Is my oscillator analysis method in SPICE valid?

[Whales]

I've been investigating Vackar oscillators on my workbench and in my spice sims (QUCS and Micro-Cap).  Take for example this triode powered oscillator from figure 5 of Vackar's 1949 paper:



My initial buildup looks approximately like this:



I promise the diagrams are identical, I've just re-arranged things and added a (slightly off) bias voltage for my mosfet.

Previously I used transient analysis in my SPICE sim to see how oscillators behaved, but this was fiddly and time consuming.  I realised I could probably get most of the information I needed by running an AC simulation instead.

Method 1: Substituting the mosfet

• Replace the transistor with a sine voltage source (ie Vds = input to control, Vgs = output to measure)
• Look for a peak in my gate voltage versus frequency graph indicating where the circuit resonates (and therefore may oscillate)
• Sweep the values of my variable capacitor so I can see different target oscillation frequencies (peaks in my graph).

Question: is this analysis technique valid, or am I making some bad assumptions here?

Method 2: Adding an AC signal in series with the transistor

I thought keeping the transistor itself in the loop would make more sense, ie analyse the oscillator as a whole (not just as a network of RLC).   (EDIT:) It seems to resonate much more sharply, floor at 0V and has an odd resonating point low down, making me think I'm doing something silly:



Playing with the MOSFET bias doesn't seem to improve this, only make it worse.

[Whales]

Misc note: I've been playing with two different pieces of SPICE sim software, QUCS (as shown above) and Micro-Cap.  Both give identical results for AC analysis of the same RLC circuit (hooray!) BUT qucs is about 10x or more faster doing it!  Same number of points and everything, it's amazingly better.  The results do diverge however if I add the 2n7000 in, but they probably have different models of its parasitics/etc so that makes sense.

Both Micro-cap and QUCS seem to have a long list of interface annoyances unique to them both, so I hate them both equally now

[Ian.M]

See Michael Tien's method for plotting the loop gain of a SPICE simulation.
You'll find a reasonable introduction at: https://sites.google.com/site/frankwiedmann/loopgain
It helps if you can run LTspice as that comes with a Tien method loop gain example.