VCO noise - phase noise increases when s-parameter is used

[hmsheng]

VCO noise

A 3GHz VCO was designed with on-chip inductor model. Two kinds of models were used, s-parameter model and spice-model that was extracted from the s-paramtere model. When I simulate the VCO phase noise with the two kinds of models there is much differenct. The phase nosie is better if I use the spice inductor model. The phase noise will worse 3dB or more if the s-paramter model is used. In the noise contribution, NPORT0 which is the inductor s-paramter component contribute 82% noise. That's abnormal!
I doubt we can't use s-parameter file to simulate VCO phase noise. Is that?

 

[Blackuni]

Re: VCO noise

Hi,

How accurate your Sparameter modelling of the inductor?

As you know when Sparameter models are used, they were treated as black box models and there will back and forth conversion between freq. domain and time domain..and i suspect it might add numerical noise to the system.

recommendations: 1. make sure that you have enough freq. points in Sparameters, this will help to avoid interpolation noise.
2.tighten the accuracy settings/error tolerance for the S parameter model

By the way we can possibly reduce /make sure that numerical noise less.

If you want to effectively model the inductor, then i would recommend modeling it directly as 2D r 3D field solver model, by the way u can include freq. dependent parameter variation.

By the way which tool u r using? How r u converting S parameter to Spice model?

Hope this helps.

thanks,

 

[hmsheng]

Re: VCO noise

Hi,

How accurate your Sparameter modelling of the inductor?

As you know when Sparameter models are used, they were treated as black box models and there will back and forth conversion between freq. domain and time domain..and i suspect it might add numerical noise to the system.

recommendations: 1. make sure that you have enough freq. points in Sparameters, this will help to avoid interpolation noise.
2.tighten the accuracy settings/error tolerance for the S parameter model

By the way we can possibly reduce /make sure that numerical noise less.

If you want to effectively model the inductor, then i would recommend modeling it directly as 2D r 3D field solver model, by the way u can include freq. dependent parameter variation.

By the way which tool u r using? How r u converting S parameter to Spice model?

Hope this helps.

thanks,

1. "Dense" mode which is the most accurate mode is used in VPCM simulation of the inductor model.
2. Frequency range of the s-paramter file is 0.1GHz-10GHz with step of 0.1GHz.
3. I use VPCM to simulate the on-chip inductor to get the s-paramter file. I use Spectre to simulated the VCO phase noise with the s-parameter file model or spice model. There is an equivilent RLC circuit of the inductor. Each RLC value is fitted with ADS optimization. Then I get the inductor spice model.

Thanks for you advice. I will try it.