Is ADS accurate for VCO simulation???

[gggould]

Recently I need to use ADS2003C to design a 5GHz VCO.
But I don’t know if the simulated phase noise/ oscillating frequency accurate.
Does anyone have experiences about that?


Thanks

 

[toonafishy]

Frequency is fairly accurate if you have the right parasitics modelled. But phase noise is underestimated 6-8 dB in my experience. This is true for IC. I haven't tried it for discrete design.

 

[BigBoss]

Generally I use SpectreRF to simulate IC oscillators both to predict oscillator frequency and phase noise. But due to some convergence problems, I switch to ADS even tough I don't prefer.

But last version of ADS which is 2003C becomes closer to the reality than Cadence. I don't know why but it is so...

In additional to , the next version (2004A) of ADS will have simpler oscillator probe that could be connected "parallel" to predict oscillator frequency.

For phase noise, neither simulator can not give the very close result but pre-results...

I think, within discrete applications, the models are more accurate, so results might be more realistic.

 

[goodboy_pl]

though it seems to be too late for to answer this problem, I think it will be good for those use the "Search" !

I think current theories for phase noise are OK and the problem for bad phase noise prediction is in modelling.
Current models are not completely physics based, therefore some phenomenas are not included, specially in transistor models(BSIM) which are the main source of the noise. modelling is a very complicated task in today technologies and designers must not rely on foundry's model specially in UDSM, new process releases and accurate circuits.

substrate coupling is another source os phase noise which is not considered during design phase. it can be another source of noise specially in SOC and highly integrated chips.

BEST!

 

[vfone]

After using a few different simulators for this topic, recently I used M*W*O v6.5. I don’t know what algorithms these guys are using to cheat with the semiconductor physics, but the results in terms of phase noise are pretty accurate. At least for a few models that I used.

 

[boy]

Is @DS accurate for VCO simulation???

I have experince about phase noise simulation. your model have to support phase noise simulation, otherwise the phase noise results is useless. There is a way to extract the phase noise parameter from the dicrete devices. I have no idea what should we do when we design ICs.

 

[vonleo]

I think ADS has a converge problem too, especially for ring osc. And in ADS, you have to add a special port in the circuit to measure the PN, I never figure out how to set all the factors in that port correctly...

 

[khouly]

in @DS 2003c , the ring oscillator converge very well even without the intianl guess of transient analysis , but u must set the simulation options well

like oversampling , the frequency "chose it very close to oscillation frequency"
and the solver try to use it "auto" but it will take longer time , but it will work


thanks
khouly

 

[caep_daniel]

Is @DS accurate for VCO simulation???

contrary to ansoft designer, I prefer ADS2003,because it can predict the result of VCO design. I have design a VCO at 2GHz which includes many microstrip. and I think there are 100MHz offset between sumilation and measurements. the phase noise simulation is not reliable I think. but the simulation result can only give u a reference.

 

[itaifrenkel]

caep_daniel - Ansoft has released a new circuit solver for these kind of simulations called Nexxim that integrates into Ansoft Designer. You can see snapshots of a PLL design in their whitepaper at **broken link removed**

Itai

 

[wccheng]

Is @DS accurate for VCO simulation???

Dear,

I have used SpectreRF simulator to simulate my LC A-MOS differential VCO. It is post-layout (RC extraction) simulation. The simulation result is extremely closed to my measurement result, such as output power, phase noise and tunning range. I think this is due to accuracy of the model and the RC extration post-layout simulation.
Also, the convengence problem could be solved by setting the step size of the simulation.

Best Regards,

wccheng