Quality factor for VCO=fc/BW=4?

Hi, all:

I am taking the course of RF CMOS IC Design and realized that normal on-die inductor can give a Q around 4 due large ESR. If I am going to have a 2.4GHz carrier signal, then based on the definition of Q.

Q=fc/BW=>BW=fc/Q=2.4G/4=600MHz.

This 3dB BW seems unreasonable large for me. I would expect a 3dB BW around 100KHz or 10KHz.

I have this assumption because normal PLL/CDR BW is around 1MHz or 10MHz. To effectively suppress the oscillator phase noise, the oscillator 3dB BW should much smaller than PLL/CDR BW. I get this conclusion from a Lorenzian shape phase noise model in optical communication. Please correct me where do I get wrong.

What's the application for a VCO with Q=4? How does the RF IC design made nowadays? I remember that I see a GPS chip has a Q over 10,000. How is that large Q achieved?

In additional to that, the Coil with Q=4 is quite lossy and your VCO will-probably-never oscillate or if it-by chance-oscillates, the PN will be horrible.Q should be equal or higher than 15 @ 2.4GHz in according to my past experiences.This is a practical value that I have found but I don't know your case.

monglebest said:

realized that normal on-die inductor can give a Q around 4 due large ESR.

Click to expand...

It depends what metal thickness your technology offers. I've worked quite a bit on on-chip inductor optimization, and 10nH with Q=15 at 2.4GHz is a typical value, with N=4 turns for reasonable size.

monglebest said:

To effectively suppress the oscillator phase noise, the oscillator 3dB BW should much smaller than PLL/CDR BW. I get this conclusion from a Lorenzian shape phase noise model in optical communication. Please correct me where do I get wrong.

Click to expand...

Within the loop bandwidth, the phase noise is mostly determined by the upconverted reference oscillator noise, so the loop effectively cleans the VCO signal.

Thanks, Volker and Bigboss. It helps me understand, especially the phase noise is determined by the upconverted reference oscillator noise. I think it's mainly flicker noise in (1/f)^3 region.

my understanding is that the BW=fc/Q is not refer to the 3dB BW as the lesson's phase noise model stated the oscillator noise flattens at half the resonator bandwidth. Due to the -30dB/dec slope, the actual 3dB BW is less than 100KHz, but the frequency where oscillator noise flattens is close to 80MHz, which leads me to believe the resonator BW = 80*2=160MHz

It aligned with the Q definition,
BW=2.4G/15=160MHz

Please correct me if you feel it is wrong.

You really can tell much about the oscillator Q if it is measured with the PLL in action.