Boss, can I have a pay raise to $7.75/hr ? A few more t-shirts also, please ? -----Original Message----- From: steve weir To: Chris.Cheng@xxxxxxxxxxxx; 'MikonCons@xxxxxxx'; si-list@xxxxxxxxxxxxx Sent: 3/18/2004 8:15 PM Subject: Re: [SI-LIST] Re: ESD is a low frequency event -really?? Chris, we are in absolute agreement about what Doug's experiment shows. The nice thing about using my original model is that structure has been around a long time and there is lots of characterization data for it. Now, unsurprsingly, what it shows is pretty much what has already been said here, but I would just like to clarify a little bit. The real situation is a matrix, but the one section structure is good enough for discussion. 1) As Mike points out, the common to differential mode conversion is determined by the match of the impedance dividers formed by the combination of R1/L1/Cy1 in one path and R2/L2/Cy2 in another, and then by the shunting of Cx. The match before Cx is unlikely to be much closer than 1-5%. 2) As you point out, the energy goes out through Cy1/Cy2. So, I completely agree that what we want to do first is make the impedance in the Cy1 and Cy2 paths: a) As low as economically possible. This means as many tie points as we can tolerate. They don't all have to be bolts. Chassis spring fingers up to pads work too. b) To the extent practical, balanced. Uniformity, particularly near the edges is helpful. Beyond that, it is a matter of modelling and/or lab verification to see if we have met our goals. If so, we are done. If not, the first choice in my mind is to: a) modify chassis ground placements, more or different locations. b) if the source of the vulnerability is physically localized work the geometry in that area. Ferrite, selective moating, spacing modifications may all be useful, and relatively cheap. The second choice is to throw money at Cx in the form of lowered plane impedance. In this regard, thinning out, is to me much more effective than changing to high Er material. It is inductance that is the primary enemy. Higher Er will extend the low frequency limit downward where we get a given CMRR. This is very visible from the model by changing the parameters of Cx. Regards, Steve. At 02:47 PM 3/18/2004 -0800, Chris Cheng wrote: >The way I see it, CY's control the absolute noise on the nodes and CX >control the fraction of the absolute noise that will appear as differential >noise across power and ground. You can either attack the fraction or the >absolute value to enhance your immunity. > >Now let's see who gives a better bang for the buck. > >Any sane designer has to start with some power/gnd pairs in a PCB (though >not BC). Without paying premiums for BC, you are looking at 3-4 mils >dielectric planes. If in a panic you switch to BC you can drop it to 1-2mil, >may be a little thinner if you pay an arm and a leg. So you are looking at a >1.5x-4x improvement in CX switching to BC. > >Now if you completely leave the chassis open wrt logic ground, your CY will >probably be close to the plane capacitance between the PCB and the back side >of chassis a few mm away. That will be in ~pf/in^2. >However, if you provide a dead short with a bolt that tie the PCB logic >ground directly to chassis, you now parallel the bolt with that plane >impedance. At ~GHz the inductance of the bolt with a few mm in diammeter and >a few mm in length (across the PCB) will be in the ~nh range, and even >better if the frequency of interest decrease. So the dead short bolt will be >compatible if not better than the impedance of CY. Evertime you add a dead >short bolt, you drop the impedance by a proportional fraction. The good news >is, the bolt is almost free since chances are you have to mechanically tie >the PCB to the chassis anyways for shock and vibe reasons. It is just a >matter of isolating the tie point on PCB or shorting it to logic ground. So >if you need to drop CY by 2x-4x, short 2-4 bolts near your area of concern, >probably for free. And I believe that's what Doug's experiment shows. > >Sometimes you get the "surprise" benefit from you disk or PC card vendors by >them shorting their logic ground to their chassis without even telling you. >Now that's the best bang for the buck. > >-----Original Message----- >From: MikonCons@xxxxxxx [mailto:MikonCons@xxxxxxx] >Sent: Wednesday, March 17, 2004 4:47 PM >To: si-list@xxxxxxxxxxxxx >Subject: [SI-LIST] Re: ESD is a low frequency event -really?? > > >Good comments and thoughts, guys. My thoughts follow. >************** >Steve's original model (with non-ideal capacitors): > >vsource 1 0 > >R1 1 12 >L1 12 13 >CX 13 23 >CY1 13 0 > >R2 1 22 >L2 22 23 >CY2 23 0 >*************** >This is a good model for dimensions where the lumped elements apply. Steve's > >comments on a more complex (distributed) model would be necessary for the >higher frequency content of the leading edge of an ESD strike. The higher >frequencies are more likely to cause soft errors from the coupled >spikes/glitches. I >have used the essence of this model configuration in differential analog >signal >conditioning circuits to reduce CMV-induced errors since 1961, and it works >beautifully. The CMV rejection (i.e., the reduction in differential, or >normal-mode, voltage conversion from injection of common-mode voltage) is >inversely >related to the relative size of Cx to Cy. This means one can have a total >imbalance in Cy1 versus Cy2 and compensate for it by a large ratio of Cx to >Cy. The >impact of differences between R1 and R2 is also beneficially reduced. > >You can model it or do the math, but the result is the same. Where Cx (in >our >case, the BC value) is substantially larger than Cy (i.e., the coupling from > >either the ground or power plane to chassis in the model(s) noted), then any > >voltage coupled to the circuit is forced to be primarily CMV. When BC is >used, >a representative ratio of Cx to Cy for a small 25 square inch PCB is ~2500:1 > >or ~68 dB. If the coupling from chassis to the power and ground planes is >the >same within 10% (i.e., Cy1 and Cy2 are within 10% of each other in value), >an >additional 20 dB of rejection is added. A larger PCB gets proportionally >better rejection (e.g., a 10 x 10 inch PCB would demonstrate a further 12 dB > >improvement). >********* >Now, given that one uses the (now infamous) chassis ground rings that I >recommend for some applications, the imbalance in Cy1 and Cy2 can easily be >held to >5% or less, which adds another 6 dB of improvement. In summary, a 100 square > >inch PCB designed with two BC sandwiches (which yields ~1 nF/square inch) >and >chassis ground rings can be expected to exhibit ~106 dB rejection of an >ESD-induced voltage (or from any other external source of EMI). >********* >To address the numbers requested by Steve, one has to estimate the values of > >Cy1 and Cy2, identify the PCB size, and calculate the expected CMV rejection > >(i.e., conversion to normal-mode voltage). I haven't provided specific >numbers >before as there are clearly some caveats and special conditions that apply >to >each unique design. Additionally, be aware that any modeling and associated >analysis will be altered by the frequency range over which the prediction >will >apply. Assuming ESD is the source of concern, the lowest frequency content >of >the strike will be beneficially affected by the bulk capacitors (as well as >the >many typical ceramic capacitors) of the PDS design. The midrange frequencies > >(up to ~50 MHz) will benefit from the ceramic capacitors. Beyond ~50 MHz and > >up into the GHz region (where most soft errors are induced in today's >designs), >the BC capacitance will be most effective. ************** >Note that many designs that are not directly exposed to the high frequency >content of the leading edge of an ESD strike (as typically applies in a >shielded >enclosure) are reasonably protected by a good PDS design that DOES NOT HAVE >TO HAVE BC content. This is most likely the reason that Chris Cheng has >demonstrated success without the use of BC techniques. The shielding effect >of a >conducting enclosure improves with increasing frequency because of the >lesser >depth penetration due to the frequency dependence of skin effect. >*********** >Hopefully, the above comments will explain why one can't simply give a >simple >guideline or a couple of formulas for prevention or correction of ESD (or >other radiated susceptibility) problems. > >Respectfully, > >Mike > >Michael L. Conn >Owner/Principal Consultant >Mikon Consulting >Cell: 408-821-9843 > >*** Serving Your Needs with Technical Excellence *** > > >------------------------------------------------------------------ >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 FAQ wiki page is located at: > http://si-list.org/wiki/wiki.pl > >List technical documents are available at: > http://www.si-list.org > >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 FAQ wiki page is located at: > http://si-list.org/wiki/wiki.pl > >List technical documents are available at: > http://www.si-list.org > >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 FAQ wiki page is located at: http://si-list.org/wiki/wiki.pl List technical documents are available at: http://www.si-list.org 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