Colin Not ganging up on you at all. Just ganging up on the fact that the "Intel/Huray" approach of performing this level of microscopic detailed modeling is gross overkill. I'd certainly love to see good complex correlation of both magnitude and phase for the Huray model in a simulator. But, it's just not necessary, having done hundreds of models correlated with measurements using the Djordjevic-Sarkar model combined with the Hammerstadt-Jensen model (or other macroscopic model). Scott Scott McMorrow Teraspeed Consulting Group LLC 121 North River Drive Narragansett, RI 02882 (401) 284-1827 Business (401) 284-1840 Fax http://www.teraspeed.com Teraspeed® is the registered service mark of Teraspeed Consulting Group LLC On 11/22/2011 4:16 PM, colin_warwick@xxxxxxxxxxx wrote: > Hi Scott, Todd, > > Ganging up on me, heh? :-) > > To be clear, SEM and AFM data aren't the only way to set the parameters of > our model. The macroscopic fitting approach you mention works too. And that > approach does indeed have the advantage that you can let the optimizer tune > the Svensson-Djordevic dielectric model parameters at the same time. But the > fact that direct microscopic measurement can be used to set the parameters > correctly gives us confidence that the model has a physical basis and is > something more than a polynomial fit. > > Best regards, > > -- Colin Warwick > > Product Manager for High Speed Digital > > Agilent EEsof EDA > > Blog: http://Signal-Integrity.TM.Agilent.com > > > > -----Original Message----- > From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] On > Behalf Of Scott McMorrow > Sent: Monday, November 21, 2011 2:54 PM > To: si-list@xxxxxxxxxxxxx > Subject: [SI-LIST] Re: surface roughness > > To pile on. The way I see it, copper surface modeling is not the problem. > Rather, it is more important to determine how to holistically determine > and model broadband dielectric constant, loss tangent and copper surface > roughness, in order to determine a realistic model for PCB structures. > Given the variability of copper foil and adhesion treatments across vendors > and fabricators, I find it much more useful to characterize a particular > dielectric material, foil and fabrication process with measurements. I > don't care about the microscopic structure of the copper, whether it be > small or large snowballs or golf balls. What I care about is the > macroscopic behavior, as Yuriy describes, which can easily be modeled with > something as simple as the Hammerstad equation for every high performance > interconnect situation I've ever found. > > I've published quite a bit on the process that I use. Probably too much, to > be frank.` Needless to say, the models I create have extremely low error > from DC to our maximum measurement bandwidth of 50 GHz thus far. I am able > to correctly model not only the attenuation behavior of an interconnect, > but also the phase behavior across the complete broadband spectrum. This > is necessary to correctly predict all interconnect resonances that will > impair performance. The models are fully de-embedded and remove all > measurement artifacts. > > We are currently working to extend our method across temperature and > humidity, with results that are "quite enlightening", and will be > presenting some of our data at DesignCon this coming year. I've simplified > the process of Generalized Modal S-Parameter (GMS-Parameter) material > identification, such that those of us that are not blessed with a > laboratory with an SEM, and the equipment to determine what the exact > material composition of the copper snowballs are, can still create scalable > models that are deadly accurate. > > Oh, and at high enough frequencies, copper surface roughness in IC packages > does make a huge difference, accounting for up to 25% of the loss budget. > > best regards, > > Scott > >> >> On Mon, Nov 21, 2011 at 7:39 AM, Todd Westerhoff<twesterh@xxxxxxxxxx>wrote: >> >>> I'm with Yuriy and Jeff on this one. The physical details of copper >>> roughness are complex and hard to measure; we can create simulation models >>> that represent arbitrarily detailed physical features, but we rarely (if >>> ever) get the measurement data from PCB manufacturers to plug into those >>> models. What good is a detailed model if we have no data from the >>> manufacturer to plug into it? >>> >>> We've found the process Yuriy describes to be very useful. Starting with >>> measured data, we adjust the trace roughness factor to correlate to >>> measured data. The process is incredibly valuable, as we gain great >>> insight in the characteristics of the same board built by different >>> suppliers (and even the manufacturing process control within the same fab >>> run). Once the models are tuned in, automated interconnect extraction and >>> simulation allows us to validate the behavior of all the links in a high >>> speed system as a batch simulation process. >>> >>> Hope that helps, >>> >>> Todd. >>> -- >>> >>> Todd Westerhoff >>> VP, Software Products >>> SiSoft >>> 6 Clock Tower Place, Suite 250 >>> Maynard, MA 01754 >>> (978) 461-0449 x24 >>> twesterh@xxxxxxxxxx >>> www.sisoft.com ------------------------------------------------------------------ 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 Old (prior to June 6, 2001) list archives are viewable at: http://www.qsl.net/wb6tpu