kwest_tech
James LoganNew gear!! Time to find out if there's anywhere I've made unintentional antennas!
With 3 switching converters, a 48MHz CMOS clock, and ethernet, there's a nonzero chance of gremlins in here...
RMII ethernet micro-to-phy routing! This is the noisiest thing so far. Might have to move those traces to an inner layer for shielding!
Sanity check: nothing at all detectable outside the aluminum enclosure, as expected!
RealGene ☣️@ponderingpothos
Shutting off the bathroom fan causes the USB camera plugged into my spouse's laptop to disconnect.
Shutting off the bathroom fan causes the USB camera plugged into my spouse's laptop to disconnect.
@RealGene Lmao that's incredibly bad. Probably a spark gap emission from a commutator?
@ponderingpothos
Something along those lines, or back-EMF with no snubbing.
The Windows laptop helpfully chimes the USB disconnect about 1 second later.
Something along those lines, or back-EMF with no snubbing.
The Windows laptop helpfully chimes the USB disconnect about 1 second later.
Snooping around with the tiny loop shows the B signal near the switching converter is essentially entirely concentrated at the inductor, as one would hope. E-field is quiet, so despite the high near-field B, it's probably not radiating much.
Deimos DAQ EMC sweep is done!
End result:
* No detectable E-field signals
* Some noticeable B-field signal, but all decays within <1cm distance of sources
* SMPS, CMOS clock, micro, and RMII traces are the notable emitters
* No notable signal near cables (power converter input filters are working!)
* Whisper-quiet near the precision voltage references, all analog frontends, and even active PWM outputs
Actionable bits:
* Move RMII routing to inner layer
* Series resistor for CMOS clock? maybe
* Maybe copper tape over SMPS for prod
* Maybe copper tape over micro and clock if damping doesn't work
Ben@ponderingpothos Buck converters couple switch noise to their inputs, adding a damped LC filter is the trick.
https://www.ti.com/lit/an/snva886/snva886.pdf?ts=1774094776683
@0h00000000 Thanks! Right now, I just have a lightly damped RC filter on the input, and am leaning on the power supply cable's ferrite for inductance.
This particular board only pulls about 100mA max, so I can probably afford to add a little inline damping from a ferrite.
@ponderingpothos You want to avoid power cable and input traces acting like an antenna and impacting your measurements. Check out the TI Web Bench tool for your part, there is a check box to add a EMI filter.
Basically more inductance very close to the VIN on the IC is best, maybe 2.2uH or something, and then if the part needs maybe 4.7uF input capacitance, then you add a damped capacitor (~1-3 ohms) that's 4x that size between the inductor and the input cap, and add bulk capacitance on the other side of the inductor with the power line. This will slow down the edges coming into the IC and force it to use more of it's input capacitor.
@ponderingpothos Edit - this will slow down current spikes/edges coming into the board.
@0h00000000 Thanks! This is awesome, they even have my exact chip and circuit in webench, and it templates out an EMI filter just like you said.
It recommends 220nH for the input inductor, which is only about 2 times the 0.1uH minimum possible inductance due to the supply trace and ground-plane return path from the connector to the converters on the other side of the board.
Webench's recommended capacitor values also match what I have in place now.
So... looks like, while I hadn't deliberately built it, I more-or-less have the EMI filter in-place already! That might explain why I haven't picked anything up on the near-field probes near the supply cable
Routed the RMII traces on IN2 and re-oriented the ethernet PHY to keep the traces shorter and cleaner.
I might regret co-routing the traces this tightly, but as far as external emissions go, this should quiet things down.