Very good contribution Robert, One enhancement to point 7:
It's the current edge that you want to reduce in speed, not the voltage edge !!! This is something very generic in EMC: Look at the current, not the voltage !! Regards, Gert Gremmen ce-test, qualified testing >-----Original Message----- >From: owner-emc-p...@ieee.org [mailto:owner-emc-p...@ieee.org]On Behalf >Of Robert Macy >Sent: Sunday, January 23, 2000 8:42 PM >To: rehel...@mmm.com; emc-p...@majordomo.ieee.org >Subject: Re: EMC Circuit Board Design > > > >Here are a set of rules worth recommending: > >Engineering design checks: >1 Avoid "spaghetti logic" block diagram. Make certain there's a good flow >of functionality across and around the PCB. This includes using >the concept >of transferring INFORMATION not ENERGY (except where one must, like with >clock distribution, but even then, if the clock goes a long way, use a >buffer) >2 ABSOLUTELY check the design for bus "contention". Not the debilitating >kind that stops the functionality, but that pesky little 1 nS of overlap >from sloppy digital design. It's like using two make-before-break switches >for a 3 way light bulb in your home. It will all work, but you will get >some dramatic action at each transition. > >PCB Layout: >1 Go on and off the board along one edge as close together as possible. >*IF* a connection must come on or off from some where else make it a high >common mode impedance connection with beads (for dc leads) and common mode >chokes for higher frequency signals. On the PCB reference the trace(s) to >CHASSIS gnd area where the traces have been isolated by going through high >impedances. Bridge between chassis referenced area to PCB gnd referenced >area using the high impedance component. >2 Interleave PWR GND and SIGNAL GND traces at the connectors to >lower their >impedances. Particularly high speed lines need their own gnd returns. >3 Use short runs, few "meandering" traces. >4 Bypass caps: use all the same value, the thermal relief pads >all the same >dimensions. >5 Have traces maintain good distances between edges and cutouts. [This to >insure a good low impedance ground return path for current.] >6 On particularly "hot" traces include the ground return path and don't >(unless absolutely necessary) "jump" layers. Use 2 traces, if possible and >if you must jump layers (like bus lines) add 2 ground vias around >the signal >via. >7 Lower the rise time to minimum speeds with SERIES source terminations. >Use a resistor 22-50 in series plus a bead if you have room. DO NOT EVER >LOWER THE RISE TIME OF A LOW IMPEDANCE SOURCE WITH A CAPACITOR. >8 Keep high speed traces inside the layers, don't route them on >the surface, >you'll reduce the radiation by around 14 dB. >9 Stack up the board with traces as close to the GND & VCC planes as >possible (like 4-5 mils even for a 4, or 6 layer board) This will >lower the >impedance of the trace *and* reduce crosstalk. >10 Use minimum number of vias to prevent turning the GND & VCC planes into >Swiss cheese. Sometimes blind vias will help. >11 Where the traces for the bus lines turn and jump layers, don't allow >your PCB designer to line up all the vias with the vias so close that they >create a cutout slot! Stagger the vias' positions, or do >something, to make >certain there's room for adequate metal between each via. >12 Discourage the use of high speed clocks such as 125MHz, 1.25GHz, etc. >with the argument that using such high speed clocks demonstrates a >designer's weakness because he can't make his design work with slower >clocks. Demand the clock frequency be no higher than 20MHz, because we all >know how to solve those problems. <g> > > - Robert - > >-----Original Message----- >From: rehel...@mmm.com <rehel...@mmm.com> >To: emc-p...@majordomo.ieee.org <emc-p...@majordomo.ieee.org> >Date: Friday, January 21, 2000 5:39 AM >Subject: EMC Circuit Board Design > > >>Dear List-Members, >> >>I am requesting information/opinions/etc. on the following: >> >>When circuit boards are designed, what are the common mistakes that the >>circuit board designers make regarding EMC (multi-layer boards in >>particular)? >> >>You can respond to me directly but I would prefer a response to >the list as >>I believe that the question is of interest to many on this list-server. In >>either event I will compile the responses and resend the >compilation later. >> >>Thanks for your time, >> >>Bob Heller >>3M Company >> > > > >--------- >This message is coming from the emc-pstc discussion list. >To cancel your subscription, send mail to majord...@ieee.org >with the single line: "unsubscribe emc-pstc" (without the >quotes). For help, send mail to ed.pr...@cubic.com, >jim_bac...@monarch.com, ri...@sdd.hp.com, or >roger.volgst...@compaq.com (the list administrators). > > --------- This message is coming from the emc-pstc discussion list. To cancel your subscription, send mail to majord...@ieee.org with the single line: "unsubscribe emc-pstc" (without the quotes). For help, send mail to ed.pr...@cubic.com, jim_bac...@monarch.com, ri...@sdd.hp.com, or roger.volgst...@compaq.com (the list administrators).