Designer’s Guide Community

chen

New Member

Offline

Posts: 7

VCO pnoise problem
Aug 06^th, 2005, 6:03pm

Hi. I tried to simulate a differential LC VCO in spectreRF. However, the one with ideal L has the same (actually a bit higher!) phase noise than the one with real L (with finite Q) I use real transistors as varactor so the LC tank has finite Q in both cases. Here is my simulation setting:

pss:
# of harmonics=10
accuracy: conservative
tstab=50n
oscilator checked, (I have tried choosing nodes across the tank and 1 side of the tank with ref. to gnd. Both gives similar results)

pnoise:
relative harmonic=1
10k-10M
15 pts/decade
max sideband=10
voltage output, nodes same as pss
input:none
noise type: sources

Anyone has idea what dumb mistakes I have made?

Thx!

Back to top

IP Logged

Andrew Beckett

Senior Fellow

Offline

Life, don't talk to
me about Life...

Posts: 1742
Bracknell, UK

Re: VCO pnoise problem
Reply #1 - Aug 7^th, 2005, 1:21am

Well, your question is a bit light on details. Try (assuming you're using IC5141 or later) setting the highorder=yes option on the PSS options form to enable the Multi Interval Chebychev method - which should reduce the numerical noise if the oscillator is quite linear. See if that makes a difference (I'm not sure it will, but I wanted to ensure that what you're seeing is not related to numerical accuracy in the PSS).

Regards,

Andrew.

Back to top

IP Logged

chen

New Member

Offline

Posts: 7

Re: VCO pnoise problem
Reply #2 - Aug 8^th, 2005, 10:30pm

Thx so much for your reply!

Unfortunately, I only have access to ic5033 right now and it seems that this version does not allow highorder for autonomous PSS.

I am not sure what kind of info will be useful to you. I am getting around 110dBc/Hz for both cases. When I looked at the spot noise summary, it does include the noise due to the inductor. (though it is not the dominant one) So I guess either it is because the MOS are too noisy or I am doing something wrong in my simulation.

Back to top

IP Logged

sheldon

Community Fellow

Offline

Posts: 751

Re: VCO pnoise problem
Reply #3 - Aug 15^th, 2005, 4:10am

Chen,

You can also try using relref=pointlocal to reduce the numerical noise
and that is available in IC5033. BTW, be careful pointlocal can slow
down the simulation time. Using the Noise Summary should tell
you where the noise is coming from. If you modify the circuit to reduce
the noise contribution from the noisy devices, what happens?

Sheldon

Back to top

IP Logged

Andrew Beckett Senior Fellow Offline Life, don't talk to me about Life... Posts: 1742 Bracknell, UK	Re: VCO pnoise problem Reply #4 - Aug 15^th, 2005, 12:37pm The finitediff option can also help in reducing the numerical noise. relref=pointlocal can help, but usually I find it just slows the simulation down and doesn't improve the accuracy... but your mileage may vary. Andrew.
Back to top	IP Logged

byang Community Member Offline Posts: 46	Re: VCO pnoise problem Reply #5 - Aug 15^th, 2005, 1:32pm Another trick that may or may not help is to set maxstep. If you are lucky, maxstep can force uniform (euqally-spaced) timesteps and the numerical noise will be low. Baolin
Back to top	IP Logged

chen New Member Offline Posts: 7	Re: VCO pnoise problem Reply #6 - Aug 15^th, 2005, 10:07pm Thx so much for the suggestions! I will give them a try.
Back to top	IP Logged