Dear All:

In the phase noise simulatin, I find that the phase noise of ring oscillator is
generally much bigger than driven circuits (for example: cml clock buffer).

Why? The only difference I can think is that osc will "integrate the phase", so
the phase noise of osc will have 1/f^3 and 1/f^2 region. The driven circuit does
not "integrate the phase, so the phase noise only have 1/f and a noise
floor.
But this can not explain that osclillator's phase noise will be much bigger than
driven circuit, I think. The "integration property" will only make phase noise
decrease more quickly at higher frequency.

So, what is the real reason behind this phenomenon?

Any reply is highly appreciated.  :)

---johnson

Dear Ken:

       Yes, the noise is integrated in ring osc.  This means that
    for frequecny < 1Hz, the noise is amplified, for frequency > 1Hz,
    the noise is decreased.
       However I see that ring osc's phase noise is much bigger than
    driven circuit even when frequecny >> 1 Hz.

       
    ---johnson

Dear Ken:
        Thanks for your immediate response, and that's was indeed helpful for me.
        So, the real reason is that ring osc not only "integrates noise, but also
        amplifies it.

        However as far as Hajimiri's theory is concerned, I still have one question.
        This well-known figure is quoted from his paper

        In this figure, I can see integration, I can see periodical modulation.
        But I can not see "amplification". May be 1/qmax is the "amplification"
        factor. If it is, what about driven circuit? I think that driven circuit
        also have this  1/qmax. So, this confuses me.


       
        (By the way, here qmax is the maximum charge displacement across the capacitor
        on osc's node. Hajimiri thinks that if deltaQ is injected into the node capacitor
        The instantaneous voltage change deltaV will be
                   deltaV=deltaQ/C
        So, hajimiri uses qmax to normalize his equations.)