4
\$\begingroup\$

I'm currently having trouble getting a product through EMC. Radiated emissions were comfortably under the limit, but conducted are failing badly.

Here are the results: enter image description here

The noise appears to be largely common mode, with the exception of the first few big spikes which are differential. The first of the big spikes is at the SMPS switching frequency. I'm guessing the next 3 are harmonics of that.

Here's a schematic of the relevant/noisy parts of the circuit (the rest of the cicuit appears to be very quiet. The noise is only a problem when these parts are enabled):

enter image description here

Anyone have any ideas on how I might tackle this?

Update:

The cable is a pair of discrete un-shielded wires, ~2 meters long, connected to the 'BATT' input and GND. The BATT input is the power supply input to the module, nominally 12V DC (vehicle battery).

The components (along with a micro and a few other bits) are all on the same PCB. PCB is 4-layer with an ground plane on an internal layer. Unused areas on top and bottom layers are GND filled. The module enclosure is plastic/ABS.

The SIM900 is a GPRS module. It was powered for the test and talking to the micro via UART, but not connected to a mobile network.

Edit Added PCB Layout after comments requested layout

enter image description here

\$\endgroup\$
7
  • 1
    \$\begingroup\$ Conducted emissions - where is the cable in the circuit diagram of the problem part? \$\endgroup\$
    – Andy aka
    May 13, 2016 at 17:52
  • \$\begingroup\$ The cable is connected to the BATT input and ground. \$\endgroup\$
    – Skov
    May 13, 2016 at 18:09
  • 1
    \$\begingroup\$ C42, C36 and C37 would appear to be doing a bad job at filtering out the conducted emissions coming from U11 - possibly due to internal inductance of capacitors? Would have thought a low value cap placed near the U11 Vin may help along with a small series inductor ( this only speculation / conjecture on my part) \$\endgroup\$ May 13, 2016 at 18:22
  • 1
    \$\begingroup\$ You really need to supply a board layout if possible and a picture of your setup. \$\endgroup\$
    – Voltage Spike
    May 13, 2016 at 18:43
  • 1
    \$\begingroup\$ Check this link. Especially the 3-part video series. I've never read anything so clearly explained. learnemc.com/emc-tutorials \$\endgroup\$ May 14, 2016 at 3:16

3 Answers 3

1
\$\begingroup\$

As a rule High frequency signals will always take the path of lowest impedance to get back to the source (or radiate). I can't speak for your whole setup because you have currently only posted a schematic.

I can tell you this, if its conducted emissions that is you problem on the battery cable, then you are making a nice antenna out of you battery cable. So applying the rule: the current's most preferred and lowest inductance path is through your cable and radiating out into the world. To find out if this is the case, increase the high frequency impedance of the cable. This can be done by adding inductance, either by ferrite or by adding inductance on the PCB with inductors that attenuate high frequency "noise".

If it is your board radiating, then you probably don't have a sufficient ground plane or you have made some nice slot or dipole antennas. But I'll digress on speculating since you haven't provided any information on the PCB.

\$\endgroup\$
1
  • \$\begingroup\$ I've just added pictures of the layout. It's definitely the lower end that's the problem, I don't have any issues above a few MHz. \$\endgroup\$
    – Skov
    May 13, 2016 at 19:42
1
\$\begingroup\$

Increase C41, I think. Try double or quad mounting this component to see the effect (you can use the switcher harmonic as a proxy, if you have anything able to measure it). You also need some inductance on the cable side of C42, this will help, particularly if you add another low impedance capacitor across C42. You could also try adding some small resistances into the circuit, this can sometimes absorb the energy you're trying to suppress.

Something else to try, add another 220uf (or 10uf smd) where C41 is.

You don't get a direct reliable answer, this is mostly trial and error until you've done it a few times with a similar design.

\$\endgroup\$
0
\$\begingroup\$

You have bigger problems than you think, and the solution requires both changes to the schematic and a proper layout.

The differential mode conducted emissions at the switching frequency and its first few harmonics should have been expected. The battery voltage is feeding what looks like a buck converter. That's going to draw power in pulses at the switching frequency. There's only one half-hearted attempt to deal with this.

The common mode conducted emission seems to be general noise in the board making its way outside via the power cable. The first way to address this is proper with grounding. You have all the high frequency currents running across the ground plane, with no apparent attempt to isolate the local loop currents of the power supply and the microcontroller. You don't have a ground plane anymore, but a center-fed patch antenna. I've gone into this many times here, so no point repeating.

The last resort to address the common mode emissions would be a common mode choke on the power input. Again though, you need to address the real cause properly first.

Addressing emissions is not something you do at the end after the tests fail. This is something you think about and address in various ways all thru the design process in both the schematic and the layout. There are usually no quick and cheap fixes after the fact.

You are now stuck in the uncomfortable position of not having done this right and ending up with exactly the result that should have been expected. Go back and do it right, then learn from this and do it right from the beginning next time. You should be able to salvage most of the schematic, but the layout needs to be a do-over. If you don't understand why or how to fix this, then get someone that does. Watch and learn, then you can do it yourself next time.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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