Strange wave-like measurement output on CW305


I am trying to test a masked design on CW305 (NAE-CW305-04-7A100-0.10-X) with Picoscope 5000 series. The design is relatively big and consumes 50% of the LUTs in CW305. I basically run some experiments, collect the measurements and then plot the t-test figure for the implementation.

I was observing wave-like structures in my t-test, which I could not explain and I decided to plot the raw measurement data. Then I observed that the wave-like pattern also appears in the measurement data after an operation causing high vol/power consumption.

In the attached plot, you will see the output of +20dB amplified SMA connector for a series of operations involving polynomial arithmetic.

In the beginning, the measured data seems ok for a short time. Then, I start an operation showing high vol/power consumption and creating a lot of spikes. After that operation, this wave-like pattern appears and it resides during the whole measurement. If I keep the circuit in idle long enough after that operation, the wave gradually diminishes.

Did anyone face this behavior before? Do you have any suggestions to remove this wave-like pattern immediately after the operation?



To some degree this is unavoidable, especially with an FPGA target which can have huge and sudden changes in power draw. Essentially, you are seeing capacitive effects. Some of it may be from CW’s analog front-end, some of it from the target itself. I don’t think it’s possible to fully eliminate this in practice. Nor should it be necessary: side-channel attacks leverage the relative differences between power measurements, not their actual values.

I agree that it can be annoying; certainly in the case of TVLA where you want to identify potential leakage at specific clock cycles, and where you do care about absolute measurements (i.e. the 4.5 threshold). You may have some success in applying post-processing to the signal, but I don’t have specific recommendations for that.