[SOLVED] Crystal Oscillator Waveform

[HardwareChap]

I've got a 32.768 kHz crystal running on a microprocessor and I'm concerned about the glitches I'm seeing on the waveform, should I be? The loading cap spec for the crystal is 12.5 pF (it's a CM8V from Golledge). I've loaded the crystal with 20 pF caps (tried to account for stray capacitances). I've attached a photo from my oscilloscope. My concern is, that if the crystal is fed into a schmitt trigger in the micro (silicon labs gecko), the glitches might cause frequency issues. Does this look normal, any tips?

 

[andre_luis]

Oscillators circuits due to its high impedance sometimes works as antennas.
Are you grounding the body of the crystal to GND ?

 

[betwixt]

Also check where you are measuring the voltage against. It is quite possible you are not measuring only the oscillator output, remember the oscilloscope measures the voltage difference between the tip and ground clip, if the ground point has other signals on it you will see a combination of them on the trace.


Brian.

 

[HardwareChap]

Thanks for your responses guys.

andre_teprom: I haven't grounded the body of the crystal. There wasn't any guidance to do this on the datasheet; is this best practice? How is this usually done? Using some kind of tab?

Brian: Very good point! *sheepish grin*. I am running off USB (so not the cleanest of supplies). I moved the clip to take a ground point closer to the crystal and it does look a lot 'cleaner'. Like I was expecting to see - a smooth sine wave.

 

[D.A.(Tony)Stewart]

Looks like excessive long probe ground clip and noisy ground and also you are looking at input, not output.

Best practise for scope is remove ground clip and probe sleeve and use only barrel tip and ring between 2 test points for sig, gnd for textbook waveforms on 10:1 mode w

Also check if calibrated probe for zero overshoot on cal. Front panel of scope. t.p. 1KHz square wave.

no need to gnd can of xtal,

 

[betwixt]

Grounding the can isn't essential but under some circumstances it might help. As long as you don't overheat it, you can solder the skirt directly to a ground underneath it or you can solder a wire to the can and use it like a third leg. I have one product where the oscillator circuit is surrounded by a coil pulsed at several hundred mA, it is unavoidable given the mechanical constraints from the customer. Without grounding the can, the oscillator waveform jitters at every pulse.

Brian.

 

[BigBoss]

They can also come from supply line that is disturbed by other digital circuits if they are connected to the same supply.Oscillator circuits want to have very clean and silent supply sources therefore this supply should be cleaned up from these unwanted interferences. Using ferrite beads, tantalum decoupling capacitors will help.

 

[FvM]

32 kHz crystal oscillators have rather high driver impedance (otherwise it won't achieve the almost pure sine waveform). It's effectively impossible that the spikes are present in the oscillator waveform itself. Similarly they won't be detected by the circuit driven from the oscillator as long it has a proper common ground reference with the oscillator capacitors.

Besides correct supply bypassing, it's suggested to keep interfering pulse signals out of the crystal and oscillator pin region. E.g. don't route them underneath the crystal.

Probing a 32 kHz crystal oscillator with a regular passive probe may even stop the oscillator in some cases.