Research activities in Liège Ir. V. Beauvois , Ir. S. Coets, Ir. M. Renard and Ir. Ph. Camus

1. Research activities in Liège Ir. V. Beauvois, Ir. S. Coets, Ir. M. Renard and Ir. Ph. Camus V.Beauvois@ulg.ac.be

2. 1stResearch interest: conducted and radiated emissions measurements on large systems Ir. V. Beauvois, Ir. S. Coets and Ir. M. Renard (Sorrento 2002 and Zurich 2003) in coll. With Johan Catrysse (KHBO, Oostende, Belgium) V.Beauvois@ulg.ac.be

3. Conducted Emission – Introduction In the low frequencyrange [150kHz - 30MHz] , conducted emission measurements are performed with:  a LISN (Line Impedance Stabilized Network) which • prevents the EUT from the noise coming from the mains  provides a defined impedance at the point of measurement  a passive voltage probe if LISN unavailable (i.e. if currents too large)

4. Conducted Emission - Measurement Setup

5. With the LISN, the measurement is always performedat point 1b  The point of measurement is ‘‘standardized’’ (fixed) Solution: introduction of a new concept: the six-pole  Conducted Emission - Problems encountered • Oppositely, the voltage probe (point 2) can be placed anywhere between points 1b and 3, depending on the ‘‘accessibility’’ of the EUT Which signal voltage is measured (distributed impedance between 1b and 3 not negligible) ?? 

6.  In this research, only single-phased situations are considered  concept of the six-pole  If three-phased situations without neutral (3P + PE)  eight-pole  If three-phased situations with neutral (3P + N + PE)  ten-pole Conducted Emission - Six-pole Concept

7. EUT emission signal voltage  H1: six-pole between mains and point of measurement (1b or 2)  H2: six-pole between point of measurement and EUT Cond. Emission - More accurate model

8. Cond. Emission - Actual and Future Works As the tests performed without LISN may not be compared with those performed with LISN, the next steps of the study are: • create a ‘‘virtual’’ power mains network similar to the LISN ’s one and perform measurements with the help of this • ‘‘LISN-equivalent network’’  for instance by the use of an EMI filter instead of the LISN  an alternating method to the LISN could then be reached

9. Cond. Emission - Actual and Future Works (contd) • time domain and frequency domain measurements • are done. • alternate measurement methods : • classical passive voltage probe (one or two • with differential method) • alternate probes and clamps (capacitive clamp, • EM clamp, current probe) •  improve the sensitivity of the measurements and the signal • processing to get more accurate results

10. Radiated emission - Introduction Theoretical test configuration:

11. Radiated emission - Problems encountered  Noise coming from the environment. • Multiple reflexions « against » the environment which leads to an over-estimation or an under-estimation of the emission of the EUT. • The measurement cannot necessarily be performed at a 10 meters distance from the EUT. • How many measurements are to be done? And where?

12. Radiated emission - Solutions to consider • Retrieve the right signal from noise by using a differential method. • As the measurements cannot be performed at a 10 meters (or 3 meters) distance, perform near-field measurements in addition with a near-field  far-field transformation. • Take into account the reflecting characteristics of the global environment

13. 2ndResearch interest: Characterization and modelling of embedded systemsemissions Ir. V. Beauvois, Ir. Ph. Camus V.Beauvois@ulg.ac.be

14. Architecture of embedded systems Main clock Data Transmission Lines Level shifters I/O Controller µcontroller Sensors and Actuators Analog and Power Section Memory Power Supply

15. Fast switching occurs on the bus and transmission lines which leads to current pulses. • Current pulses produce electromagnetic emissions radiated through the P.C.B. traces, integrated circuits pads and connected cables. • For a given architecture and software, the current waveforms on the board connections can be evaluated - they are related to electromagnetic emission. • By mean of Fourier transform the spectrum can be computed and compared with EMC limits.

16. Boucle while(1); durée =100 µs 7 6 5 4 3 2 1 0 0 50 100 150 200 250 300 350 400 t (µs) 5000 4000 3000 2000 1000 0 0 20 40 60 80 100 120 140 160 f (kHz) Typical waveforms – Data bus Data bus, address bus and control signal are combined (summation) in time domain. As signals are synchronous (one main clock) phase is the same for each signal at a given frequency -> spectrum can be easily computed. One simple loop on 80C320 Dallas

17. Typical waveforms – Power supply • White filtered noise • for analog parts • Simple pulse noise • for switched circuits 7805 linear regulator (white noise with cut off frequency near 1 MHz) and MAX232 level shifter (245 kHz and harmonics)

18. Typical waveforms – Analog and Power section • Classification into families of circuits • with same noise signatures •  White filtered noise and switching noise.

19. First results and actual works • Conducted and radiated emission of a 80C320 board • were measured for different codes and compared • with computed spectrum : • computed spectrum components occur • at the same frequency as in the measurement on a real circuit ; • variation of amplitude follows a similar envelope.

20. Actual and Future Works • Characterization of a great number of boards : same CPU with different clocks and different peripherals, … • Better modelling of switching process and relationship with conducted and radiated noise. • VHDL modelling of CPU to gain a better comprehension of processor noise sources, synthesis into an FPGA and measurements in anechoic chamber. • Development of software tools to predict noise behaviour of embedded systems.