←back to thread

PicoEMP – A low-cost Electromagnetic Fault Injection (EMFI) tool

(github.com)
55 points transpute | 9 comments | 25 Jun 25 03:05 UTC | HN request time: 1.349s | source | bottom
Show context
fxtentacle ◴[25 Jun 25 05:54 UTC] No.44374012[source]▶
Are there any affordable open source projects in the opposite direction?

I’ve built a rather complex robot with Nema steppers. Now something is messing with the control board and causing the CPU to glitch. Physically apart from the robot, the CPU board works fine for weeks. It also has a completely separate power circuit with optocouplers. So my guess is that it might be the power supply or one of the motors emitting electromagnetic interference. But how would I measure that?

replies(3): >>44374194 #>>44379027 #>>44382919 #
1. colechristensen ◴[25 Jun 25 06:34 UTC] No.44374194[source]▶
First step is an oscilloscope so you can actually look at what's happening on data lines.
replies(1): >>44374480 #
2. fxtentacle ◴[25 Jun 25 07:20 UTC] No.44374480[source]▶
I have one and the power supply lines (to the glitching CPU) seem stable with minimal noise. The CPU-integrated USB2 PHY also retains its connection even during the CPU glitch. A connected Linux workstation will not log any USB connection or disconnection events. And a CDC serial connection will remain open. The power levels to the stepper driver chips also remain stable during the glitch.

So to me, it looks like the glitching happens exclusively inside the CPU. It appears that I’m randomly experiencing the exact same issue that the PicoEMP in the original article can induce.

replies(3): >>44374742 #>>44375399 #>>44376978 #
3. nicman23 ◴[25 Jun 25 08:01 UTC] No.44374742[source]▶
are you sure it is not just the nemas causing a under voltage situation?
replies(1): >>44375409 #
4. ddeck ◴[25 Jun 25 09:49 UTC] No.44375399[source]▶
Just a reminder to ensure that you have brownout detection enabled on the CPU/MCU (assuming it's available) and are checking for it in your firmware. Supply drops can be very brief.
5. fxtentacle ◴[25 Jun 25 09:50 UTC] No.44375409{3}[source]▶
Yes.

1. Stepper and CPU use separate power supplies and are optically decoupled.

2. I’ve monitored min max voltage on the CPU caps and they are fine the entire time, even during glitches.

replies(1): >>44379435 #
6. colechristensen ◴[25 Jun 25 13:16 UTC] No.44376978[source]▶
Next steps: EMI (near field) probes for your scope.

https://www.rohde-schwarz.com/us/products/test-and-measureme...

replies(1): >>44384902 #
7. nicman23 ◴[25 Jun 25 16:54 UTC] No.44379435{4}[source]▶
optical decoupled uart?
replies(1): >>44384867 #
8. fxtentacle ◴[26 Jun 25 06:53 UTC] No.44384867{5}[source]▶
No UART at all: Control flow is through USB directly into the CPU, so the serial interface there is purely virtual. And then from CPU to stepper drivers it's Enable, Direction, and Step pins, all of them optically decoupled. The stepper drivers then have their own power supply and caps so that they can't interfere with the CPU power circuitry.
9. fxtentacle ◴[26 Jun 25 06:58 UTC] No.44384902{3}[source]▶
My scope is a SDS1204X-E and I don't believe there are any EMI probes that will work with it out of the box. The scopes that can do EMI (in addition to all the other stuff) are $5000+ which is why I was wondering if there are any more affordable single-purpose EMI measurement devices. Amazon is full with devices claiming to measure EMI, but since they usually advertise with people wearing tinfoil hats, 5G shielding burkas, and copper capes, I don't think those are useful for science.