Andreas, Ray and Istvan, A more detailed study on the "circle of influence" or the "effective radius" of decoupling capacitors" was published in last year's EPEP conference. The paper is titled "Effective Decoupling Radius of Decoupling Capacitors". It is found in the paper that 1/10 of the wavelength, where the wavelength corresponds to the capacitor resonant frequency, may not yield a reliable estimate of the effective radius. For example, the effective radius of a capacitor is found to depend strongly on the separation between planes. Also, due to the interaction between the capacitor impedance and power plane impedance, the frequency where the capacitor has the largest effective radius is not necessarily the series resonance frequency of the capacitor. The paper and the presentation slides are available at http://www.sigrity.com/SUPPORT/support_tech_doc.htm Best Regards, Raj Raghuram Sigrity, Inc. "Achieve what others can't" raghu@xxxxxxxxxxx http://www.sigrity.com 4675 Stevens Creek Blvd. , Ste 130 Santa Clara, CA-95051 PH: 408-260-9344 x116 CELL: 408-390-7614 FAX: 408-260-9342 > -----Original Message----- > From: si-list-bounce@xxxxxxxxxxxxx > [mailto:si-list-bounce@xxxxxxxxxxxxx]On Behalf Of Istvan Novak > Sent: Tuesday, August 06, 2002 5:55 AM > To: Raymond.Anderson@xxxxxxx; si-list@xxxxxxxxxxxxx > Cc: Andreas_Lenkisch@xxxxxxxxxx > Subject: [SI-LIST] Re: Antwort: Re: Placement of Decoupling Caps > > > > Ray, > > I beleive your generic statement about bypass methodolgy > "We select decaps with specific SRF's (series resonant > frequencies) such that the composite impedance profile that is > created by the superposition of the individual decap's impedance > profiles is at or below a calculated target impedance." > is correct, and describes all possible implementation flavors. > > I think Andreas's comment about inductance was also correct, because: > If we take your above generic definition, and some further > statements, in which there seems to be agreement among > most of the si people: > - it is not very practical to count on bypass capacitors' SRF > above a few hundred MHz > - significant PCB resonances can occur above several hundred MHz > > then from all of the above it follows that if we want to smooth out the > impedance profile above the highest SRF of bypass caps (say because > we now that we have fast noise sources that would create a noise > voltage exceeding our target), we like it or not, whether we count on it > or not, the above-SRF behavior (namely the inductance) of bypass > capacitors > is an integral part of our power distribution network. > > I think one key differentiator among the various bypass methodologies is > the Q of bypass capacitors we use or the bandwidth we want to cover with > a capacitor value. The two extremes are: placing relatively narrow-band > capacitor coverages along the frequency axes, versus the widest possible > frequency coverage with a single capacitor value. However, the inductive > part of the capacitor's response is a key in both. > > Regards > Istvan Novak > SUN Microsystems > > > ----- Original Message ----- > From: "Ray Anderson" <Raymond.Anderson@xxxxxxx> > To: <si-list@xxxxxxxxxxxxx> > Cc: <Andreas_Lenkisch@xxxxxxxxxx> > Sent: Friday, August 02, 2002 12:23 PM > Subject: [SI-LIST] Re: Antwort: Re: Placement of Decoupling Caps > > > > > > Andreas wrote: > > > > >Ray, > > > > > >I agree with your numbers of 1/10 or 1/8 of wavelength and term "sphere > of > > >influence", but I would relate this number to the highest frequency of > > >interest (approx. 0,3xx/trise), not to the self resonant > frequency (SRF). > > >In a very simple approach, at higher frequencies you bypass anyway with > the > > >inductive part of the cap, the capacitive part is only used to > block DC. > > >Two different caps of different capacitance but the same package will > have > > >different SRF's, but (nearly) the same body inductance. Above the SRF > they > > >behave in the same manner. Looking from the point of > application, I would > > >make no difference in placing, but I would choose in the highest > > >capacitance in the smallest package (due to lowest inductance) > > > > > >regards > > >Andreas Lenkisch > > > > > > Andreas- > > > > Thanks for your comments. > > > > Regarding using the 0.35/Tr criteria for determining the radius > > for effective decoupling I have a few comments: > > > > I think if you use that criteria you will find that effective placement > > radius (circle of influence) is constrained to be very small for all > > decaps regardless of value. As an example, assume you have a system > > with a 500 MHz processor, 100 MHz buss, and rise times on the order > > of 350psec. (this would equate to a frequency of 1 GHz utilizing > > the 0.35/Tr criteria). So if we use a freq. of 1GHz to calculate the > > size of 1/10 wavelength on FR4 it would be about 1.5 cm. It would > > seem you are saying that all decaps need to be within 1.5 cm of the > > current sinks for them to be effective. While this would certainly > > work from an electrical perspective, it is way too conservative and > > in practice would be very difficult to achieve. Decaps that are > > targeted to decoupling lower frequencies can be much farther away > > from the decoupled device and still be effective. > > > > The second part of your response hits an area that is controversial > > amongst the SI community. The approach that I like to take is > > a bit different. We select decaps with specific SRF's (series resonant > > frequencies) such that the composite impedance profile that is > > created by the superposition of the individual decap's impedance > > profiles is at or below a calculated target impedance. We have found > > this technique to be very effective in practice. I think you will > > also find that utilizing decaps above resonance ends up requiring > > many more parts to achieve the equivalent performance of many > > fewer decaps with carefully chosen resoant frequencies. > > > > Just off the top of my head I can think of at least 3 decoupling > > philosophies advocated by various members of the SI community: > > > > 1 Selecting Decaps with specific SRF's > > 2 Utilizing controlled ESR caps for Edge Termination > > 3 Utilizing decaps above resonance > > > > I'm sure there are at least a couple others. Each method has > > pros and cons that can be debated, but as I said the methodology > > that uses targeted SRF's has been shown to be very effective in > > our designs. For more info on the methodology I am advocating, > > go to http://groups.yahoo.com/group/si-list/files/ and follow > > the links to Signal Integrity Documents and Published SI Papers > > from Sun for a number of conference papers which address the > > method. > > > > Best Regards, > > > > Ray Anderson > > Staff SI Engineer > > > > VLSI Engineering Dept > > ECIT SI Group > > Sun Microsystems > > > > ------------------------------------------------------------------ > > 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 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 > > > > > > > > ------------------------------------------------------------------ > 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 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 > > > ------------------------------------------------------------------ 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 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