Nucleo H743, large spikes in Idd in regular pattern

[KMach.1]

I have the same problem as described and unanswered here: https://community.st.com/s/question/0D53W00000IUzefSAD/nucleo-h745ziq-drawing-current-spikes-from-usb-power-port-producing-noisy-5v i.e. bursts of current spikes each 18 ms.

The particular board is Nucleo STM32H743ZIT6, old chip revision Y.

It does no go away even when holding the reset button. Not even when the st-link is kept in reset by installing SB107 and the ethernet PHY is held in reset by removing SB177. There is no SMPS on the board.

I measured with two different oscilloscopes, either with voltage probe between Vdd and Vss as well as with a current probe (on a short wire jumper across the Idd header). I tried powering the board from USB or from a benchtop power supply. I tried heating up and cooling down the MCU, but the pattern never changes. The same pulse sequence repeats every 18 ms.

Is it some kind of bootloader or a debugging interface that is trying to establish a connection?

I would ignore it if it did not make any difference in performance, but unfortunately it is ruining the quality of the ADC data.

Attached are the captured waveforms.

Thanks for any suggestions.

[Uwe Bonnes]

Is VCAP decoupled as needed? Look with a probe at VCAP, is the disturbance eventually synchronous with the curreen surges?

[KMach.1]

Yes, the disturbance on VCAP (and everywhere else) is synchronous with the current spikes. The unsettling thing is the perfect regularity of the noise. When i set the scope to normal mode with 5 ms holdoff, the waveform can be zoomed and panned with barely visible changes. Its period is exactly 18 milliseconds, each burst consisting of 3 pulses 65 us apart, then 3 pulses 130 us apart, then 9 pulses etc., as pictured in the first figure above. This made me believe it is not caused by a voltage regulator, but something inside the MCU clocked with a relatively precise oscillator.

[Uwe Bonnes]

If you see it everywhere, I suspect some external strong disturbance. Do the same measurement on some other board w/o any CPU a nd perhaps in onaother plave far away from where you measured now.

[KMach.1]

Okay, now I feel dumb. It was all caused by the Ethernet PHY which I thought was disabled. But to keep it resetted one has to remove SB177 and add a bodge wire to GND.