Random vs deterministic modulations in power converters

...or why you sometimes cannot claim that random solves your EMI issues

In my last paper,  we argue against using random modulation as a cure to achieve electromagnetic compatibility in systems with serial communication devices. In this blog I want to outline the main reasoning steps. 

First of all, it suffices to mention that there is a multitude of papers regarding random modulation, which show that it can be used to achieve the electromagnetic compatiblity (just have a look at the paper's references). And in fact, in some cases, such as when we are using our converters in a proximity of devices which communicate through PLC, we are in fact able to satisfy the EMC requirements, as some of our ESRs showed in their papers. 

However, we had some insights from in situ work: having a system with ac/dc fast charging station and an electric bus, the CAN communication between the station and the battery management system was disrupted. It was due to the fact, that the transmission wires of a single-ended unshielded CAN were laid along with power wires in the charging cable. After applying random modulation in this case, we got a significant reduction in the EMI spectra, however it did not help reduce the disturbances! So we started building intuition on why this was the case.

Recall the demonstration already published on this blog, which shows how picking out a particular time frame or the number of samples will lead to different results, when the time domain signal is tranformed into frequency domain. Well, the EMI receiver is not performing an FFT, but we can picture the phenomena, bearing in mind this example. What matters is that by changing one parameter (time or number of samples) we get signifcantly different results. The same can be thought when considering EMI receivers. If we play around up with the dwell time of the equipment, we might catch the amplitudes and frequencies that we are not able to catch using the standard setup, frequencies and dwell times. If you struggle to think about this, think about something simpler: a TV monitor recorded with a phone. How is it possible that your eyes can watch TV and enjoy your favourite TV series, but as soon as you record it on a camera you cause yourself only suffering?

Random modulation spreads the energy over a larger range of frequencies, therefore it provides a "smoother" frequency plots, than the "spiky" one produced with deterministic modulation. Okay, okay, but when you look at the signals from a time-domain perspective, the only thing that changes between the deterministic and random modulation is the times at which the interferences occur... The EMI receiver is simply not fast enough to catch all of them!

We have now built that intuition, but why actually do we care? As soon as the device passes the tests, everyone is happy, right?  Well, tell that to the people who used CAN in the aforementioned example!

In order to prove our point, we developed two mathematical models: one to describe the effect of random modulations onto the serial communication system and one to describe the impact of deterministic modulations on such system. The models were verified experimentally, and further developed. Moreover, using the models in simulation, and changing the parameters (frequency, time of the interference and the number of bits sent in a frame) we were able to compare the models for the very first time. Without getting into much details, I need to mention that the models represent the phenomenon from a statistical point of view.

As the results, we showed that on average, there is no difference between using random and deterministic modulation when it comes to their impact onto serial communication systems. A takeaway message is therefore: do not claim that random modulations solves your EMC problems, when you apply it in such, serial communication systems.

...unless you have counterargument, and then join the discussion!

© Copyright 2022 SCENT