8 Replies Latest reply on Sep 14, 2009 7:15 AM by Kenneth_Wood

    Quarter wavelength rule?

    bruce_qin

            I know something about the mysterious quarter wavelength rule, it is usually to be said as: when routing high speed signal trace, the trace length must be avoided to become integral multiple of quarter wavelength (such as 1/4, 1/2 ,1, 3/4, 5/4 and etc.) so to avoid the standing wave, which cause seriouse EMI problem.

            Have I got the correct meaning? If not, what is it? Or if so, my questions:

           1. How to calculate the wavelength?

              considering a material with Dk=4, so I can got the velocity of the signal by 11.8/squar(4)=11.8/2=5.9in/ns, so for a  transmission line with 6Gbit/s signal on it, the wavelength is

                wavelength=velocity*period=velocity/frequency=5.9/6=inch?

           

           2.Even the material is constant, the signals travell in microstrip  and stripline have different velocities, the signal in microstrip will be faster. It means the wavelength for the same bit rate is not constant.  So should I use one wavelength for microstrip and another for microstrip? How about the same net changing from outer layer to inner layer?

       

           3. If there is a capacitor or a via or a connector in the net, should I add its length to the whole net length? Or just calculate each net segment?

        • 1. Re: Quarter wavelength rule?
          yu.yanfeng

          Hi Bruce,

           

          1/4 Lamda is a concept in Microwave passive circuit, which mean you can implement a passive component  to acting as coupling,directional or shorting, by using a simple copper line. However, It must be desinged and arranged based the theory. the connection line between driver and receiver is out of this scope! please don't get confused.

           

          1) You should calculate the Lamda based the Vp which is equals  3X100,000,000/ root of Epslon according to interest frequency

           

          2) Vp in the microstirp is slow than Vp in the stripline(determined by Epslon of Air and Epslon of FR4).

           

          3) Only Allegro support Z length, Expeditonpcb doesn't support Via length. And I think you better simulate it and don't manually add via length into your delay formula.

           

           

          Yanfeng

          1 of 1 people found this helpful
          • 2. Re: Quarter wavelength rule?
            strangd

            Good day Bruce:

             

            If you have terminated your transmission lines properly, you should not have enough reflected energy cause you EMI problems.  However, if you allow a reflected signal to a arrive at the source at the same time as the source is high and the reflection is high, you could over voltage your driver's output stage and, of course, design in a latent failure for your board.  If the reflected signal returns to the source when your driver is low and the reflection is low, the voltage at the driver's output will go negative and it will eventually take out the diodes that protect the driver's output stage.  Again you have designed in a latent failure into your board.

             

            Now if your worry is that wires that are a quarter wave length of the signal that is being sent through makes a good antenna and not matching that length degrades the antenna's ability's to radiate, true enough.  But please remember that you are not in the business of eliminating fields.  You are in the business of generating fields.  In a well designed board half and the fields are going positive at the same time as the other half of the fields are going negative.  Your job is to "null" the effects of these fields.  Then you pass your EMI tests.  Remember that quarter wave length antenna may well have had a length of coax between it and the transmitter.  The signal wire was radiating all of the way from the transmitter to the tip of the antenna, but the portion that was covered by the coax shield had an opposite field radiating from it and a nulling resulted.  Design you board so that your 6Ghz lines will induce your reference planes to generate nulling fields.  (continuous, unbroken planes)  If you need to match the length of each side of the diff pair, do your best to see that for most of the signals' travel, the signals are side by side. ( Don't make up all of the difference at one end.)

             

            You are at 6Ghz; make sure that your via stubs are less than 0.032".

             

            If you are going more than a couple of inches at 6Ghz or greater, see yu.yanfeng's posting on routing at odd angles in Expedition.

             

            have fun

            Dwain

            • 3. Re: Quarter wavelength rule?
              yu.yanfeng

              Hi Dwain,

              For EMI issue,Your explaination is perfect. I think Bruce now get understanded that  a well- terminated trasmission line does'nt genrated too much EMI to outside air.

              Yanfeng

              • 4. Re: Quarter wavelength rule?
                bruce_qin

                Dear Dwain,

                        Thanks for your detail explaination.

                        Since the 6G signal on my board travels more than 10 inches total, and I think the connectors and vias also will impair the integrity of the signal. So could you give some advices on the layout of this situation?  Or just a hyperlink for more reference?

                        Beside that, you mentioned minimum 32mil via stub, why 32mil?  The fact is that all my vias has a stub longer then 32mil, anything I can do to eliminate the harm caused?

                • 5. Re: Quarter wavelength rule?
                  yu.yanfeng

                  Hi Bruce,

                   

                  Why keep the via stub less than 32mil?  Here is the anwser:

                   

                  1 You mentinoned the data rate is 6G Bit, and We suppose it's NRZ signaling, so the base frequency is 3GHz, and We at leaset we shoud take care  to 3X the frequencey according to Nequist Sampling Law

                   

                  2 You manethioned you are serious to 1/4 Lambda issues. For 9 GHZ, the 1/4 Lambda is (300,000,000 (M/S)/900,000,900(Hz )/4, that is 320mil. So this size of stub make the stub as a excellent microwave component now, and it can constribute anything you can't get really know before the pro-type.

                  3 To  avoid the stub's effect , So we have to control to 1/10 of the size. That is 32mil.

                   

                  I hope you can understand the basic and run a simulation use HL GHz. It's an exllent tool to design and I believe You may get benifit by the simulation.

                   

                  Yanfeng

                  • 6. Re: Quarter wavelength rule?
                    Steve_McKinney

                    Hi Bruce,

                     

                    Just to add to what Yanfeng and Dwain have said - one of the things to remember is that this is a digital design, not an RF/Microwave design.  The 1/4 wavelength rule applies to RF/Microwave because you have a narrowband signal at a very specific frequency.  If you were to take an iFFT of a sinusoid, you'd see a single impulse at the frequency of your signal (in your case, if this was RF, there would be energy at 3Ghz) so it becomes much easier to create a powerful antenna.  However, in the digital world, we have wideband signals. If we did an iFFT of the digital signal, we'd see energy across a broad spectrum of frequencies.  Sure, there may be more energy at your fundamental frequency of 3Ghz but overall, the energy is spread out - that's what makes up our digital switching edge. It becomes much more difficult to create an antenna because there are many wavelengths.

                     

                    So what am I getting at?  Unless you have a 3GHz clock signal (I assume your 6Gbps link that is 8b/10b encoded which isn't really going to look like a clock signal at all), then you should have little to worry about in the way of 1/4 wavelength EMI propogation because it would be really tough to design an antenna with this type switching signal.

                     

                    -Steve

                    1 of 1 people found this helpful
                    • 7. Re: Quarter wavelength rule?
                      strangd

                      You can control the via stub length by designing with blind vias.  On backplanes, via back drilling is often employed to remove the stubs.  That is, after the board is built using through hole vias, the via barrel is drill out from the surface of the board down to the layer of the first active trace for that net.

                       

                      Dwain

                      • 8. Re: Quarter wavelength rule?
                        Kenneth_Wood

                        Hi,

                        Our PCB Toolkit should help you out with this.

                        Use the "Signal Properties" tab for these calculations.

                        Ken

                         

                        www.saturnpcb.com