Verifying Simulation Results with Measurements Scott Piper General Motors
EM Simulation Software Can be easy to justify the purchase of software packages even costing tens of thousands of dollars Upper management tends to be more forward thinking The potential benefit of these tools are clear Difficult to implement in a company There are doubts Many will ask for real world data to validate simulation results 2
The Real World 3
The EMC Real World 4
EMC Component Level Measurements A form of modeling and simulation The component s actual environment is replaced by a test chamber Efforts have been made to remove the influence from other devices Cable harness lengths may be altered The priority of compliance testing is repeatability over accuracy 5
Measurements vs. Simulation Measurement Techniques Used and relied upon for over 50 years Numerical Modeling Rarely used in industry until 7-8 years ago Measurement devices are physical objects Measurement devices can be infinitely small Ideal conditions are difficult to produce Ideal conditions are easy to produce (non-ideal is difficult) Information about the product is not necessary Only as accurate as the information provided 6
Can EM simulation provide useful information on EMC performance of complex electronic products? You Need to convince Your Company? & Yourself 7
Design of Experiment An experiment (or more likely a series of experiments) needs to be designed to show that the simulation tool correctly handles the EM principles for the intended application The experiment must be: Simple (shows only one principle at a time) Relevant to your application Verifiable (a correct answer is known) 8
Example: Crosstalk on a Printed Circuit Board A signal travelling on a conductor has the ability to transfer energy to an adjacent conductor This phenomenon is due to electromagnetic fields radiated by the signal Cross coupling Aggressor Signal Victim Line 9
Electromagnetic Cross Coupling A simple circuit board was designed to show the principle 10
PCB Model 11
Test Setup 12
Initial Results Measured Simulated 13
Which is Right? Inductive and capacitive coupling contributions are direct functions of the excitation frequency. Therefore they increase linearly with an increase in frequency Clayton R Paul (Introduction to Electromagnetic Compatibility) Simulation data doesn t follow this it must be incorrect 14
PCB Model 15
Starting Discretization of Volume 16
Finer Discretization of Volume 17
Final Results Measured Simulated Finer Cells Simulated Larger Cells 18
Crosstalk PCB The dominant feature in this investigation is cross coupling This example would work well for issues dealing with printed circuit boards The results of this example can be checked with a network analyzer and textbook equations 19
Test Setup simulation 20
Test Setup Simulation EMC tests try to mimic ideal conditions as much as possible but the ideal is unattainable When attempting to simulate a test setup, certain non-ideal elements will need to be remembered and included Some aspects of the test setup will make a large difference in the results Other aspects of the test setup will only add complexity to the model and consume resources 21
Wire above a Metal Plane 22
Measurement Setup 23
Experiment Check A known signal was exciting a known geometry An aerospace standard test setup was used The goal of this simulation is to reproduce the measurement 24
Model from 1-30 MHz 25
Unterminated Wire Measured Simulated 26
Comparison Results Frequency Measurement Simulation Difference 150 khz 68 dbµv/m 64 dbµv/m -4 db 1 MHz 52 dbµv/m 47 dbµv/m -5 db 10 MHz 40 dbµv/m 37 dbµv/m -3 db 15 MHz 21 dbµv/m 22 dbµv/m 1 db Minimum (19.7 MHz measured vs 19 MHz simulated) Maximum (30 MHz) 11 dbµv/m 9 dbµv/m 2 db 40 dbµv/m 37 dbµv/m 3 db 27
Wire Terminated into 100 Measured Simulated 11 db difference NOT GOOD 28
Test Setup Simulation When verifying the accuracy of a simulation tool, one must consider the setup specifics When these specifics change, re-verify! If the goal is to support test data, certain aspects of the test setup (wires, support equipment) should be considered even though it goes against the intent of the test Insight can be gained into test setups by attempting to simulate their results 29
Extra Coaxial Cables Added Coax shield for antenna connection Without Cable 22 MHz With Cable 30
Comparison with Coax Cables 31
Test Setup Simulation If the goal is to support test data, certain aspects of the test setup (wires, support equipment) should be considered even though it goes against the intent of the test Insight can be gained into test setups by attempting to simulate their results 32
Limitations of Measurements Exploring 33
PCB Shield 34
Field Probe Setup - Simulation 35
Plane Wave Excitation 36
Electric Field Plot 37
Shielding Effectiveness vs. Frequency At Different Locations 38
Real World Data? 39
Real World Problems Point electric field sensors are typically too large to fit underneath the shield A noise generating circuit attached to a small antenna could be placed inside the shield The circuit would be battery powered Emissions from the circuit could then be measured 40
PCB Shield 41
Noise Generating Circuit 42
Noise Generating Circuit Model 43
Shielding Effectiveness vs. Frequency 44
Measured Results Shielding Effectiveness = Unshielded-Shielded Unshielded Shielded 45
Simulated Results Shielding Effectiveness 46
9V Battery Added to Model 47
New Results 48
Measured Results 49
Electric Field Plot at 480 MHz Shield 50
Battery Shielded 51
Measured Results with Shielded Battery Measured Emissions (dbm) 0-10 -20-30 -40-50 -60-70 -80-90 No Shield Full Shield Ambient -100 1 10 100 1000 Frequency (MHz) 52
Shielding Effectiveness of PCB Shields There are some investigations that are difficult to use measurement equipment While measurements can be a good way to verify the results of a simulation, simulations can also be used to verify the results of measurements! 53
Clues from Measurements Simplifying Simulations: 54
Lumped Elements A capacitor is typically represented as a series LRC circuit Parasitic R and L is typically provided on a vendor s datasheet A circuit simulator would require all three of these values in order to produce a correct answer When performing EMC simulations using a full wave solving method, is it important to include these values? 55
Experiment Design - ESL A test PCB will be designed such that capacitor impedance can be measured easily by a network analyzer The design will then be simulated with and without capacitor ESL and compared with the measured result 56
Simulation Setup FIT Solver SMA Connector 4 Layer PCB 10 nf 0402 capacitors 57
Measurement Setup 58
Results Measured ESL Dominates Impedance Simulated With Lumped ESL Simulated Without Lumped ESL 59
Plates with Ceramic Material in-between Capacitor ESL Leads 60
Simulated Capacitor Lumped Element 61
Different Simulation Tool 3D finiteelement method solver intended for PCB problems 62
Results Simulated With Lumped ESL Measured Simulated Without Lumped ESL What is right for one simulation tool is not always right for another! 63
Conclusion - ESL A setup was used that isolated one variable A multi-layer PCB with ceramic SMD capacitors was used for this investigation The results were in harmony with VNA measurements and existing knowledge 64
Product Immunity Investigations 65
Measurements are the Standard If you want to know if a product will operate normally in a field of 100 V/m at 400 MHz what do you do?. Build the product, then expose it to 100 V/m at 400 MHz and see if it works of course! Can we replace this with a simulation? 66
Example Pair of wires Transmitter Receiver What is the optimal configuration for the wire pair in order to carry the signal from the transmitter to the receiver? 67
To Twist or Not to Twist? Twisted Separated by 2cm 68
To Twist or Not to Twist? S21 Comparision for Wiring Configurations S21 (db) 0-2 -4-6 -8-10 -12-14 -16-18 -20 0 5 10 15 20 25 30 35 40 45 50 Frequency (MHz) Side by Side Twisted Separated by 2cm 69
To Twist or Not to Twist? Measurement/Simulation Comparison 0-2 -4-6 -8-10 -12-14 -16-18 -20 0 5 10 15 20 25 30 35 40 45 50 Twisted Wires Untwisted Wires Separated by 2cm Measurement Untwisted Measurement Twisted Measurement Separated 70
To Twist or Not to Twist? 0 S21 Comparision for Wiring Configurations S21 (db) -2-4 -6-8 -10-12 -14-16 -18-20 0 5 10 15 20 25 30 35 40 45 50 Frequency (MHz) Side by Side Separated Twisted Does this prove that if wires are separated by 2 cm it won t work? What if you tried this wiring configuration and the communication failed? 71
Immunity Simulation Can Still be Valuable Measurements typically provide YES or NO answers Simulations fill in the gaps when conducting troubleshooting investigations Root cause is the circuit weak at this frequency? Or is there a large amount of RF coupling into the circuit caused by a resonance? Is this relating to the test setup or product geometry? Often one can find a fix to the problem while the product is still in the lab. This usually requires cuts and jumpers as a lab fix. If a problem can be reproduced in simulation, then the lab fix can then be simulated along with a production fix. Therefore we have a better idea that the production fix works just as well or better than the lab fix. 72
Measurements & Simulations 73
Fishing Someone may throw a product layout over the wall and this person would like some simulation done of their product Simulation tools are not very effective at discovering problems More useful information can be obtained by keeping a narrow scope for the analysis Measurements can help bound the investigation 74
Actual Product 2 Layer PCB Had sensor malfunctions during immunity testing from 385-940 MHz Near-field probe investigation has determined that the area near the sensor s data trace is susceptible to EMI 75
Near Field Probe Position Initial Position Moved 5cm 76
Initial Position Antenna Hidden Moved 5cm 77
Magnitude Comparison Initial Position Shifted 5cm 78
Antenna 90 with Trace 79
Comparison Initial Position Shifted 5cm Antenna @90 80
Simulation Setup Replicating a near-field probe investigation would take much time to perform using a simulation tool A diagnostic method that is convenient for laboratory measurements is not always best with simulation A plane wave source will first be used in the model to see if the problem can be duplicated without constructing the near-field antenna 81
Simulation Result of Baseline PCB 379 MHz (3pF load assumed at trace ends) Sensor Processor The product had sensor malfunctions during immunity testing from 385-940 MHz 900 MHz 82
Product Product PCB vs. Ideal PCB Ideal 83
Comparison Product PCB Ideal PCB 84
Make Ideal PCB Product Size Product PCB Product Size Infinite PCB 85
Comparison Finite Size Trace near PCB center Finite Size Trace near PCB Edge 86
Comparison Results Product PCB Close to Edge Far From Edge 87
Conclusion The immunity problem appears to be caused by EM coupling from the edge of the PCB near 900 MHz The Immunity problem would be reduced if the trace is moved away from the PCB edge The problem at 380 MHz was able to be reproduced in the full PCB simulation but the cause has not been found 88
Schematic Sensor Processor Resistors will make trace lossy 89
Resistors Product Sensor Trace on Ideal PCB 90
With and Without Resistors R= 0 R= 47 Resistors appear to do a great job with dampening resonance why is there a problem with this trace? 91
Electric Field at 379 MHz on Product PCB Sensor Trace It appears at 379 MHz the power trace is resonant and runs parallel to the sensor trace 379 MHz Power Trace 92
Comparison with Power Trace Removed With Power Trace W/O Power Trace 93
380 MHz Immunity Problem Upon further investigation it is shown that the immunity problem at 380 MHz is caused by cross coupling from a nearby resonant power trace. Either the resonance can be dampened or the trace can be moved further away from the sensor trace to solve this problem 94
Summary Simulation can be used to find the root causes of problems that may exist in products in the testing phase of development by breaking down the problem into a simpler situation These simulations are more effective if guided by measurement data 95
EMC Simulation and Measurements EMC simulation and measurement methods both have their strengths and weaknesses but they do not have many weaknesses in common Using EMC simulation and measurement together can provide useful data more than what one or the other can provide by itself. 96
References / Acknowledgements Piper, S. and Teune, J., "Verifying Electromagnetic Simulation Results with Measurements," SAE Int. J. Passeng. Cars - Electron. Electr. Syst. 5(1):2012, doi:10.4271/2012-01-1160. Paul, Clayton. Introduction to Electromagnetic Compatibility. Wiley, 2006 Test PCBs were provided by Gentex Corporation Test facilities were provided by Gentex Corporation and Woodward MPC, Inc. 97