Announcement

Collapse
No announcement yet.

Guard Trace for Radiated Emission Test

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Guard Trace for Radiated Emission Test

    Hello Robert,
    I am facing an issue for PCB Layout. As you have good PCB knowledge, so I thought to approach for your help and guidance as well.
    Could you please help me?

    Problem:

    We are having a failure in RE test, as two frequencies of 336 & 208 MHz are crossing the limit line. The clock frequency is 16 MHZ. The clock signal is highlighted in green color.

    The changes in 2 PCB revision is attached for your reference. Compare to previous revision, I changed the guard trace only. Earlier, it was continuous around clock signal while in next revision, it is broken. The clock signal and guard signal is in inner layer 4, which is referenced with Vcc on layer 5. The PCB is a six layer stack-up. Layer 2 is GND and 3 is used for routing the signals.

    Query:

    1. Even I broken the guard loop around clock, but connected the broken guard track to GND via on bottom side. Will it work or not to control radiation test?
    2. As left side guard traces is still same as earlier revision to some extent, so will it work to prevent radiation.
    3. Or the broken guard is the strong reason for this failure.

    As learned from experience, guard traces are used to avoid cross talk to adjacent trace. Does guard traces help to reduce radiated emission? Should we use guard traces on inner layer as well?

  • #2
    Query:

    1. Even I broken the guard loop around clock, but connected the broken guard track to GND via on bottom side. Will it work or not to control radiation test?
    best is to keep it connected on both sides and with multiple via's , now you have made an antenna for the signal to propagate in.

    2. As left side guard traces is still same as earlier revision to some extent, so will it work to prevent radiation.
    Radiation of the signal is confined by the two adjacend layers.and the gaurd traces of course but since layer 3 is the signal layer, i would suggest to use the option add shielding to net.

    3. Or the broken guard is the strong reason for this failure.
    in my opnion the gaurd trace does nothing at this moment.

    Comment


    • #3
      Dear Paul,
      Thanks for your reply,

      1. I didn't use the guard traces with multiple vias in previous version as well. so was it not effective guard tracks to prevent radiation.

      2. I have used only one GND via at the point, where the clock is changing the layer from layer 4 to layer 1. is it the effective way to control radiation?

      3. In 3rd Picture shared by you, Guarding is not a complete loop. it is individual guarding with multiple GND. So i learned that a creating loop with just via is not effective way. Its better to have multiple stitching.

      Comment


      • #4
        Dear Paul,

        If guard trace with one via works as antenna, why radiation is not for other signals. i have used guarding for other clock signals in the same way. the same guard traces with one via is used.

        Comment


        • #5
          for highspeed signals it is wise to add multiple via's to gnd when changing layers.. keep them parrallel to the transistion via (so equal distance) see image
          I am assuming that with radiation you mean the crosstalk it will put onto other signals.

          https://www.youtube.com/watch?v=N1TIjxdZIoE
          above might be helpfull

          1. I didn't use the guard traces with multiple vias in previous version as well. so was it not effective guard tracks to prevent radiation.
          in my opnion not.

          2. I have used only one GND via at the point, where the clock is changing the layer from layer 4 to layer 1. is it the effective way to control radiation?

          always put via's near transmission layer changes if it is for high speed signals. it will help with transient responses and gnd loop issues.

          3. In 3rd Picture shared by you, Guarding is not a complete loop. it is individual guarding with multiple GND. So i learned that a creating loop with just via is not effective way. Its better to have multiple stitching.

          a complete loop is not possible in most cases since you have two end points. best way is to keep a fixed distance from the main trace and enough via's to shield other signals. there is an optimum here.. more via's is not always better.. pcb layout is always a trade off in design rules in my opninion.

          Comment


          • #6
            also check your power supplies... having a noise power plane will introduce some nasty stuff on you clock signals..

            Comment


            • #7
              I am not normally using guard traces - and no problems passing EMC even for very complex boards.

              My very first questions would be, if you are 100% sure, that it really is the clock making problems. Did you try to switch the clock off and measure your board?

              As Paul van Avesaath is suggesting, a lot of EMC problems are caused by power supplies - what is on the right of the clock - these look like power transistors?

              Comment


              • #8
                I only do this with single ended highspeed 3GHz-12GHz then gaurd traces become pretty mandatory when having a lot of signals running side by side for long distances

                Comment


                • #9
                  Hello Folks,

                  Nice points above.
                  About the guard traces these works differently in the digital high speed design since these days there are solid reference planes which are included in the design,these reference planes provide the best return path for the current just beneath the track which is always recommended.
                  For your design, if the distance between the signal and guard trace is smaller than the distance between the signal layer and the reference layer then only guard track will be helpful or else it might not even work. You may check this bit.

                  On crosstalk, since it is almost proportional to the square of trace height and inversely with square of separation you may alter these 2 quantities to get crosstalk reduced.

                  Thank you.

                  Comment


                  • #10
                    Hello Robert,
                    Sorry for late reply. Yes you are right, as there are power transistor near by. these power transistors are used to drive the stepper motor. But the previous version was pass with same placement.

                    Regards
                    Meet Payak

                    Comment


                    • #11
                      Originally posted by meet View Post
                      Hello Robert,
                      Sorry for late reply. Yes you are right, as there are power transistor near by. these power transistors are used to drive the stepper motor. But the previous version was pass with same placement.

                      Regards
                      Meet Payak

                      what changed since the previous release? if that one passed you have to take a close look at things changed since then that might cause the issue now?

                      Comment

                      Working...
                      X