Oscillator design

[init]

Hey guys, newbie to RF here. I've given a go at building an oscillator that was designed to operate (on paper) at 100 MHz. Since I'm new to this I decided to start with a simple colpitts oscillator design shown below. I was aware that by operating at 100MHz there would be significant parasitics from the BJT junction capacitances as well as stray inductance etc. so my circuit would oscillate in the vicinity of 100MHz which it did (~90MHz). I have heard that colpitts are generally less stable and more jittery compared to other types of oscillators which is apparent when I change the parasitics in the environment by moving the board, touching surfaces or moving things closer to the circuit. From what I've built, what sort of metrics would I use to "rate" the quality of the oscillator? Would it be better for my next attempt to build a crystal based feedback loop with PLL? Any other types of oscillators that are good to have a go at?

Thanks for the feedback 

[MrSlack]

Congratulations - that works pretty well

Improvements to consider:

1. The board you are using is an array of small capacitors. Grab some single sided FR4 and solder everything mid-air so-called "deadbug" style. This will kill a lot of stray capacitance and give the entire thing a nice hefty ground plane. That will increase stability considerably.
2. Lose the electrolytic capacitors entirely and use NP0/C0G rated capacitors for the whole thing. A lot of drift will occur due to temperature effects on cheap ceramic capacitors. To demonstrate this, poke your finger on one of the caps for a few seconds and watch the frequency drift off as it warms up.
3. You don't need big electrolytics here; a 10n ceramic across the emitter resistor is probably fine as is a 100pF or so coupling capacitor at 100MHz.
4. Add an actual trimmer in the final circuit, set it to oscillate at 100MHz exactly and play to see if it drifts and if so how quickly.
5. Add a low-pass filter on the output to suppress harmonics.
6. Instead of using an RF choke as the inductor, grab a suitable Micrometals toroid and wind one. More fun and considerably more stable.

To rate quality, the main parameters are phase noise and harmonic content. I've not worked out how to measure phase noise yet. Harmonic content is visible in your FFT.

Edit: also watch out for crappy 2n2222a's. I've seen a few of them with a very low transition frequency recently. Managed to bag 50 Motorola ones off ebay which all work nicely.

[Earendil]

Instead of a PLL you could also try using an integrated VCO, like the MAX2606. It has a decent phase noise performance.
Minicircuits also has an interesting appnote about VCO test methods: https://www.minicircuits.com/app/VCO15-15.pdf
You can get some ideas from there how to rate your oscillator.

[AF6LJ]

You are off to a good start and you have some good recommendations.

[Earendil]

This is also a very useful summary for various oscillator concepts: http://www.ittc.ku.edu/~jstiles/622/handouts/Oscillator_package.pdf

[MrSlack]

Instead of a PLL you could also try using an integrated VCO, like the MAX2606. It has a decent phase noise performance.
Minicircuits also has an interesting appnote about VCO test methods: https://www.minicircuits.com/app/VCO15-15.pdf
You can get some ideas from there how to rate your oscillator.

Very interesting - thanks for posting this.

[init]

Thank you for the replies, lots of good stuff to take on board with the next design 

[init]

Thought I would add to this thread with a follow up question. I've designed a matching network with a bandwidth of 10MHz for a 100MHz center frequency and from what I can see in LTSpice, the bandwidth always appears to be twice what I calculated when I simulate it. Am I interpreting the -3dB cutoff correctly? The obvious answer is that I've confused the 10log and 20log forms of decibel but I'm not sure.

[Fank1]

The display shows a voltage being measured.
Use 6db with volts and amps.
Use 3db with watts.
also?
The bandwidth has a great deal to do with the Q of the circuit.
What value of Q did you assign to the inductors in the matching network?

[Z80]

Nice job, can I ask though if there is any reason you want to build it at 100MHz?  If you are just experimenting then you will have a much easier time dropping the frequency right down to something like 4MHz.  There are 2 main reasons for this: 1) It will be much more stable and less affected by temperature and the environment.  2) I'm assuming your scope is a 100MHz Rigol?  Measuring your oscillator is way beyond its capability, sure it will show you a waveform but that's not the actual waveform of the circuit.  Dropping the frequency will allow you to measure the harmonic content too which is important if you want to experiment.  Keep up the good work regardless.

[rfbroadband]

the key parameters for oscillators are
- tuning range
- phase noise
- harmonic content

tuning range and harmonic content are easy to measure.

Phase noise is totally different story. Best case your spectrum analyzer can be used if your phase noise if your DUT is not too good. If your phase noise is really low you need a phase noise analyzer as even the best SAs on the market (eg R&S  FSW)  are not good enough (the internal LO will be too noisy). Many companies can't justify a phase noise analyzer since it is too specialized and really expensive.

It looks like you are off to a good start. If you are interested to learn more about phase noise, I would suggest you read a few Keysight or R&S application notes.

[KJDS]

It's fairly easy to make a phase noise measurement system.

Make two of your oscillator design and mix them together down to DC.

[nugglix]

It's fairly easy to make a phase noise measurement system.

Make two of your oscillator design and mix them together down to DC.

And then you have 2 oscillators with unknown attributes.
And I bet they won't be the same anyway.

[KJDS]

It's fairly easy to make a phase noise measurement system.

Make two of your oscillator design and mix them together down to DC.

And then you have 2 oscillators with unknown attributes.
And I bet they won't be the same anyway.

Or if you can't make two the same, then just run it through a sensitive FM discriminator.