6

I am working on a battery powered 2 layer RF design using a 32.768kHz Crystal oscillator with an Atmega328p micro controller. The crystal will run a timer in asynchronous mode as an RTC clock.

My pcb board is fairly small, and the crystal is pretty big. The easiest way for me to run traces using my current layout would be to run my VCC trace under the crystal from the battery. At this low of frequency would this cause problems in my power domain (oscillating/ noise)?

Trace under Crystal

I do not have it on the board yet, but the bottom layer of the board will be a solid ground plane (for RF return path), and the top layer traces will be surrounded with a ground pour as well. (With via stitching to eliminate any impedance between layers).

Would I be safe to run this power line under the crystal and into a 47uF capacitor. Then do a star power distribution?

placeholder
  • 29,982
  • 10
  • 63
  • 110
apaul
  • 169
  • 1
  • 2
  • 10
  • Datasheet says? – Ignacio Vazquez-Abrams Oct 29 '14 at 02:03
  • 3
    I do not see anything in the [datasheet](http://www.abracon.com/Resonators/abs25.pdf) or [drawings](http://www.abracon.com/Support/PackageDrawing/Resonators/ABS25-2D.PDF) about keeping traces out from underneath the component. – apaul Oct 29 '14 at 02:14

2 Answers2

5

I would never put anything but a ground plane directly under a crystal, especially a high-Z crystal like a 32.768kHz type. And preferably an isolated flag tied to the GND on the chip, with no current running through it, if it's a 2-layer board.

Why don't you put a ground plane under the chip and run the power connection on the opposite side (preferably not directly under the crystal)?

Spehro Pefhany
  • 376,485
  • 21
  • 320
  • 842
  • 2
    Agreed, the easiest thing to do would be to run the trace on the bottom layer, and through a ground plane directly under the chip. This is for a very small 900mHz device, and I am trying to keep almost everything off of the bottom plane. I ended up running power around the crystal. – apaul Oct 29 '14 at 17:44
3

This is OK if you aren't concerned about the accuracy of the oscillator.

At 32kHz, an insignificant amount of noise would couple to your power domain from the crystal oscillator.

Noise coupling from the power trace to your oscillator is more of a concern. If your circuit demands power at a 32kHz rate or some harmonic of it, it could bias the oscillator's frequency. This can be minimized by placing the power trace on a layer furthest from the crystal and its pads and traces.

ken
  • 411
  • 3
  • 7
  • The clock is 32kHz, but the rising/falling edge may much faster, does this matter? – diverger Oct 29 '14 at 09:14
  • 3
    The OP linked "32.768kHz Crystal oscillator" to a crystal, not an oscillator module. The voltages and currents on its pads and traces should be close to sinusoidal. – ken Oct 29 '14 at 09:28
  • Umm, maybe you are right, it should be close to sinusoidal. – diverger Oct 29 '14 at 09:59