3
\$\begingroup\$

I've been doing some simple RF boards as part of an internship in a lab, mainly to test particular ICs and prototyping, and mainly using coplanar waveguides.

I've now been tasked with designing a board to test an ADC with two differential inputsfunctioning in the 1GHz range.

Now, I know how to calculate coupled microstrip impedance, etc, but I'm kind of at a loss as to how to properly do the transitions (or even how much of an impact said transitions have on the overall impedance).

The following picture is the particular situation I'm now stuck in:

Example of the type of transition I'm talking about.

This would vaguely be my idea of what the connection for the inputs should be (for discussion purposes, lets assume the coupled microstrips are properly dimensioned). However, I have 2 types of transitions that will (I think) give me some troubles.

  1. Transition between pad and track: I usually manage this just fine, just by adding a teardrop at a certain angle (according to the info I've found around the web); then I just vary the distance between the pad and ground plane to keep the same impedance as the microstrip. I'm curious though as this is the most used approach.
  2. Transition between microstrip and coupled microstrip: This is where I'm more at a loss. Do I just vary the width of the traces and place a teardrop between them? Sound kind of a hamfisted approach, though I can't really say why it wouldn't work. I've seen a couple of boards that place some vias and change layer for this kinds of transitions, but I couldn't find justification for this.

For reference, this is my stack up:

Top: 1-oz copper.
   6.7 mil FR-408 prepreg (Er = 3.66 @ 1 GHz).
Ground: 1/2-oz copper.
   47 mil FR-408 core.
Power: 1/2-oz copper.
   6.7 mil FR-408 prepreg.
Bottom: 1-oz copper.

Thank you all in advance for your answers!

EDIT: I'm adding an example of what I'm trying to achieve from the board we are currently using to get our measurements. It's a reference design by National Semiconductor, link for the app note here. There are design files available if anyone's curious, that's where I took the picture from.

As you can see, there are pink tracks on a 3rd layer, that are coupled but then separate in the direction of the pads. Finally, there's a via right underneath the pads to connect with them. I find it interesting that there is no transition at all done between the coupled microstrip and microstrip, and that the via to connect with the top layer don't generate problems.

enter image description here

\$\endgroup\$
  • \$\begingroup\$ Please give a little more details on your problem. For example: Are the tracks routed to different connectors and if yes, why? Where does that signal for the ADC come from? Is the connector GND connected to your PCB ground plane and how? ... \$\endgroup\$ – Stefan Wyss Sep 30 '18 at 18:50
  • \$\begingroup\$ @StefanWyss Thank you for the response. As to your questions: - Yes, the tracks are routed to different connectors, as the ADC has a differential input and I'd like to take advantage of it. - The board is supposed to be as general as possible. One of the things I'm gonna use it for is to measure the output of an I/Q Demodulator, for example. -The connector GND is gonna be connected to the ground plane with several vias. I may also use CPWG instead of microstrips for the tracks. I thought all these details were independent of the transitions themselves. Why is it important? \$\endgroup\$ – G. Alcaraz Sep 30 '18 at 20:17
1
\$\begingroup\$

The most important design goal is to keep the wave impedance as homogenous as possible. This impedance is defined between the track and GND plane (microstrip) or between two tracks (differential microstrip pair) or between two tracks with GND plane (3 conductor differential pair). It is not easy to change from one to another.

The 1. transition between pad and track is not a problem, because it is 1) just tiny and 2) electrically short compared to your 1GHz wavelength.

But the 2. transition is hard to do and most likely unnecessary in your case. I have never seen a setup where the wave impedance was defined between two traces on different cables, so you might want to reconsider that.

Try to find a way to route your differential signal over one cable instead. Take a 1GBit LAN cable as a reference example, because this has multiple high speed capable twisted differential pairs with well defined impedance.

\$\endgroup\$
  • \$\begingroup\$ Thanks a lot for your response. Please check, if you can, the EDIT I made to my original post. I give a link to the reference board we are currently using to get our measurements and that the board I'm making is meant to replace. I'd like to achieve something similar to this if I can. We are currently using SMA conns and cables in all our boards... wouldn't a LAN cable get worse performance? Gotta admit I'm just starting at RF design and so there's a lot of basics I skipped over, while there's some rather advanced stuff I did learn :). \$\endgroup\$ – G. Alcaraz Oct 1 '18 at 18:38
  • \$\begingroup\$ Well, this gets interesting. I have seen in the appnote that the differential pair must have 100 Ohms impedance, and I‘m also puzzled of how to achieve that with this forking type of routing. The only explanation I can think of is that there must be a solid GND plane at the forking point, so that the single traces have 50 Ohms to GND, which adds up to 100 Ohms for both. Can you check this in the ref layout? \$\endgroup\$ – Stefan Wyss Oct 2 '18 at 3:54
  • \$\begingroup\$ Indeed, it's a 10 layer stack up and this differential pair is on layer 3, sandwiched between a solid GND plane on top, and a power plane below. I fail to see how the tracks would work though... Shouldn't the impedance change when they are next to one another and once they separate? If you are curious about the design of this board, it's supposed to be found here: ti.com/tool/adc12d1800rb though for some reason the page doesn't work for me. The files are in an Allegro format, you can view them with Allegro physical viewer 16.6 (other versions didn't work for me). \$\endgroup\$ – G. Alcaraz Oct 2 '18 at 4:39
  • \$\begingroup\$ How near are these planes? I think that in the differential section, both diff tracks are very tight and the planes are far, so that the impedance is mostly defined between the tracks, whereas in the single track section, there could be another plane right under each track (very near), so that there is 50 Ohms between track and plane? \$\endgroup\$ – Stefan Wyss Oct 2 '18 at 4:45
  • \$\begingroup\$ Actually, no. The GND plane on top, and the power plane below, are both covering the diff and single tracks. There aren't any other planes between as far as the stack up is concerned. I'm kind of stumped too... \$\endgroup\$ – G. Alcaraz Oct 3 '18 at 1:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.