slots

Don't Use 'Em!

Douglas Brooks

Has this ever happened to you? The board is finished.The traces are tightly routed and you have done a magnifi-cent job of confining them to the fewest possible number ofsignal trace layers. Then the engineer calls and tells you hehas forgotten one little thing--just one little net that hap-pens to run horizontally across the entire board! It will takehours to tweak hundreds of other traces to fit this one in.But, just one little slot in an adjacent ground plane wouldmake room for it and you would be done in minutes. And,after all, it is the engineer's fault, not yours, that you have tocreate this little slot!

Tempting, isn't it? Well, don't do it! This article willgive you three reasons why slots in planes are to be avoidedon high speed boards.

Signal 1Signal Return 1Signal 2Signal Return 2SlotTrace LayerGround PlaneImpedance Control:

Signal traces begin to look like transmission lines tosignals with very fast rise times. The problem with trans-mission lines is that reflections, and therefore noise andfalse triggering, can occur if the characteristic impedance ofthe line is not controlled over its entire length. If there is adiscontinuity in the impedance of the line, a destructivereflection can be caused by that discontinuity.

The characteristic impedance of a trace is determinedby its geometry, one element of which is the distancebetween the trace and an adjacent plane. If all other thingsare constant, but the distance to the plane changes, theimpedance will change at that point and a reflection islikely to occur.

Consider Signal 1 in Figure 1. It is referenced to theground plane along its length except where the trace crossesthe slot in the plane. In high speed designs, the returnsignal for Signal 1 will be on the ground plane directlyunder the trace for reasons I gave in my January column(Footnote 1). But the slot interrupts the path of the returnsignal, and it must find a way around the slot. This discon-tinuity causes an obvious change in geometry that in turncauses a change in impedance. An easy analogy would be ifwe cut a coaxial cable and spliced it as shown in Figure 2.We intuitively know that this is bad practice! So is allowingan impedance controlled trace to cross a slot in a plane!

Figure 1.Allowing signals on trace layers to cross over slots on adja-cent reference planes can cause problems in impedance con-trol, EMI radiation, and crosstalk.keep loops as small as possible. The case of a signal tracedirectly over a plane is an excellent example of controlled looparea. The return signal is tightly coupled to the trace and theloop area is very small.It is clear from Figure 1, however, that if there is adiscontinuity in the plane, the signal return path must neces-sarily move away from the trace. Where it actually goes canbecome an interesting effort in speculation. The return signalmight, for example, find its way through nearby bypass caps toa different plane. In this case we might get lucky and thepractical effect of the slot might be small. On the other hand,the return signal may have to travel all the way around the slot.In this case the loop area could be relatively large, causingserious EMI problems. The point is that the path of the returnsignal is uncontrolled, with subsequent unknown conse-quences. That's why we don't want to create the situation in thefirst place.Crosstalk:When twotraces are adjacentto each other, theycan couple un-wanted (noise)signals into eachother. This cou-EMI Noise:

In my January column I also pointed out that onesource of EMI radiation is the \"loop area\" defined by asignal trace and its signal return path. Since destructiveradiation can be directly related to loop area, we want to

Figure 2Splicing a coax cable in this mannerwould likely cause severe reflection prob-This article appeared in Printed Circuit Design, a Miller Freeman publication, March, 1999(c) 1999 Miller Freeman, Inc. (c) 1999 UltraCAD Design, Inc.

pled noise is called \"crosstalk\" (Footnote 2). Thedegree of coupling is inversely related to the square ofthe distance between the traces. To a large extent, thefurther the traces are spaced apart the better.

In Figure 1, Signal 1 and Signal 2 are spaced wellapart from each other. In a well controlled designtheir return paths would be directly underneath theirrespective traces and would necessarily be spaced thesame distance as their signals. But if there is a slot inthe reference plane, then the return paths have to finda way around the slot. Over this distance they mightbe very close together, even possibly congruent! Thiscreates obvious potential for crosstalk between thesignal returns, and under certain circumstances thecrosstalk coupling might be very high.

Tough Troubleshooting:

It sometimes is not apparent or recognized that there is a slotin an internal plane, especially to an engineer or technician whodidn't design the board. After the prototypes are built, the engi-neer discovers that he has unwanted reflections on a trace, EMIradiation, and crosstalk problems. He checks the traces andverifies that the impedance control guidelines (trace thicknessand spacing) have been met, the routing is seemingly good, andthe crosstalk control guidelines (trace separations) are correct.The effects of slots are really subtle and difficult to recognize. Itdoesn't occur to the engineer that the problems are related to theplanes rather than to the traces. Some engineers spend anenormous amount of time and, unfortunately, never find the realproblem in the board design.

Responsible designers don't do this to their associates!

FOOTNOTES:

1. See \"Loop Areas: Close 'Em Tight,\" PC Design Magazine, January, 1999

2. See \"Crosstalk, Part 1: The Conversation We Wish Would Stop,\" November, 1997, and \"Crosstalk, Part 2:How Loud Is It?\