Make a quick-turnaround PCB for RF parts
Parts get smaller, but the laws of physics don't change. Your designs must change to compensate.
Steve Hageman, AnalogHome.com, Windsor, CA; Edited by Martin Rowe and Fran Granville -- EDN, December 15, 2010
That precision is fine for low-frequency circuits, but RF circuits usually need 50Ω traces for proper circuit operation. Parts get smaller, but the laws of physics don’t change. Thus, a microstrip trace on a 0.062-in.-thick standard prototype board that was calculated to be 0.11 in. wide 30 years ago is still 0.11 in. wide today. Many surface-mount parts are far smaller than their predecessors, however, so it would seem that low-cost, two-layer prototype boards for RF prototyping are unsuitable for today’s small SMT (surface-mount-technology) parts.
You can use a CPWG (coplanar-waveguide-over-ground) structure to build 50Ω RF traces on PCBs. A CPWG structure lets you make the required trace width smaller than that of a microstrip structure.
| Read More Design Ideas |
How can you design the CPWG structure for a low cost and a fast PCB process? You can find many online CPWG calculators, but they often fail when the ground-plane gap gets less than approximately 30 to 50% of the trace width because the height of the copper traces on the board becomes a significant factor. It adds more capacitance than the calculators assume. Hence, the lines these calculators design have too much capacitance, which reduces their impedance to less than 50Ω. The equations date back many years to IC design.
The equations in many calculators fall apart because today PCBs differ physically from ICs. The best way to properly design a CPWG on a PCB with a narrow gap-to-center-trace ratio is to use a full 3-D electromagnetic simulator. This Design Idea provides the values for a few common structures.
In keeping with the minimum trace-to-trace spacing of 6 mils, I simulated, built, and tested a CPWG structure. For a common 0.062-in.-thick FR-4 PCB material, a trace width of 0.032 in. with a gap of 0.006 in. is as close to 50Ω as you can get. It provides better than 40-dB return loss on the trace at 6 GHz.
This approach is better than using
a 0.11-in.-wide trace and is compatible
with SMT-sized parts. A 0603-sized SMT part and a common SMA
(surface-mount-assembly) edge-launch
connector fit the line perfectly. Figure
1 compares several common RF-type
parts with the fabricated PCB. For parts
with larger pad dimensions than the
0.032-in. trace width, just increase the
spacing to the top ground plane to compensate.
For instance, increase the spacing
to the top plane of a 0805 SMT pad
to approximately 0.008 in. and increase
the top-plane spacing for a 1206 SMT-component
pad to 0.012 in. to keep the
pad from being too capacitive.In keeping with common design rules, I pulled back the copper planes on the tested PCBs 0.01 in. from the routed board edge. This pull-back and the edge-launch connector both add a slight amount of inductance to the transition, however. The big center pin of the edge-launch connector on top of the trace adds extra capacitance, providing built-in capacitive compensation. Cutting the pin to about half its original length yields about equal capacitance to balance the transition inductance.
The CPWG structure needs a solid ground plane under the trace; leaving cutouts in the bottom ground plane under the topside trace adds a significant inductance to the structure, which degrades high-frequency performance. You also need to “stitch” the top ground plane to the bottom ground plane with vias. Place the stitching vias less than one-eighth of a wavelength of the highest frequency that your circuit will use. Note that 0.1-in. spacing works well at frequencies greater than 10 GHz.
Spacing of the stitching vias to the center trace follows the same spacing rules. You can easily get enough vias in and around the trace to make it work.
If you don’t have enough vias, you will see a slight but rapid 0.5- to 1-dB drop in the S21 transmission characteristics instead of a linear loss slope with frequency. You can instantly see this effect by using a VNA (vector network analyzer). Measuring the test board shows approximately 0.25 dB/in. of loss at 3 GHz and 1 dB/in. of loss at 10 GHz, including two edge-launch connectors.
To interface to an SMT part or an IC with narrower pads than 0.032 in., narrow down the center conductor as needed as close to the part as possible. If the discontinuity is physically small, it will have little effect until very high frequencies.
Talkback
-
Hi Steve,
Thanks, it does answer my question. I couldn't tell any kind of relationship from the picture as I can't see the edge of those ground strips. My main concern was how much board real estate this would take. It would have been kind of bad if those ground traces needed to be 1" wide! :)
Larry L - 2010-20-12 13:33:43 PST -
Larry L Asked - How wide do the top ground plane straps need to be? Larry I model them at 2 times the center trace width - mostly to keep the side walls in the EM simulator out of the picture. Most of the current flows to the via and then directly down to the bottom ground plane. Very little current will flow past the via. I am sure that in practice a width that is the same as the trace width would work just fine. Just be sure to stitch the vias in there. I hope this answers your question.
Steve Hageman - 2010-20-12 07:33:36 PST -
Very useful with some good practical tips.
A good reminder that microstrip is just one
of the techniques to use.
I'd like to hear more about the various
microwave layout techniques.
Tom Christian - 2010-18-12 04:46:43 PST -
Hi Steve,
Nice article. How wide do the top side ground strips need to be to have this effect? That is, do the ground strips on either side of the center trace need to have a certain fixed amount of width regardless of the center trace width or do they need to be at least x times the center trace width?
Larry L - 2010-17-12 18:30:42 PST -
The edge launch connector referred to in the article is a Johnson P/N: 142-0711-821, in case anyone is interested.
Steve Hageman - 2010-16-12 19:51:13 PST





















