The Designer's Guide Community Forum
Design >> RF Design >> VCO with ring oscillators

Message started by skt on Oct 14th, 2003, 8:24pm

Title: VCO with ring oscillators
Post by skt on Oct 14th, 2003, 8:24pm

Hi all,
          I am designing a 3 stage VCO using ring oscillators  in spectre.  I am trying to find the frequency of oscillation by doing a transient analysis.
        the problem is the circuit never oscillates. spectre always seems to find a valid bias point.
       Do i need to trigger the circuit using some sort of impulse ( i know  how to do it in pspice, but not in cadence) or is something wrong in my circuit.  
        Any suggestions ......

Title: Re: VCO with ring oscillators
Post by Paul on Oct 14th, 2003, 11:44pm

Hi Surya,

in reality (on a chip), an oscillator always starts up from noise. So I would say you must probably trigger your oscillator to start up with some small perturbation. If I remember correctly, a 3-stage oscillator has a worse start-up than higher order ring oscillators. So maybe you can in a first time try to increase the number of stages to compare, although this will reduce the frequency of oscillation.


Title: Re: VCO with ring oscillators
Post by Ken Kundert on Oct 15th, 2003, 1:19am

I have a whole section in my book discussing this topic. But the simple answer is yes, you need to perturb the oscillator to get it to start. With ring oscillators, the easiest way to do that is to apply an initial condition to the oscillator.


Title: Re: VCO with ring oscillators
Post by circuit_master_RT on Oct 26th, 2003, 12:12am

The above answers are good, but no one has mentioned the usage of a current source with a pulse or step waveform.  The current source as an oscillation starter is nice since it does not change the impedance of the node you attach to, and you can put a very small amplitude "kick" to get the oscillator started, and observe the oscillation growing and then limiting in amplitude.

Title: Re: VCO with ring oscillators
Post by Ken Kundert on Oct 27th, 2003, 9:46am

In my experience, a sinusoidal source that drives the circuit at the resonant frequency and that has a exponential decay works quite well, particularly for resonant oscillators. For ring oscillators, I find that using an initial condition on one stage works best. But either approach should be adequate.

One thing I have seen people try on ring oscillators that I don't recommend is to jiggle the supply voltage.  This injects a signal into all stages symmetrically, which does not do a good job of stimulating the oscillation.


Title: Re: VCO with ring oscillators
Post by storm on Feb 11th, 2004, 6:46am

I am also working on ring oscillator, with 2-stage. Hard to start oscillation. I just put an additional vsin in the circuit separately instead of a ipulse to kick-start the circuit. i mean, there is no any physical connection between such vsin and the circuit under test.

It works.

Title: Re: VCO with ring oscillators
Post by emad on Feb 21st, 2004, 9:49am

I think the best starter is the use of initial conditions. As Ken mentioned and elaborated in his book, kicking through the supply lines is not advisable whether it is ring or LC based oscillator. The supply node is typically slow and kicking the oscillator through it can cause long-lasting transients, especially for high Q ones.
Most people like the current source that decays. I don't. Some people think it is the only method that works for crytal oscillators, I don't think so. Initial conditions seem to produce the fastest convergence in my experience.


Title: Re: VCO with ring oscillators
Post by eggleg on Mar 10th, 2004, 11:45pm

I just wonder what are initial conditions here.
In my case, I have four ring oscillators.
And .ic v(voutp)=pvdd v(voutn)=0
are my initial condition.
Is it a correct way?
I have no problem to get starting oscillation.

- Jacky Z.Lin

The Designer's Guide Community Forum » Powered by YaBB 2.2.2!
YaBB © 2000-2008. All Rights Reserved.