Transient noise

Dear Nalaeg,

Nalaeg said:

When I run the transient, the oscillator does not oscillate. Then when I define the transient noise as Noise Fmax = 30G, the oscillator oscillates. Can you provide me with some insight into this Transient noise?

You have not provided any information on your oscillator topology ((i.e., ring oscillator, harmonic oscillator, etc...), your transient simulation settings, nor your transient noise settings other than you set noisefmax to 30 GHz and you observed oscillations in the transient noise simulation. However, I can easily surmise a likely hypothesis.

Any oscillator with sufficient gain (or whose sustaining amplifier negative resistance) to sustain oscillations requires noise to initiate the oscillation. In the case of a harmonic oscillator, the amplitude of the internal oscillator waveforms will build until ithey are limited by some circuit dependent parameter that reduces the magnitude of the sustaining amplifier's negative resistance to exactly equal the real impedance of the resonator.

When you perform a transient simulation of an oscillator, the only noise present is due to numerical noise. Hence, if there is not sufficient  numerical noise in the course of the simulation, the gain of your sustaining amplifier may not be sufficient to build up oscillations. Further, in a simulator with tilmestep control, if an oscillation is not detected with a given time step, the time step will increase and oscillation will not occur as the tilmestep may then exceed the desired period of your oscillator. Hence, in a conventional transient simulation, it is wise to set the maximum integration time step (maxtstep) to, say something between 1% and 10% of your oscillator period. Did you set maxstep in your transient simulation to something much less than your desired period? Alternately, did you choose a very conservative simulator setting? I do not know what version of spectre you are using. However, if it is not spectre X, then choosing an errpreset of "conservative" is recommended.

When you perform a transient noise simulation, your simulation now includes the noise sources of your circuit. This will result in noise that likely overcomes any numerical noise and will make it more likely that an oscillator will show sustained oscillations. When you set  noisefmax to 30 GHz, that forces a maximum time step corresponding to 1/(2*noisefmax). In your case with noisefmax of 30 GHz, the minimum integration tilmestep is 16.667 ps. I don't know the design frequency of your oscillator, but it may be that a single period of your desired frequency is much greater than 16.67 ps.

Nalaeg said:

Can you provide me with some insight into this Transient noise? What are Noise Fmax, Noise Fmin, Noise Seed and Noise Factor? 

There is a wealth of information on transient noise simulations and the meaning of noisefmax, noisefmin, etc.. in the documentation contained in your spectre installation, from the Help menu, as well as on the Cadence On-line support site. If you do not have an account to access the Cadence On-line Support site, I would highly suggest you consider obtaining an account. A quick search will provide many documents detailing these parameters and more. 

Nalaeg said:

In addition, can I check the spectrum of this transient noise I add at the oscillator? If yes, how can I do that?

As an example, you might review Section 4 of the note referred to in the Forum post at URL:

https://community.cadence.com/cadence_technology_forums/f/custom-ic-design/56811/modelling-and-design-of-an-optical-receiver/1388646#1388646

I cannot include the entire link as my response will be classified as spam. However, if you type the prefix through "....com/" into your web browser, you may copy and paste the remainder of the URL shown below to access the note.

s/5za1ilay6sqa06b/phase_noise_response_081422v1p0.pdf?dl=0

Section 4 computes the phase noise of a ring oscillator using both a Transient Noise simulation and compares it to that obtained using a pss/pnoise simulation.

Shawn