The usual advice for high speed chips is to place some bypass capacitance as close to the chip "as possible". However, any chip designer who doesn't include at least a few tens of pF at the VDD bond pads is a fool. In other words, dI/dt at the PCB interconnect should already be slowed by the on-chip capacitance, the bond wire and lead frame inductance, and adding at least some hundreds of pF immediately adjacent to the chip. This completes a pi filter that cascades into the further filtering created by connection inductance and even larger capacitances further from the chip (these reactances are often distributed in the power/gnd plane and embellished by discrete caps that serve all nearby chips. The most critical bypass treatment though, is that inside and immediately adjacent to the IC. Distances are short enough and the filtering should be good enough at microwave frequencies that the actual propagation delay should not be a material issue. The distance of the larger, additional caps, while not trivial, is less critical. The issue is one of impedance in series with your energy storage, not distance per se. We worry about distance for signals because they are not filtered and do have significant energy at frequencies high enough to make delay an issue. Properly implemented power pins are not like signals. Orin Laney On Sun, 8 Jul 2007 19:47:46 -0700 (PDT) Bill Owsley <wdowsley@xxxxxxxxx> writes: > "Eddy how far away a capacitor can be depends on the interconnect > between > the capacitor and the load." > > Well distance does seem like a factor. The cap is meant to be > close enough to maintain the voltage at an acceptable level and to > do this it has to be within a distance determined by the speed of > propagation in the substate of interest in order to supply charge an > enough to maintain the voltage within the defined levels/limits, say > for example 1/20 of the supply. > So how far away is 1/20, or some fraction of the voltage, given a > speed of propagation, then the caps should not be any further away > than that distance, and large enough to supply the charege needed. > > > steve weir <weirsi@xxxxxxxxxx> wrote: > Eddy how far away a capacitor can be depends on the interconnect > between > the capacitor and the load. > > Ideally this is something that we engineer > by design. Unfortunately it is often something that is handled by > rote > practice. In the past rote practice often overengineered the power > system. These days more often than not it underengineers it. I > suspect > that failure analysis will find power and thermal issues at the > heart of > Microsoft's current $1.15 billion recall. > > Steve. > Eddy wrote: > > "a temporary power feeder in case of power shortage" > > I think that is nice job description for a decoupling > > capacitor. :-) > > The problem is impedance of power lines. Distance > > means inductance and inductors resist fast changes of > > current. When a CMOS buffer changes state, it goes > > together with a current spike in the power supply. > > Depending upon the impedance of the power supply, > > there will be a "negative spike" (dip) in the power > > supply voltage. This dip not only slows down the > > transition of the CMOS buffer itself but also affects > > other circuits tied to the same power supply nearby. A > > current spike is not just "a frequency" but rather a > > wide spectrum of frequencies. Most chips have lots of > > different circuits all creating total chaos ("noise") > > on the power supply. For most chips it is vital to > > have a decoupling capacitor as close as possible > > between the power supply pins of your circuit. > > Sometimes 10mm distance is already too far. The most > > used decoupling capacitor value has to be 0.01uF or > > 10nF. > > > > Eddy > > Fremont CA > > > > --- M Sridhar wrote: > > > >> I have a doubt about Decoupling capacitor, I > >> understand that decoupling capacitor is used to > >> decouple power supply to the device, so it acts as a > >> temperery power feeder in case of power shortage. > >> My doubt is how to know at what frequency the power > >> fluctuation would happen? > >> From were we may get this information. > >> > >> Thanks, > >> Sridhar. ------------------------------------------------------------------ To unsubscribe from si-list: si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field or to administer your membership from a web page, go to: //www.freelists.org/webpage/si-list For help: si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field List technical documents are available at: http://www.si-list.net List archives are viewable at: //www.freelists.org/archives/si-list or at our remote archives: http://groups.yahoo.com/group/si-list/messages Old (prior to June 6, 2001) list archives are viewable at: http://www.qsl.net/wb6tpu