Why do we need capacitance between the power and ground planes? The power planes are just two different levels of voltages. Power is typically at five or three volts, which powers the chips, and ground is zero volts or the return path for the voltage. When you think of voltage, remind yourself it must be a loop. What you send out to power the chip, you have to get back to ground to complete the path. Basic electrical law states that every power or signal line needs a ground or return.
The reason we need capacitance on the power and ground plane is the planes are full of electrical noise, which interferes with the signals. Every time we make the voltage signal turn on or off quickly, such as in a chip or power transistor, we get reflections and noise spikes in the power traces. The quickly rising voltages are driving signals into a low impedance. This causes an initial dip in the power line, then we get a spring-like voltage snap return, which causes a spike, and this electrical noise is embedded right in the power and ground plane signals. To get rid of this electrical noise, we can use the ability of a capacitor to absorb and fill in the electrical voltage variations.
Capacitance acts like a small battery. The capacitor absorbs voltage increases and then releases the stored voltage during a decrease in the present voltage condition. Because the capacitor is like any battery, it can only absorb so much and so fast. The smaller the capacitance value, the faster it can absorb small spikes in voltage (i.e., a small battery is quicker to charge), but the smaller the capacitance, the less voltage it can accept before it is fully charged. This relates to the frequency of the spike pulses on the power plane.
Small, faster (high-frequency) pulses affecting and creating the noise on the power planes are better absorbed by a small capacitor, such as 200–1000 pF. Larger, lower frequency bumps and spikes from big chips and driver transistors may need larger 0.1–1-nf sized capacitors. Big, very low-frequency pulses, such as power line ripple, are handled by larger capacitors in the 100-mF size.
The capacitor size is its ability to absorb and release energy and is rated in farads. One pF (picofarad) is 10-12 farads, nanofarads (nF) is 10-9 farads, and a millifarad (mF) is 10-6 farads. A small AA-sized battery would be equivalent to a few farads in its ability to absorb and fill in power spikes. As we are trying to reduce the faster noise spikes from switching in the chips, we require the smaller 0.001–0.01-nF sized capacitors.
To put the discrete little capacitors at every point needed to absorb chips’ pulses would mean mounting hundreds or possibly thousands of small bypass capacitors all over the PCB. These capacitors cost money to put on and QA after assembly, as well as using up valuable PCB real estate. We have a few choices in embedding capacitance in the power/ground plane, and each has its own good and bad points.
The first is the use of additive films, which have capacitance; they can be imaged and etched. FaradFlex MC25ST, 3M, and DuPont HK04 are some of the most used. A second option is the use of resin-coated copper, such as the RTFoil, or flipped double-treated foil. The thinner laminate layer creates a higher capacitance. The capacitance film applied to the power plane has a very high Dk, which enables higher and better capacitance ratings but still only in the pF range.
Remember that Df is a rating of how well two parallel copper conductors separated by a dielectric retain voltage (i.e., its loss value). Thin 2-mil polyimide films have a further advantage of being lower loss than typical FR-4. The Df of flex PI material is 0.002. The Dk factors of 3.2 and the Df loss factor tells us how fast the stored electrons are leaked off. The high loss Df of FR-4 (0.012) is not great for capacitors, as they leak voltage so quickly as to be more of an absorber than a battery. We call that type of capacitor a snubber.
The most used method to create capacitance between the power and ground plane is to use a very thin layer of FR-4 or flex PI dielectric between the copper layers. The capacitor is made by putting a dielectric (insulator) between two parallel conductors. That is exactly what we have in the normal multilayer ground/power plane core. To make the capacitance large enough to effectively remove the spikes and pulses, we need to increase the capacitance over what is obtained using a typical 5-mil FR-4 or PI core.
To increase capacitance, we have three variables: area, which is predefined by the size of the PCB; increase Dk, which is defined by the laminate used (i.e., FR-4 prepreg at 4.2 Dk, and PI has a Dk of 3.2); or we decrease the distance between the two parallel conductive plates (i.e., thinner laminate between planes).
The easiest to implement is to decrease the spacing between the power and ground copper planes while maintaining reliability and voltage ratings. A 2-mil FR-4 core between two copper conductors has a capacitance of 50–250 pF per square inch. There have been many patents over the years pertaining to different applications of this technology; however, most have expired. Presently, there are quite a few vendors selling thin 2-mil FR-4 cores with the dendrites facing out to increase voltage rating, where the PI film has a much higher dielectric rating and no need for outward-facing dendrites.
The thin FR-4 cores are dry-filmed, imaged, developed, and then etched as normal, resulting in a power ground plane, which is then laminated into a normal multilayer. The buried capacitance is spread out evenly between the planes. The FR-4 or PI capacitive layer is very effective in suppressing spikes right at the source. The further away the capacitor is from the power pin of the chip, the less effective it is in suppressing the spikes or noise. Buried capacitive FR-4 or PI laminate cores are an easy, low-cost, and effective solution to reduce power plane spikes and noise.
This column originally appeared in the December 2020 issue of Design007 Magazine.