Analysis and Modeling of Coplanar Strip Line

Size: px

Start display at page:

Download "Analysis and Modeling of Coplanar Strip Line"

Gyles Howard
5 years ago
Views:

1 Chapter - 6 Analysis and Modeling of Coplanar Strip Line 6. Introduction Coplanar Strip-lines (CPS are used extensively in the MMIC and integrated optics applications. The CPS is a alanced transmission line and makes more efficient use of the wafer area than CPW. The CPS is used for the design of filters, antennas, mixers and opto-electronic devices. It has een adopted to the non-planar surfaces also. In this context, the elliptical coplanar strip lines (ECPS and cylindrical coplanar strip lines (CCPS have een examined. The quasi TEM parameters for ECPS and CCPS have een analyzed and analytical expressions for the losses of the planar symmetrical CPS are reported [48, 93-94, 7, 43, 54]. The losses and dispersion of these non-planar CPS are not reported. The full wave methods like SDA [0], variational method [53] and mode matching [34] have een used to otain the frequency dependent characteristic impedance and effective relative permittivity of the CPS line. In this chapter, we present the modeling of conductor thickness and frequency dependent line parameters, dispersion and losses, of different configurations of CPS. All developed models are verified against availale experimental results and the results otained using the EM- simulators over wide range of parameters. The effect of strip asymmetry on the characteristics of CPS is also analyzed. All the developed models are further extended to the multilayered planar and non-planar (elliptical and circular cylindrical CPS. Finally a circuit model, applicale to these structures, is also presented. It accounts for the effect of finite strip conductivity on the propagation constant and also on the characteristic impedance at lower range of microwave frequency. 69

2 6. Effect of Conductor Thickness on Characteristics of CPS The symmetrical coplanar strip (CPS line, shown in Fig.(6., consists of two finite widths w metallic strips separated y a slot-gap of width s, sustrate thickness h and conductor thickness t with relative permittivity ε r. The conductor thickness independent static expressions to compute the effective relative permittivity ε eff and the characteristic impedance Z 0 of the CPS with finite sustrate thickness are otained, using the conformal mapping method, y Ghione and Naldi [45]. The expressions are summarized elow: Fig.(6.: Cross section of planar CPS on finite thickness sustrate. ( ε ' r K( k0 K( k6 ε eff = (6. ' K( k K( k6 0 0π K( k Z 0 0 = (6. ε ' eff K( k0 where, modulus k 0 and k ' 6 along with their complementary modulus k 0 and ' k 6 are defined in equation-(3. and equation-(3.69 respectively. On modifying these expressions empirically to take into account the finite strip conductor thickness t [73], the equivalent slot-width s eq is reduced and the equivalent strip-width w eq is enlarged. s eq = s w ( i w eq = w w ( ii t w = πε r 8π s ln t ( iii (6.3 70

3 The modified aspect ratio k0, t and, t k 6 are given elow: seq ' k0,t = ( i k0,t = k 0, t ( ii (6.4 seq weq πseq tanh 4h k 6,t = ( i k 6,t ' = k 6, t π tanh ( seq weq 4h ( ii (6.5 The effective relative permittivity of the CPS with conductor thickness is computed from the following empirical relation [73]. 4[ εeff (t = 0 ]t / s ε eff (t = εeff (t = 0 (6.6 ' [ K(k / K(k ]. 4(t / s 0 0 The characteristic impedance of the CPW with conductor thickness t is computed from the equation-(6., using conductor thickness dependent effective relative permittivity and modified aspect ratio: Z 0 ( t = 0π ε K( k 0,t ( t ' eff K( k0,t (6.7 Both εeff and Z0 decrease with increasing conductor thickness. 6.3 Dispersion in CPS In this section, the dispersion expression for the CPS due to Frankel et. al. [9] is used to include the finite conductivity and finite conductor thickness in propagation characteristics. Like CPW, we have incorporated skin-depth dependent factor (S with 7

4 static effective relative permittivity as well. This factor will account for the dispersion at low frequency due to the field penetration in the finite conductivity strip conductors Effective Relative Permittivity The dispersion empirical expression for the CPS due to Frankel et. al. [9] is modified to include the conductor thickness up to mm-wave range, using the same methodology of formulation employed for CPW in Chapter-5. The modified closed-form expressions for CPS dispersion with skin-depth dependent factor (S and finite strip thickness is summarized elow: ε eff ( f,t, δ = S ε eff ( f = 0,t ' K( k K( k 0, δ 0,t where S = ' K( k K( k,t 0, δ 0 [ ε S ε ( f = 0,t ] r eff r m( f / fte ( a ( (6.8 where r.8. The cut-off frequency of the lowest order TE mode is given y equation- (5. and the parameter m is otained from equation-(5.. Fig.(6.a shows characteristics of ε eff in CPS on different sustrates with ε r =.5, 9.8,.9 and 37; s = 8 µm, w = 4 µm, h = 670 µm and t = 5 µm. The dispersion expression for the CPS indicates that the dispersion is very small up to hundred GHz for the structure on GaAs ( ε = 3. and Sapphire ( ε = 0. 5 sustrate having thickness 500 µm. When r r compared against the EM-simulators in the frequency range 0. GHz 00 GHz, the model has 3.6% average deviation for ε r = 37. At low frequency, 0. GHz, the results of CST are much higher than others for all the sustrates. The model does account for the dispersion due to the skin-effect after incorporating the factor S and shows small peak at the lowest end of the frequency. Overall, the model closely follows the results of CST. 7

5 (a ( (c (d Fig.(6.: Comparisons of ε eff ( f, t computed y closed-form model against EM-simulators and measurement data as a function of: (a & ( Frequency, (c Conductor thickness, and (d s/(sw ratio for CPS on various sustrates. The experimental results of Frankel et. al. [9] over frequency range 5 GHz -00 GHz for ε r = 0.5 are also shown in Fig.(6.a. Fig.(6. shows comparison of the phase constant β computed y the model and the EM-simulators against experimental results of Kiziloglu et. al. [74] over frequency range GHz-8 GHz. The CPS is taken on 670 µm thick GaAs ( ε r =.9 sustrate with t = 0.5 µm, 3 µm and 9 µm; w = 4.5 µm and s =7 µm. With an increase in conductor thickness, there is decrease in effective relative 73

6 permittivity which has resulted in susequent decrease in phase constant. The results of the present model follow the experimental results, as well as the results of Sonnet and HFSS faithfully. The % average and % maximum deviation in the results otained from the comparison against HFSS results are summarized in Tale-6.a. For. 5 ε r 37, 0.5 µm t 9 µm and 0. s / ( s w 0. 6, the closed-form model has % average and % maximum deviation of (.9%, 5.4%, (.5%, 6.5% and (4.7%, 7.% against HFSS, Sonnet and CST respectively, as shown in Fig.(6.c and Fig.(6.d Characteristic Impedance Some authors have used the voltage-power definition for CPS [0], whereas others have used the current-power definition [53]. The voltage-current definition for the characteristic impedance of a CPS line monotonically decreases with increasing frequency. However on using power-current definition, it shows initial decrease and then increase in characteristic impedance with increasing frequency. The dispersion in the characteristic impedance of the CPS could e estimated from the following voltagecurrent definition ased expression: Z 0 ( f,t = 0π S ε ( f,t eff ( a where S K( k = K( k ' 0, δ δ 0, K( k K( k 0,t ' 0,t ( (6.9 The aove expression accounts for the low frequency dispersion in characteristic impedance due to field penetration. It causes increase in characteristic impedance at low frequency due to increase in the internal inductance. Fig.(6.3a shows the dispersion in Z 0 of CPS on different sustrates with ε r =.5, 9.8,.9 and 37; s = 8 µm, w = 4 µm, h = 670 µm and t = 5 µm in the frequency range 0.GHz 00 GHz. 74

7 (a ( (c (d Fig.(6.3: Comparisons of Z 0 ( f, t computed y the closed-form model, EM-simulators against the measurement data as a function of: (a & ( Frequency, (c Conductor thickness, and (d s/(sw ratio for CPS on various sustrates. The closed-form model is within 3.4% average deviation against the results otained from EM-simulators till 70 GHz and eyond that the results of the model deviate from that of simulators with 5.4% average deviation. It is due to different definitions of the characteristic impedance adapted in the model and in the EM-simulators. Fig. (6.3 shows the results on characteristic impedances of the planar CPS for conductor thicknesses: t = 0.5 µm, 3 µm and 9 µm against experimental results of Kiziloglu et. al. [74]. The characteristic impedance decreases with increase in conductor thickness. There 75

8 are large deviations in the results of the model at the lower end of the frequency w.r.t. the experimental and EM-simulated results. This could e improved y the circuit model presented in section-6.8. The average and maximum % deviation of the present model and Sonnet are summarized in Tale-6.. For. 5 ε r 37, 0.5 µm t 9 µm and ( s w s /., the closed-form model has % average and % maximum deviation of (3.%, 6.6%, (3.%, 5.8% and (.%, 5.3% against HFSS, Sonnet and CST respectively, as shown in Fig.(6.3c and Fig.(6.3d. The CPS line provides much higher value of Z 0 as compared to the CPW and with increase in aspect ratio (s/(sw, Z 0 of the CPS increases while that of the CPW decreases. Tale- 6.: % Average and maximum deviation of closed-form model, Sonnet and experimental results [74] against HFSS for CPS. [Data range: t = 0.5 µm -9 µm; Freq = GHz - 60 GHz; ε r =.9; tan δ = 0.00; σ=4. x 0 7 S/m; h=670 µm] a Effective relative permittivity Models t = 0.5µm t = 0.5µm t = 3µm t = 6µm t = 9µm Av. Max. Av. Max. Av. Max. Av. Max. Av. Max. Kiziloglu [74] Sonnet Closed-form Model Characteristic impedance Models t = 0.5µm t = 0.5µm t = 3µm t = 6µm t = 9µm Av. Max. Av. Max. Av. Max. Av. Max. Av. Max. Kiziloglu [74] Sonnet Closed-form Model

9 In summary we can say that changes due to conductor thickness in the characteristic impedance, especially in the higher frequency range are more significant as compared to the changes in the effective relative permittivity. Thus for the variation in the normalized strip thickness from t/h = 0 to 0., ε eff is changed y 6%, whereas the Z 0 is changed y 9% and % at h/λ = 0.0 and h/λ = 0. respectively. The effect is more significant for the small strip-slot ratio CPS on the low permittivity sustrate. 6.4 Computation of Losses in CPS In this section, we present the improved closed-form models for computation of the dielectric loss and the conductor loss of the CPS, suitale for the microwave CAD purpose Dielectric Loss The following standard expression is used to compute the dielectric loss of the CPS structures [46]: α ε f t r ε eff (, tanδ = 7.9 db unit length (6.0 ε eff ( f, t ε r λ 0 d / where, λ 0, tan δ and ε r are free-space wavelength, loss tangent and relative permittivity of the sustrate respectively. The frequency and conductor thickness dependent ε ( f, t of CPS is computed using equation-(6.8. Fig.(6.4 compares the computation of eff α d y the closed-form model 77

10 Fig.(6.4: Dielectric loss as a function of conductor thickness for CPS. against the EM-simulators w.r.t. conductor thickness and the model is within 0.4% average deviation. It is oserved that for the same loss tangent and the same CPS structure, the dielectric loss increases with increase in the relative permittivity of the sustrate. Again it is due to more field confinement in the dielectric region for the high permittivity sustrate Conductor Loss The conductor loss of the CPS is computed using two closed-form models: (i Wheeler s incremental inductance formulation (ii Improved Holloway and Kuester model (IHK The models are compared and validated against the results from experimental data and EM-simulators for wide range of data. 78

Wheeler s Incremental Inductance Formulation We have extended Wheeler s incremental inductance formulation [3] to CPS for conductor loss computation.

λ0 Z 0( weq,seq,h, f,t, ε r = where, λ 0 is free space wavelength. The expressions for the ε and ( w,s,h, f,t parameters (,w,s,h, f,t eff ε r (6.8 and equation-(6.9 respectively.

11 Wheeler s Incremental Inductance Formulation We have extended Wheeler s incremental inductance formulation [3] to CPS for conductor loss computation. Therefore, the conductor loss of a CPS is computed from the following expression: π Z( ε r =,w,s,h, f,t, δ s α c = ε eff ( ε r, w, s,h, f,t Np/m (6. λ0 Z 0( weq,seq,h, f,t, ε r = where, λ 0 is free space wavelength. The expressions for the ε and ( w,s,h, f,t parameters (,w,s,h, f,t eff ε r (6.8 and equation-(6.9 respectively. Z0 eq eq, ε r = are otained from equation- (a Decrease in slot-width due to finite conductor thickness ( Increase in slot-width due to field penetration Fig.(6.5: Application of Wheeler s inductance rule to compute conductor loss in CPS. Fig.(6.5a shows that the slot-width of CPS is reduced due to the finite conductor thickness. The change in width due to strip conductor thickness is calculated y using equation-(6.3. Fig. (6.5 shows the skin-depth increases the slot-width y δ s due to the field penetration all around the strip conductor. We note that due to the field penetration in the conductor y δ s all around, the strip width has decreased y δ s and the slot width has increased y δ s. Likewise, the sustrate height also increases yδ s. We further note 79

12 that due to the skin-depth, the conductor thickness is also decreased yδ s. Thus the field penetration modifies equation (6.3 as follows, t' 8π s w' = ln πε t' r (a where, t' = t δ s ( (6. The difference characteristic impedance, Z is given y: of a CPS with and without field penetration Z ( ε =,w,s,h, f,t, δ r s δ s = Z0 w w' δ s,s w' δ s,h Z0 ( weq,seq,h, f,t, ε r =, f,t, ε r = (6.3 The characteristic impedances for the parameters shown in equation- (6.3 are computed from the model discussed in previous sections. Improved Holloway and Kuester model The second closed-form model, for computation of conductor loss of the CPS, is ased on the standard perturation method [38],[05] along with the concept of the stopping distance [], [5]. Fig.(6.6: Infinitely thin CPS conductors with stopping distance (. 80

13 8 The perturation expression to compute the conductor loss of the planar shown in Fig.(6.6 is given as dl I J,t f ( Z R sm c 0 α (6.4 where, Z 0 (f, t is the conductor thickness and frequency dependent characteristic impedance of the CPS structure, otained from equation-(6.9. The surface resistance R sm of the strip conductor of finite thickness t is computed from equation-(5.8. In CPS, the current density (J over the strip conductors is defined y the following function [7] ( ' (,, ( s w k K I A where a x s w s x A J = = (6.5 The ratio of current density (J to the longitudinal current (I is integrated along whole of the strip as follows = dl I J dl I J s w s (6.6 The stopping distance is computed using equations-(5.4-(5.5, for the rectangular and trapezoidal strip cross-sections [4]. The final expression for the conductor loss of the CPS structure is ( ( ( = α s w w s w ln s w s w w s ln s s ( w ' ( k K,t f ( wz s w R sm c (6.7

14 The total loss of the CPS structures can e computed y adding the conductor loss and dielectric loss together, i.e. α = α α db / unit length (6.8 T c d Fig.(6.7: Comparison of total loss in CPS against results of Kiziloglu et. al. [74]. Fig.(6.7 shows the comparison of the total loss computed y the closed-form model # (= Wheeler α d, closed-form model # (= IHK α d, two EM-simulators - Sonnet and HFSS against the experimental results [74]. The loss has een computed for strip conductor thickness t = 0.5 µm 9 µm. The CPS is on the sustrate with ε r =.9, h = 670 µm, w = 4.5 µm, s = 7 µm. The loss is computed over the frequency range GHz - 60 GHz. The experimental results are availale up to 8 GHz only. The closed-form model # eing ased on Wheeler s incremental inductance model, is only applicale for the conductor thickness, t. δ. The closed-form model # comfortaly computes for s oth thick ( t. δ s and thin ( t <. δ s strip conductors. The computed conductor loss, using Wheeler s model, is less for thicker conductor. Tale-6. presents the % 8

15 average and maximum deviation in the results otained from experiment, Sonnet and the closed-form models. The results of HFSS are taken as the reference. Tale- 6.: % Average and maximum deviation of closed-form models, Sonnet and experimental results [74] against HFSS for CPS. [Data range: t = 0.5 µm -9 µm; Freq = GHz - 60 GHz; ε r =.9; tanδ = 0.00; σ=4. x 0 7 S/m; h=670 µm] Models t = 0.5µm t = 0.5µm t = 3µm t = 6µm t = 9µm Av. Max. Av. Max. Av. Max. Av. Max. Av. Max. Kiziloglu [74] Sonnet Closed-form Model# Closed-form Model# The effect of frequency on the characteristics of αt in CPS, on different sustrates with ε r = 3.78, 9.8 and 0; s = 8 µm, w = 4 µm, h = 670 µm, tan δ = 0.00 and t = 5 µm, is shown in Fig.(6.8a. The closed-form model # is used here for the computation of and compared against the EM- simulators with 4.8% average deviation till 60 GHz and eyond that it increases upto 8.9%. Fig.(6.8 shows the comparison of the computed α T y the model against the EM-simulators as a function of s/(sw ratio with 6.8% average deviation. Fig.(6.8c shows the effect of frequency on the characteristics of Q u in CPS on different sustrates with ε r = 3.78, 9.8 and 0; s = 8 µm, w = 4 µm, h = 670 µm, tan δ = 0.00 and t = 5 µm in the frequency range 0. GHz 00 GHz with 6.9% average deviation till 5 GHz. Fig.(6.8d compares computation of Q u for ε r = 3.78 and 0; f = 40 GHz as a function of s/(sw ratio with % average and % maximum deviation 0. s / s w.. of 8.4% and 5% respectively for the range ( 0 6 α T 83

16 (a ( (c (d Fig.(6.8: Total loss, as a function of (a Frequency and ( s/(sw ratio and Q factor, as function of (c Frequency and (d s/(sw ratio for CPS on various sustrates. 6.5 Closed-form Dispersion and Loss Models for Multilayer CPS The cross-sectional view of a CPS on multilayer dielectric sustrates is shown in Fig.(

Fig.(6.9: Multilayered CPS structure. The conformal mapping technique proposed y Ghione et.al.[45] to analyze CPS has poor accuracy for large inter-strip spacing and sustrate thickness, which are often used in modern microwave circuits.

17 Fig.(6.9: Multilayered CPS structure. The conformal mapping technique proposed y Ghione et.al.[45] to analyze CPS has poor accuracy for large inter-strip spacing and sustrate thickness, which are often used in modern microwave circuits. The new corrected transformations are suggested y Gevorgian et.al. [] for analysis of multilayer sustrate CPS, whose results are consistent with experiments and numerical simulations as well. The method employed is generalized enough to handle any numer of lower or upper layers efficiently, ut for the sake of convenience, we discuss and present results for three-layer CPS shown in Fig.(6.9. For multilayer CPS, ε eff and Z 0 are given y: ε eff ε r ε r = K( k0 K( k' ε r K( k0 K( k' ε r3 K( k0 K( k3' K( k0' K( k K( k0' K( k K( k0' K( k3 0π K( k0 Z 0 = ( (6.9 ε eff K( k0' (a where modulus k i (i=,,3 and k 0 along with their complements can e computed using equation-(3.0 and (3. respectively. The sustrate thickness h in equation (3.0 will e replaced y h i (i=,,3 accordingly. Then we have extended the conductor thickness and frequency ased closed-form models for ε eff ( f, t, Z 0 ( f,t, αc and α d of single-layered CPS to multilayer CPS y first converting multilayer into single-layer of finite dielectric thickness using SLR: 85

18 Effective relative permittivity ε eff ( f,t, δ = S ε eff ( f = 0,t K( k where Skin effect factor, S = K( k Characteristic impedance [ ε S ε ( f = 0,t ] m( f / f δ K( k K( k, δ ' 0, 0 req 0,t ' 0,t eff TE r (a 0π Z0 ( f,t = S ( ε ( f,t eff Wheeler s incremental inductance rule 7. 9 Z( ε req =,w,s,heq, f,t, δ s α c = ε eff ( ε req,w, s,heq, f,t db/m (c λ Z ( w,s,h, f,t, ε 0 0 eq eq eq req = Improved Holloway and Kuester (IHK α c R = sm 6wZ0( f,t K s w ln s w s ( k0' ( w s ( w s w ln ( w s w s ( w s Np/m (d Dielectric loss α d ε req ε eff ( f,t tanδeq = 7. 9 db/m (e (6.0 εeff ( f,t ε λ req 0 where heq is the total sustrate thickness etween strip conductors and ottom layer of the multilayer sustrate. ε req and tan δeq of the equivalent single-layer sustrate CPS are otained from equation-(

19 (a ( (c (d Fig.(6.0: Multilayer CPS: (a Effective relative permittivity, ( Characteristic impedance, (c Conductor loss, and (d Dielectric loss. Fig. (6.0a (6.0d shows comparison and validity of SLR-ased computed line parameters of multilayer CPS against EM-simulators for frequency range GHz 60 GHz. The computed ε eff and Z 0 y the model has % average and % maximum deviation of (0.7%,.8% and (.%, 6.5% respectively against oth the EM simulators, as shown in Fig.(6.0a and (

20 In Fig.(6.0c and (6.0d, variation in α c and α d of multilayer CPS for ε r= 0 and.9. The % average and % maximum deviation in IHK and Wheeler are (4.4%, 6.8% and (7.8%, 6.% respectively against HFSS and Sonnet. It is oserved from Fig.(6.0c, that the conductor loss decreases with increasing ε r, which is explained y the fact that the field is pulled out from the lossy sustrate into the high permittivity layer, where the losses are smaller []. The computed α d y the model is in close agreement with HFSS and MoM ased LINPAR, for frequency range GHz 60 GHz, with % average and % maximum deviation of.8% and.5% respectively, as shown in Fig.(6.0d. 6.6 Effect of Asymmetry in Characteristics of CPS An asymmetric coplanar strip (ACPS waveguide is flexile in practical device design. Along with small sizes, the impedance, effective dielectric constant and microwave refractive index can e adjusted y changing the width of one of the strips while keeping the width of the other strip and of the gap fixed. In Fig.(6., ACPS has sustrate thickness h with permittivity ε r, w is the width of the strip conductor, s is the space etween signal and ground conductors and w is the width of ground conductor. Fig.(6.: Structure of asymmetrical CPS (ACPS In our study, we have used the conductor thickness independent static expressions for ε eff and Z 0 of the ACPS otained from equations-(3.78 and (3.79 respectively: 88

21 ε eff K( k = ( ε r K( k 7 ' 7 K( k K( k ' 8 8 (a 60π K( k7 Z0 = ( (6. ε ' eff K( k7 ' where, modulus k 7 and k8 along with their complementary modulus k 7 and ' k 8 are defined in equations-(3.75 and (3.80 respectively. However, these expressions are modified empirically to take into account the finite strip conductor thickness and the skin-depth penetration using equation-(6.8 and (6.9. Fig. (6.a and (6. compare ε eff and Z 0 computed y the closed-form model on the sustrate with ε r =.9, t = 3 µm, w = 0 µm and h = 635 µm against HFSS and Sonnet for three cases of asymmetry i.e. w /w =, and 4 for frequency range GHz - 60 GHz. The computation of ε eff and Z 0 y the model has % average and % maximum deviations of (.%, 4.8% and (.4%, 5.6% respectively against oth the EM simulators. The dielectric loss of ACPS is computed using equation-(6.0 in which the frequency and conductor thickness dependent ε ( f, t is computed using equation-(6.8. On eff accounting asymmetry in Wheeler s incremental inductance formulation given y equation-(6., the expression ecomes: π Z( ε r =,w,w,s,h, f,t, δ s α c = ε eff ( ε r,w,w,s,h, f,t Np/m (6. λ Z ( w,w,s,h, f,t, ε 0 0 eq eq eq r = Holloway and Kuester have not accounted asymmetry in conductor loss computation of CPS [5]. The improved Holloway and Kuester (IHK closed-form expression for computation of the conductor loss of ACPS is otained y modifying equation-(

22 (a ( Fig.(6.: Comparison of (a Effective relative permittivity, and ( Characteristic impedance of ACPS with three cases of asymmetry for ε r =.9 and t = 3 µm. Due to asymmetry, the longitudinal current I and the current density J over the strip conductors from equation-(6.5, is defined y the following function: J = where, A x < a ( a J = ( x a ( x ( x ( a x I A = 4K( k ( a( a ( c A y > ( (6.3 s where a = ; = w a ; = w a ; =. The ratio of current density (J to the longitudinal current (I can e integrated along whole of the strip as follows: J I dl ( a J J = dl ( dl (6.4 I I strip a strip 90

23 9 On integrating to the whole range, the final expression for the conductor loss of the ACPS structure is α a a ln ln a a a ln a a a a a ln ln a a a ln a a a (k' K,t f ( Z R sm c 0 3 (6.5 Fig.(6.3 compares the results for computed T α of ACPS y closed-form model #, closed-form model #, Sonnet and LINPAR against experimental results of Garg et.al. [93]. The ACPS is considered on a p- type sustrate with sustrate thickness h = 50µm. The Al metal layer of thickness µm lie on top of SiO uffer layer of thickness 7.4 µm that is deposited on p-sustrate. The strip conductor has conductivity 3.7 x 0 7 S/m. The relative dielectric constant of SiO is 3.9 and tan δ = The operating frequency range is 0. GHz - 50 GHz. The effective relative permittivity for multilayer ACPS structure realized with the sustrate stack has een calculated from equations-(6.9a and (6.0a. Fig.(6.3a - Fig.(6.3c compare the results for three cases targeting approximately the same characteristic impedance i.e. Z 0 =50Ω: (a w = 5 µm, s = 5 µm, ( w = 0 µm, s = 0 µm, and (c w = 30 µm, s = 30 µm. In this case w was fixed at twice that of w. The narrow line has more series resistance and hence more resistive loss is oserved at low frequencies. At high frequencies, the wider line exhiits more dielectric loss due to increased spreading out of the electromagnetic field into the sustrate. The closed-form models show etter agreement with the results from experiment and Sonnet except at narrow strip conductor and narrow gap for lower frequency range. The LINPAR has more deviation from the experimental results all over the range. The line widths of around 0 µm give the optimum comination of low and high frequency loss over the andwidth of the distriuted mixer [93].

24 Fig.(6.3d further shows the line characteristic impedance dependent αt at 30 GHz as computed y oth the closed-form models and HFSS. The loss increases with increase in characteristic impedance and line aspect ratio. Outcome of the overall comparison in terms of % average and % maximum deviation is summarized in Tale-6.3. (a w =5µm, s=5µm ( w =0µm, s=0µm (c w =30µm, s=30µm (d f =30GHz, s =30µm Fig.(6.3: Total loss of ACPS as function of: (a-(c Frequency, and (d Line characteristic impedance. 9

25 Tale 6.3: % Average and Maximum deviation of models against Garg et. al. [93] [Data range: t = µm; f = 0. GHz - 50 GHz; ε r =3.9; tanδ = 0.00; σ = 3.7 x 0 7 S/m] % Deviation Closed-form Model # Closed-form Model # HFSS Sonnet LINPAR Average Maximum Closed-form Dispersion and Loss Models for Non-Planar CPS The elliptical coplanar strip lines (ECPS can e used as adapters and slotlines as well as antennas [7-43]. The closed-form expressions for dispersion and losses presented in the previous sections for planar CPS could e adopted to the CPS with finite strip conductor thickness on the circular and elliptical cylindrical surfaces. In this section we will present the conductor thickness and frequency dependent closed-form models of line parameters for oth single-layered and multilayer non-planar CPS, shown in Fig.(6.4 and ( Single-Layer Case Fig.(6.4 presents the CPS on the elliptical and circular cylindrical surfaces with two different finite ground plane widths. The structural parameters of SC - transformed ECPS/ CCPS into the corresponding planar CPS are given y equation-(3.84: a a3 3 s = Ψ ( i w = θ (ii h = ln (iii t = ln (iv (6.6 a a and the detailed definition of the aove mentioned parameters are given in Chapter-3. 93

26 In our study, we have used the conductor thickness and frequency independent expressions for ε eff and Z 0 of the non-planar CPS otained from equations-(3.8 and (3.83 respectively: ε eff ε r K( k0 K( k' = K( k0' K( k (a Z 0 0π K( k0 = ( (6.7 ε eff K( k0' where, modulus k 0 and k ' ' along with their complementary modulus k 0 and k for finite ground plane widths ( π θ ψ and ( π θ ψ, are otained from equations-(3.4 - (3.44 and equations-(3.55-(3.56 respectively after replacing θ with θ ψ. (a ( (c (d Fig.(6.4: CPS on the curved surfaces: (a Elliptical CPS (ECPS,( Circular Cylindrical CPS (CCPS, (c Semi-ellipsoidal CPS (SECPS and (d Semi- circular cylindrical CPS ( SCCPS. 94

27 In order to account the strip conductor thickness for these line structures, we have to modify equation- (6.3 - (6.7 y using equation-(6.6. The equivalent strip width ( θ eq and equivalent slot width ( ψ for the ECPS / CCPS are given as eq ψ = ψ θ eq ( i θ = θ θ eq ( ii where, θ = πε r a ln a ln πψ a 3 3 ln a ( iii (6.8 and the modulus k 0 and k will e modified into k0, t and k, t along with their complementary modulus, in which ψ and θ will e replaced y ψ eq and θ eq respectively. The aove equations - (6.8 is inserted in equations - (6.8 and (6.9 to compute the conductor thickness and frequency ased ε ( f, t and ( f,t of the ECPS and CCPS lines. eff Z 0 The computation of αd is done using equation - (6.0 in which the frequency and conductor thickness dependent ε eff ( f, t of ECPS and CCPS is used. For α c computation of non-planar CPS, firstly Wheeler s incremental inductance formulation is modified using equation-(6.6: αc = π λ 0 a ε eff ε r, θ, ψ,ln a, f,ln a Z ε r =, θ, ψ,ln a3 3 a a a Z 0 θ eq, ψ eq,ln a, f,ln a3 3, δ s a Np/m a3 3, ε = r a, f,ln (6.9 95

28 Then IHK model for ECPS/CCPS with ground plane width ( π θ is otained y using equation (6.6 with equation - (6.7: R θ Ψ Ψ θ θ Ψ θ α = sm ( ( ln ln 4Z θ θ Ψ Ψ θ Ψ ( θ Ψ θ Ψ 0( f,t K (k0'. ( c Np/m (6.30 When π is replaced y π in the aove equation, IHK model for non-planar CPS with ground plane width ( π θ ψ is otained. We have tested the accuracy of the closedform models developed for propagation characteristics of non-planar CPS, for oth ( π θ ψ and ( π θ ψ ground plane width, against the results otained from EMsimulators- HFSS and CST, as shown in Fig.(6.5 and Fig.(6.6. Fig.(6.5 presents comparisons of performances of CPS on the circular and semi-circular cylindrical surfaces in respect of effective relative permittivity, characteristic impedance and losses. It also presents such comparisons for the CPS on the elliptical and semi- ellipsoidal cylindrical surfaces. The results are otained over the frequency range, GHz 60 GHz. For simulation, we have taken the sustrate with ε r =.5, θ = 40, ψ = 5, h = 635 µm and t = 3 µm. For ECPS, the ellipticity c = 0.7 is considered. With increase in the ellipticity and decrease in the ground width, there is increase in oth ε eff and Z 0.So y controlling ellipticity, ε eff and Z 0 of the line can e controlled as well. We have oserved that in oth HFSS and CST, increase in the value of effective relative permittivity starts after 00 GHz; whereas the closed-form model shows continuous increase after 0 GHz. The experimental results are not traceale to verify the nature of dispersion. However the model has an average deviation of.9% against the results of oth HFSS and CST over whole frequency range. 96

29 (a ( (c (d (e Fig.(6.5: Comparison of different line parameters of non-planar CPS w.r.t. frequency. (f 97

30 Like CPW, the effective relative permittivity increases for the semi- circular and semiellipsoidal CPS as compared against the CPS on the circular and elliptical surfaces. The characteristic impedances of semi - circular and semi- ellipsoidal cases are also higher as compared to the CPS on the circular and elliptical surfaces. The total loss of nonplanar CPS is computed for ε r αt =.9, t = 3µm, θ = 8, ψ = 5, c = 0.7, h = 635 µm, tan δ = and σ = 4. x 0 7 S/m etween frequency range GHz 60 GHz. On comparing, the closed-form model # and closed-form model # have % average and % maximum deviation of (6.5%, 6.7% and (3%, 8% respectively against oth the EMsimulators, excluding results at GHz due to large deviations. The closed-form model # is in close agreement with HFSS at higher frequencies. The losses are almost identical for oth the cases. The nature of dispersion is identical for oth cases. The results are summarized in Tale- (6.4. Tale - 6.4: % Change in characteristics of models against simulators on non-planar CPS with different ground-plane widths (a: % change in characteristics of semi-ellipsoidal CPS over elliptical CPS. Closed-form Model HFSS CST Characteristics Increase in Effective relative permittivity Increase in Characteristic impedance Decrease in Total loss % av. change % max. change % av. change % max. change % av. change % max. change

31 (: % change in characteristics of semi-circular CPS over circular cylindrical CPS. Closed-form Model HFSS CST Characteristics Increase in Effective relative permittivity Increase in Characteristic impedance Decrease in Total loss % av. change % max. change % av. change % max. change % av. change % max. change (a ( (c Fig.(6.6: Comparison of different line parameters of non-planar CPS w.r.t. structural parameters. 99

32 Fig. (6.6a and (6.6 show the effect of conductor thickness on εeff and Z 0 of ECPS and CCPS with varying angular strip width θ/h for ε r =.5, ψ = 35, h = 635 µm, c = 0.7, and f = 0 GHz.The two conductor thicknesses t = 3 µm and 6 µm are used in the investigation. Both εeff and Z 0 decreases with increase in the conductor thickness. Fig. (6.6c further compares the total loss of the ECPS and CCPS for ε r =.9, θ = 8, ψ = 5, f = 0GHz, h = 635 µm, tan δ = and σ = 4. x0 7 S/m for conductor thickness range 0.5µm - 0µm. Both the models are in close agreement with EM-simulators. The closed-form model# fails to compute for t <. δ. s 6.7. Multilayer Case The mechanical strength and the average power handling of a CPS is increased y using another dielectric sustrate elow the fragile sustrate. In this section, we have extended the improved closed-form models for the line parameters of single-layered, equations- (6.7-(6.30, to multilayer non-planar CPS y applying SLR technique and using equation-(6.0. The structural parameters of multilayer structure otained from SC-transformation (discussed in Chapter-3 are: s = Ψ ( i w = θ h a3 3 = ln a ( iv h ( ii a3 3 = ln a ( v t a 4 4 = ln a3 3 a5 5 h3 = ln a3 3 ( iii ( vi (6.3 We have computed ε eff and Z 0 of multilayer CPS y using equation-(6.9. The modulus k 0 and k i (i=,,3 along with their complementary modulus for finite ground plane widths ( π θ ψ and ( π θ ψ, are otained from equations- (3.4-(

33 and equations-(3.55-(3.56 respectively after replacing θ with in equation (3.45 will e replaced y h i (i=,,3 accordingly. θ ψ. The parameter H (a ( (c (d Fig.(6.7: Multilayer CPS on the curved surfaces: (a Elliptical CPS (MECPS,( Circular Cylindrical CPS (MCCPS,(c Semi-ellipsoidal CPS (MSECPS and (d Semi- circular cylindrical (MSCCPS. Fig. (6.7 shows comparison and validity of SLR-ased computed line parameters of multilayer non-planar CPS with ε r =.9, ε r =0, ε r3= 0, θ = 6 and ψ = against EM-simulators for frequency range GHz 60 GHz. Fig. (6.7a shows that effective relative permittivities of semi-ellipsoidal CPS and semi-circular CPS are more than elliptical and circular CPS. Also effective relative permittivity of CPS on elliptical surface is higher than that of on circular surface. Fig.(6.7 shows that characteristic impedance on these surfaces follows inverse rule. The computation of ε eff and Z 0 y the model has % average and % maximum deviation of (4.9%,7.6% and (5.%,.9% respectively against oth the EM simulators, as shown in Fig.(6.7a and (6.7. 0

34 (a ( (c Fig.(6.8: Multilayer non-planar CPS: (a Effective relative permittivity, ( Characteristic impedance, and (c Total loss. In Fig.(6.7c variation in αt of multilayer non-planar CPS for conductor thickness range 0.5µm - 0µm at f = 5 GHz is shown. The loss on semi-ellipsoidal CPS is the least and it is maximum on the circular CPS. The results are very close to each other. The % average and % maximum deviation of closed-form model # and closed-form model # against EM-simulators are (5.8%, 5.5% and (4.4%,.4% respectively. 0

35 6.8 Circuit Model of CPS In this section, the accuracy of the circuit model of the CPS are compared against four softwares- HFSS, Sonnet, CST and MOM-ased software LINPAR on the effective relative permittivity, real and imaginary characteristics impedances and total loss. We compare accuracy of our parameters extraction against the extracted line parameters using EM-simulators- HFSS, Sonnet, CST and MOM-ased software LINPAR. We have taken 0.0 GHz 0 GHz frequency range and the CPS is considered on the sustrate with ε r =.9, s = 8µm, w = 4µm, h = 00µm and t = 0.5µm. The extracted frequency dependent RLCG line parameters are shown and compared in Fig.(6.9. The line resistance R, presented in Fig.(6.9a shows increase with frequency. As frequency increases, the skin-depth ecomes smaller than the conductor thickness and surface resistance increases as the square-root of frequency. The results otained from circuit model are closer to LINPAR and HFSS. The line inductance L, presented in Fig.(6.9, shows decrease in its value with increase in frequency from 0.0 GHz to GHz. It is due to decrease in the internal inductance of the CPS. Aove GHz change in line inductance is insignificant. The results of the circuit model follow results of EMsimulators closely. The line capacitance C, presented in Fig.(6.9c, also shows decrease in its value with increase in frequency from 0.0 GHz to 0 GHz. The non-causal circuit model closely follows results of HFSS. It follows physical ehavior of a capacitor. The results of circuit model and EM-simulators, presented in Fig.(6.9d, show increase in line conductance G with frequency. The circuit model closely follows EM-simulators. 03

36 (a ( (c (d Fig.(6.9: Extraction of RLCG parameters of lossy planar CPS using circuit model and EM-simulators. Finally we otain frequency dependent effective relative permittivity, characteristic impedance - real and imaginary parts, and total loss from the RLCG parameters. These results otained from the circuit models are presented in Fig.(6.0 and compared against the results of EM-simulators. The results of circuit model follow results of EMsimulators; whereas results of our previous closed-form individual models do not follow simulators, especially at lower end of frequency inspite of our accounting for the internal inductance effect. It is true for all the parameters - effective relative permittivity, losses and characteristic impedance. 04

37 (a ( (c (d Fig.(6.0: Comparison of circuit model, closed-form model against EM-simulators for line parameters of lossy planar CPS in respect of ε ( f,t, α T ( f,t, Re( Z * 0 ( f,t and Im(Z* 0 ( f,t. eff Tale-6.5 consolidates the comparison of the models against HFSS for line parameters of CPS. The circuit model accurately predicts and improves the dispersive nature of the CPS at the lower frequency range for computation of ε eff ( f, t, Z 0( f, t and α T ( f,t with % average and % maximum deviation of (3.%, 7.5%, (3.7%, 6.% and (.6%, 5.5% respectively against EM- simulators. The circuit model is applicale to all the line structures presented in this chapter. However for sake of revity, we are not presenting the results. 05

38 Tale - 6.5: % Deviation of models against HFSS [Data range: t =0. 5 µm; f = 0.0 GHz - 0 GHz; ε r =.9; tanδ=0.000; σ=4.x 0 7 S/m] Model ε eff Z 0 (Ω α T (Np/cm Av. Max. Av. Max. Av. Max. LINPAR Sonnet CST Closed-form Model Circuit Model

Application of EM- Simulators for Extraction of Line Parameters

Chapter - 2 Application of EM- Simulators for Extraction of Line Parameters 2. 1 Introduction The EM-simulators-2D, 2.5D and 3D, are powerful tools for the analysis of the planar transmission lines structure.