Steel-Plate Power Planes

A thin coating of steel applied to the inside-facing surfaces of a power- and ground-plane pair may help damp power-plane resonance. This inspiration struck me last Friday night in the office of my good friend Sergio Camerlo at Cisco, during a discussion about power-supply noise issues.

My previous article on Power-plane resonance (EDN, Sept 1, 1998, pg 22) points out the existence of certain resonant modes within power and ground planes. These modes arise due to the limited propagation velocity of electrical disturbances between the planes. Power-plane resonance occurs at a few hundred megahertz in many FR-4 circuit boards, exaggerating the measured power-supply voltage noise at the resonant frequency.

The degree of noise exaggeration you will experience due to power-plane resonance depends on many factors, including the current waveforms demanded by your chips, the placement of those chips, the efficacy of your bypass capacitors, and any dielectric or skin-effect losses associated with the interplane cavity. Any changes you make that increase the losses within the power-plane cavity will act in a helpful way to reduce the Q of the power-plane resonance, thus reducing the worst-case noise.

For example, at 250 MHz, the skin-effect losses in a 5-mil copper-clad FR-4 core amount to roughly 1.5 dB/m, whereas the dielectric losses are only a third of that amount. If you want to obtain a significant reduction in Q, attacking the skin-effect loss seems the right approach. To change the skin-effect losses, you have two choices: change the conductivity of the power and ground planes or change the spacing between them. Decreasing the spacing pays immediate benefits because the Q changes in direct proportion to the spacing. This is one of the many reasons you should always use the smallest power-to-ground spacing your manufacturing team thinks it can reliably produce at a reasonable cost.

Decreasing the conductivity (that is, increasing the resistivity of the material) at first seems less fruitful because the Q scales only with the square root of conductivity, a less-than-direct relationship. Also, decreasing the conductivity would leave the planes less able to handle the heavy dc currents required in today's digital products. A general change to a less conductive material, therefore, seems ill-advised.

How about plating the copper? Coating the planes with a low conductivity material would certainly render them more absorptive of high-frequency energy. Unfortunately, achieving a layer thick enough to absorb significant amounts of power is a big problem. The basis of this problem is that the skin depth of a material actually grows as you decrease the conductivity. Therefore, a lower conductivity material must be more thickly plated to maintain its effectiveness. For example, the skin depth of tin at 250 MHz is 10.6 microns, or 2.6 times the skin depth of copper. A layer of tin thick enough to be effective (a couple of skin depths thick) would actually exceed the thickness of a 1/2-oz copper sheet, exacerbating your lithographic difficulties.

Perhaps you can use a high-permeability material, such as iron or steel, to solve the layer-thickness problem. A large value of magnetic permeability shrinks the skin depth to an incredible degree, making possible the application of a layer perhaps only 1-micron thick. Such a layer would knock down the Q of a power-plane resonance by a factor of 100, leaving it perfectly well damped.

As with any new process technology, problems with the plating approach exist, especially in etching chemistry and rust prevention. It's also possible that steel creates too much loss and that it might prevent the flow of ordinary currents required to make your system function. I don't have all the answers, as this problem cuts across many layers of technology, but that's the exciting thing about an inspired idea-it gets you out there talking to a lot of people about how to improve the way we build high-speed digital products.

I have an additional idea for your iron power ground plane: any magnetic material will do. Every now and then we encounter someone climbing the microwave learning curve using magnetic nickel somewhere in the RF ground path and asking, "Why are my losses so high?" I also understand that magnetic nickel is plated on plastic cases to reduce EMI. Point you might want to mention is that, in the skin effect region, the equivalent surface resistance increases with the sqrt(mu). Thus, the conducted EMI in the magnetic ground path gets significantly attenuated, while DC currents are unaffected by mu. However, the designer should be very careful not to introduce magnetic metals into the ground current return path for any high speed or high frequency signals, or they will find themselves climbing the same learning curve us microwave guys have.

Sincerely, Jim Rautio