being in a small start up, we don't have the fortune to hire any emc engineer so we decided to use our common sense and experience to do the decoupling scheme. the first thing i did is to NOT use any decoupling that has value less than 10nf. i don't have a single capacitor on our system used for decoupling with values than 10nf. i don't believe they work and i don't believe they are needed for a properly designed electrical system. i have miles of 1,2g fcals, high power ghz cpus, loads of memory banks and high speed backplanes and they all pass class b first time. the fud about highspeed caps are way overblown. like i said before a) in a properly design packages, core noise > 100mhz will not pass out to the pcb b) the highspeed noise observed on a pcb is mostly due to imperfect return of i/o switching current. this has more to do with signal referencing to proper power/gnd planes than decoupling i practice what i preach and i don't use any caps less than 10nf in my systems. i believe the place for exotic capacitors are on the packaging, not on the pcb. at least you have some chance of using low esl/esr caps on good packages if you design them properly. -----Original Message----- From: FLOWERDEW, Peter [mailto:peter.flowerdew@xxxxxxxxxxxxxxx] Sent: Wednesday, August 01, 2001 5:51 AM To: 'mgreim@xxxxxxxxxxxx'; 'khalida@xxxxxxxxxxx'; Larry Miller; si-list@xxxxxxxxxxxxx Subject: [SI-LIST] Re: Placement of decoupling capacitors Hi Khalid, Your intuition is correct, but as other people have indicated, you have other options on your kind of pcb. However: I work on emc immunity of products, particularly against mobile phones- I frequently need something like 2n2 to cover below 80MHz and 15pF to cover around 1GHz - if you piggy-back the capacitors they work better than putting them side-by-side on the PCB. Two capacitors more than a decade apart in value will produce two notches in the attenuation curve. Syfer and Johansson offer this as a special packaged component, but it is not cheap. They also stack same-value components to produce high value low-inductance capacitors for compact switch-mode converters. You can also get what are called X2Y caps that are for decoupling a differential pair to a ground plane: C from line A to GND, C from line B to GND, and c/2 from line A to line B. Three capacitors in one package. The structure has currents flowing in opposite directions so the package is low inductance. Syfer and others have been selling these for motor suppression and I have been educating them to telecom applications. They are generally in 0805 packages, but I have some low values for hf in 0603. There is also a 'filter capacitor' which is a surface mount version of a feed through, but all it really is, is a low inductance capacitor. Low inductance is good - more farads at the same self-resonance or higher operating frequency for the same value. Just to show you the mileage in your idea: there is a special capacitor manufactured by AVX which has multiple resonances - The trick is that the resonances are at the frequencies used by mobile telephones so this one component acts as three tuned filters to remove the interference from phones. you can chose three of - 900MHz, 1.8GHz,1.9GHz and 2.45GHz. e.g. W1T15A248A - an 0603 component filtering 900MHz, 1.8GHz and 2.45GHz. Regards, Peter -----Original Message----- From: Greim, Michael [mailto:mgreim@xxxxxxxxxxxx] Sent: 01 August 2001 13:06 To: 'khalida@xxxxxxxxxxx'; Larry Miller; si-list@xxxxxxxxxxxxx Subject: [SI-LIST] Re: Placement of decoupling capacitors Hi Khalid, Well, I think that mfg would take great umbrage with piggy backing capacitors. For a prototype this might be possible but not for production. You might want to make use of plane capacitance to help at high frequencies. Think of what you are proposing, by piggy backing the caps, you are minimizing the foot print but not the loop inductance to the plane. Essentially you are just presenting a larger cap to the PDS. This will move your resonant frequency and give you perhaps unexpected behavior. You may get some additional decoupling through the use of smaller footprint caps like 0402 and via sharing with the bga if in fact there is no other way to get the decoupling in. I can't believe that the board is so packed, that there isn't a better alternative to piggy backing caps. If the board is that packed, routing will prove to be your next problem I suspect. Do you have a .brd file that I could take a look at and offer some suggestions on what options you might have. best regards, MG -----Original Message----- From: Khalid Ansari [mailto:khalida@xxxxxxxxxxx] Sent: Tuesday, July 31, 2001 5:11 PM To: Larry Miller; si-list@xxxxxxxxxxxxx Subject: [SI-LIST] Re: Placement of decoupling capacitors Larry, >I think you have it right on placing the smallest caps (highest frequency) >closest to the pins, but these are very small (we use 0603's), so the next I just have enough space to place this one close to the pin, the rest have to go quite further away due to space constraint. Placing them too far away I have to worry about the loop inductance. Are there any issues with piggy backing capacitors? >bigger (lower frequency) ones are not that much further away. By the time >you get to the bulk caps, they only have to be in the general neighborhood. > >No tricks or piggybacks needed-- the lower the frequency range you need to >cover the less critical the placement. This is especially true where you >have power and ground planes that are essentially zero ohms at lower >frequencies. > >My opinion, anyway, and we haven't had any problems following this idea. > >Larry Miller > >-----Original Message----- >From: Khalid Ansari [mailto:khalida@xxxxxxxxxxx] >Sent: Monday, July 30, 2001 2:19 PM >To: si-list@xxxxxxxxxxxxx >Subject: [SI-LIST] Placement of decoupling capacitors > > > >I looked through various emails from Larry Miller, Larry Smith and the rest >who had contributed to a discussion earlier on decoupling. Those >emails helped me select the values for decoupling caps but now the >challenge is, how do I place these? I don't have enough room on each of my >power supply pins to place these caps around. If I try to place the >cap that decouples the highest frequency components out of the selected >ones close to the power pin and the one that decouples the lowest frequency >components the farthest, the caps are being placed too far away from the >pin. >Can I piggy back these caps? What are some of the tricks to overcome this >problem? > >Thanks in advance, > >Khalid > > > > > > > >------------------------------------------------------------------ >To unsubscribe from si-list: >si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field >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 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 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 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 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