Hi all,
recently I saw a MCU PCB, that commands Relays(mounted like 2inch away from MCU), that switch on and off incandescent or LED lamps. Sometimes when the load is big enough like 100W or more the MCU stops responding. It seems the pcb has problems with too long traces, that act as an antenna for EM disturbances. This made me realize how big the problem with EMC is.
So my question is having a badly traced PCB, is there any way to make it EMC proof like transils, GDTs, MOVs or LC RC low pass filters or ferrite on the 5v DC supply line, where I measured up to 2 - 2,5v attenuating spikes with 100 -250 ns period (what frequency is this : 1-4MHz ?)?
Does such a voltage spike for such a small time cause the problem with MCU?
Or may be some of the I/O pins also acts as an antenna and needs filtering?
So what is best to use as filter on power supply line and on input pins?

Thanks in advance for the help!

 

Hello,

Do you have decoupling capacitors on the board near the MCU?
Here is a thread with information on them:
Decoupling or Bypass Capacitors, Why?

Bertus

 

Bertus,
its a really good thread you suggested, Thank you!
Yes, there are decoupling caps 100nF near the MCU and other ics

 

Hello,

Do you use diodes accross the coils of the relays?
If so, what kind of diodes?
Best use fast diodes there.

Bertus

 

A regular 4007 and 4148 are used on the coil side of the relay and the contact side has RC snubber

 

Hello,

How much current does the MCU take?
You could try to "isolate" the power for the MCU from the main power, as shown in the schematic below:



The following page of the EDUCYPEDIA will have links with more info om EMC/EMI:
http://educypedia.karadimov.info/electronics/emc-emi.htm

Bertus

 

It is done,
the relays operate at 12v separate power driven by mos transistor connected to a pin of the MCU, which operates at 5v.

I think its more like electromagnetic disturbance, because traces are made too much longer than necessary, there is no grounding planes on the second layer.

I tried putting MOV on the relay contacts with no noticeable effect - I still see a spark inside the relay and the oscilloscope reads again some 1-2 volts spike for several dozens of nano seconds. Maybe RC snubber and the MOV reduce the energy of the spark, and so reduce the EMI effect to the other parts on the PCB, but the spikes n the 5v are still visible, so I thought I need something like low pass filter filtering all harmonics for the 5v or something to shunt them to ground like Transient voltage suppressors. I was thinking of buying 5,5v MOV and transil diode because their response time is in the nano seconds range. Or should I use a LC RC filter instead?

 

Hello,

Did you see my last post?
I added a link that might help.

Bertus

 

no I missed it...
I'll check it out

Thanks again for the useful guidance!

 

In my experience, the problem almost always stems from ground loops where the inductive load current flow in the same traces/wires as the ground wiring associated with the MCU. The solution is to separate the two as much as possible. Bypassing will not fix this problem.

Techniques to do this include:
1.Use a separate power supply for the inductive loads.
2.Keep MCU analog ground bus separate from MCU digital ground bus separate from the load ground bus. There is only one place where the respective ground busses touch each other ("Single Point Ground").
3. The ultimate way of keeping the load currents away from the MCU wiring/traces is to use opto-isolation between the two subsystems.