to establish a reliable electrical connection between the snaps and the conductive textile. Silicone paste is chosen for its elastic properties compared to conductive silver epoxy adhesives, which were shown to be unreliable when exposed to washing. The width of the conductive stripes is 6 cm. They are folded to 2 cm before sewing onto the non-conductive textile which acts as a substrate, see Figure 3. The width of 6 cm is taken in order to achieve adequate DC resistance for power transmission. Taking the value of surface resistivity supplied by the manufacturer, the estimated value of the CTTL's DC resistance is 2.3 Ω.
Using a network analyzer (NA) HP8714ET, measurements are performed on the CTTL stretched in the air 10 cm above a lab table. The experimental testing atmospheric conditions were measured to be 50±5 per cent for the relative humidity and 22±1°C for the temperature. As a part of experimental study, the constructed CTTL is subjected to washing. A washing cycle consisted of a standard 150 min cycle in a commercial off-the-shelf washing machine (Elektro Helios TF 1002). The washing cycle included pre-washing at 40°C, main washing at 60°C, three rinsing sub-cycles and centrifuging. Washing was conducted without laundry detergent. After washing, the CTTL is dried at room temperature.
Measurement procedure
When designing a given communication system based on TL, the frequency dependent characteristic impedance of the TL determines the quality of communication. To ensure maximum power transfer between the transceivers, their input impedance must be matched to the TL characteristic impedance. The signal power at the receiver reaches a maximum when the transmitter, TL and receiver impedances are matched. A complete knowledge of the complex characteristic impedance of the TL is required in order to design an impedance matching network in the transceivers. Another reason for the measurement of the TL's characteristics is to study the degradation of the conductive textile when it is washed.
The parameters of main interest are R (resistance per-unit-length in Ω/m), L (inductance per-unit-length in H/m), G (conductance per-unit-length of the dielectric medium between conducting surfaces in S/m) and C (capacitance per-unit-length between conducting surfaces in F/m). Together, they form an RLGC model of the TL, see Figure 4.
These parameters define propagation constant γ and characteristic impedance Z C as equations 1 and 2, (Eisenstadt and Eo, 1992): Equation 1 Equation 2 where ω=2πf, and f is the frequency.
In order to estimate unknown per-unit-length parameters of the CTTL, two reflection measurements are done for two different test setups: one end of the TL is always connected to the output port of the NA while the other end of the TL is short-circuit for the first measurement and open-circuit for the second measurement, see Figure 5. When a balanced TL is measured (as it is the case for the studied CTTL), it is necessary to connect a balance-to-unbalance balun (transformer) between the instrument and the TL. A balun is required for measuring balanced TL because the NA's output terminal is unbalanced. Open/short/load calibration is done when the balun is connected to ensure accurate measurements. The measurements are done for the frequency range from 300 kHz to 10 MHz.
The measured parameters, Γ s and Γ o , are the voltage reflection coefficients for the short- and open-circuit measurement setup, respectively. Letting Z 0 denote the NA's impedance and l the length of the TL, Z C and γ can be obtained from equations (3)-(6) (Agilent Technologies, 2002): Equation 3 Equation 4 Equation 5 Equation 6 Combining equations (1), (2), (5) and (6), the four per-unit-length parameters can then be extracted as shown in equations (7)-(10) (Eisenstadt and Eo, 1992): Equation 7 Equation 8 Equation 9 Equation 10 where Re( ) and Im( ) denote the real part and imaginary part, respectively. The characteristic impedance can be approximated theoretically from the geometric configuration of the TL (Figure 6), (Gevorgian and Berg, 2001). Data provided in Table II gives a theoretical value of the characteristics impedance 194.7 Ω.
Considering the fact that the CTTL is designed for installation on a human body, the TL will be bended in different directions and therefore the impedance will vary with time. However, studying the characteristics for the TL with a rigid geometry still gives a pointer on how conductive textile degrades when subjected to washing cycles.
Equivalent circuit of the CTTL
Having performed the reflection measurements the resistance and inductance per-unit-length are extracted, see Figure 7. R wc0(f) and L wc0(f) are the extracted parameters from measurement before washing while effective parameters R wc1(f) and L wc1(f) are extracted from measurement made after washing. The large variation in the extracted inductance per-unit-length at low frequencies is due to variation from NA measurements. The capacitance per unit length (C) is found to be constant with a value of 22 pF/m. In addition, the conductance per unit length (G) at the frequencies considered approaches zero and is therefore neglected.
The extracted data for R wc0(f) and L wc0(f) in Figure 7, is used to derive the RLGC model (Figure 4) of the CTTL before washing. By examining the measurement data for frequencies up to 10 MHz, R wc0(f) is fitted with a linear function of frequency f and L wc0(f) is fitted with a nonlinear function of √f. Since, the length of the studied TL is 1 m, the extracted parameters are considered as the per-unit-length parameters. The analytical expressions for per-unit-length parameters R wc0(f) and L wc0(f) are given in equations (11) and (12), respectively: Equation 11 Equation 12 The CTTL changes characteristic behaviour due to washing. The new behaviour can be simulated by loading the RLGC model with an electrical equivalent circuit shown in Figure 8. This circuit consists of a resistor and two RC sub-circuits placed in series. The resistor R 1 represents the fact that part of the conductive layer is washed out resulting in a reduction of effective conductivity and therefore higher resistance. The capacitances C 2 and C 3 represent the cracks that have built up within the conductive layer. These cracks force the current to take alternative paths and thus further increasing the resistance of the line (R 2 + R 3). As the frequency increases the decreasing impedance of the capacitors C 2 and C 3 causes the current through R 2 and R 3 to decrease, respectively. At sufficiently high frequencies these RC sub-circuits are more or less shortcut and the only resistance seen is R + R 1. This explains the negative slope of the curve for parameter R wc1 and the positive slope of the curve for parameter L wc1 in Figure 7. The per-unit-length parameters corresponding to the original RLGC model are kept as they are to make it possible to load the line with the equivalent circuit in the subsequently created TLM model. The introduction of two RC sub-circuits proved to be sufficient to capture the CTTL's behaviour when fitting the model to the measurement data. Furthermore, these two sub-circuits represent the fact that the displacement current through the various cracks is not attributed to a specific frequency. It rather increases within a frequency band.
The loaded line can be represented as impedance X in series and admittance Y in parallel with the line conductors, see Figure 8. Admittance Y is not altered after washing due to the fact that the geometrical configuration of the CTTL and the material between the lines remain unchanged. Impedance X can be calculated from equation (13): Equation 13 The values of components R 1, R 2, R 3, C 2 and C 3 can be extracted from the measurement by solving a system of nonlinear equation (14): Equation 14 It is worth noting that parameters R wc1 and L wc1 do not stand for fundamental resistance and inductance, rather being the real part and the imaginary part of impedance X, respectively, with the imaginary part normalized to 2πf. Therefore, they are attributed as effective parameters. Taking f 1=1 MHz, f 2=5 MHz, f 3=10 MHz, the equation system is solved. The frequencies are chosen to capture boundary conditions and the dynamics in between. Using the values obtained from measurement results in the following values for the equivalent circuit in Figure 8: R 1=174.5 Ω R 2=5.8 Ω R 3=13.9 Ω C 2=17.1nF, C 3=1.1nF.
TLM model of CTTL
Constructing a TLM model requires beside the geometrical data, the electrical properties of materials used, i.e. conductivity, permittivity and permeability. Using the method proposed in (Henn, 1998) to approximate surface resistivity of conductive textiles, the effective conductivity of the textile can be calculated as follows. Let d f denote the diameter of the conductive filament and δ the thickness of the Ni/Cu layer, then the area of the metal layer, A m , can be calculated according to equation (15): Equation 15 Every yarn is composed of n filaments. Let ρ eff denote the effective resistivity of the Ni/Cu layer, then the resistance per-unit-length, r, of one yarn can be calculated according to equations (16) and (17): Equation 16 Equation 17 Having the yarn density, N, which describes the number of yarn per-unit-length of textile, the surface resistivity of the textile, r s , can be approximated as in equation (18): Equation 18 The effective conductivity, σ eff, can then be calculated according to equation (19): Equation 19 Taking into account the values given in Table I, results in the effective conductivity of the studied conductive textile, 1.50 × 106 S/m.
A commercially available electromagnetic simulation tool, Flo/EMC from Flomerics is used in TLM simulations. The TLM tool solves Maxwell's equations in time domain. The frequency domain data is then computed. The computational mesh is composed of cuboids of different size. Absorbing boundary conditions represent open faces of the computational domain. Convergence tests have been performed to determine the effect of the open boundary location on the simulation results. Convergence criterion is satisfied when initial system energy decays by at least 60 dB. The conductive textile in the CTTL is modelled as thin plates of isotropic dielectric material with fixed frequency independent relative permittivity and permeability and a fixed conductivity. These thin plates are penetrable to electromagnetic fields and the skin effect is simulated with acceptable accuracy. Wires are used to establish electrical connection between the two conducting plates. The NA is simulated by a wire port, including a voltage source and a 50 Ω impedance. For the wire port the TLM tool includes the wire current in the output, and associates with it the load impedance. A post-processing tool determines the signals entering and leaving the model through the port, resulting in the voltage reflection coefficient.
In order to properly define thin plates in TLM simulation, their equivalent thickness t eq has to be specified. It is calculated from the surface resistivity of the conductive textile provided by the manufacturer and from its effective conductivity σ eff. Every stripe in CTTL is made of three layers of conductive textile and therefore t eq is calculated according to equation (20): Equation 20 Equation (21) (Chen and Fang, 2000) approximates the resistance of a rectangular cross-sectional conductor at a given frequency: Equation 21 where μ eff and w are the effective permeability and the width of the stripe, respectively. Although this formula takes only into account the skin effect of the TL conductors, it still gives an initial approximation of the permeability of the Ni/Cu layer. Using the measured value of resistance at 10 MHz, the effective relative permeability μ r is predicted to be 350. In TLM simulations where other physical effects are included such as the proximity effect (Figure 9), μ r is adjusted to 325 to get better agreement with measurement data. The input parameters used in the TLM simulations are summarized in Table III.
As can be seen in Figure 9 the distribution of the surface current is frequency-dependent. Higher frequencies cause the surface current to flow on the innermost edges of the CTTL. This represents the proximity effect and together with the skin effect tends to cause the TL's resistance to rise with increasing frequency.
Modelling versus measurement
Voltage reflection coefficients are obtained from measurements and from TLM simulations performed for the open- and short-circuit setups (Figure 5). Effective per-unit-length parameters are then extracted according to equations (7)-(10). Results from measurement, equivalent circuit and TLM computation are shown in Figure 10. In Figure 10 (a), (c) and (e) the characteristic impedance and the extracted per-unit-length parameters (resistance and inductance) are presented for the TL before washing. Figure 10 (b), (d) and (f) presents the respective parameters after washing.
Measured characteristic impedance Z C before and after washing is in good agreement with modelling results. Z C exhibits different frequency depending behaviour: in low-frequency region Z C is frequency dependent while in high-frequency region Z C is rather constant. After washing cycle, the high-frequency region is shifted up in frequency. Furthermore, the impedance plots reveal that the theoretically predicted value for the high frequency impedance (194.7 Ω) of the CTTL matches the experimental results when it is not subjected to any washing cycle.
The resistance per-unit-length before and after washing is shown in Figure 10(c) and (d), respectively. The offset between the TLM model and measurement data before washing is due to the contact resistance that is not included in the TLM model. The curves however have the same slope. The resistance per-unit-length of the CTTL deteriorates dramatically after washing (from ∼3 Ω/m to ∼200 Ω/m at 300 kHz). The DC resistance of the TL is too high resulting in a TL that is no longer feasible for DC signals. Use of different types of protective top coatings should therefore be investigated to protect the conductive layer and enhance its robustness to washing.
The inductance per-unit-length before and after washing is shown in Figure 10(e) and (f), respectively. The difference between the TLM model data and measurements seen in the results is due to the fact that permeability of the metallic layer is modelled as constant instead of being frequency dependent. The maximum error between measurement and TLM is bounded to approximately 15 per cent. The shape of the curves shows correlation between the TLM model and measurement results.
There are several sources of uncertainty that contribute to measurement error. One major source is the ground plane that is not well defined in the measurement setup due to the fact that the CTTL is stretched 10 cm above the table. Another source of uncertainty is the snap buttons that are connected to the TL. The snap buttons add bias resistance to the measurement. While the error from the latter is small, the error from the former source is hard to predict and therefore sensitivity analysis based on TLM simulation is made and shown in Figure 11. The introduction of the ground plane leads to a relative change in extracted parameters of approximately 1 per cent, which is negligible.
Conclusion
Experimental and modelling-based study of frequency dependent (up to 10 MHz) electrical parameters of the TL made of Ni/Cu plated polyester ripstop fabric subjected to washing cycles led to the following conclusions:
- Efficient combination of equivalent circuit modelling and TLM modelling was introduced in order to study deterioration of frequency dependent per-unit-length parameters of CTTL. The developed equivalent circuit was used to load the original TLM model of the non-washed TL and thus enabling the construction of the TLM model of the washed Tl. The efficiency of such approach is confirmed by good agreement between measurement and simulation data.
- The developed approach can be of assistance during virtual prototyping in design phase of wearable applications. For example, computational experiments with different geometries that affect the characteristic impedance of the line can be conducted and thereby partly replacing real experiments.
- The studied conductive textile was found not to be appropriate in wearable application due to severe degradation of the Ni/Cu layer when washed. Different types of protective top coatings from the textile industry such as Polyurethane and Silicone coating should be investigated to improve robustness of the conductive textile to environmental stresses introduced by washing cycles.
Equation 1
Equation 2
Equation 3
Equation 4
Equation 5
Equation 6
Equation 7
Equation 8
Equation 9
Equation 10
Equation 11
Equation 12
Equation 13
Equation 14
Equation 15
Equation 16
Equation 17
Equation 18
Equation 19
Equation 20
Equation 21
Figure 1Flowchart of the experimental and modelling steps
Figure 2Experimental sample of the textile TL
Figure 3Architecture of the CTTL (left) and a scanning electron microscope picture of the conductive textile (right)
Figure 4RLGC model of a transmission line
Figure 5Short- and open-circuit measurement setup for the textile TL
Figure 6Schematic of the CTTL cross-section
Figure 7Extracted per-unit-length parameters R
wc0, R
wc1 (left) and L
wc0, L
wc1 (right) before and after washing
Figure 8CTTL model after washing cycle (left) and a representative simplified circuit (right)
Figure 9Surface current distribution in the CTTL TLM model at 300 kHz (left) and at 10 MHz (right)
Figure 10Comparison between the measured and modelled characteristic impedance and per-unit length parameters for CTTL before and after washing
Figure 11Sensitivity of simulation data to presence of ground plane
Table INi/Cu polyester ripstop characteristics
Table IIDimensions of the CTTL
Table IIITLM model input parameters
References
Agilent Technologies, Co. Ltd (2002), Agilent Technology Impedance Measurement Handbook, 2nd ed., Agilent Technologies, Co. Ltd, St Louis, MO, .
Chedid, M., Leisner, P. (2005), "Roadmap: wearable computers", Proceedings of IMAPS Nordic Conference, 11-14 September, Tønsberg, Norway, pp.146-54.
Chen, H., Fang, J. (2000), "Modelling of impedance of rectangular cross-section conductors", Proceedings of IEEE Conference on Electrical Performance of Electronic Packaging, 23-25 October, Scottsdale, USA, pp.159-62.
Cottet, D., Grzyb, J., Kirsten, T., Tröster, G. (2003), "Electrical characterization of textile transmission lines", IEEE Transactions on Advanced Packaging, Vol. 26 No.2, pp.182-90.
Eisenstadt, W.R., Eo, Y. (1992), "S-parameter-based IC interconnect transmission line characterization", IEEE Transactions on Components, Hybrids and Manufacturing Technology, Vol. 15 pp.483-90.
Gevorgian, S., Berg, H. (2001), "Line capacitance and impedance of coplanar-strip waveguides on substrates with multiple dielectric layers", Proceedings of European Microwave Conference, 25-27 September, London, UK, pp.153-6.
Gorlick, M. (1999), "Electric suspenders: a fabric power bus and data network for wearable digital devices", Proceedings of the Third International Symposium on Wearable Computers, 18-19 October, San Francisco, CA, USA, pp.114-21.
Henn, A.R. (1998), "Calculating the surface resistivity of conductive fabric", Interference Technology Magazine ITEM Update, available at: www.rbitem.com, .
Jayoung, C., Jihye, M., Gilsoo, C., Keesam, J. (2005), "An exploration of electrolessly Cu/Ni plated polyester fabrics as e-textiles", Proceedings of the Ninth International Symposium on Wearable Computers, 18-21 October, Osaka, Japan, pp.206-7.
Wade, E., Asada, H.H. (2006), "DC behavior of conductive fabric networks with application to wearable sensor nodes", paper presented at International Workshop on Wearable and Implantable Body Sensor Networks, 3-5 April, Cambridge, pp.27-30.
Corresponding author
Michel Chedid can be contacted at: michel.chedid@sts.saab.se