Cycle-to-cycle jitter on Colpitts oscillator

[JeanTech]

Hi there

I'm struggling to get rid of jitter on my Colpitts oscillator. It is currently oscillating at 15 kHz. The problem is that I am getting cycle-to-cycle jitter of up to 384 nS. Is it possible to get a lower cycle-to-cycle jitter with this kind of oscillator. What is a typical cycle-to-cycle jitter for a very stable Colpitts oscillator?

Thanks in advance
JeanTech

 

[MikeMl]

Are you certain that the jitter is in the oscillator and not in the zero-crossing detector you are using to measure the period?
The intrinsic noise in said detector is likely to be the reason for the jitter you see.

 

[BobW]

Also, how clean is the oscillator's power supply?

 

[JeanTech]

Thank you for the replies.

I have measured the signal before my Schmitt trigger (which I am using to convert the sine wave into a square wave) and I can also see the jitter there, so it is not coming from the detector. Also, the oscillator's power supply is very clean. At the moment I am using a battery as the power supply.

Is it possible to get jitter-free oscillations from the Colpitts oscillator, or is there going to be a certain jitter. If so, what is the typical jitter one can expect?

 

[BobW]

Sounds like you may have some high frequency parasitic oscillations going on.

 

[OBW0549]

The inductor in your oscillator may be picking up external EMFs from the mains or from nearby switching power supplies or CFL or LED lighting, and that interference may be causing the jitter.

 

[Cicero]

Posting your circuit could help, as we dont know whether you're using opamps/transistors or anything other than its a Colpitts.
Are you isolating the oscillator for these measurements or is there a downstream circuit attached, or something else on the power rail?
What kind of caps are you using?
Is this a PCB boarded design, or are you breadboarding or prototyping?

I've been involved in some Colpitts designs for inductive loop detection, they're usually very stable. As a side note, for the square wave generation we usually used high speed comparators (+-4ns) to minimise jitter.

 

[tomizett]

I have measured the signal before my Schmitt trigger (which I am using to convert the sine wave into a square wave) and I can also see the jitter there, so it is not coming from the detector.

To clarify, I think what Mike was getting at in post #2 (and correct me if I'm wrong) but whatever you are using to measure the frequency - a hardware frequency counter, a DSO or whatever - will be detecting a zero-crossing point to either time each period of the input signal or to count how many input cycles fall within its gate period. That means that amplitude noise on top of the oscillator's output could be altering the point at which the signal crosses zero, and therefore the measured period.

 

[ci139]

fuzzy science https://www.google.com/search?q="co...eedback"+"frequency+stability"+opamp+-crystal

 

[Diver300]

I think that you would get far lower jitter with a crystal oscillator. The frequency would be too high, but you could divide down to 15 kHz.

If you have a sinewave at 15 kHz, then a variation of the trigger point of just a few percent of the peak to peak voltage will give you the jitter that you are seeing. The voltages are changing very slowly with a 15 kHz signal.

 

[MikeMl]

Question still is if the oscillator has jitter, or if your measurement technique has the jitter?

I would listen to it on my VLF receiver with the BFO on (Narrow CW reception mode). The audible beat would instantly tell me if the oscillator has jitter or not. I'll almost bet it is your measurement method...

 

[JeanTech]

Posting your circuit could help, as we dont know whether you're using opamps/transistors or anything other than its a Colpitts.
Are you isolating the oscillator for these measurements or is there a downstream circuit attached, or something else on the power rail?
What kind of caps are you using?
Is this a PCB boarded design, or are you breadboarding or prototyping?

I've been involved in some Colpitts designs for inductive loop detection, they're usually very stable. As a side note, for the square wave generation we usually used high speed comparators (+-4ns) to minimise jitter.

It is a Colpitts oscillator that uses an op-amp. I am currently isolating the circuit by just measuring the oscillator - no other circuitry attached. I am using X7R caps with a PCB design.

Did the high speed comparators reduce the jitter coming from the input sine wave? The problem is that the jitter on my sine wave causes the Schmitt trigger to trigger inconsistently, since its threshold voltage is triggered incorrectly by the jitter on the sine wave.

 

[JeanTech]

Question still is if the oscillator has jitter, or if your measurement technique has the jitter?

I would listen to it on my VLF receiver with the BFO on (Narrow CW reception mode). The audible beat would instantly tell me if the oscillator has jitter or not. I'll almost bet it is your measurement method...

The final circuit feeds the Schmitt-triggered square wave into a Atmel SAM microcontroller which triggers an interrupt. I then use this interrupt to determine the period between each rising edge. Using this measurement method, I can see the period varying randomly due to jitter. To double check, I then measured the signal before and after the Schmitt trigger with an oscilloscope and I still get the same results.

 

[RichTheDude]

I have seen Schmitt triggers introduce terrible supply noise when people have left the remaining gate inputs floating. Remember, no-one likes to see a floater nearby.

 

[Cicero]

The final circuit feeds the Schmitt-triggered square wave into a Atmel SAM microcontroller which triggers an interrupt. I then use this interrupt to determine the period between each rising edge. Using this measurement method, I can see the period varying randomly due to jitter. To double check, I then measured the signal before and after the Schmitt trigger with an oscilloscope and I still get the same results.

I'm sceptical of the sine wave having such jitter, but we just have to take your word for it.

I'd also double check your measurement techniques, and also make sure you're measuring correctly using the scope. You've got a 15kHz signal, generating a 66us wave period, half wave of 33us and trying to measure ns jitter.
Completely isolate the oscillator without the schmitt trigger and check jitter. Using a micro to determine jitter down to a ns range can also be a tricky thing to get right, interrupts can easily not be deterministic.

Also what bias are you using, how stable is it?

If your schmitt trigger has any hysteresis (which a schmitt trigger has by definition virtually) then you're gonna have cycle jitter, granted this will manifest after the square wave generation where you dont believe the issue exists though.