Does this 16MHz crystal oscillator layout advice seem like total bullshit?

[danieljfarrell]

I am reading this app note from NXP regarding the PCB layout of a 16 MHz Colpitts Oscillator, https://www.nxp.com/docs/en/application-note/AN2500.pdf

Screenshot attached.

They recommend "no ground or power plane under oscillator components to minimise parasitics*" with the huge asterisk "ECM considerations should also be taken into account".

I hate stuff like this. All design is a trade off, this seems to be attempting to minimising parasitics at the expense of ECM. Using this calculator (https://www.daycounter.com/Calculators/Plate-Capacitor-Calculator.phtml) and assuming the area under the components is 50 sq. mm, spacing between the PCB layers is 1mm and dielectric constant is 4.4, the capacitance formed between two planes is about 2pF. OK so that is like 10-20% the value of the recommended capacitance, which is a reasonable proportion to warrant attention.

How bad is the EMC situation, do we now have 50 sq. mm of board radiating because the high speed signal fields are not contained?

Dan

[Geoff-AU]

50mm^2 is pretty small for a 16MHz antenna, so I wouldn't expect much from the oscillator itself.

But if that ground cutout forces return currents to flow around it, what's happening on the rest of your board might start to matter.

[bpiphany]

It's fun that there is no googleable information to be found on the two suggested crystals from the app note. Isn't that a very peculiar pinout for a crystal? Was that layout just cobbled up in paint and not actually tested in any way in real life?

[iMo]

The problem is the high frequency currents flow in a solid ground-plane following their shortest path, thus they create "current paths" similar to individual wires (and radiating)..

[coppercone2]

I think their trying to tell you how to make it work for sure and that you should try to ground it


before you pass EMC you need a working design.

If you can't even use the prototype then you are in more trouble then if it makes some radiation.

[nctnico]

Looking at the appnote, I can't help but thinking somebody at Motorola messed up the oscillator design big time. The phase margin to make the oscillator work is probably incredibly thin. Also, the i.MX1 series is rather old. Why use such an old chip which is likely to become obsolete any day? I'd use a more modern SOC which likely has a better oscillator. Another option is to use an external oscillator instead of a crystal just to be sure.

Either way, in this case I'd try to adhere to the suggested layout to avoid having problems with the oscillator. Maybe put a ground plane on the bottom layer and stitch it together. I don't think EMC is a big problem as this processor is rather old and not very fast by modern day standards. A single 4 or 6 layer board is hard to get wrong bad enough to cause EMC problems from digital circuitry.

[thm_w]

You can see the ST app note recommends local ground: https://www.st.com/resource/en/application_note/an2867-guidelines-for-oscillator-design-on-stm8afals-and-stm32-mcusmpus-stmicroelectronics.pdf
Mentioned here: https://www.eevblog.com/forum/projects/routing-crystal/msg4889336/#msg4889336

Remember to add it as part of stray capacitance: https://ecsxtal.com/crystal-load-capacitance-calculator/