RF noise from microcontroller oscillator

[maniana]

I have designed a board with a microcontroller (FX2) and 1.8GHz receive chain (LNA, I/Q downconverter, PLL, IF amplifier).
I was trying to follow all the guidelines for low EMI radiation but still I get a lot of noise at the output of my receiver, which I identified as coming from microcontroller. This noise is largely reduced when I touch with a finger the leads of a 24MHz crystal that is a clock for the microcontroller. As I said I payed a lot of attention to PCB design, so ground is well routed, power supply separate for uC and RF , there is a ground ring around that crystal oscillator, The separation between uC and Rf is about 10cm, etc...
Still, At some frequencies I get spurs in the receive spectrum spaced at 1kHz over several MHz.
I am generally an RF eng. and only occasionally deal with uC,
but how **** do I get rid of that noise from the microcontroller in my RF part ????

 

[FvM]

Only a really worse crystal oscillator design will expose spurs in the GHz range, but it's not completely impossible for a crystal oscillator with high drive level. I thought this kind of oscillators has been abandoned by chip designers at least 10 years ago when people started to worry about EMC. It should be possible to tune the oscillator for lower crystal overdrive with smooth waveforms.

The other possibility is that your RF design is quite succeptible for conducted interferences distributed by the PCB, that somehow manage to enter the IF part.

 

[maniana]

I am not sure whether noise is introduced in the GHz range. I think it is more likely that this noise interference happens in the baseband frequency range. When I connect an antenna the noise floor rises and the spurs don't. I don't know whether the interference is with the fundamental freq of the crystal or with harmonics. There are ferrites (BLM series) on every chip on the pcb.
The oscillator is a standard crystal with 2x22pf capacitors connected to the uC. I tried to put 220ohm resistor on one of the tracks between the crystal and uC to smooth the clock signal, but that made no difference.
Could that noise transfer over the air too?
And if it really is conducted on the PCB, how to fight that, how to identify in which way it gets transferred ?

 

[FvM]

You observations suggest low frequency/base band interfernces indeed. In this case, it's somewhat more complicated. You say you have a lot of decoupling stuff, but the ground plane itself may become "hot", depending on how potential interfering signal sources and return pathes are connected to it.

It's not only the crystal oscillator, also all digital logic supply and ground pins and possibly existing periodical signals, e.g. data busses and clocks that may source the spurs.

Searching the interfering nets with a inductive or capacitive probe and a spectrum analyzer and checking for entrance points in the RF design by artificially injecting interferences with a signal generator are possible steps to debug the design.

 

[Element_115]

I have designed a board with a microcontroller (FX2) and 1.8GHz receive chain (LNA, I/Q downconverter, PLL, IF amplifier).
I was trying to follow all the guidelines for low EMI radiation but still I get a lot of noise at the output of my receiver, which I identified as coming from microcontroller. This noise is largely reduced when I touch with a finger the leads of a 24MHz crystal that is a clock for the microcontroller. As I said I payed a lot of attention to PCB design, so ground is well routed, power supply separate for uC and RF , there is a ground ring around that crystal oscillator, The separation between uC and Rf is about 10cm, etc...
Still, At some frequencies I get spurs in the receive spectrum spaced at 1kHz over several MHz.
I am generally an RF eng. and only occasionally deal with uC,
but how **** do I get rid of that noise from the microcontroller in my RF part ????

It's hard to say with out seeing the layout, but I would suggest that you put a ground shield around
the BB fltr(I/Q path) , and if you route any digital lines around the RF path make sure they
are 90° to the RF path ( minimize coupling).

Is the RF Vcc the same as the Digital Vcc ? Maybe add some RF Chokes to each sensitive component.

Cheers

 

[Joel-Tang]

It's hard to say with out seeing the layout, but I would suggest that you put a ground shield around
the BB fltr(I/Q path) , and if you route any digital lines around the RF path make sure they
are 90° to the RF path ( minimize coupling).

Is the RF Vcc the same as the Digital Vcc ? Maybe add some RF Chokes to each sensitive component.

Cheers

Good points from Element_115

I would try and understand if your problem is mostly conducted or radiated. This will determine your requirement for a shield or chokes. It's easy to try RF chokes compared to using a shield on a layout that never had this originally in mind.
You could also do some measurements using E and H field probes to determine from which it is you are suffering.
I'll share a small collection of App-notes for using and making probes and how to use them ( LINK: here ) that I recently collected while troubleshooting a similar issue.

Good Luck!

 

[maniana]

Joel, very good collection on the subject !

 

[BigBoss]

Sometimes sharing GND layer with a high noisy signals can create interference between RF and Digital (because of Common Mode) .In this case, seperating GND layers ( using GND layers for each block) may contribute something into solution.