4
\$\begingroup\$

I am looking for insight as to whether this would even work on paper.

I am a amateur radio hobbyist and am still learning. I built a broadcast FM superhet but my colpitts oscillator is unstable and drifts within minutes changing stations; it will jump around if people enter the room. I would like to create a more stable oscillator. Quartz is stable, I would like to design a variable series-crystal oscillator. Schematic

The idea is that a 10MHz crystal would generate harmonics in the 1xx MHz range that I can select with an LC circuit as the collector load via a variable capacitor (C32).

Before I break out the soldering iron to test it, I am having some suspicion about a series LC and parallel LC in the same circuit.

On one hand, the parallel collector LC will have very high impedance at its resonant frequency (1xx MHz). The crystal will have a resonant frequency of 10MHz and present a very high impedance to a far away 1xx MHz frequency. The output capacitor will present a low impedance to this 1xx MHz signal. Lets assume that oscillation will occur; it makes sense that I can get an output signal from C7 at a harmonic of the crystal. However, would this harmonic be amplified/is there a path for this high frequency to be fed back to input capacitor C8? Would a second 1xx MHz amplifier stage and buffer be a reasonable idea?

On the other hand, would the transistor output capacitance || crystal capacitance || C32 as well as L1 || crystal inductance form a giant LC circuit and throw everything way off?

Would the circuit even oscillate as drawn with a 270° phase shift between C8 and the crystal lead?

Thanks for your insight!

\$\endgroup\$
9
  • 1
    \$\begingroup\$ If you did manage to generate harmonics of 10 MHz, these would be integer multiples of 10 MHz (look up Fourier Series). This is why the ham radio bands are harmonically related (80m; 40m; 20m etc).The crystal frequency can be ‘pulled’ a little by connecting a variable capacitor in parallel with the crystal. But the fact is that oscillator drift was a major problem before the phase locked loop was developed. \$\endgroup\$
    – Chu
    Nov 11, 2022 at 7:51
  • \$\begingroup\$ Why don't you use a conventional oscillator rather than fashion your own? Have you thought about using a simulator to determine if it will work. Simulators are free and accurate these days. It's a no-brainer to me. \$\endgroup\$
    – Andy aka
    Nov 11, 2022 at 9:56
  • \$\begingroup\$ Your goal seems to be 100MHz oscillator stabilized by a 10 MHz crystal...that's quite a stretch for one stage. The older traditional approach might attempt X5 followed by a X2 multiplier. Or perhaps X3 followed by X3. A more modern approach uses a phase-locked-loop. And a current approach might use a chip (SI5351) to multiply a 25 MHz crystal to almost any frequency (a microcontroller would assist). \$\endgroup\$
    – glen_geek
    Nov 11, 2022 at 14:10
  • \$\begingroup\$ @Andyaka I originally did use a colpitts oscillator with a hand made air core inductor but it really likes to drift; stations would change after a few minutes if left alone and waving my hand near the radio would disrupt it. I do not know much about oscillator design, perhaps the component values I chose could be optimized. I will post a schematic of it. \$\endgroup\$ Nov 11, 2022 at 14:50
  • \$\begingroup\$ @Chu You make a good point about drift being a fundamental issue before PLL. I figured a crystal would not drift as much so that would be the next step. I will try making C32 static and putting a variable capacitor in parallel with the crystal. \$\endgroup\$ Nov 11, 2022 at 14:53

1 Answer 1

1
\$\begingroup\$

However, would this harmonic be amplified/is there a path for this high frequency to be fed back to input capacitor C8?

A typical AT-cut crystal (most common type) has more than one resonant frequency. These are harmonically-related, mostly odd-harmonics. A 10 MHz crystal will be resonant at 10 MHz, 30 MHz, 50 MHz. 70 MHz...and so on. But each harmonic resonance is weaker than the fundamental (10 MHz).
So yes, the crystal provides a resonant feedback path to the input through C8 for these harmonics. But the OP's simple circuit will prefer to oscillate at 10 MHz, since the crystal resonance at this frequency is strongest. To encourage oscillation at a harmonic, L1, C32 should be resonant at that harmonic frequency.

Would a second 1xx MHz amplifier stage and buffer be a reasonable idea?

Yes, a frequency-multiplier chain was a method used in the past. For example, the oscillator might be coaxed to oscillate at the crystal's 3rd harmonic. It would feed a 3X frequency to the next stage, which might be a frequency doubler, or frequency tripler.

On the other hand, would the transistor output capacitance || crystal capacitance || C32 as well as L1 || crystal inductance form a giant LC circuit and throw everything way off?

Yes, this is a risk. Since crystal resonance at high harmonic numbers are so much weaker than fundamenal, it is somewhat difficult to arrange the external LC components to attenuate low harmonics (as well as 1X fundamental) and at the same time make the desired high harmonic dominant.
Along with this problem, the transistor gain falls off at higher frequencies where you want it to oscillate.

Would the circuit even oscillate as drawn with a 270° phase shift between C8 and the crystal lead?

It may oscillate, but at a slightly different frequency where loop phase shift is 0 degrees. The collector L & C, along with crystal internal resonance would be involved in causing phase to change rapidly where resonances occur.

Here's an example Colpitts oscillator that's been adapted for harmonic operation:

schematic

simulate this circuit – Schematic created using CircuitLab

The series L1, C3 resonator is a low impedance at the crystal's 10 MHz fundamental frequency, frustrating oscillation at 10 MHz. The next strongest crystal resonance is near 30 MHz, so oscillation likely occurs there. This example "harmonic oscillator" is taken from SA612 data sheet from NXP.


The preceding approach with chained stages of multiplication is not often used today. One of OP's desires is variable frequency. Harmonic oscillators don't do this well, if at all.
Some chips contain complete fractional-division Phase-Locked-Loops that can output frequencies above 100 MHz. They are attractive, because fractional-division allows output frequency to be a non-integer multiple of the reference crystal's frequency - so variable frequency is possible, even with a crystal reference.

\$\endgroup\$
1
  • \$\begingroup\$ Interesting! I like the idea of the "10Mhz trap", a low impedance shunt to the crystal fundamental frequency to encourage harmonic oscillation---I learned something new today. I am simultaneously working on a transmitter, I will use the frequency multiplier chains that you mentioned to generate a stable carrier frequency. Thanks for answering the rest of my questions, Iooks like experimentation and measurement is the way to go. I am also glad my concerns weren't baseless, to a non-engineer RF seems like witchcraft at times, really hard to build an intuition as to whats going on. \$\endgroup\$ Nov 12, 2022 at 0:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.