Making a Clean Crystal Oscillator

[jim_griff]

Hi all.

So, I have consulted the Google and read books for the past few months, trying to learn about oscillators of all different kinds. I have been experimenting on my breadboard and some copper clad board, prototyping all different kinds of oscillators, but I have yet to produce a sine wave that does not contain tonnes and tonnes of harmonics. The best I have come to producing a nice looking sine wave is taking a feed from the crystal itself instead of the transistor. Even then, it has awful harmonic content. Nothing like the images I have seen on Google where people have -90dBc and better!

Most of the circuits on Google Images do not function or produce horrible waveforms.

Question 1: How do I reduce harmonics to an acceptable level? (before the output filtering stage)
Question 2: Is -40dBc an acceptable level for the first and second harmonic?

Thanks, Jim.

[kg4arn]

Your results are consistent with what I see in my humble radio shack.

Before you added the 100pF cap from base to emitter, the BJT's own BE capacitance was acting as a voltage divider with the 100pF emitter capacitance.  The external cap probably swamped the less linear BE capacitance of the BJT and provided a more linear Colpitts feedback divider.  I would have thought that the output frequency would change slightly when you did this.  Did you have to trim it back with VC1 after adding the cap?

BJT emitters and collectors are full of harmonic energy.  If you want the cleanest output, get it from the high Q crystal side (like you did).  Sometimes the harmonics don't matter much (like when you are driving a diode ring mixer, since they are going to be generated inside the ring anyway). 

You could clean it up more by buffering the output and adding a low pass filter.

I am curious about your collector bypass circuit.  Try adding a 0.1uF or 0.01uF from collector to ground and change the 1K to say 47ohms at the collector.  See if this reduces the harmonic content at the emitter any (by reducing modulation from changes in Vce)?  Also, in my experience the RFC, 470pF and 1uF  components are unnecessary and can be eliminated.

[kg4arn]

"but the transistor is in cutoff and saturation simultaneously"

I don't understand what you mean by that??

It is an emitter follower, so the collector can sit at the positive supply potential, no problem.
It won't be saturated because Vce will be about 10V.
It won't be cut off because Ic is about 13 mA.

The dissipation may be an issue at about 130mW but if your transistor can handle that, ok.

Anyway, thanks for sharing and I am off to work now.  Good soldering my friend.

[kg4arn]

I was thinking about what you said.  The transistor during operation will cycle between cut off and saturation in the Colpitts.
I was focused more on the DC bias conditions.

Sorry to hear you are giving up on it.  Crystal oscillators are very interesting.  I have built a few, but I do not understand their intricacies very well at all.  I have found a good source of information in Crystal Oscillator Circuits by Robert J. Matthys.  In it he states the Colpitts is "Physically,... very simple, but analytically, it is very complex."

You have stimulated my interest in learning more about them, but it will be the weekend before I can solder one up.

Are you enjoying the RIGOL Spectrum Analyzer?


Ed