# Should long RF traces be differentially routed?

I'm designing a phased array controller and as such includes a dozen splitters . The board dimensions are approximately around 10 - 11 cm and the RF traces are about 8-9 cm in length . Would differential pair routing offer significant improvement ? Would using a balun to convert the diff to single ended signal introduce any noise which defeats the purpose of using diff ended routing ? Operates in C band (4.8 - 5.4 Ghz)

• What frequencies are you using? – The Photon Mar 23 '17 at 18:28
• @ThePhoton added – Aakusti Mar 23 '17 at 18:31

So, in general: For RF signals, you'll have a properly matched transmission line, so that the only thing affected by the line length is the phase of the output signal relative to the input signal.

Since these lengths don't change during operation, you can calibrate that out once, and not care about it.

If your job, however, is to have a defined phase relationship between in- and outputs, you'd of course have to take care to properly choose and match lengths (but I guess you knew that).

Then, you forgot to mention the frequencies you're operating at. Of course, 9 cm means something different for the phase of a 10 GHz signal compared to a 100 MHz signal going through the same length.

Then: Your Balun question doesn't make much sense. You can have single-ended transmission lines on a PCB (typically, something like a single microstrip line, or a coplanar wave guide with ground forming the outer conductor), or balanced transmission lines, like a coupled coplanar line. You'd typically choose the transmission line type based on the thing you need to do, and the abilities of your production process and substrate material.

• The board operates in C band (4.8 - 5.4 Ghz). I asked about balun because some of the IC's i'm using accept only single ended signal , so would a balun help ? – Aakusti Mar 23 '17 at 18:28
• Now they've given a frequency band, I think they need to worry about attenuation as well as phase. @Aakusti, is this standard FR4 board material, or are you using some high-performance material? – The Photon Mar 23 '17 at 18:47
• @ThePhoton its 185HR from Isola(isola-group.com/wp-content/uploads/2016/11/…) . I guess its slightly better than FR4 and they have given specifications up to 10Ghz . I'm also ordering it from OSHPark and they use FR408 – Aakusti Mar 23 '17 at 18:54
• @ThePhoton while I do agree that $\tan\delta$ isn't great for cookie-cutter FR4 (which doesn't specify any dielectric properties), my experience is that, meh, you can build pretty nice devices from good 4 layer FR4 – Marcus Müller Mar 23 '17 at 19:07
• @Aakusti, you should be good then. You definitely don't need to use Rogers or anything like that for 5 GHz and 10 cm these days. – The Photon Mar 23 '17 at 19:58

No, a balun will not change the performance you achieve with your circuit. The phase of each path will be static, plus or minus the impact of PCB manufacture tolerance, loading variation and component placement.

Using one single-ended transmission line or a balanced pair might result in some averaging of the errors, or might increase them - I expect the difference might be minimal unless you're cutting PBCs by hand...