EMC 30-50 MHz emssions

Question

I was trying to pass EMC tests for my 48V 5A BLDC driver and failed in 30-50 MHz range. I think I found the problem in my drive circuit, but I have a few questions that got me thinking, maybe someone can help me to answer.

1. If my PCB is not too large (15cm x 5cm) 30-50 MHz emissions are mainly radiated through power cables? If 30-50 MHz 1/4 wavelength is 2.5-1.5 m then I need to mainly look at cables with length around 2.5-1.5 m?

2. If I have 48V and GND which radiates and I want to decrease emissions. What is the difference if I place ferrite around 48V and GND together or I place two separate ferrites one for 48V and one for GND? If I place one ferrite it will only filter common mode emissions?

3. Will inductor chokes will work better then ferrites for 30-50 MHz range? Because there is no ferrite to work in such low frequencies.

Answer 1

If my PCB is not too large (15cm x 5cm) 30-50 MHz emissions are mainly radiated through power cables?

Yes, that is a general truth and, that means you need to prevent conducted emissions because on a long(ish) cable, conducted emissions become radiated emissions. The good news about conducted emissions is that you can actually simulate the results (with a little care) so, set up your circuit in a simulator and mimic the circuit of a LISN and see what the differential conducted emission levels look like (basically the current flow in one power wire).

You'd probably be more interested in CM conducted emissions because they are more likely to becomes radiated emissions however.... CM conducted emissions will very much be influenced by high-speed differential currents produced by your board/circuit and that is prime work for a simulator (the current flow in one power wire).

Guessing about what inductor or ferrite to fit here or there is really just guessing. Set up a simulation and then introduce capacitors, inductors and ferrites to reduce differential conducted emissions.

Answer 2

If my PCB is not too large (15cm x 5cm) 30-50 MHz emissions are mainly radiated through power cables?

Maybe. Maybe not! Point is, a shorter conductor can still emit – maybe not as efficiently, but will still do. Since we don't know how much energy is not emitted, we can't assume it's not also shorter conductors/components.

If 30-50 MHz 1/4 wavelength is 2.5-1.5 m then I need to mainly look at cables with length around 2.5-1.5 m?

No, see above.

What is the difference if I place ferrite around 48V and GND together or I place two separate ferrites one for 48V and one for GND?

In a first approximation, the ferrite helps against high-frequency currents flowing through it – and that flow is calculated over its whole cross-section.

Will inductor chokes will work better then ferrites for 30-50 MHz range? Because there is no ferrite to work in such low frequencies.

Probably, yes. Shielding, and most importantly: figuring out where that power comes from help better than ferrites here, too, most likely.

Answer 3

Let us consider resonant parasitics.

1uH and 1pF will resonate at 150MHz.

10nanoHenry and 100pF will resonate at 150MHz.

1u and 100pF will resonate at sqrt(100) lower, or 15MHz.

100nanoHenry and 100pF will resonate at 45MHz.

And 4" of PCB trace (not over Ground) will provide about 100nanoHenry.

Thus PCB traces, and power driver junction parasitics, in resonance, may be the source of the irksome energy.

=================================

A product_development team once came to me with "Our LIN transceiver PCB has failed FCC". All looked good, about the PCB. But there was a 6 dB violation of the Pass/Fail mask, up at 600MHz.

Eventually I realized the PCB trace, for VDD bypassing, ran an extra centimeter around the LIN IC. With a few milliMeters of gap under that VDD trace, as the Ground Plane was cut out in one location. And that ---- the additional 5 or 10 nanoHenries, PLUS the onchip VDD/SUBSTRATE junction capacitance, was the moderate_Q resonator.

To resonate at 150MHz, the Luh * Cpf product must be 1. Thus 10nH and 100pF are resonant at 150MHz.

To resonate at 600MHz, the Luh * Cpf product must be 1/16, thus 10/2 nH and 100/8 = 12 pF are resonant at 600MHz.

Apparently the LIN IC has many many VDD_well ties/contacts, and many many Ground/substrate ties/contacts, and the myriad paths thru the reverse_biased junctions provides a LOW_LOSS onchip capacitance.