0
\$\begingroup\$

I tried to look up this kind of question in the archives but didnt find much usefull information. I also have another question in regards to EMI but I will ask this later in a seperate post.

Currently I am designing a PCB which is actually pretty basic, it has some switches(around 10), some leds, a buzzer and a few other outputs and a microcontroller and it is powered by 2 AA alkaline batteries in series, thus 3V. Product will be tested at a notified body like TUV rheinland, so far nothing really troubling. I have done this before, some issues were discovered and resolved. But this was for EMC according to CE guidelines and was tested with an EMI field strength of 3V/m. Since it is a product to be used in a car I now have to test against automotive EMC and thus against other test levels up to 30V/m. Previous products I tested had to comply with 3V/m, I also did some testing at 10V/m and both products I tested failed but since only 3V/m was required that was ok. Since I am now designing for levels of 30V/m I do expect more issues.

My question actually involves around this issue, I am actually struggling to comprehend how much my design will be affected at these levels of EMI(30V/m). How much will the digital inputs be changed by the field, how about my outputs? Are my pull down/up resistors to weak? How much will my supply voltage be affected? Is there any way I can guess how good my countermeasures to the EMI interference will perform?

To make it a little more concrete I will give an example. I have several digital IO with a pullup of 100k to VCC and a switch to ground. I also have some outputs which need to be able to switch up to 2Mhz so added an RC filter of 1k and 100pF. I can imagine a RC filter of 100R and 1nF might be a stronger filter and harder to affect but it is more a gutt feeling than that I can explain it with hard numbers or facts. Since frequency isnt really important for my switches and thus digital IO should I settle for 1k-100pF filter/100R-1nF RC filter or go high. Perhaps 100R/100nF?

\$\endgroup\$
  • \$\begingroup\$ If you go outside naked you might get sunstroke or you might freeze to death. So put reasonable clothes on and carry a coat in case the weather gets worse. If you have problems with your inputs and they could be "more protected" with better earth plane usage and bigger value resistors and capacitors (without a detriment to performance) then what are you waiting for? \$\endgroup\$ – Andy aka Sep 1 '17 at 16:20
2
\$\begingroup\$

30 V/m are a lot, also to generate them (!). I make immunity tests following EU automotive directive on trucks, trolley buses, buses, and in a really few cases they got into troubles: 1) when the direction light relay was clearly a chinese product (with small "c" because there are superb Chinese products); unprotected inputs of the internal timer; replaced with compatible relay from a nearby Mercedes, everything went ok; 2) displays to show bus number, destination, etc.; long cables from head to tail pick up noise, and incomprehensible and unbelievable changes during production of the connected equipment did the rest: one type worked, the other absolutely not.

Experience confirms that the problems arise at two frequency intervals: a) about 70-100 MHz if applying the e.m. field with a biconical antenna, that might resonate with the ground, however good its balun; this increases local field intensity; b) 250-400 MHz coupling well even with short cables (1 m or so). c) probably at higher frequency cable losses are higher and the resulting disturbance is not so dramatic; never seen issues, but, I haven't tested new cars with a lot of smart/fast electronics on-board.

So, going to the question: - protecting your inputs with low-pass RC filters is advisable; - if your signal bandwidth is large, you cannot reduce the RC cutoff frequency: so, you may put a ferrite on the incoming cable (common mode), useful especially when things turn worse at high frequency; - pull-up are a good thing; 100k is to keep consumption low; on I2C (possibly going to a display or keyboard far away) I normally put 4.7k; for switches and remote contacts it is ok, maybe going down a factor of 5 to 22k might be useful; - to pass tests of conducted disturbance (RF or transients, as per ISO automotive std), put a transient voltage suppressor that clips disturbance to the minimum needed to let the signals do their job, but reduces the energy of incoming disturbance; - if your cables are long, they pick up a lot of noise (of course): common mode noise, usually; if you do not have a good ground on the pcb, bonded together with the enclosure, you may want to terminate the cable shield on the vehicle chassis; diff mode noise may occur in case you have unsatisfactory terminations (CANbus is extremely susceptible to bad terminations, that may cause not only stationary waves, but a strong common-to-diff mode transformation; I saw problems also on locomotives, where it was used to make two signaling systems communicate).

Last: think of doing some pre-compliance testing using old VHF radios (140-160 MHz) and a short-range radio (433 MHz-868 MHz), jerking around your product at close distance, with cables laid down in a more or less realistic fashion. If you have a broadband e.m. meter you can measure the applied field, otherwise you can calculate it. With nominal transmitted power P and gain G (linear), plus distance d, you know, E is given as E=sqrt(30*P*G)/d [V/m]

\$\endgroup\$
1
\$\begingroup\$

Use a Ground Plane. Do not cut slits in the plane. Slits serve as antennas.

Do not make the keepback distances on vias so large that numerous adjacent via keepbacks join together and form a large slit.

For MCUs with many pins on a side, stagger the vias used to connect the pins down to inner layers, so the vias don't form slits that serve as antennas.

Did I remind you to use a Ground Plane?

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.