Hi everyone,
I’m testing a custom STM32-based control board that reads a pressure sensor through the ADC. The readings are stable until a relay energizes, then the ADC value shifts noticeably for a few milliseconds before settling again.

The relay coil is powered from the same 12V input, and the MCU uses a 3.3V regulator. I already added software averaging, but it only hides the symptom.

My questions:

1. Would this be caused by supply dip, ground bounce, or EMI from the relay coil?
2. Is adding more local decoupling near the MCU ADC pins usually enough?
3. On the PCB side, should relay current return paths be completely isolated from analog ground areas?

 

hi eva,
Without a circuit diagram and a photo of the project it is difficult to give a detailed answer to your questions.

E

 

1. Supply dip, ground bounce, or EMI?

Most likely a combination, but ground bounce + supply disturbance is the main suspect.

When the relay energizes:

• Coil current creates a sudden surge on the 12V rail
• Return current can flow through shared ground impedance .... ground bounce
• That shifts the ADC reference (or MCU ground) ..... readings jump temporarily
• EMI from the relay is usually secondary unless routing is very close

So the ADC shift is often not “noise in the signal”, but movement of your reference/ground level.

2. Is adding local decoupling near ADC pins enough?

Helpful, but not sufficient alone. Decoupling near MCU/VDDA helps:

• Smooth fast transients
• Stabilize ADC reference locally

But if the issue is ground movement or 3.3V rail dip ..... capacitors at the MCU won’t fully fix it. You also need:

• Proper bulk capacitance near relay supply entry, check this: https://www.ti.com/lit/an/slvaft0/slvaft0.pdf
• Possibly separate analog supply filtering (ferrite + cap for VDDA)

3. PCB: should relay return paths be isolated from analog ground?

Short answer is Yes ..... but with correct structure. Best practice:

• Keep relay coil current loop completely separate from analog/ADC return paths
• Use a single-point connection (star point or controlled tie) back to power ground
• Do NOT let relay current flow through analog ground copper
• Maintain a solid ground plane, but control high-current return routing, read this on grounding system: https://www.aivon.com/blog/pcb-manufacturing/how-to-design-an-efficient-pcb-grounding-system/

You don’t need separate ground planes necessarily ...... you need separation of high-current return paths, not fragmentation of ground. This will help: https://www.pcbway.com/blog/Enginee...CB_signal_return_and_cross__segmentation.html

Your symptoms strongly point to shared ground impedance + transient supply disturbance .... not ADC noise itself. Fixing return paths and supply integrity will be far more effective than software filtering or local decoupling alone.