next topic »

[on:]

After various posts and discussions about decoupling capacitors, I decided to see how much EMF an Arduino might be radiating. Operating at 16 MHz clock it is in the "high" radio frequency range (3 MHz to 30 MHz). So I set up an Arduino to output 8 MHz on pin 9, through a 1K load resistor, and then measure with an oscilloscope clipped to another nearby resistor (however, not touching), like this:



The scope showed a 8 mV peak-to-peak signal at that frequency, which decreased as I moved the probes away:





Reference: http://en.wikipedia.org/wiki/Radio_frequency

[Reply #1 on:]

Good work! Just goes to show how important the oh so often missed out decoupling caps really are!

[Reply #2 on:]

 

Operating at 16 MHz clock it is in the "high" radio frequency range (3 MHz to 30 MHz).

And as the device is mostly digital the harmonics of the square waves radiated signals will go well past 30Mhz, as a square wave is comprised of it's fundamental and all it's odd harmonic frequencies.

Lefty

[Reply #3 on:]

Can you explain what your setup and measurement have to do with decoupling capacitors.
It seems what you're doing here is just building an 8-Mhz RF generator. What size R across
the scope probe?

[Reply #4 on:]

Can you explain what your setup and measurement have to do with decoupling capacitors.

I was hoping no-one would notice that.

It is probably more of a demonstration that signals are likely to stray from one part of a circuit to another. I had previously noticed that just adding long connection wires, or even connecting up a wire for debugging, especially on the high-frequency lines like the SPI lines, can introduce errors that weren't there without them. I speculated that the extra wiring runs were either radiating, or picking up, EMF radiation.

It seems what you're doing here is just building an 8-Mhz RF generator. What size R across the scope probe?

10K.

[Reply #5 on:]

If it's not too much work, can you show us what difference it makes with a capacitor?

[Reply #6 on:]

What value do you want? I presume over the scope probes? Or across the output?

[Reply #7 on:]

A real good example would be to show the noise on say a floating analog input, while next to like a flourescent bulb, that would show just how much noise is in an everday enviroment, lol

[Reply #8 on:]

I was hoping no-one would notice that.

It is probably more of a demonstration that signals are likely to stray from one part of a circuit to another.

Good, I was afraid of possibly yet another up-braiding today, :-).

Yeah, RF is RF. This is one good reason to keep input lines terminated, and also use low
values Rs [impedances] in a digital ckt. Eg, < 10K rather than 50K - 100K. The latter are
about as good as tying an antenna to an input pin. Using ground planes under high-speed
digital circuitry is a good idea. I always use a ground "island" wrapped around the crystal
circuitry, as that is usually the highest frequency on most embedded controller boards.

[Reply #9 on:]

RF is also a reason to keep wires in parallel. One way to keep RF emissions lower is to keep the area between the wires low.

[Reply #10 on:]

One way to keep RF emissions lower is to keep the area between the wires low.

Yes, said another way, no inductive loops, they act like loop antennas.

[Reply #11 on:]

What value do you want? I presume over the scope probes? Or across the output?

If this is a demonstration of the importance of decoupling, I suppose a capacitor across the scope probes (the "load") would be the most relevant. Maybe just a typical decoupler, 0.1uF or so. I'm just curious to see the difference, and I don't have a scope to try it myself.