Hi, I think one thing might have been missed: The DF variation between the batches (lots) of the manufactured raw PCB material sheets. We and the mentioned articles were talking about batches of manufactured printed PCB boards. I didn't find indication for anyone making sure that the PCB fab did/didn't order a new batch of material between making two batches of printed PCBs for the same motherboard product. I think the fabs keep larger stock of their few most usual materials, to guarantee lead time of their own product. This stock might last for a very long time, maybe multiple batches of the same board design. Right now I need to select materials for 10Gbps/signal. One of the big components we use has a design guide stating channel loss requirement, I might apply the same numbers for other components running interfaces at the same speed, although I know it's not really correct. If there is a difference in the silicon IO requirements, it might be possible to adjust the loss requirement based on the difference in electrical voltage level values in the datasheets. We can also calculate loss requirement for a total channel (and divide it into segments) based on driver output voltage and receiver input voltage specs, as I used to do it. I don't know how correct/incorrect that is. Best regards, Istvan Nagy Sr. Hardware Development Engineer Fortinet, Sunnyvale CA -----Original Message----- From: Loyer, Jeff Sent: Thursday, August 16, 2012 8:06 AM To: Istvan Nagy ; si-list@xxxxxxxxxxxxx Subject: [SI-LIST] Re: DF spec for PCB materials Good questions, some of which we haven't worked out all the details for every design/vendor yet. But, I'll try to clarify some things, based on what I've experienced. Do you do this measurement on every batch of PCBs or every panel? We have found that loss doesn't vary much between lots, so doesn't need to be measured at anywhere near the frequency of impedance. Once you have found the recipe that gives adequate loss, it seems to hold steady. Chu-tien (Jerry) Chia, Richard Kunze, David Boggs, and Margaret Cromley did an excellent paper on this entitled "A Study of PCB Insertion Loss Variation in Manufacturing Using a New Low Cost Metrology", available on the web. One thing we're finding on some stackups, however, is more variation, even between layers within a stackup. For instance, on an 8 layer stackup, we have seen layers 3 and 6 have significantly different loss (and/or coupling and/or propagation velocity!). We didn't note this previously, and are working to understand it more. On the same coupon as the impedance? The SET2DIL coupon can be used to measure impedance, but we continue to have dedicated impedance coupons, since they are much easier to measure quickly. Who pays for the manufactured panels where the loss does not meet the specified value? Ideally, it will be handled just like impedance, but that might be a bit optimistic at this point. Currently, we typically go through some iterations with board vendors, building SET2DIL test boards which mimic our proposed stackup and collaboratively modifying that stackup based on the SET2DIL results. Once we find a "recipe" which meets our needs (with some tolerance), we "lock down" that recipe and apply it to our product boards. Only a change in the recipe should cause the loss value to change significantly, so I would think the fab vendor would be responsible at that point (though it hasn't been an issue, as far as I know). Soon, we hope that vendors will have the tools and experience to be able to predict/monitor/control loss before building boards (just as they predict impedance) and the onus will fall to them to meet the loss requirement. Often when a new type of requirement comes up, a lot of people especially at the fab try the requirement to look illegitimate or ridiculous. We have been working vigorously with vendors throughout the world to acquaint them with the concept of insertion loss and how to control, measure, and monitor it. You shouldn't have to go far to find a vendor who is intimately familiar with insertion loss and/or SET2DIL. I'm sure there are some (perhaps critical) vendors we've missed, but I believe the majority have some competence in this area. We have had a spec. of approximately -0.8dB/inch @ 4GHz on many designs for quite a while; if your spec. is no tighter than that, many folks should be able to accommodate it w/o angst. Regarding the channel between me and the fab and/or assembly houses - that varies, depending on what "hat" I'm wearing at the moment, and who I'm dealing with. Usually, however, I (or someone in my position) am (are?) working with the fab vendors and assembly houses directly to work out details. But that varies - Jerry Chia often provides invaluable translation/clarification/follow-through for us when dealing with vendors, for instance. One indication of how it's done is that my wife is not surprised anymore by late night phone meetings speaking engineering nonsense for hours on end... Note: these are my own personal experiences/observations, and aren't intended in any way to represent an "official" Intel position. In fact, we are learning so many new things so quickly, I don't think it's possible to articulate rigid policies at this point (for instance, periodicity of loss measurements). P.S. We try not to refer to it as "Intel's" SET2DIL method, though we originally conceived it. It has now been adopted as an IPC standard (IPC-TM-650 Method 2.5.5.12, though I don't know if the rev. has been published yet), at least 2 vendors are supporting tools to measure it in HVM environments, and there are other companies providing measurement services. It's been a wonderful child, but it's time for it to leave the nest and visit the rest of the industry... :-) Cheers, Jeff Loyer -----Original Message----- From: buenos@xxxxxxxx [mailto:buenos@xxxxxxxx] On Behalf Of Istvan Nagy Sent: Wednesday, August 15, 2012 8:49 PM To: Loyer, Jeff; si-list@xxxxxxxxxxxxx Subject: Re: [SI-LIST] Re: DF spec for PCB materials Hi, Thanks, actually it does help. So I will focus on the typical value and try to get the FAB measure the loss per unit length. Eric Bogatin had an article where he referred to you and intel's method about SET2DIL measurements. Do you do this measurement on every batch of PCBs or every panel? On the same coupon as the impedance? The other interesting thing is who pays for the manufactured panels where the loss does not meet the specified value... Often when a new type of requirement comes up, a lot of people especially at the fab try the requirement to look illegitimate or ridiculous. Wel, its a fight. I will try to come up with some spec values for this. I would not use microstrip for high speed, since all my designs are real products and very dense as well, unlike reference designs where they have a lot of space for routing on outer layers. I am working with far east assembly company who is now standing between me and the PCB fab, so the communication channel for this issue seem to have too much dropped packets... I used to work at smaller companies in Europe, and always dealt with the fabs myself. Do you do it directly, or through another company, broker, other department... Regards, Istvan Nagy Fortinet ------------------------------------------------------------------ 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 forum is accessible at: http://tech.groups.yahoo.com/group/si-list 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 *** Please note that this message and any attachments may contain confidential and proprietary material and information and are intended only for the use of the intended recipient(s). If you are not the intended recipient, you are hereby notified that any review, use, disclosure, dissemination, distribution or copying of this message and any attachments is strictly prohibited. If you have received this email in error, please immediately notify the sender and destroy this e-mail and any attachments and all copies, whether electronic or printed. Please also note that any views, opinions, conclusions or commitments expressed in this message are those of the individual sender and do not necessarily reflect the views of Fortinet, Inc., its affiliates, and emails are not binding on Fortinet and only a writing manually signed by Fortinet's General Counsel can be a binding commitment of Fortinet to Fortinet's customers or partners. Thank you. *** ------------------------------------------------------------------ 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 forum is accessible at: http://tech.groups.yahoo.com/group/si-list 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