STEP: A small EMC puzzle

Part 4: Fixing the clear DP

So now that we know the data products cause the spikes, and that the "clear" data product exceeds the limit, we thought of how to mitigate that issue. Apparently every DP command to the sensor creates a spike, so what if we partition the "clear" command into various smaller commands. We want to clear the same amount of memory, but with multiple DPS product commands. The result of partitioning the "clear" 16 times is shown on the scope picture here:

So apparently we can cut down the peak by partitioning the "clear" 16 times. We are already compliant, but to get further into a compliant regime, we partition the "clear" 32 times:

With this workaround we are compliant!

But some questions remain:

The spikes have smaller amplitude, but the amount of power didn't change, or even increased. Is that OK?
What actually causes those spikes? It looks like the spikes are generated at the points in time when the DPS command is issued, and apparently do not stay high long during the time when the DPS is actually processed. Why?

I'd really like to hear your ideas!