Ground planes

[Onvinyl]

Hi,
sometimes I think I'm the only one that dares to post the real silly and dump questions, with dozens of lurkers watching from a distance...

You have two sayings:
1) Make sure your current loops are as short as possible and don't share loops over the same wire.
2) Use groundplanes

Obviously, that is contradictionary, unless you don't use the plane as a return path for the decoupling caps.
What is it I don't get?
Rüdiger

[AndrewT]

Hi,
your not alone.

I don't understand ground planes at any frequency, but I have read many times that they are virtually mandatory at RF.

I have never read a satisfactory explanation of any advantage at audio frequency. It is this failure to understand the logic for introducing all that complexity for NO advantage that loses me.

Now leave the ground plane in and split it up into discrete tracks returning to one or two star grounds (power and signal and digital if needed). That I can understand.

[pinkmouse]

I quite frequently use both, starring for signal critical grounds, and a plane for power decoupling etc, (though I have never built designed anything with digital on it yet). Sometimes my GP has no connections to components on it at all, in which case it just serves to isolate tracks from each other, and save on etching fluid as I make my own boards. As for the added complexity, I don't mind, it forces me to think very closely about layout, which is a good thing.

See pic below for an example:

[Eva]

What I'm going to explain is an oversimplification, but it may do the trick:

Imagine the classic air cored inductor used in the passive crossover network of most speakers. It has cylindrical shape, an inner diameter, an outer diameter, a thickness and a whole lot of turns. Something like that:


Imagine that we also have a thick flat copper sheet, much bigger than the outer diameter of the inductor, that is placed in the same plane as the turns of the coil and at some distance from it.

Now, what is going happen to the inductance value as we approach the inductor to the copper sheet?

(Tip: Isn't the copper sheet going to act like a big shorted additional coil turn, thus reducing dramatically coil inductance as the inductor is closer to it?)

[AndrewT]

Hi,
tip, I don't know.

[pinkmouse]

A good book that explains this, and a lot more well is "The Fields of Electronics" by Ralph Morrison.

[Eva]

That's because you have to understand air-cored inductors first

Doing again a simplification, the point is that every loop of wire has an inductance porportional to the area enclosed by it and to the square of the number of turns (if more than one). However, if the loop is placed in the same plane and in close proximity to a conducting sheet of some non-ferrous material with low resistivity (copper for example), the inductance is dramatically reduced. That happens because a current flowing in the opposite direction as the one from the wire loop is induced in the path that results from projecting the wire loop into the conducting sheet, thus cancelling most of the magnetic flux that the initial current would otherwise produce.

In each PCB, AC current loops are formed (whose area is porportional to the degree of clumsiness of the designer). Each of these loops works like an inductor and all the loops in the PCB end up magnetically coupled. Thus, a voltage component proportional to the frequency and to the magnitude of the AC current flowing in each loop is induced in the rest of the loops. That effect is already measurable in the high range of audio frequencies, but it turns really messy as operating frequencies approach the Mhz range.

A continuous ground plane (without gaps) reduces PCB track leakage-inductances by at least an order of magnitude, thus improving inductive cross-talk (and PSRR) dramatically. On the other hand, the classic ground plane full of gaps found on most audio PCBs is essentially cosmetic and hardly produces any PCB inductance reduction.

[Onvinyl]

Hi,
Eva, thanks for that elaborate answer!

Originally posted by Eva
However, if the loop is placed in the same plane

That means in parallel, I guess (or even levelled, when look at the 'side', so you have to get as close as you can?)
But if I got it, then it is essential to *not* return e.g. the decoup. caps to that plane, because the plane would become part of the loop, correct? Rather, the plane connects only to common ground?
Rüdiger

[Eva]

You may make other currents to flow through the plane and it still won't lose its inductance cancellation properties. However, when high AC currents are involved, both in the plane or in the PCB tracks, you can expect voltage gradients to appear across the plane, so it will be no longer a good ground reference for small signals.

Also, you may place the plane closer and closer to the PCB tracks, but this will increase capacitance from everything to ground. A compromise has to be found.

[Onvinyl]

Hi,
I've found the AN-47 of Linear tech. that deals with high-speed opamp breadboarding and circuit construction. From the photos, you see that the decoup. caps are solderd on the plane just were they are. I just can't see how that fits to the 'no-shared-loops' impetus. Even if currents take the shortest/lowest impedance way on the large plane, there might be 'crossroads', second, according ohms law, there will be some current on higher impedance paths.
If my guesses are right:
Is it a matter of trial&error whats is best for a given circuit? No shared loops vs. low inductance? Or are there rules?
Rüdiger