Minimising EM noise from PWM switched fairly lights?

Cheapest, simplest, quickest: get an old-school battery powered pocket AM/FM radio. Find a quiet spot on the AM broadcast band and start sniffing your unit under test for RFI.

piglet

---------------

Analogue TVs with rabbit-ears are also good at the same task. Built a simple, 18V linear reg once using a BD139 while watching CH9 on TV in the same room.

When energised, it wiped out the pix - a 10nF ceramic soon came to the rescue.

Another time, I was getting regular, slow impulse noise spikes breaking into my scope and guitar amps.

Tracked it down with a AM portable to a local Pharmacist, 65 metres away, with it a dodgy neon sign in the window.

Who woulda thunk ?

..... Phil

Why bother for a private project ?

Thanks for that :)

I could baseline the current controller on the same lights with it first to get some idea of what level is acceptable.

I'd like to not piss the neighbours off when their TVs or radios stop working :)

Or worse, interefere with some comms system that matters. I'm in England, my plot is 100x75 feet so we're all fairly close to each other.

You want a small capacitor across the load to provide a local path for higher frequencies to attenuate them further. Try this simple URL:

If you lowpass it to below 30kHz you probably won't annoy any hams. (or jam local 60kHz radio clocks)

AM radio tuned to an otherwise quiet spot. You would have to be doing something really silly to end up broadcasting MHz harmonics though.

It's the edges that generate the high frequency content, so minimising the PWM frequency isn't all that helpful.

The aim is to stop the high-frequency content getting out into the string o f lights (which acts like an aerial broad-casting and high-frequenes you ma y push aronud the loop), so put the choke as close as possible to the H-bri dge, and immediately follow it with a capacitor to the return connection of the H-bridge.

This is an L-section filter. As second one right up against it will attentu ate any high frequency content even further.

Any wound inductor has interwinding capacitance, and a self-resonant freque ncy. It doesn't do much to attentuate frequencies higher than the self-reso nant frequency. Ferrite beads and chips aren't wound, but don't have much i nductance.

Transistor radio - as everybody else has pointed out.

Hi Martin,

That's great - thanks! I'll play with some numbers and see what looks sane.

Sounds like a job for a bit of sold fashioned tagboard. I'll have 4 modules (filter, DC-DC, H-Bridge and Pi) to mount on an acrylic plate to sit inside of an airtight food/fridge box which is ideal for the job.

Gland the cables in and out and it'll sit on the grass right next to the lights it's driving. The plan is to use a Pi Zero/W and access it over WiFi - 3 sets in total, one for each fairy string.

It's more for my son who wants to dabble with some coding and asked if we could make a controller for our existing Xmas lights.

Thank you - good design limit. I have 60kHz radio clocks, so hopefully any badness will be apparant to me fairy quickly.

I'd hope it would be hard :) Actual drive frequency depends on whether the Pi's 2 hardware PWMs are phase locked which which case I will use 2 to drive a single H-Bridge. If not, I'll have to do it in software and it will be really slow. I have an el-cheapo logic analyser USB widget so I'll test that soon.

If this works, next will be to make a 4th box but drive some neopixels in fairy string format which will add a dash of colour to the whole thing.

Many many thanks for your kind suggestions :)

At a low PWM rate, say 100 Hz, anything radiated in the AM band will be ballpark the 10,000th harmonic of the switch frequency. The energy will be nil.

Slow down the switching edges a bit if you like. Add mosfet gate resistors to soften the edges.

Why not deal with the RFI by not creating it? I'm not sure what your setup is at the moment but driving with a DC current should be pretty much as si mple as PWM. You talk about adding a choke. I'd be willing to bet if you look at the resulting circuit it is a short stone's throw from being a DC/D C converter.

Drive with DC and you only need to deal with residual high frequency noise you can stop with a simple filter.

I'm one of those people who see and are very annoyed by the blinking tailli ghts on so many cars. Cadillac was the first car I noticed it on many year s ago and have always been the worst offenders with huge taillights on some of their vehicles. More recently they seem to have changed their design, either by bumping up the frequency or by using DC.

Interesting that you can't effectively google this effect because there is a distinct effect from replacement LED turn signals blinking too fast becau se they don't draw enough current for the blink circuit to work correctly, called hyper flashing.

Do cars get EMI tested?

Easiest: use a low frequency like 1kHz to avoid flicker, and a switching speed that's ponderously slow like 10us to avoid much radiation. Maybe still put an LC across it (with a cutoff of some 100s kHz), with damping so it doesn't ring, to be sure.

Slow switching can be done by putting an R+C across the MOSFET D-G and a relatively generous series gate drive resistor. The R's prevent oscillation (capacitance from G to D/S is a no-no, you'll make an oscillator at 100s MHz that way) and the feedback network sets bandwidth as if it were an amplifier (which it clearly is during switching edges :) ).

Best efficiency: use an adjustable DC converter. Filtered input and output, high switching frequency, steady output. Typically you'd use a buck or boost converter (depending on supply, you might need a flyback or SEPIC configuration rather than boost, but, close enough) and either control the current setpoint node (external compensation type) or tease the voltage feedback pin with a resistor (thus shifting Vout).

In the first case, you try to avoid generation of harmonics; in the second case, you go all in, then filter them as close as possible to the source.

Additional filtering may still apply, on account of the first case, the impedances between all the nodes mean that the ~5ns edge you get from the MCU pin driver or IO expander or whatever it is, can feed forwards through everything and get into the wires. An RC in front of the gate R can help with that, for example. Or in both cases, there may be ground-loop noise, due to return currents from logic signals, or switching currents.

Tim

That was about as much as I understood about the whole thing :)

Interesting... Sadly won't be able to do that as I'm using pre built H-Bridge modules based on the L298N

Sorry for wimping out there ;->

Interesting idea - so I'd need a programmable bi directional constant current drive to do that.

However, not really something I could use here as I'd still have to strobe as the LEDs are in 2 sets, back to back wired on a single feed pair of wires - hence the H-Bridge.

As mentioned in another reply just now, I won't be able to influence the switching speed as I'm using an L298N prebuilt H-Bridge module (because H-Bridges are hard for simple plebs like me - I'm more about the software :)

But interesting all the same - thanks :)

That's very considerate of you.

My only addition to the other suggestions (AM radio to snoop, try to keep repetition frequency low) is to put a CM choke (ferrite torus, snap-on choke, or even wrap a turn on a rusty iron item) in one or more places on the long string, including near the power-source end. A slew-rate-limiting Miller capacitor or emitter-series-inductor can be very effective, as well (and for mass production, would usually be employed).

Putting the Rasberry Pi in a metal container is also useful (if some RF is being generated, the OTHER wiring deserves some attention as well).

That and I don't want Ofcom paying a visit :)

Thanks for those other suggestions :)

Hard to beat for the price!

I don't think there will be detectable EMI from a low frequency PWM. But you could add some sort of lowpass if you want to, a small inductor or ferrite bead or something. I wouldn't bother.

I have some under-counter dimmable LED lights that switch at about 4 KHz. Doesn't radiate much, doesn't show up in my bench testing.

But then, we're two miles from Sutro Tower.

Not true. AM transmitters are delectable for miles. Even a small high frequency component from a radiator next door can be big enough to create problems.

Which increases the switching losses. If you slow down the edges enough, you will cook your MOSFETs.

John Larkin doesn't think very hard.

John Larkin doesn't mind irritating his neighbours by screwing up their AM reception.

John Larkin doesn't seem to spend much on bench testing

Not everybody can rely on a nearby radio-transmitter to swamp the consequences of their incompetence.