2
\$\begingroup\$

I'm building a light device with LEDs connected in series. One device could hold up to 200 LEDs in a strip of 1.5m.

I'm also supposed to take care about the good choice of a PWM frequency (using a microcontroller) to drive the LEDs (via LED driver) as if it's to low you may have flickering (show up easily with video cameras) but if it's to high I'm afraid of EMC so it cannot be certified for selling. Two questions arises then:

1) Who is the biggest problem here? Should I be concerned with EM wave generation somehow or just EMI via circuit?

2) Who should be the main EMI generator? The LED driver (switching), the LED itself, some other component or all together?

Thanks

\$\endgroup\$
7
  • 1
    \$\begingroup\$ The main EMI is from anything carrying a switching current (or voltage) of significant length. \$\endgroup\$
    – Andy aka
    Commented Dec 1, 2016 at 15:06
  • 1
    \$\begingroup\$ EMC will likely come from too fast edges rather than too high frequencies, it isn't hard to do a useful 40kHz switching frequency \$\endgroup\$
    – PlasmaHH
    Commented Dec 1, 2016 at 15:08
  • \$\begingroup\$ Have you considered adjustable constant current output? \$\endgroup\$
    – winny
    Commented Dec 1, 2016 at 15:12
  • \$\begingroup\$ Adjustable constant current shifts the LED color so we opted for PWM. So the whole microcontroller+traces+driver+LED circuit is a concern? Not specifically the LED or the transistor inside the driver? Are there work arounds since the edges are fixed by the microcontroller+driver switching speed? \$\endgroup\$ Commented Dec 1, 2016 at 15:32
  • \$\begingroup\$ Your power supply an particular will radiate EMI when you connect it to your antenna/strip. But everything you connect will contribute. Yes, you can slow down your flanks, add filters and CM chokes. \$\endgroup\$
    – winny
    Commented Dec 1, 2016 at 17:36

2 Answers 2

2
\$\begingroup\$

1) EMI is caused by switching currents. Look at the edge of your PWM signal, sharp, fast edges cause higher spike in current, and worse EMI issues. The frequency is pretty much irrelevant. If each LED is routed well so there is a good current path (low induction for the frequency used), there will be no issues with EMI. If you need to know how to design a good current path, I would need a lot more than this answer to explain it to you.

2) The EMI will be caused by current. Look at the current loop, check the full current path to and from every current source, which includes capacitors, ICs and the LEDs themselves.

I'm pretty sure you're over thinking this EMI issue. At low frequencies (10kHz when looking at distances of <50mm) the current path is pretty predictable, as long as there is reasonable decoupling and good PCB traces, you shouldn't have an issue.

\$\endgroup\$
2
  • \$\begingroup\$ Thanks for your answer. I have some theoretical background on wave theory and electromagnetism but not much practice in good design rules. Do you by any chance have any recommendation of good material/website to read about it (specifically for EMI issues, if possible)? \$\endgroup\$ Commented Dec 6, 2016 at 9:59
  • 1
    \$\begingroup\$ The best material I've found is going on some EMC courses: York EMC, University of Oxford and TUV all do some good two or three day courses for around £1000. Alternatively, there's the famous book by Tim Williams (EMC for Product Designers). or the less famous PCB Design for Real World EMI control by Bruce R Achambeault, about £100 each, though some websites will have them cheap. If you want something less expensive, then I have no recommendations for you, you get what you pay for. Websites are not a good source in my experience, though asking very specific questions here may be your best bet. \$\endgroup\$
    – Puffafish
    Commented Dec 6, 2016 at 10:20
1
\$\begingroup\$

Typically EMI/Radiation is going to be a factor of the Current Loop area, the Edge Rate and the magnitude of the switching currents. For instance if you are modulating the LED's with a 1KHz clock/PWM signal but that switch is changing state in 1ns, you have a 1GHz signal to worry about (and all the headaches associated with high speed signals). Make sure that your edges are as slow as they can be for your target efficiency and modulation frequency, and make sure that the current loop area is as small as possible. Best is to have an unbroken return path on an adjacent layer of the PCB to limit loop area, and try a slew limited switch. the parasitic capacitance of the whole design may limit this to some extent if you have some series resistance on the output of the driver circuit.

\$\endgroup\$

Your Answer

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

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