I have seen a lot of PCBs with this issue. Yet many people simply dismiss it. But it is an important EMC issue that has the potential to cause severe troubleshooting headaches when sensitive analog front-end is also in the system, or your products have to pass third-party EMC testing for UL/CENELEC certifications.
Besides, in a big project where many engineers are involved, it is next to impossible for the poor analog guys to “prove beyond reasonable doubt” that certain noise problems are caused by such seemingly very mundane layout “mishaps”, above all other “bigger” possibilities.
I know, I know. I am sounding a little un-team-player here as if I am trying unduly hard to teach an engineer to dodge a blame. But, you know what I mean? You do not want such a thing blowing up at your face at a later stage of the project. It is a potential 20dB improvement that is easy to achieve.
So, what is the “current returning paths” I want to talk about?
Let’s take a look at the drawing in Fig 1.
The pink dotted line shows the ground current return path via the ground plane of the PCB. So, what’s the problem?
Dr. Archambeault at IBM has done some extraordinary experiments over this issue. The figure on the right shows his results at 1Mhz, where returning current of his 1Mhz signal ignores the shortest path on the ground plane(blue) to return to the source. Instead it follows the same route of the signal path(red trace).
Going back to our drawing in Fig 1, the behavior Dr. Archambeault has shown us means that our return current will follow the edge of the ground copper cut until it gets back under the signal trace. We have hence created a “nice” loop antenna with “edges” to radiate/bounce these signals. In fact, other experiments by Dr. Archambeault further show that over a frequency range of 20-1000Mhz, the near field radiation caused by a grounding split is much greater than that of no split, in excess of 20 dB !!
There are many other EMC implications as well, such as high frequency ground return current that accidentally expose other elements of the circuits to the interferences.
Now, you may like to say that your signals are of low frequencies. So this should not affect you.
Is that really? Not so fast, mister. When digital pulses are involved, the harmonic content of the signal have more to do with the rising and falling speeds of the pulses than the base frequency of the signal itself.
The base frequency may not be very high, but a very fast rising/falling edge may force its harmonics far into the your view, like it or not.
Well, now that will be the subject of my possible future blog.