Dr. Johnson, I would like to ask you a question regarding bypass capacitor techniques and to also praise you on your well written published article in September 1997 Printed Circuit Design Journal.
The question happens to be one that was posed by the journal's Brookspeak Perspective author Douglas Brook, within the same issue. Mr. Brook was discussing a technique I have used for many years as an engineer working with CMOS DRAM modules regarding the placement of capacitors from the chip Vcc to ground. I have always thought it was best to snug the capacitor very very close to the Vcc pin and not the ground pin. Mr. Douglas suggests that capacitor placement is best when half-way between Vcc and ground for CMOS devices. After reading his article, I am now really wondering what is correct???
Thanks for your interest in High-Speed Digital Design.
The function of a bypass capacitor is this: to help returning signal current get from the board back into the driver. Keep in mind that current always makes a loop. When a driver switches HIGH, sending current out one of its signal pins, that current must eventually re-enter the driver chip through its Vcc pins, making a complete loop.
When a driver switches LOW, sending (negative) current out one of its signal pins, this current must also eventually return the driver chip, this time entering through one or more ground pins.
Now, let's look at how this returning signal current flows on the board. In a board with solid Vcc and ground planes, returning signal current flows either on the Vcc plane or the ground plane, *whichever is closest to the signal trace*. This is a consequence of the fact that high-frequency current always seeks the path of least inductance. It doesn’t matter whether the signal is switched high or low, the return current flows (mostly) on the nearest plane.
So, imagine a surge of returning signal current, flowing on one particular plane, as it approaches the driver. Depending on whether the driver is switching high or low, this current must re-enter the driver through either the Vcc pins or the ground pins, respectively. Sometimes the appropriate return- current pins are already connected to the plane carrying the return current, in which case the return current can flow smartly right back into the driver. In other cases, when the return-current pins and the current-carrying plane are of opposite polarity, the return current must divert through a bypass cap, onto the opposite plane, and from there back into the return-current pins of the driver. These returning currents, as they surge through your bypass capacitors, create power supply noise.
As you can see, sometimes current is diverted from the ground plane, through the cap, to the Vcc pins on the driver. Other times it is diverted from the Vcc plane, through the cap, to the ground pins on the driver. Both mechanisms are important. For best performance, both loops need to be small.
It matters little whether a cap is placed particularly close to a power pin, a ground pin, or halfway between the two. What matters is the total size of the diversionary loop.
What you don't want to do is place bypass caps at a distance much further away from the chip than the natural Vcc-Gnd pin spacing of the chip itself.
Dr. Howard Johnson