Ground plane EMI project solution - SOLVED

[system]

This project has been ongoing, partly due to the learning curve and getting back to it after a long hiatus, working on solving the analog/digital EMI issues. Camera electronics seem to be very sensitive to digital switching. I would like guidance on some aspects of ground plane usage and connections to reduce EMI - see attached general layout/concept.

First, a general description - this setup does 3 things.

1. Cools a DSLR camera sensor to a sepoint point temperature (PWM switching through MOSFET)
2. Heats the sensor glass face for defogging - always on (Linear)
3. Heats the camera lens for dew protection -may be turned off (Linear)

A 12V SMPS (low ripple) supplies power to a TEC module and to a DC-DC convertor, which supplies 5V to the microprocessor and heaters.

Temperature control is a function of ambient air temperature vs system cooling capability - at start-up, setpoint is calculated to remain within the limits of the system given variations in ambient air temperature. A PWM value is estimated for the calculated setpoint/differential and is updated continually. Effectively, PWM dithers either side of setpoint. Maximum differential ~ 26C. Setpoint is maintained by reference to a temperature sensor attached to a copper plate conductor, which is fixed to the sensor back.

EMI

Several must do’s…
The camera chassis, which is its ground, must be connected to the system ground.
MOSFET Source must be connected to Arduino/Teensy GND
High speed diode Source to Drain handles large current spikes - or so it appears

This leaves the low side switching EMI, which is proving difficult to resolve. I have tried lots of different snubbers and switching timing strategies to no effect. The most consistent method is a 22uf cap, Gate to Source. But it’s not perfect. BTW it doesn’t make a noticable difference whether the Gate is driven by a Gate driver or directly by PWM.

Is the EMI associated with my setup a ground loop problem (discussed on this forum many moons ago)? I thought it was solved, but with so many dead ends, I’m not convinced.

There is room in the electronics box to add PCB ground planes. It’s tight, but doable. There would be 3 planes - 2 x GND, digital and analog connected by a high impedance wire, and, a +ve plate for incoming 12V from the power supply and distribution to the TEC and DC - DC convertor. However, I am a little unsure of how to set up the GND plates: that is, incoming and outgoing connections, so as to avoid bypassing any benefit provided by separating the GND sources. Eventually, all roads lead to GND.

Is it simply a matter of connecting each ground plane to separate GND points on the Arduino/Teensy? What about the power supply and DC - DC convertor GND connections. Finally, the camera chassis has to be connected to GND as well.

I’ve added a few pics of the project for context as well as some wave forms with and without the 400mohm resistor, which is intended to smooth/delay Gate switching - not sure how effective with and without the resistor. Electronics compartment attached to camera bottom.

OSC timebase 10us, high side resolution 10mv (red), low side, 1V resolution (blue).

The 30pH inductor - low side, removes spikes/stray inductance in the negative node of the cap - if that makes sense. Still learning about these concepts…

Fixed the layout diagram which showed low impedance between ground planes - should be high impedance…

[MarkT]

You ideally need a star-ground setup, keep all the high current spikes on different wires from
the sensitive stuff. If you can't arrange that you'll need faraday cage style screening
like you suggest.

You need to reduce the inductance of your high current switching circuit first though,
ie use twisted pair for all power and high current wiring, avoid current loops, they just
radiate loads.

I'm confused, you claim a 30pH inductor - that's impossible to make as even a few mm of
wire has orders of magnitude more inductance. Perhaps you meant 30nH ?

[system]

Twisted pairs is not something I had considered and will look at rewiring the power side.

The inductor is 30pH and was designed using an on-line calculator. It's 32mm long, 1 turn and 1mm diameter - 5v silvered wire. Not sure how accurate. Came in at 0.030nH.

[system]

Success! After a bit of reading on twisted pairs and more on ground planes (see pics attached), I am driving the Gate from the pwm pin without capacitors, snubbers, low pass filters, with no discernable switching noise evident in 30 second dark images (with the lens cap on). The dark frames appear identical to those taken without cooling activated. I am very, very happy with the result. Temp control is rock solid.

As can be seen from the images, there are two boards. A smaller one for the mosfet with 3 planes. Gate, Drain and Source, with a high power, (high speed diode between source and drain), and the large one with power, analog and digital ground planes. The analog and digital planes are linked by a 100R resistor. The analog plane connects to one side of the Teensy and the digital, the second ground point.

I suspect that there is an optimal area for each plane, but for now it’s verified as the best solution to date. For imaging purposes, I don’t need to pursue this further.