Presented by: Allen Hollister Author of Wideband Amplifier Design textbook VP Engineering at: Besser Associate How to Design T-Coils April 3 rd , 2008 Page 2 Overview T-Coils Bandwidth Improvement factor of 2.72 Good step response 50 Ohm input impedance 2007 Allen Hollister This Webinar will prove these assertions and show you how to create and use these circuits to solve your problem How to Design T-Coils April 3 rd , 2008 Page 3 The ProblemMust Drive a Load Capacitor Through a Source ResistorCreates a LPF 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 4 We Use Peaking Circuits to Get As Much Bandwidth as Possible 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 5 I 1 V out R 1 C 1 V 1 V out R 1 C 1 Simple RC 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 6 Assume that R 1 = 50 and C 1 = 1pF Choose 1 pF because it is a nice number for scaling Choose 50 because it is commonly used as a transmission line termination Bandwidth Theoretical Bandwidth of an RC Single Pole Filter 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 7 I 1 L 13 R 1 C 1 V out L 23 C b M Z in 1 2 3 L s =L 13 L r =L 23 If the values of L s , L r , M, and C b are selected appropriately for the values of R 1 and C 1 then: The Bandwidth Improvement Factor will be 2.72 The filter will be two pole and have an MFED step response Z in will equal R 1 for all frequencies (until the parts no longer look like Rs, Ls and Cs) Can be used as a transmission line termination Other kinds of filters (MFA, etc) can also be created with this same input impedance property T-Coil Peaking 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 8 Symmetric T-Coil Parameters 2007 Allen Hollister Definitions are shown in the yellow box After a lot of algebra (not shown in this webinar) we will find that if the values for L s , L r , and C b are defined by the equations shown in the pink boxes, then the equations at the bottom of the page shown in green will be true! How to Design T-Coils April 3 rd , 2008 Page 9 Standard Forms for Second Order Systems 2007 Allen Hollister Shown is a standard form for second order systems that we will use heavily This form will allow us to develop a generalized theory that can then be applied to specific circuits The damping factor , is a parameter that we will use heavily How to Design T-Coils April 3 rd , 2008 Page 10 Compare Standard Form To T-Coil 2007 Allen Hollister This exercise will allow us to derive the T-Coil parameters in terms of ,R 1 ,andC 1 How to Design T-Coils April 3 rd , 2008 Page 11 T-Coil Parameters as a function of , R1, C1 2007 Allen Hollister The equations in the pink boxes give us the complete set of T-Coil parameters as a Function of R 1 , C 1 , and How to Design T-Coils April 3 rd , 2008 Page 12 Solving for bandwidth in this way gives: For the MFED case (=Sqrt(3)/2),f 3dB is: Bandwidth Improvement Factor One can find the bandwidth of any second order system using this equation Solving for Bandwidth 2007 Allen Hollister We can solve for the bandwidth of the normal equation by substituting j2f for s, Then finding the magnitude by taking the square root of the sum of the real part squared and the imaginary part squared And then solving this equation for the point it is 0.707 as a function of f (frequency) How to Design T-Coils April 3 rd , 2008 Page 13 What is MFED Anyway 2007 Allen Hollister Maximally Flat Envelope Delay (MFED) is the kind of filter required if the desire is to have good step response in the time domain. By good, it is meant that there will be little to no overshoot when a unit step function is applied at the input, while achieving maximum possible bandwidth under that constraint. We will very shortly show that this requires a value of = Sqrt(3)/2 = 0.8666 for a second order Low Pass Filter How to Design T-Coils April 3 rd , 2008 Page 14 The Ideal Time Delay Function 2007 Allen Hollister It is acceptable in almost all circumstances to trade delay time for more bandwidth. However, we must delay all frequencies the exact same amount to prevent overshoot and ring in a time domain step response. The time domain expression for an ideal time delay is given by the DiracDelta function: How to Design T-Coils April 3 rd , 2008 Page 15 Transforming the time domain function into the frequency domain gives: Not a rational function so it cannot be realized with standard RLC components However, a realization can be approximated Ideal Time Delay 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 16 Phase & Phase Delay of the Ideal Time Delay Function 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 17 Standard Form 2 Pole Low Pass 2007 Allen Hollister Solve for phase, phase delay, and envelope delay of the normalized two pole filter gives Substitute s=j Find real part And Imaginary part Definition of Phase Definition of Phase Delay How to Design T-Coils April 3 rd , 2008 Page 18 Phase Delay, Simple RC 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 19 The Meaning of Phase & Envelope Delay 2007 Allen Hollister Phase Delay Phase delay ties frequency response to the step response. It is the time corresponding to the phase lag, i.e. a phase lag of 45 degrees at 100 MHz corresponds to 1/8 of the period of 10 ns or 1.25 ns. Phase delay gives some idea as to what happens to the high-frequency components of a step response. If the phase delay peaks (exceeds the low-frequency value) you can expect to see high-frequency components late in the step response. This causes ringing. Envelope Delay Envelope delay (also known as group delay) is not as intuitive, but it is easier to understand mathematically. It is a more sensitive measure of aberrations than phase delay. If envelope delay is flat with increasing frequency, then the phase delay will be also. It is the envelope delay that is used as the prime measure to approximate a constant delay. How to Design T-Coils April 3 rd , 2008 Page 20 Envelope Delay Phase Delay Envelope Delay Envelope Delay & Phase Delay =0.866, T=1 2007 Allen Hollister Observe that there is no peaking in the response as becomes large Low Frequency delay is 2 T = 1.732 How to Design T-Coils April 3 rd , 2008 Page 21 Envelope Delay Phase Delay Envelope Delay and Phase Delay =0.5, T=1 2007 Allen Hollister Now there is peaking in the response as becomes large Causes Ring in the time domain Step Response Low Frequency delay is 2 T = 1.0 How to Design T-Coils April 3 rd , 2008 Page 22 Maximally Flat Envelope Delay (MFED) Functions 2007 Allen Hollister A class of transfer functions called Maximally Flat Envelope Delay (MFED) has been determined: When low-pass filters are implemented to match the MFED response, they will provide maximum bandwidth with minimum overshoot and ring They are an attempt to make the envelope delay completely flat, with no variation with respect to frequency, so that an input step response will be delayed in time but not distorted in shape Filters implemented to these criteria are not perfect, but they are the best that can be achieved The derivation of these filters is quite complex, involving Bessel polynomials, but the results are easily used The step response of these functions is good, with overshoot of less than 0.76% How to Design T-Coils April 3 rd , 2008 Page 23 Suppose a low-pass filter exists as defined in the equations below Then the resulting filter will be MFED Low Frequency Phase Delay is the phase delay when =0 The MFED Function 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 24 n a 0 a 1 a 2 a 3 a 4 a 5 1 1 1 2 3 3 1 3 15 15 6 1 4 105 105 45 10 1 5 945 945 420 105 15 1 MFED Table of Coefficients 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 25 To be MFED requires = MFED Second Order Filter 2007 Allen Hollister Using the previous table, a second order MFED filter is F(s) shown below Comparing this against the standard form for a two pole filter and solving for gives How to Design T-Coils April 3 rd , 2008 Page 26 Determines the Kind of Filter (MFED, MFA) 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 27 Let R 1 = 50 and C 1 = 1 pF, then Bandwidth should equal 8.66 GHz (2.72 times the bandwidth for just the RC3.18 GHz) Input impedance should equal 50 Step response should be MFED Example Using Agilent Genesys 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 28 L=0.833nH [Lrr] Lr L=0.833nH [Lss] Ls C=1pF [CL] C1 PORT=3 ZO=5000000000000 Port_3 F0=200MHz TF=0ns PW=2.5ns TR=0ns TD=0ns I2=20mA I1=0mA R=50 PORT=1 INP_IPULSE_INPUT_1 PORT=2 ZO=50e6M Port_2 C=0.083pF [Cb] C2 R=50 [RL] R4 MUTIND K=0.5 [k8] L2='Lr L1='Ls MUTIND_MUTUAL_INDUCTOR_1 Component values calculated in equation editor Design Example Schematic in Agilent Genesys 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 29 The tune button allows a variable to be changed dynamically while you watch. Component values calculated in equation editor Design Example Schematic in Agilent Genesys 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 30 Equations to calculate the component values automatically Resultant values calculated by the equations Setup Design Formulas Using the Equation Editor in Agilent Genesys 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 31 Genesys Frequency Response for T-Coil Peaked Circuit Observe: Genesys simulated bandwidth is 8.66 GHz vs. theoretical value of 8.66 GHz 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 32 Genesys Step Response for T-Coil Circuit 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 33 Z in & S 11 for T-Coil Circuit (Z in is 50 Ohms) 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 34 VSWR for T-Coil Circuit 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 35 Group Delay for T-Coil Peaked Circuit 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 36 What about other values for Sofar,wehavelookedatthespecialcaseofanMFEDfilterresponse Itisusefultoconsiderothervaluesof thatwillproducedifferentkindsoftwo polefilters Comparethefollowingkindsoffilters: StraightRCfilternopeaking MFED =SQRT(3)/2 MFA(MaximallyFlatAmplitude)otherwiseknownasaButterworth =SQRT(2)/2 Elliptic =0.5 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 37 PORT=5 ZO=50M Port_5 L=0.833nH [Lrr] Lr L=0.833nH [Lss] Ls C=1pF [CL] C1 PORT=3 ZO=5000000000000 Port_3 F0=200MHz TF=0ns PW=2.5ns TR=0ns TD=0ns I2=20mA I1=0mA R=50 PORT=1 INP_IPULSE_INPUT_1 PORT=2 ZO=50e6M Port_2 PORT=11 ZO=5000000000 Port_11 C=1pF [CL] C8 R=50 [RL] R1 F0=200MHz TF=0ns PW=2.5ns TR=0ns TD=0ns V2=.5V V1=0V R=.00050 PORT=10 INP_VPULSE_INPUT_10 C=1pF [CL] C4 R=50 [RL] R2 PORT=6 ZO=50000000 Port_6 PORT=8 ZO=50M Port_8 C=1pF [CL] C6 R=50 [RL] R3 PORT=9 ZO=50000000 Port_9 C=0.125pF [Cb7] C3 L=0.938nH [Lss7] Ls1 L=0.938nH [Lrr7] Lr1 C=0.083pF [Cb] C2 R=50 [RL] R4 C=0.25pF [Cb5] C5 L=1.25nH [Lss5] Ls2 L=1.25nH [Lrr5] Lr2 MUTIND K=0.5 [k8] L2='Lr L1='Ls MUTIND_MUTUAL_INDUCTOR_1 MUTIND K=0.333 [k7] L2='Lr1 L1='Ls1 MUTIND_MUTUAL_INDUCTOR_2 MUTIND K=0 [k5] L2='Lr1 L1='Ls1 MUTIND_MUTUAL_INDUCTOR_3 F0=200MHz PW=2.5ns TD=0ns I2=.02A I1=0A R=50 PORT=7 TF=0ns TR=0ns INP_IPULSE_INPUT_12 F0=200MHz PW=2.5ns TD=0ns I2=.02A I1=0A R=50 PORT=4 TF=0ns TR=0ns INP_IPULSE_INPUT_4 Four Circuits with Different 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 38 Equations for the Four Circuits 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 39 Frequency Response of the Four Circuits 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 40 Take the Derivative of f 3dB_T-Coil with respect to , set it equal to 0 and solve for gives a maximum at =1/SQRT(2) Maximum Bandwidth occurs here Graph of Bandwidth vs. for the Example 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 41 Step Response of the Four Circuits 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 42 Z in for the three T-coil cases 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 43 Z out for the three T-Coil Circuits 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 44 Group Delay for the four cases 2007 Allen Hollister Monte Carlo Transient Response with 5% part Variation How to Design T-Coils April 3 rd , 2008 Page 45 Monte Carlo Frequency Response with 5% Part Variation How to Design T-Coils April 3 rd , 2008 Page 46 Monte Carlo Analysis for Group Delay with 5% Component Variation From this graph, we see that the parameter variation mainly affects the low frequency delay At high frequencies, there is little peaking above the low frequency delay This is why the step response continues to look good while there is significant variation in bandwidth How to Design T-Coils April 3 rd , 2008 Page 47 Monte Carlo Analysis for Input Impedance with 5% Component Variation How to Design T-Coils April 3 rd , 2008 Page 48 Monte Carlo Analysis for VSWR with 5% Component Variation How to Design T-Coils April 3 rd , 2008 Page 49 Monte Carlo Analysis for S 11 with 5% Component Variation How to Design T-Coils April 3 rd , 2008 Page 50 How to Design T-Coils April 3 rd , 2008 Page 51 Physical Design of T-Coils Components: Spiral Inductors All inductors have some stray capacitance between windings. This capacitance shows up in parallel with the bridging capacitance C b If you know the self resonant frequency of the inductors, you can estimate the stray capacitance using the formula below Once you know the capacitance, subtract it from C b How to Design T-Coils April 3 rd , 2008 Page 52 Momentum GX: 3D-Planar Electromagnetic Simulation for Genesys Sub Eddy Currents Displaceme nt Currents Use Momentum GX to accurately model: Self & mutual inductance Self-resonance frequency Displacement currents from spiral to substrate thru oxide capacitance Eddy Currents in substrate due time varying magnetic fields Spiral to Substrate Insulation Skin effects How to Design T-Coils April 3 rd , 2008 Page 53 A Bond Wire T-Coil 2007 Allen Hollister How to Design T-Coils April 3 rd , 2008 Page 54 Additional Resources For a copy of this presentation and to download the T-Coils Genesys design workspace template, go to: www.allenhollister.com or email Allen@AllenHollister.com Interested in learning more about wideband amplifier design, see Allens course on Wideband/HF Amplifier Design Techniques offered through Besser Associates You can order your copy of Wideband Amplifier Design from Amazon or SciTech Publishing To download your Free evaluation copy of Agilent Genesys and Momentum GX software, please follow this link: http://eesof.tm.agilent.com/prod ucts/genesys/ 2007 Allen Hollister