simulation of an oscillator in frequency domain

hi
i want to simulate an oscillator and see the frequency response, which analysis i should use? i used .AC analysis but i get wrong answer!

It's definitely possible to check the oscillation condition in AC analysis. But how can we know that you did it right?

I can see the time domain response but i want to see the output in the frequency domain
the circiut image and hspice program are shown bellow.
did i use the red line correctly to see the frequency domain output of oscillator?

**ring oscillator using differential stage**

.lib 'C:\synopsys\Hspice_G-2012.06-SP1\ADDED LIB by me\mos.35u\logs355v.l' TT

.param lmin=0.35u

vdd vdd 0 3.3

m1 x1 x3 p12 0 nmos w='100*lmin' l='lmin'
m2 y1 y3 p12 0 nmos w='100*lmin' l='lmin'
m3 x2 y1 p34 0 nmos w='100*lmin' l='lmin'
m4 y2 x1 p34 0 nmos w='100*lmin' l='lmin'
m5 x3 y2 p56 0 nmos w='100*lmin' l='lmin'
m6 y3 x2 p56 0 nmos w='100*lmin' l='lmin'

r11 vdd x1 12k
r12 vdd y1 12k
r21 vdd x2 12k
r22 vdd y2 12k
r31 vdd x3 12k
r32 vdd y3 12k

Iss1 p12 0 150u
Iss2 p34 0 150u
Iss3 p56 0 150u


.ic v(x1)=1
.TRAN .001n 10n
.ac lin 10 100 100meg

.end

That's not a reasonable way to use AC analysis for oscillator circuit analysis. You don't see an output signal in AC analysis because there's no input signal.

You can e.g. measure the oscillator loop gain.

O.k,thanks
So how i can simulate this oscillator? If i measure the loop gain , then what i can do? I don't get it really !

As FvM mentioned you can see the oscillations in AC analysis. Try applying Barkhausen stability criterion to the circuit.
So, perform a setup by which you get loop gain and the phase information and you can tell whether it's oscillating or not.

P.S,

For the ckt, I see the differential ring oscillator has 3 stages. Double check the theory of diff. ring oscillators because odd no. of stages won't oscillate I guess (not sure though).

I think, odd number of inverters is just O.K. Negative DC gain is necessary to prevent latching into a non-oscillating state. Circuit phase shift will allow to fulfill the oscillation condiction if the gain is suffcicient at the same time.

thanks, FvM and BMR.
odd number of differential stage will oscillate and as i have said before the circuit is oscillating and i can see the time domain output. my problem is that i want to see the variation in oscillating frequency with temperature, so i should plot a graph in which the vertical axis is frequency and the horizontal axis is temperature, that is my main goal.

You can't derive the exact oscillation frequency from AC analysis for > 1 gain margin because it depends on large signal behaviour. The best way would be an autmated frequency measurement in a set of parameterized transient analyses.

Again as mentioned by FvM what you can do is invoke your propagation delay as a variable and sweep it against temperature and see the results .

i read your answer and thanks for your answers but unfortunately i don't understand them ! can you tell me what should i exactly to do ? maybe you means i must take the simulation result into matlab? i thought maybe i can use an transient analysis for each temperature and then i take the transient result into the matlab and then i use FFT to see the frequency of oscillator in each temperature.

Yes, you need to use transient analysis, as suggested. I presume you determine the oscillation frequency by a .MEASURE statement without using fft.