Passive components in MMIC technology

Transcription

1 Passive components in MMIC technology Evangéline BENEVENT Università Mediterranea di Reggio Calabria DIMET 1

2 Introduction Design cycle of passive components in MMIC technology Passive components in MMIC technology Distributed components Inductor, capacitor Localized components Resistor, capacitor, inductor Microwave characterization of passive devices S-parameters Extraction of device s characteristics from measurements De-embedding and calibration 2

3 Introduction Design cycle of passive components in MMIC technology Passive components in MMIC technology Distributed components Inductor, capacitor Localized components Resistor, capacitor, inductor Microwave characterization of passive devices S-parameters Extraction of device s characteristics from measurements De-embedding and calibration 3

4 Introduction Maxwell s equations All electromagnetic behaviors can ultimately be explained by Maxwell s four basic equations:. D = ρ. B = 0 B E = t H = D j + t However, it isn t always possible or convenient to use these equations directly. Solving them can be quite difficult. Efficient design requires the use of approximations such as lumped and distributed models. Why are models needed? Models help us predict the behavior of components, circuits and systems. Lumped models are useful at lower frequencies, where some physical effects can be ignored. Distributed models are needed at higher frequencies to account for the increased behavioral impact of those physical effects. 4

5 Introduction Two ports models Two-port, three-port, and n-port models simplify the input/output response of active and passive devices and circuits into black boxes described by a set of four linear parameters. Lumped models use representations such as admittances (Y) and resistances (R). Distributed models use S-parameters (transmission and reflection coefficients). Limitations of lumped models At low frequencies most circuits behave in a predictable manner and can be described by a group of replaceable, lumped-equivalent black boxes. 5

6 Introduction Limitations of lumped models At microwave frequencies, as circuit element size approaches the wavelengths of operating frequencies, such a simplified type of model becomes inaccurate. The physical arrangements of the circuit components can no longer be treated as black boxes. We have to use a distributed circuit element model and S-parameters. S-parameters S-parameters and distributed models provide a means of measuring, describing, and characterizing circuits elements. They are used for the design of many high-frequency products. 6

7 Introduction Design cycle of passive components in MMIC technology Passive components in MMIC technology Distributed components Inductor, capacitor Localized components Resistor, capacitor, inductor Microwave characterization of passive devices S-parameters Extraction of device s characteristics from measurements De-embedding and calibration 7

8 Design cycle of passive components in MMIC technology Choice (or no choice!) of the substrate respect to the application or the specifications Choice of additional key materials such as dielectric and magnetic materials Analytical models approximated size and performance of passive components EM simulation (numerical modeling) performance of passive components Performance = specifications? NO YES Fabrication of a prototype, Characterization, Test DESIGN COST! Performance = specifications? NO YES GOOD JOB! 8

9 Introduction Design cycle of passive components in MMIC technology Passive components in MMIC technology Distributed components Inductor, capacitor Localized components Resistor, capacitor, inductor Microwave characterization of passive devices S-parameters Extraction of device s characteristics from measurements De-embedding and calibration 9

10 Distributed components From transmission lines, it is possible to realize low values passive components like capacitances or inductances, provided that the length of the line is less than λ/10. Theory of transmission lines: A section of a transmission line without losses (or with low losses), with a length l, with a characteristic impedance Z c, and terminated by a impedance (load) Z L presents an impedance Z(l), on the input, equal to: Z c Z L Z( l) = Z c Z Z L c + + jz tg( βl) c jz tg( βl) L l [1] C. Algani, Composants passifs, Support de cours du CNAM, Spécialité Electronique-Automatique. 10

11 Distributed components If the length of the transmission line is small respect to the wavelength: βl < π/6 or l < λ/12 Then: Z( l) = Z c Z Z L c + + jz βl c jz βl L This input impedance is a complex impedance so: If Re(Z(l)) << Im(Z(l)): Z(l) pure imaginary One can realize a capacitor or an inductor! 11

12 Distributed components Inductor: If Z L = 0 or Z L << Z c tg(βl): Z( l) jz tg( βl) c By identification: Z = jlω The synthesized inductance L (H) has a value equal to: L Zc ω tg ( βl) This inductance can be realized by a short-circuited line or by a line with a characteristic impedance Z c high respect to the impedance of the load. 12

13 Distributed components Real realization of distributed inductor: Series inductance: l Z 01 Z 02 Z 0 >> Z 01, Z 02 Z 0 Shunt inductance: Z 01 Z 0 l Short-circuit 13

14 Distributed components Capacitor: If Z L = or Z L >> Z c tg(βl): Z( l) Z c jtg( βl) By identification: 1 Z = jcω The synthesized capacitance C(F) has a value equal to: tg( β l) C = ω Z c This capacitance can be realized by a open-circuit line or by a line with a characteristic impedance Z c low respect to the impedance of the load. 14

15 Distributed components Real realization of distributed capacitor: Series capacitance: Z 0 Z g 0 Shunt capacitance: Z 0 << Z 01, Z 02 l Z 01 Z 02 Z 0 15

16 Introduction Design cycle of passive components in MMIC technology Passive components in MMIC technology Distributed components Inductor, capacitor Localized components Resistor, capacitor, inductor Microwave characterization of passive devices S-parameters Extraction of device s characteristics from measurements De-embedding and calibration 16

17 Localized components The localized components have higher values than the distributed components. However, due to parasitic elements at high frequency, the dimensions of localized components must be small compared to the wavelength (l < λ/30). In this way, the variations of phase are negligible. Localized components can be described by analytical models which take into account the frequency-dependent parasitic effects and different kinds of losses by adding other localized components. 17

18 Localized components Resistor Structure: resistor metallization substrate ground plane The resistance R (Ω) of a conductor strip is defined by the following equation: R = 1 σ l W t where σ is the conductivity of the conductor, l the length, W the width, and t the thickness of the conductor strip. If the conductor strip is square, the resistance does not depend on the dimensions of the strip and the square resistance (Ω/square) is equal to: 1 1 Rs = σ t 18

19 Localized components Resistor At high frequency, the current circulates only in a thin thickness of the resistive layer called skin depth and not in the total thickness t. The skin depth δ (m) depends on the frequency: δ = 2 ωµ 0 µ c σ where ω=2πf is the pulsation (rad/s), µ 0 and µ c are the conductivities of the vacuum and conductor respectively, σ is the conductivity. The square resistance becomes: 1 1 Rs = σ δ Typical values: 20 to 500 Ω/square. 19

20 Localized components Resistor C 3 Resistor model: R (f) L C 1 C 2 R is the resistance, depending on the skin-effect, The distributed nature of the resistor is taken into account with the series inductance L, C 1, C 2 are the parasitic shunt capacitances to ground of the resistor and its contact pads, C 3 is the end-to-end feedback capacitance. [2] Frank Ellinger, RF Integrated Circuits and Technologies, Springer,

21 Localized components Capacitor Interdigital capacitor The capacitance increases with the number of fingers. Port 1 Port 2 21

22 Localized components Capacitor Interdigital capacitor Equivalent circuit: C R L C 1 C 2 C is the interdigital capacitance. R corresponds to the resistive losses. L is the parasitic inductance of the fingers. C 1, C 2 are the parasitic capacitances to the ground. 22

23 Localized components Capacitor Interdigital capacitor Advantages: Only one metallization plane, Easy to manufacture. Drawback: Too small capacitance: typically C = 0.5 to 2 pf/mm². 23

24 Localized components Capacitor MIM (Metal-Insulator-Metal) capacitor C MIM ε 0ε rs ε 0ε rwl ( F) = = e e ε 0 is the vacuum permittivity. ε r is the relative permittivity of the insulator. W is the width of the capacitor. l is the length of the capacitor. e is the thickness of the insulator layer. 24

25 Localized components Capacitor MIM (Metal-Insulator-Metal) capacitor MIM capacitor model: L 1 C R L 2 C 1 C 2 C is the MIM capacitance, R corresponds to the losses of the capacitor, C 1, C 2 are the parasitic capacitances to ground from bottom, top plate, L 1, L 2 are the parasitic inductances of bottom, top plate. 25

26 Localized components Capacitor MIM (Metal-Insulator-Metal) capacitor Choice of the dielectric material: The higher the relative permittivity of the material is, the higher the value of the capacitance is (C = ε dielectric.c 0 ). So one can choose a high permittivity material. But in a MMIC circuit, the capacitors must support various DC polarization voltages. So one have to also consider the breakdown voltage (or breakdown electric field). [3] C. Rumelhard, MMIC Composants, Techniques de l Ingénieur. 26

27 Localized components Capacitor MIM (Metal-Insulator-Metal) capacitor For example, in order to the titanium dioxide (TiO 2 ) supports the same voltage than the tantalum pentoxide (Ta 2 O 5 ), it is necessary to multiple the thickness by five, so to reduce the capacitance by five. Dielectric material Relative permittivity Breakdown electric field (V/µm) Capacitance density for V max = 50 V (pf/mm²) SiO 2 (silica) Si 3 N 4 (silicon nitride) Al 2 O 3 (alumina) Ta 2 O 5 (tantalum pentoxide) TiO 2 (titanium dioxide)

28 Localized components Capacitor MIM (Metal-Insulator-Metal) capacitor This is summarized in the third column with the capacitance density for a maximum voltage. Regarding this parameter, the best dielectric material is now the tantalum pentoxide instead of the titanium dioxide. Dielectric material Relative permittivity Breakdown electric field (V/µm) Capacitance density for V max = 50 V (pf/mm²) SiO 2 (silica) Si 3 N 4 (silicon nitride) Al 2 O 3 (alumina) Ta 2 O 5 (tantalum pentoxide) TiO 2 (titanium dioxide)

29 Localized components Capacitor MIM (Metal-Insulator-Metal) capacitor Because of the leakage area, in real topology, it is necessary to add an air bridge. Air bridge (deck) Air bridge (pillar) 2 nd thick metal First metal 2 nd thick metal Leakage area Silicon nitride Si 3 N 4 Silicon nitride Si 3 N 4 First metal Substrate Substrate 29

30 Localized components Inductor Rectangular plate inductor 2l W + t L = 2µ 0 l ln W + t 3l l t W In order to reduce the area occupied by the inductor, one can: Fold down the conductor, Make loops. 30

31 Localized components Inductor Loop inductor L = l. l ln 1.76 W + t W is the width of the conductor, t is the thickness of the conductor, l is the circumference of the loop equal to: l = 2πR 31

32 Localized components Inductor Meander inductor L = l. l ln W + t + W + t l W is the width of the conductor, t is the thickness of the conductor, l is the length of the meander. Typical values: 0.4 to 4 nh. 32

33 Localized components Inductor Circular spiral inductor 394a² n² L = K 8a + 11c W W K = ln > h h D o + D a = i 4 D o D c = i 2 n is the number of turns, W is the width of the conductor, h is the height of the substrate, D o is the outer diameter, D i is the inner diameter. Typical values: 0.2 to 15 nh. 33

34 Localized components Inductor Square spiral inductor There are many ways to layout a planar spiral inductor. The optimal structure is the circular spiral. This structure places the largest amount of conductors in the smallest possible area, reducing the series resistance of the spiral. This structure, however, is often not used because it is not supported by many mask generation systems. Many of these systems are able to only generate Manhattan geometries (and possibly 45 angles as well). Manhattan-style layouts only contain structures with 90 angles. [4] R.L. Bunch, D.I. Sanderson, S. Raman, Application Note, Quality factor and inductance in differential IC implementations, IEEE Microwave Magazine, June

35 Localized components Inductor Square spiral inductor So a simple solution is to approximate a circle by a polygon. An octagonal spiral as a Q that is slightly lower than the circular structure but is much easier to lay out. Octagonal inductor The square spiral structure does not have the best performance, but it is one of the easiest structure to lay out and simulate. 35

36 Localized components Inductor Square spiral inductor 2µ 0 n² d avg L = ln ρ ρ ² π ρ n is the number of turns, d avg represents the average diameter of the spiral, ρ represents the percentage of the inductor area that is filled by metal traces. 36

37 Localized components Inductor with magnetic material When a high permeability material is placed near a conductor carrying electrical current, the inductance of the conductor is know to increase. Ideally, if a conductor is enclosed in an infinite magnetic medium, the inductance is increased by a factor of µ r, the relative permeability of the medium. If µ r is purely real (no magnetic loss) and large, then the inductance as well as the quality factor Q of the structure are significantly enhanced. It also means that, for the same inductance value, a much smaller substrate area would be needed. Furthermore, since the magnetic flux is confined within the magnetic material, cross-talk between the inductors on the same chip would be reduced. [5] V. Korenivski, R.B. van Dover, Magnetic film inductors for radio frequency applications, J. Appl. Phys. 82 (10), Nov. 1997, pp

38 Localized components Inductor with magnetic material Thin film solenoid with a magnetic core What s a solenoid? 38

39 Localized components Inductor with magnetic material Thin film solenoid with a magnetic core Cross section of thin film rectangular solenoid W conductor/coil magnetic core µ r t m t s insulator µ 0 t i t c 39

40 Localized components Inductor with magnetic material Thin film solenoid with a magnetic core Top view of thin film rectangular solenoid W c N turns W l 40

41 Localized components Inductor with magnetic material Thin film solenoid with a magnetic core Inductance: NΦ L = = I µ µ N² W. r t m Quality factor: ωl ωµ µ Nt W Q = = R 2lρ 0 l t 0 r m c c where N is the number of turns, Φ the magnetic flux, µ 0 the vacuum permeability, µ r the relative permeability of the magnetic material, t m its thickness, W the width of the solenoid, l its length. where W c is the width of the conductor strip, t c its thickness, ρ its resistivity.. 41

42 Localized components Inductor with magnetic material Thin film solenoid with a magnetic core Parasitic capacitance: WcW C 2Nε t i where ε = ε 0.ε r is the permittivity of the insulator, t i its thickness. Resonance frequency: f r 1 = 2π LC 8π ² µ = ε t l 3 0 µ r N W ² tm i W c 1/ 2 42

43 Localized components Inductor with magnetic material Magnetically sandwiched stripe inductor Magnetic layer Conductor strip µ r µ 0 t m t c = g L = µ µ 0 rl tm 2W 2K W 1 tanh W 2K K = gt m µ r 2 Where µ 0 is the vacuum permeability, µ r the relative permeability of the magnetic material, l the length of the strip, tm the thickness of the magnetic, W the width of the structure, g the gap between the two magnetic layers. 43

44 Localized components Inductor with magnetic material Magnetically wrapped stripe inductor L = µ 0µ rl tm 2W Magnetic layer Conductor strip µ r µ 0 t m t c = g The magnetically wrapped stripe inductor is an improved version of the magnetically sandwiched stripe inductor as the factor: 2K W 1 tanh W 2K was removed. This is due to the enclosure of the magnetic flux in the wrapped version. 44

45 Introduction Design cycle of passive components in MMIC technology Passive components in MMIC technology Distributed components Inductor, capacitor Localized components Resistor, capacitor, inductor Microwave characterization of passive devices S-parameters Extraction of device s characteristics from measurements De-embedding and calibration 45

46 S-parameters Two-port model Any device can be described by a set of four variables associated with a two-port model. Two of these variables represent the excitation (independent variables), and the remaining two represent the response of the device to the excitation (dependent variables). If the device is excited by voltage sources V 1 and V 2, the currents I 1 and I 2 will be related by the following equations: I + 1 = y11v 1 y12v2 I + 2 = y 21V 1 y 22V2 Port 1 Two-port device Port 2 V 1 I 1 I 2 V 2 [6] Test & Measurement Application Note 95-1, Hewlett Packard, S-parameters techniques for faster, More accurate network design,

47 S-parameters Two-port model In this case, with port voltages selected as independent variables and port currents taken as dependent variables, the relating parameters are called short-circuit admittance parameters, or y-parameters. Four measurements are required to determine the four parameters y 11, y 12, y 21, y 22. Each measurement is made with one port excited by a voltage source, while the other port is short-circuited. For example: y 21 I2 = V 1 V2 = 0 At high frequencies, lead inductance and capacitance make short and open circuits difficult to obtain. So the characterization of microwave devices by S-parameters is more convenient. 47

48 S-parameters Using S-parameters Scattering parameters which are commonly referred as S-parameters, are a parameter set that relates to the traveling waves that are scattered or reflected when an n-port network is inserted into a transmission line. S-parameters are usually measured with the device imbedded between a 50 Ω load and source. Z S a 1 a 2 V S Two-port device b 1 b 2 Z L 48

49 S-parameters Incident and reflected waves The independent variables a 1 and a 2 are normalized incident voltages: a 1 V1 + I1Z0 voltage wave incident on port 1 V = = = 2 Z Z Z 0 0 i1 0 a 2 V2 + I2Z0 voltage wave incident on port 2 V = = = 2 Z Z Z 0 0 i 2 0 The dependent variables b 1 and b 2 are normalized reflected voltages: b 1 V1 I1Z0 voltage wave reflected from port 1 V = = = 2 Z Z Z 0 0 r1 0 b 2 V2 I2Z0 voltage wave reflected from port 2 V = = = 2 Z Z Z 0 The parameters are referenced to Z 0 generally equal to 50 Ω 0 r 2 0 (supposed real and positive) 49

50 S-parameters Definition of S-parameters S 21 a 1 a 2 b 1 S 11 S12 S 22 b 2 The linear equations describing the two-port device are then: b + 1 = S11a1 S12a2 b + 2 = S21a1 S22a2 Under the matrix form: b1 S = b2 S S S a a

51 S-parameters Definition of S-parameters S 11 is the input reflection coefficient with the output port terminated by a matched load (Z L = Z 0 sets a 2 = 0): S 22 is the output reflection coefficient with the input port terminated by a matched load (Z S = Z 0 sets V S = 0): S 21 is the forward transmission coefficient with the output port terminated by a matched load (Z L = Z 0 sets a 2 = 0): S 12 is the reverse transmission coefficient with the input port terminated by a matched load (Z S = Z 0 sets V S = 0): S S S S b = a b = a 1 1 a2 = 0 2 b = a b = a 2 a1 = a2= a1 = 0 51

52 52 Passive components in MMIC technology S-parameters Cascade of several two-port devices The ABCD matrix: = b a D C B A a b Two-port device 1 [A 1 B 1 C 1 D 1 ] a 1 a 2 b 2 b 1 Two-port device 2 [A 2 B 2 C 2 D 2 ] Two-port device 3 [A 3 B 3 C 3 D 3 ] a 1 a 2 b 2 b 1 Equivalent two-port device 2 [ABCD] = = b a D C B A D C B A D C B A b a D C B A a b

53 S-parameters Two other matrix are widely used: Y-matrix (admittance matrix) I1 = I 2 Z-matrix (impedance matrix) V V 1 2 = V V [ Y ]. = 2 I I Z Z I I [ Z]. = 2 Y Y Y Y Z Z V V 2 2 I 1 I 2 Port 1 Two-port device Port 2 V 1 V 2 53

54 S-parameters Relations between the matrix of a two-port device 54

55 S-parameters Relations between the matrix of a two-port device 55

56 S-parameters ABCD matrix and S-parameters matrix of useful two-port devices Z Y Z 1 Z 2 Z 3 56

57 S-parameters ABCD matrix and S-parameters matrix of useful two-port devices Y 3 Y 1 Y 2 α Z c, γ l 57

58 Introduction Design cycle of passive components in MMIC technology Passive components in MMIC technology Distributed components Inductor, capacitor Localized components Resistor, capacitor, inductor Microwave characterization of passive devices S-parameters Extraction of device s characteristics from measurements De-embedding and calibration 58

59 Extraction of the device s characteristics from measurements How to compare analytical and experimental (measurements) results? Analytical study: Measurements: Propagation constant γ Characteristic impedance Z c conversion Comparison is now possible! S-parameters S 11, S 12, S 21, S 22 conversion S-parameters S 11, S 12, S 21, S 22 Propagation constant γ Characteristic impedance Z c 59

60 Extraction of the device s characteristics from measurements Experimental results: During measurements, the device is placed between two 50 Ω ports. 50 Ω port Two-port device 50 Ω port 1-Γ T 1+Γ 50 Ω port Γ -Γ -Γ Γ 50 Ω port 1+Γ T Graph of fluency 1-Γ 60

61 Extraction of the device s characteristics from measurements Extraction of the characteristic impedance and the propagation constant from the S-parameters (for a reciprocal device): Transmission coefficient: S21 T = 1 S 11 Γ Propagation constant: γ = 1 ln( l T ) Reflection coefficient: S11² S21² + 1 Γ = K ± K² + 1 K = Γ 1 2S Characteristic impedance: Z c = Z 0 1+ Γ 1 Γ 11 61

62 Extraction of the device s characteristics from measurements Calculation of S-parameters from analytical evaluation of the propagation constant and the characteristic impedance (for a reciprocal device): Transmission coefficient: T = exp( γl) Reflection coefficient: Z Γ = Z c c Z + Z 0 0 S-parameters: S 11 = S 22 Γ(1 T ²) = 1 T ² Γ² S 12 = S 21 T(1 Γ²) = 1 T ² Γ² 62

63 Introduction Design cycle of passive components in MMIC technology Passive components in MMIC technology Distributed components Inductor, capacitor Localized components Resistor, capacitor, inductor Microwave characterization of passive devices S-parameters Extraction of device s characteristics from measurements De-embedding and calibration 63

64 De-embedding and calibration Vector network analyzer (VNA) Vector Network Analyzer are commonly used to measure S-parameters of a DUT (Device Under Test). VNA are available for measurements from 45 MHz up to 220 GHz. The DUT is excited on one port by a sinusoidal signal of constant magnitude and a frequency range defined by the user. The transmitted and reflected signals are measured by the VNA. The operation is repeated for each port, and then the scattering matrix (S-parameters) can be evaluated for each point of frequency. [7] B. Bayard, Contribution au développement de composants magnétiques pour l électronique hyperfréquence, Thèse de Doctorat,

65 De-embedding and calibration Vector network analyzer (VNA) In case of a two-port device, the VNA automatically excites first the port 1 of the DUT and measures the parameters S 11 and S 21, and second excites the port 2 and measures the parameters S 22 and S 12. In this way, it is not necessary to reverse the DUT. When one of the two port is excited, the VNA divides the signal in two parts. The first one will be the excitation source of the DUT, the second one will be needed as a reference. The reflected and transmitted signals should be compared to this reference. The DUT is linked to the VNA by coaxial cables. The bandwidth of the cables and the VNA must be greater than the frequency range study of the DUT. 65

66 De-embedding and calibration Vector network analyzer (VNA) Thin frequency sweeping Screen display Digit keypad Port 1 Port 2 Command buttons 66

67 De-embedding and calibration Measurement benchmark VNA Coaxial cable Port 1 Port 2 DUT GSG coplanar probes Substrate Ground Signal 67

68 De-embedding and calibration Calibration permits to suppress the parasitic effects of the cables, the probes and the VNA. VNA Coaxial cable Port 1 Port 2 DUT GSG coplanar probes Substrate CALIBRATION 68

69 De-embedding and calibration Calibration: Two categories of errors: Random errors: Can not be corrected, Supposed negligible respect to the systematic errors, Example: noise, temperature drift, user manipulation To use the maximal power source without saturate the DUT to optimize the SNR (Signal/Noise Ratio). Systematic errors: Reproducible errors, Must be corrected by the calibration. 69

70 De-embedding and calibration Calibration Systematic errors: Directivity error: error due to the imperfect separation of reflected and transmitted signals. Impedance mismatching of the generator output: a part of the signal reflected by the DUT is reflected by the generator. Impedance mismatching of the load: a part of the signal transmitted by the DUT to the load is reflected by the load. Tracking error: this error is due to the path difference between the measured (external) signals and the reference (internal) signals. Error due to the dissymmetry of the switch that orients the signals from the generator to the ports 1 or 2. Insulation error: this is due to the coupling between the two ports. 70

71 De-embedding and calibration Calibration The goal of the calibration is to obtain a perfect measurement system by removing the errors introduced by the experimental benchmark. The calibration consists on the measurement of special components called standards in order to obtain data to evaluate the elements of the error model. The standards take place in a calibration kit or a calibration substrate. Three models exist: The model with 12 error elements, The model with 10 error elements, The model with 8 error elements. Complexity Accuracy 71

72 De-embedding and calibration Calibration The model with 12 error elements: Forward (F) model 6 elements E IF S ij : intrinsic S-parameters of the DUT E IF : insulation error Incident wave Transmission measurement E GF : generator impedance mismatching 1 S 21 E TF E LF : load impedance mismatching E DF : directivity error E DF E GF S 11 S 22 E LF E RF : reflection error E TF : transmission error Reflected wave E RF S 12 72

73 De-embedding and calibration Calibration The model with 12 error elements: Reverse (R) model 6 elements S ij : intrinsic S-parameters of the DUT E IR : insulation error S 21 E RR Reflected wave E GR : generator impedance mismatching E LR : load impedance mismatching E DR : directivity error E LR S 11 S 12 S 22 E GR 1 E DR Incident wave E RR : reflection error E TR : transmission error Transmission measurement E TR E IR 73

74 De-embedding and calibration OSTL calibration (open-short-thru-load calibration) Commonly used calibration based on the model with 12 error elements. 4 standards are required: open, short, thru, and load. Large bandwidth calibration. OST (open-short-thru) or OSL (open-short-line) calibration Based on the model with 8 error elements. 3 standards are needed instead of 4. The standard load is eliminated, this is the hardiest to manufacture. 74

75 De-embedding and calibration TRL calibration (thru-reflect-line) High accuracy calibration. Initially based on the model with 8 error elements. Standard thru : the two ports are directly linked together. The standard thru must be perfect. Standard reflect : each port is connected to a unknown device with a high reflection level. Standard line : the two ports are linked by a transmission line. The length of the line could be unknown. LRL calibration (line-reflect-line) Identical to the TRL calibration, but it could be convenient for the calibration of planar lines since they can not be directly linked together (the standard thru is not feasible). 75

76 De-embedding and calibration Schematic of a calibration substrate Probes positioned on a calibration substrate Transition from probes to coaxial cable 2 probes on the same support GSG probes positioning on the contact pads of an inductor Probes positioning on the contact pads of an IC 76

77 De-embedding and calibration Manual probe system 77

78 De-embedding and calibration De-embedding permits to suppress the parasitic effects of the transition between the probes and the DUT, and the access of the DUT. VNA Coaxial cable Port 1 Port 2 DUT GSG coplanar probes Substrate DE-EMBEDDING 78

79 De-embedding and calibration De-embedding techniques fall into two broad categories: Modeling based approach, Measurement based approach. The de-embedding approach starts with the knowledge (by measurements or simulation) of the S-parameters of a structure containing the discontinuity to be studied and other auxiliary part such as traces, adapters, etc. The S parameters of these parts are evaluated by means of simulation or measurements. The S matrix of the discontinuity is extracted from the S matrix of the complete structure by means of the information on the auxiliary parts. More exactly, the ABCD matrix is used for the calculation. [8] S. Agili, A. Morales, De-embedding techniques in signal integrity: a comparison study, 2005 Conference on Information Sciences and Systems. 79

80 De-embedding and calibration De-embedding step-by-step: Measurement of the S-parameters of the complete structure and conversion in ABCD matrix. Evaluation by measurement or simulation of the S-parameters of the auxiliary parts and conversion in ABCD matrix. Evaluation of the ABCD matrix of the DUT, and conversion in S- parameters: A C DUT DUT B D DUT DUT A1 = C1 B1 D 1 1 A. C total total B D total total A. C 2 2 B D DUT Substrate [A 1 B 1 C 1 D 1 ] [A DUT B DUT [A 2 B 2 C 2 D 2 ] C DUT D DUT ] [A total B toital C total D total ] 80

81 De-embedding and calibration De-embedding Example with a planar spiral inductor: [9] S. Couderc, Etude de matériaux ferromagnétiques doux à forte aimantation et à résistivité élevée pour les radio-fréquences, applications aux inductances spirales planaires sur silicium pour réduire la surface occupée, Thèse de Doctorat,

82 De-embedding and calibration On-wafer de-embedding: Short-open de-embedding The open-short de-embedding method is a two-step de-embedding method and is considered as the industry standard. Shunt and series parasitic elements are removed by using open and short dummy structures, respectively. Consequently, a short and an open circuits are added on the wafer for each device to be measured. [10] M. Drakaki, A.A. Hatzopoulos, S. Siskos, De-embedding method fro on-wafer RF CMOS inductor measurements, Microelectronics Journal 40 (2009) [11] T.E. Kolding, On-wafer calibration techniques for GHz CMOS measurements, Proc. IEEE 1999 Int. Conf. on Microelectronic Test Structures, Vol. 12, March 1999, pp

83 De-embedding and calibration On-wafer de-embedding: Short-open de-embedding From the measurement of the short-circuit, the open circuit, and the twoport device, it is possible to extract the intrinsic characteristics of the DUT: Y DUT ( Y Z ) ( Y Z ) 1 = total short open short Z short Z short DUT Y open Y open 83

84 De-embedding and calibration On-wafer de-embedding: Short-open de-embedding Example of a coplanar transmission line: De-embedding reference planes Open circuit Short circuit 84

85 De-embedding and calibration On-wafer de-embedding: Short-open de-embedding Example of a planar spiral inductor: 85

86 Passive components in MMIC technology Evangéline BENEVENT Università Mediterranea di Reggio Calabria DIMET Thank you for your attention! 86