Negative resistance oscillator works with and without resonator, is it a problem?

[Terminator3]

While designing negative resistance oscillator I found a solution, when active device is oscillating when matched to 50 Ohm load (on input and output) in range around 9-11GHz. When microstrip ring resonator is coupled to input microstrip line at certain phase angle (to match 1/S11 angle), then oscillator "locks" to resonator center frequency of 10GHz. My question is it normal that negative resistance oscillator works without resonator, or can it be a problem somehow? I tried many different solutions, where main problem was wrong frequency of oscillation (4GHz ,6,12,16 GHz, etc.). Compared to those new solution looks very promising, but as it oscillates even without resonator, i have some doubts.

 

[BigBoss]

So there is another resonator that is consisted with some other parasitic elements/components.

 

[biff44]

yeah that is fairly normal. But with the resonator attached, it should NEVER oscillate at those other frequencies.

 

[BigBoss]

yeah that is fairly normal. But with the resonator attached, it should NEVER oscillate at those other frequencies.

Not every time..
If an oscillation conditions appear with some parasitic elements, the resonator may not maintain this condition ( Closed Loop Gain and all over Phase Shift) anymore at certain frequency.The oscillator either won't start to oscillate or it will be between these two ( or more ) resonance points and it may swallow.Therefore oscillators are chaotic devices if they have not been carefully designed.

 

[Terminator3]

Is it true, that with lumped components (R, L, C) it is easier to avoid unwanted frequency of oscillation?
I am observing that distributed elements (microstrip lines, stubs) tend to repeat reflection coefficient angle over a frequency range. This results in a phase matching oscillation criteria at multiple frequencies.
In other words, when we use microstrip lines, there are many cases where load or source reflection coefficient goes from 0 to 360 deg multiple times at range 0...20GHz, but with lumped elements i do not observe such thing.

 

[BMR]

Is it true, that with lumped components (R, L, C) it is easier to avoid unwanted frequency of oscillation?
I am observing that distributed elements (microstrip lines, stubs) tend to repeat reflection coefficient angle over a frequency range. This results in a phase matching oscillation criteria at multiple frequencies.
In other words, when we use microstrip lines, there are many cases where load or source reflection coefficient goes from 0 to 360 deg multiple times at range 0...20GHz, but with lumped elements i do not observe such thing.

Well, that's quite opposite in my case. I have always had issues with lumped components. Especially in the PSU units.

 

[Terminator3]

I can imagine there are problems with lumped components if they have unknown parasitic inductance, capacitance and resistance at high frequencies.
But with distributed components there is another problem:
Sweeping frequencies in wide range shows, that reflected phase shift through microstrip line/stubs goes from 0 to 360 deg multiple times, and it looks like many circles on a smith chart. If we also draw active device (1/S11) curve on a smith chart, it will be obvious, that those circles intersect (or located pretty near) oscillation criteria curve (1/S11) multiple times. I also calculated distance between matching stubs distributed element reflection coefficient and 1/S11 of active device depending on frequency. If matching stub length is large, then there are many dips in distance curve. Also i build curves for negative resistances of active device. In some cases i was able to predict wrong frequency of oscillation just by looking at dips in distance curve and reflection coefficients of active device which >1 at certain frequencies.
Some of problems with distributed elements can be avoided if all ports are terminated with 50 ohm resistances, and only microstrip resonators are coupled to those ports. As 50 ohm termination does not provide reflected "phase sweep" through a wide frequency range, there are less wrong frequency points match oscillating criteria. Surprisingly, with lumped components reflection coefficient is more "stable" and does not swirl through smith chart over wide frequency range. Usually it gives small curve which intersects required point only once, and almost no other phase matching of oscillation criteria.