Steph, Rules of thumb are OK but I'd recommend creating a PCB/Package/Die model of your IO power delivery and determine the delay and noise impact of insufficient capacitance. Your supply will sag which will delay the output. Add enough capacitance to achieve any critical IO timing you have. When you model the on-die capacitor make sure you estimate the ESR of it. For that matter make sure you properly model the ESL and ESR of all the components in your power delivery network. Depending on the IO and the speed you may find that on package capacitance is needed because having PCB capacitance is not effective give the bandwidth of you package. Make sure to run the simulations over multiple cycles to observe any sagging that comes from insufficient capacitance at any point in the delivery network. Don't just do one transition to determine the capacitance! At a 2 Gbps toggle rate there is no way the package will allow you to recover with one cycle you will need enough capacitance to handle multiple cycles of the data. Without an on-package capacitor you will need a huge on die capacitor because the package bandwidth from PCB to die may achieve 150 - 250 MHz. The die to package substrate bandwidth can be higher which is why on-package capacitance will help as well. The issue with too much capacitance will be yield and leakage depending on the process your in Steve p.s. also take a look at both more than one Voltage, process corner. Supply voltage sage for a slow device may have more SSN induced delay than the fast corner which will have more SSN. Make sure you look at both Noise and Delay when evaluating decoupling. -----Original Message----- From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] On Behalf Of Lee Ritchey Sent: Wednesday, April 09, 2008 2:12 PM To: Chris Cheng; Stephane Tremblay; si-list@xxxxxxxxxxxxx Subject: [SI-LIST] Re: On-die caps for IO supply A common number I have seen used for 3.3V drivers is 100 pF per I/O. Seems to work on many memory die. Simulation would determine if this is good enough. > [Original Message] > From: Chris Cheng <Chris.Cheng@xxxxxxxx> > To: Stephane Tremblay <strembl1@xxxxxxxxxx>; <si-list@xxxxxxxxxxxxx> > Date: 4/9/2008 10:56:40 AM > Subject: [SI-LIST] Re: On-die caps for IO supply > > I don't think you will like my answer but my experience is fill it up = > with as much capacitor your technology people allow. > If you are core limited, you are SOL to begin with. > If you are pad limited, just fill the gap between the core and pad area = > with decoupling cap before the yield police or the core power = > distribution guy wants to take it for himself. > It will take a lot more than "XXX from Si-list recommended this" to grow = > your die to fit anything more than the above. > > -----Original Message----- > From: si-list-bounce@xxxxxxxxxxxxx > [mailto:si-list-bounce@xxxxxxxxxxxxx]On Behalf Of Stephane Tremblay > Sent: Wednesday, April 09, 2008 7:33 AM > To: si-list@xxxxxxxxxxxxx > Cc: 'Stephane Tremblay' > Subject: [SI-LIST] On-die caps for IO supply > > > Hi SI-Listers, > I am looking for a rule of thumb on the required on-die = > capacitors needed for proper operation of a given IO (a DDR mem IO in my = > case). I am looking for a number of "x" capacitance per IO. > > My IO can be programmed to 18 Ohms of drive and the measured slew-rate = > is slightly higher than 6 V/ns for both rise and fall. The parasitic = > capacitance of this IO is about 3.5 pF. The toggle rate could be as high = > as 2 Gbps so I expect the IO rail to recover within reasonable limits = > within half a period. > > I could always start by assuming the current I need just for the = > transmission line (being a 50 Ohms one): > > dI/dt =3D 6 V/ns / 50 Ohms =3D 120 mA / ns (per IO) > > but this basic current demand calculation neglected the parasitic = > capacitance of 3.5 pF. > > Some will say it is also greatly dependant of my power-gnd loop = > inductance from IO to on-package decaps. On the other end, if my noise = > spectrum is in the hundreds of MHz, I can't do much at the package level = > to clean my die supply. > > So from your experience and knowledge, what would be a good start as a = > required on-die capacitance required. I want to avoid my on-die IO rail = > collapsing at a frequency that my package could not keep up. > > Thanks for the answers, > Steph. > =20 > > > ------------------------------------------------------------------ > 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: =20 > //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 > =20 > > ------------------------------------------------------------------ > 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 > ------------------------------------------------------------------ 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 ------------------------------------------------------------------ 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