Mini Mainboard design flaw

[johu]

I'm currently working on the Zoe inverter, more specifically on the rotor exciter controller. While testing I noticed a strange, periodic sound from my little test inductor. Scoping the PWM showed it dropping out every second

To cut it short it turned out to be caused by the SDO bursts from the wifi module. The CAN transceiver pulls so much power from the 3.3V rail that the ADC reference voltage dips and thus all ADC readings are off:

Regular periodic CAN messages have the same effect, just not as pronounced:

I also checked the 5V rail but it is not or hardly at all affected.

I think the problem never showed because nobody used oic or ESP for periodic data polls. Also the other channels use more of the ADC dynamic range. The 50 mV/A from the exciter current feedback completely vanish in the noise floor and the P controller diligently tries to counter the seemingly excessive current by shutting down the PWM.

I will look for various hardware fixes, probably just more decoupling for the CAN transceiver. I'm thinking not all hardware may be affected because mostly a stronger 3.3V LDO is used and also more decoupling. Most of all Tesla boards have chunky tantalum caps on the 3.3V rail. The Mini mainboard is a bit minimalist in that regard.
Also I think the sine firmware won't care at all as it only uses heavily filtered ADC readings.

For now the recommendations are, mostly for Mini Mainboard based designs:
- Do not poll large amounts of data via OIC or the ESP wifi module while driving
- Try to use 100ms CAN send interval if practical
- Try to stuff most data into a single CAN message

[johu]

Added 110 uF of decoupling, still 30 mV

It could be the 3.3V LDO is just too weak

[johu]

Ok, decoupling doesn't really help. What yields much better results is installing a separate LDO for the CAN transceiver:

It's the super cheap JLC Basic part, so no sweat
https://jlcpcb.com/partdetail/TorexSemi ... 2MRG/C5446

Here's how I did it

[janosch]

Good timing, we were about to test the mini board with our little motor on the desk for the EV kit.

Saved me a headache, surprised this didn't show up on previous mainboard iterations, or did you re-work the power supply for mini?

[johu]

No, it's been like that for ages. It doesn't seem to noticeably affect operation otherwise, that's why nobody noticed.

The previous larger V3 mainboard still had the larger 1117 800mA regulator. Also all of Damiens designs use that one. So I reckon the problem might not be as pronounced on those

[rstevens81]

this problem would also probably made worse by the esp8266? although its obsoleted by the lillgo tcan now.

[bexander]

In my car (Yaris OI-board) I have patched in a MCP2562 CAN-tranciever. It is powered from 5V and can have the I/O signals from 3,3V.

MCP2562

I also use a separate 3,3V LDO to power the WiFi-module.

All this to avoid having spikes in the current and angle measurements due to fluctuating voltage to the ADC-reference.

The history can be found here:
viewtopic.php?p=19522#p19522
viewtopic.php?t=1085

[manny]

It is not the current handling abillitie of the regulator but the load transient response.

Had a look at the XC6206 datasheet cant realy beleave what it shows but.

For a load increase from 1mA to 40mA the voltage sags 0,5V

The 1117 on the other hand specs the drop in mV

With a load jump from 100mA to 800mA voltage sags 13mV

[johu]

Oh wow, that's a huge difference. So will go back to the 1117 then.

this problem would also probably made worse by the esp8266? although its obsoleted by the lillgo tcan now.

Fortunately on the mini mainboard there is no wifi module. Actually that was the reason to go for the smaller regulator. Never knew there could be such differences in transient response. Something new learned