2
\$\begingroup\$

I'm designing my first PCB layout at 2.4GHz, I have done a couple of designs at 868MHz, but never 2.4GHz.

I have due to limited space routed some components in 90 degree, but know that the guys on the review team will complain about this, because they thinks it is bad practice.

I have read numerous of articles that says it should be ok, but the articles never mention a frequency limit, so because this is my first 2.4GHz design, I just wanted to ask here, if my layout is not optimal.

Shown below is the PCB TOP layer, below this layer is a solid GND plane.

RF output circuit from Bluetooth IC CC2650

\$\endgroup\$
3
\$\begingroup\$

"High speed signal propagation" does not have an issue with right angle bends at those sorts of frequencies, and in any case your 'bend' is really a component pad, so not much you can do about the discontinuity there.

If you had a right angle in tracking (not at a component) than clipping off the corner is not horrible practise, but it needs to be fairly extreme and is more of a hard core microwave sort of thing to do, if you have not been forced to use teflon board to control the losses, don't bother. Most modern cad systems can do this for you.

I do wonder about that place where the signal seems to split in two before connecting to the chip, mostly because it does not look like any microwave structure I am familiar with and is physically too small to be a wilkinson divider.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ The split is a balun made from passive components. See the CC2650 datasheet for circuit. Btw. will a discontinue (impedance mismatch) increase EMC radiation, and if so, why? \$\endgroup\$ – JakobJ Jun 9 '17 at 10:14
  • \$\begingroup\$ Maybe it's interesting to note that the return path always takes the value of least impedance. With high-frequency this means that your return current is actually going just below the signal. Meaning that your return current will take this bends, although it would seem counterintuitive. This is because inductance dominates return. Take a look at this: learnemc.com/identifying-current-paths \$\endgroup\$ – Andrés Jun 9 '17 at 11:30
  • \$\begingroup\$ Yes, but am I correct that the return path for the above circuit would be equal, even if the components did not have the right angle turn. The signal would still have to travel beneth the length of the two components in the bend, even if they where mounted on a straight line. ? \$\endgroup\$ – JakobJ Jun 9 '17 at 14:33
  • \$\begingroup\$ Yep, what you have drawn is NOT noticeably worse then a straight line at 2.4GHz. The only real improvement would be to pick smaller discrete parts so the land geometry matched the trace width and even that is marginal at 2.4. Move the frequency up by a decade and you have to start sweating this stuff, but 2.4GHz rf with short connections, meh. \$\endgroup\$ – Dan Mills Jun 9 '17 at 15:19
1
\$\begingroup\$

The only concerns with 90 degree traces are 1) acid traps and 2) impedance discontinuities. The first, acid traps, is mostly outdated because of modern PCB fabrication processes and is not really a concern anymore. The second is potentially still a concern, though I have not seen any actual data suggesting that sharp corners have a noticeable impact on signal quality. The impedance of the trace does change slightly at a 90 degree corner (the cross section is slightly wider at a corner than on a straight section) but it isn't really any worse than 45 degree angles (especially two of them).

While I still avoid right angled routing just because I don't like the way it looks, personally, in all reality they probably won't cause any noticeable problems in your design. In fact, in the image you posted I don't see any traces routed like this, other than at T junctions which is perfectly fine. You might want to look up teardrops though -- they reinforce the junction of traces to pads.

| improve this answer | |
\$\endgroup\$
  • 1
    \$\begingroup\$ Rick Hartley did an experiment with 90 degree, chaffered, and rounded corners and from what I remember (don't have the notes in front of me), it didn't make that much of a difference. \$\endgroup\$ – efox29 Jun 9 '17 at 13:55
  • \$\begingroup\$ I think the question is about component placement at right angles, not trace bends. \$\endgroup\$ – The Photon Jun 9 '17 at 14:17
  • \$\begingroup\$ @ThePhoton I think you're right, I re-read the initial post and understood it differently the second time \$\endgroup\$ – DerStrom8 Jun 9 '17 at 14:37
1
\$\begingroup\$

Such component arrangement is OK at 2.4GHz. You can model them as lumped elements, and get good results. However, you seem to have a via right before the antenna. Is it intentional?

Regarding the balun, I would have used a ceramic balun instead of a discrete solution. With the discrete solution component tolerances and parasitic effects become very critical.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ Yes the vis is intentional, there is an RF test pad on the BOTTOM, which I know is a small open stub. You wrote at 2.4GHz there should not be a problem, at what frequency would I have to take special care. ? \$\endgroup\$ – JakobJ Jun 9 '17 at 14:37
  • 2
    \$\begingroup\$ For 0402 and 0201 devices, lumped element design becomes problematic around 8GHz, mainly due to self-resonance of inductors and parasitic capacitance of the pads. \$\endgroup\$ – Lior Bilia Jun 9 '17 at 17:22
  • \$\begingroup\$ Discrete lumped element that is. \$\endgroup\$ – Lior Bilia Jun 9 '17 at 17:42
1
\$\begingroup\$

Charges needing Return Paths always explore ALL POSSIBLE PATHS, at all frequencies, proportional to susceptance (reactive conductance).

Beware of proclamations "The return current takes the shortest path, or takes the lowest resistance path, or takes the path of least impedance."

That path ----- 40dB down ----- will still be driven with the stimulus voltage and -40dB down will be the energy of that path. Ditto for -10dB or -5dB or -95dB.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ Ok, so some current will flow in other parts of the copper, then just beneth the signal trace. But when could this present a problem? \$\endgroup\$ – JakobJ Jun 9 '17 at 17:15
  • \$\begingroup\$ That spreading of current causes even far-away paths to contain a predictable (using Inductance-of-loop computations) fraction of the total GND (RETURN) current. In building radios, or frequency synthesizers, energy of -100dBc or -120dBc or -140dBc (walky-talky closein desensing) levels are part of successful PCB planning. Or the 16-bit ADC (-90dBc) or the 12-bit ADC (- 66dBc) need control of return currents. \$\endgroup\$ – analogsystemsrf Jun 10 '17 at 3:11

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.