LSM6DSO SPI Bug for high volume project

[ZarvanM]

Using the LSM6 over SPI with the RP2040 Microcontroller for a mission-critical project (flight controller for UAVs).

The SPI works well. Well in theory. In practice, 1 in every 20-100 reads returns corrupt SPI data (WHOAMI reads as 0x68/0x58). This issue is somewhat reduced when moving the sensor (worst when still) and when adding decoupling capacitors (reduces corruption dramatically).

The project itself is ultra high volume (10k+ units), and we have already invested a very large amount of time and resources into testing.

[TDK]

> In practice, 1 in every 20-100 reads returns corrupt SPI data (WHOAMI reads as 0x68/0x58).

> we have already invested a very large amount of time and resources into testing.

These statements are somewhat at odds. A 1-5% failure rate should be pretty easy to check.

 

No magic bullet here. Some things that can help:

• Lower clock speed and/or slower edges.
• Better shielding from EMI.
• Terminating resistor on SPI line with or without controlled impedance.
• LVDS or similar differential transceivers at each end.
• Decoupling capacitors.
• Reduced signal/wire length.
• Better grounding (avoid loops).

 

A UAV will have higher EMI than a normal application. Consider decoupling capacitors a must rather than a recommendation.

[ZarvanM]

Hi!

Thanks for the reply.

Yes, we have followed best practices, and the failure occurs on with bench testing, not yet on a UAV.

The 1-5% failure rate is influenced by:

1. Whether the IMU is untouched

2. Adding lots of electrolytics at input

3. Changing SPI speed (lower sometimes causes MORE problems)