Factors affecting EM radiation from pcb

[fonograph]

What are the factors that impact how much photons does radiate from pcb? I am talking about far field radio wave or microwave electromagnetic waves.

I did some brief reading on antenas and I read that the longer the wavelenght/ lower the frequency,the longer the antena needs to be to efficiently radiate.Thats why the switch mode power supplies radiate away the high frequency transients at the edges of square waves.

I also rear that impedance of free space is 377 ohm,so I assume 377 ohm is ideal antena impedance so it couples to the air?

With my current knowledge,if I designed electronics and didnt want it to radiate away photons,causing interference,then I would make sure there is no long straight conductor,so everything is short as possible.I would also try to make everything as low frequency as possible and make every impedance far away from 377ohm,either lower or higher,like 50 ohm or 1K ohm.

Is my understanding as I presented it in this post correct? What are other factors and strategies to combat energy being radiated away into far field in form of photons?

[T3sl4co1l]

Close.  The ideal antenna (a self-similar, self-dual shape) has input impedance Zo/2.

Transmission lines have a characteristic impedance proportional to Zo, but geometric factors almost always make it lower.  Loading with materials further decreases (dielectric) or increases (permeable magnetic) this, as well as reducing the wave velocity.  (600 ohm ladder line is about as high an impedance as you can get: each wire is nearly Zo to space, so the two, differential, are twice that.  Needless to say, such line is prone to radiation, if you don't have it shielded, or rotating in such a way that radiation is prevented.)

Transmission lines don't necessarily radiate.  Coax, for example.  You can always shield around an active conductor.  Radiation is a best-case scenario, and you can always do worse.  (Or better, depending on your perspective!)

FYI, it's not meaningful to speak of photons, not at this energy level.  The energy per quanta is so far below ambient thermal noise that quantization is utterly immeasurable.  Classical E&M is, by a long shot, perfectly accurate and descriptive here.

The very broad strokes of your post, aren't far off -- they just need a lot of refinement before being put into practice.

For example, when a snubber is added to a switching converter, the high frequency content, the bandwidth, is being reduced.  This can be done independently of the switching (fundamental) frequency.

There can be other good reasons for reducing the switching frequency, at least by a modest amount.  This is where "keep frequency low" becomes: "well, how low?"  You don't want to go "as low as possible", because that's expensive and inefficient.  You only want to go as low as needed.

How much is needed?  Well, most EMC regulations cut off at 150kHz, or have a higher limit below 150kHz.  140kHz is a very common choice for operating frequency, because it just skirts this cutoff, without much consequence for efficiency or cost.  The second harmonic (at 280kHz) needs to be filtered, and so on, but these are naturally at a lower amplitude than the fundamental, so you have less work to do!

The alternative is to boost the frequency higher so that you save space (and hopefully cost) on the transformers and inductors and filter chokes.  Higher frequencies are easier to filter.  So you wouldn't normally choose, say, 200kHz, you're not gaining much relative to 140kHz -- but 300 or 400kHz or more, now you're getting somewhere.

This relates to another common misconception with switching converters: "minimize loop inductance" (or loop length, same idea).  While this is the correct step in many cases, it is not the general case.  The general case is this: match the loop inductance to the commutation speed and switching impedance.  (Commutation speed is how much time it takes to go from "on" to "off", and vice versa.  Switching impedance is peak switch voltage divided by peak switch current.  Z = V/I, so it's an impedance quantity.)

Whenever you have an impedance and a time or frequency, you also have an inductance and a capacitance.  These are related like so:
T = 1/F
Zsw = Vpk / Ipk
Z = sqrt(L/C)
Fo = 1 / (2*pi*sqrt(L*C))
and all the algebraic rearrangements of them.

You want Fo > F, or if you can't achieve that, then you need Z ~= Zo.  The traditional advice, to minimize L, has the effect of minimizing Z and maximizing Fo.  (Minimizing Z has the effect of reducing peak voltage, at the expense of more peak current.)  This was acceptable, back in the days of bipolar transistors and fat MOSFETs, where commutation times of hundreds of nanoseconds was typical.  With T < 10ns in many applications today, this is not possible, and one must consider the full situation.

The other thing that "minimize L" gets you, is less loop area for radiating that noise (see, it's still on topic ).  In that case, if you are in a situation where you need more L, you can use an inductor component, rather than trace length.  This keeps the stray field confined inside the component (more or less).



The general engineering lesson is this: whenever you hear "minimize", as in, make as nearly zero as possible, you're probably hearing shit advice.  Instead, look for underlying reasons that might be related, and optimize (find the best compromise or fit, not some arbitrary zero) instead.

Tim