EE155/255 Green Electronics

Transcription

1 EE155/255 Green Electronics Thermal and EMI 11/7/16 Prof. William Dally Computer Systems Laboratory Stanford University

2 No Date Topic HW out HW in Lab out Lab ck Lab HW 1 9/26/16 Intro (basic converters) 1 1 Intro to ST32F3 Periodic Steady State 2 9/28/16 Embedded Prog/Power Elect. 3 10/3/16 Power Electronics - 1 (switches) AC Energy Meter Power Devices 4 10/5/16 Power Electronics - 2 (circuits) 5 10/10/16 Photovoltaics PV MPPT PV SPICE 6 10/12/16 Feedback Control 7 10/19/16 Electric Motors Motor control Matlab Feedback 8 10/21/16 Isolated Converters 9 10/24/16 Solar Day 5/PP Motor control - Lab/ Isolated Converters 10 10/26/16 Magnetics 11 10/31/16 Soft Switching 6 5/PP 6 5 PS Magnetics and Inverters 12 11/2/16 Project Discussions 13 11/7/16 Inverters, Grid, PF, and Batteries 6 P 6 Project YAH 14 11/9/16 Thermal & EMI 15 11/14/16 Quiz Review C /16/16 Grounding, and Debugging Q 11/16/16 Quiz - in the evening 11/21/16 Thanksgiving Break C2 11/23/16 Thanksgiving Break 17 11/28/ /30/16 Martin Fornage, enphase C /5/16 Colin Campbell, Tesla 20 12/7/16 Wrapup 12/15/16 Project presentations P 12/16/16 Project webpage due 39

3 PE Tech Times, Nov 9, 2016

4 Thermal Design

5 Thermal Design Many components in green-electronic systems dissipate considerable power FETs, IGBTs, Diodes conduction and switching losses Magnetics core and copper losses Resistors, capacitors, etc Typically from 2-20% of system power (2W 5kW) This power must be removed without the components overheating Heat-transfer mechanisms Conduction, convection, radiation, and phase change Thermal solutions dominated by the first two

6 Power (W) Flows through thermal system (like current) Thermal Resistance (K/W) Quantities and Units Temperature gradient drives power flow through conduction Thermal Conductivity (W/(m K)) Determines resistance of a material Diamond 1000, Copper 400, Aluminum 205, Water 0.58, Plastic 0.2, Air 0.02 Specific Heat (J/(g K)) Capacity of a fluid to carry heat through convection Determines time constant for conduction (like a capacitor) ~1 for air, ~4 for water,.39 for copper Volumetric Heat Capacity (J/(ml K)) ~10-3 for air, ~4 for water, 3.5 for copper (it takes 4,000 times as much air (by volume) than water to carry heat) Heat transfer coefficient (W/(m 2 K)) Heat transferred to cooling fluid per unit area 50W/(m 2 K) for air at 200 lfm, 5kW/(m 2 K) for water at 10lfm.

7 Typical System PCB w/thermal Vias (0.5 K/W) SMT Transistor 10W, 0.5 K/W J-case Thermal adhesive 0.1 K/(W cm 2 ) 0.4 K/W under FET Air flow 200 lfm (1 m/s) 4 x 10-4 (m 3 /s) Heat Sink 1cm x 4cm x 4cm 128 cm 2 fin area 4cm 2 cross section

8 Questions What is the temperature rise of the FET above the ambient air temperature? How hot is the exhaust air temperature?

9 Determine Thermal Resistance of Heat Sink Heat Sink Fin Area Coeff Prod Res m^2 50W/m^2 K 0.64W/K 1.56K/W

10 Sum Resistance Junction to Ambient FET 0.5K/W PCB 0.5K/W Adhesive 0.4K/W Heat Sink 1.56K/W Total 2.96K/W Power 10.00W Temp Rise 29.6K

11 Determine Temp Rise of Air Air Temp Rise Power Air Flow E/Vol Capacity of Air Temp Rise 10W 400ml/s 2.50E-02J/ml 1.00E-03J/ml K 25K

12 So If w.c. inlet air is 60C, how hot does the FET get? How hot is the outlet air? Only the front of the heat sink sees 60C air. Is our approximation in calculating its thermal resistance correct?

13 In Practice This is done by finite-element modeling (ANSYS, etc ) Answer is as good as the model. Backed up by experiment

14 Interfaces matter Even a tiny air-gap is fatal Upstream devices get cool air Details are Critical Downstream devices get air heated by upstream devices Air follows path of least resistance Need plenums to guide it to where its needed

15 Littlebox Thermal Design

16 Fan Selection - Minimum Airflow to Mantain Outlet Temperature 12V 6W 38mm x38mmx25mm PF38281BX-000U-S99

17 Full Box Simulations

18 Windtunnel test config pictures Anemometer Wind tunnel testing area V_ds power supply and multimeter Fan controls Thermocouple readout

19 Ultrastick Grease Heat sink Shim + solder Thin Pak Vias A-Pli Gap Filler

20 Thermal Stack - Need 6K/W or less Resistance/Source Estimate ~0.5 (Application note) 0.98 Max (Datasheet) 1-3 (Apllication note) 1.2 (1mm board, application note) Our board will be 0.8 mm thick Using L/KA calculation, <0.05 for all shims less than 1mm thick. But what about solder interface? psi (datasheet) 0.1 ~2 with forced convection (datasheet) Interstitial, non-uniform heat application, unknown airflow 2-5?

21 Results

22 OOPS

23 Thermal Design Summary Remove heat from power components by conduction and convection Sum thermal resistances to find total resistance K/W Determines temperature rise of component With multiple components, solve equivalent circuit Add heat capacities to determine transient response Determine heat sink resistance from surface area Heat transfer coefficient depends on airflow Determine cooling fluid temperature rise from heat capacity Validate with simulation and experiment Details matter interfaces, fluid routing, etc

24 EMI

25 EMI Electromagnetic Interference Noised caused by high currents switching Affects your circuits Large voltages induced across inductance Currents and voltages coupled into unrelated wires Radiates B and E fields Affects other circuits Violates regulations

26 Cause Fast Current Transients L P M 2 C 2 C 1 M 1 C 3

27 Cause Fast Current Transients L P M 2 C 2 C 1 M 1 C 3

28 Cause Fast Current Transients L P M 2 C 2 C 1 M 1 C 3

29 Fast Current Transients Littlebox example: C2, C3 ~ 200pF DV = 400V, Dt = 5ns Q = 80nC I peak = 32A + I norm di/dt = 13GA/s C 1 L P M 2 C 2 M 1 C 3 i t

30 13GA/s Induces 13V across every 1nH of inductance 50V difference from one end of a ground plane to another Induces a fraction of this in nearby wires (flux coupling) Everything is an inductor All capacitors have inductance Every foil trace on a PCB has inductance Everything is a transformer Flux coupled with inductor seeing current transient There is no such thing as a uniform ground plane Faster devices don t help

31 To Minimize EMI Keep the critical loop very small Place Components to minimize trace runs Supply current from multiple nearby low-inductance caps Minimum loop area Minimum inductance Use one or more ground planes Loop is between signal and plane Isolate noisy area with ferrite beads on supply Don t put ferrite beads on ground Causes huge voltages between different grounds

32 Symptoms 1 Measurement Noise Analog measurements become very noisy 30mV signal mixed with 10V of noise To fix: Take differential measurements And pick reference carefully Use a good instrumentation amp need to reject high-frequency common-mode noise Sample after the ringing has settled Use a snubber so the ringing settles quickly

33 Gate driver switches spuriously Symptoms 2 Actuator Noise 5V gate drive signal corrupted by 50V ground noise To Fix Use isolated gate drivers High-side drivers even for low-side FETs High transient immunity Use Kelvin source connection Decouple gate driver from di/dt noise across source inductance

34 Symptoms 3 Components Fail Voltage stress on nearby components microcontroller, A/D converters, amplifiers, level shifters, gate drivers Input signals exceed legal range High currents coupled into output signals To Fix Protect and Isolate Diode/capacitor (not just diode) protection on all inputs and outputs Series resistance on outputs or buffer with robust driver Power supplies filtered at point of use Optical communication where needed

35 Protection Circuit

36 Symptoms 4 You violate FCC Regulations Radiated and/or conducted emissions exceed specifications To Fix Use snubbers to minimize noise Use compact layout to minimize loop area Use ground planes to minimize loop area Package noisy circuits in a Faraday cage Make sure to gasket edges slots radiate EMI filters on all inputs and outputs Ferrite beads to attenuate high frequencies Bulk inductors for lower frequencies Common and differential mode filtering Avoid corrupting signals after the filter

37 An EMI Filter

38 Ferrite Bead EMI Filter

39 3.3uH Inductor Filter

40 Power switching generates EMI 10GA/s or larger switching transients EMI Summary Induces 10s of V of noise across local parasitics 50V from one end of a ground plane to another observed Take differential measurements at quiet times Directly across component of interest With a good instrumentation amplifier Use isolated gate drivers and Kelvin source connections Protect inputs and outputs Filter inputs and outputs with ferrite beads to block high frequencies and large inductors for lower freqeuncies

41 No Date Topic HW out HW in Lab out Lab ck Lab HW 1 9/26/16 Intro (basic converters) 1 1 Intro to ST32F3 Periodic Steady State 2 9/28/16 Embedded Prog/Power Elect. 3 10/3/16 Power Electronics - 1 (switches) AC Energy Meter Power Devices 4 10/5/16 Power Electronics - 2 (circuits) 5 10/10/16 Photovoltaics PV MPPT PV SPICE 6 10/12/16 Feedback Control 7 10/19/16 Electric Motors Motor control Matlab Feedback 8 10/21/16 Isolated Converters 9 10/24/16 Solar Day 5/PP Motor control - Lab/ Isolated Converters 10 10/26/16 Magnetics 11 10/31/16 Soft Switching 6 5/PP 6 5 PS Magnetics and Inverters 12 11/2/16 Project Discussions 13 11/7/16 Inverters, Grid, PF, and Batteries 6 P 6 Project YAH 14 11/9/16 Thermal & EMI 15 11/14/16 Quiz Review C /16/16 Grounding, and Debugging Q 11/16/16 Quiz - in the evening 11/21/16 Thanksgiving Break C2 11/23/16 Thanksgiving Break 17 11/28/ /30/16 Martin Fornage, enphase C /5/16 Colin Campbell, Tesla 20 12/7/16 Wrapup 12/15/16 Project presentations P 12/16/16 Project webpage due 39