Hello ST Community,
I am working on a commercial vending machine controller and would like to get feedback from ST engineers and experienced STM32 designers regarding a system architecture change.

Problem Statement

The current single-MCU design is constrained by:

• Timer availability, particularly for stepper and DC motor control
• GPIO interrupt count, driven by sensors and safety inputs
• GPIO pin availability, limiting further expansion
• Increasing real-time jitter when multiple loads operate simultaneously

These constraints are no longer addressable through software optimization alone and are limiting both scalability and reliability. This has led us to consider a multi-MCU architecture.


Proposed Architecture
Master MCU

• Device: STM32H563ZIT6
• Responsibilities:
  • All sensor acquisition (ADC and digital)
  • Control logic and system state machine
  • Recipe execution and sequencing
  • Safety checks and fault handling
  • Communication with UI/HMI
• The master MCU does not generate direct motor timing signals.

Slave MCUs (2x)

• Dedicated load-handling controllers
• One MCU for stepper motor control
• One MCU for DC motor / PWM-based load control
• Slave MCUs perform only deterministic actuation
• No sensor acquisition and no business logic are implemented on slaves

Inter-MCU Communication

• Primary option: CAN (FDCAN), chosen for arbitration, fault handling, and robustness
• Secondary option: RS485, considered for lower-priority or legacy expansion

A block-level architecture diagram has been attached for reference.





Points for Discussion

1. Is this master + load-slave MCU architecture a suitable long-term solution for vending machine or appliance-class products?
2. From an STM32 architectural perspective, are there recommended design patterns for systems with heavy timer and interrupt usage?
3. CAN vs RS485 for multi-MCU control:
   • CAN appears preferable for arbitration and fault handling
   • Are there any STM32H5-specific considerations or limitations?
4. What are the common pitfalls in such systems (boot sequencing, synchronization, fault recovery, debugging) that should be addressed early?
5. From a product lifetime perspective (8–10 years), does this architecture appear robust and maintainable?

I would appreciate feedback from:
Developers with multi-MCU STM32 designs
Anyone with experience in CAN-based industrial / appliance systems
Thank you in advance for your insights.
Best regards,
Praveen Kumar G,
Boatload Minds Pvt Ltd.

 

Welcome to AAC!

Looks like a good plan, in theory. In practice, it comes with its own complications.

I would go back and examine each one of the limitations of a single MCU solution. There are ways of overcoming each limitation.