The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> RF Simulators >> Some puzzles on Crystal oscillator design
https://designers-guide.org/forum/YaBB.pl?num=1290495700

Message started by jlee on Nov 22^nd, 2010, 11:01pm

Title: Some puzzles on Crystal oscillator design
Post by jlee on Nov 22^nd, 2010, 11:01pm

Hi guys,

As the title, there are following puzzles on crystal oscillator design.
1. I have got the following element parameters in the crystal equivalent model: Cs(motional cap), Cp(parellel cap) and Rs(ESR). Are they enough to calaculate the remaining parameter, say Ls with a given oscillation frequency?
2. I have tried to run simulation by PSS and tran, but the oscillator can normally work without any initial condition under PSS and it doesn't work under tran, or the waveform becomes wired. What's the difference between these two analysis?
3. Why the simulated oscillation frequencies are different by changing tstab in PSS analysis, although oscillation has become a steady state?
4. For example, a frequency of 11.1111MHz is obtained from PSS. If an accuracy of 11.111111MHz is preferred, how to set up in simulation?

Thanks for advance!

Regards

Jlee

Title: Re: Some puzzles on Crystal oscillator design
Post by sheldon on Nov 23^rd, 2010, 3:47am

jlee,

Given the available information, it would be difficult to respond to your
questions. Assuming that you are using Spectre and Spectre RF, I would
like to share one observation. You might want to try using harmonic
balance when analyzing a crystal oscillator: set tstab=0 and use osc
initial conditions=linear. This approach may be more suitable for your
application, crystal oscillators.

Best Regards,

Sheldon

Title: Re: Some puzzles on Crystal oscillator design
Post by wave on Nov 23^rd, 2010, 5:04pm

jlee wrote on Nov 22^nd, 2010, 11:01pm:

For Question 1, If you know the Freq and the R/C, you can calc. L.
You need to read some basic app notes on crystal design and resonant frequency equations between LRC and Freq.
Vittoz has a classic paper.

For Q2, transients are golden for startup sims.
You may need to use Trap, as explained in Ken's Spectre book, and
you may need to De-"Q" the model (change L/C); that will give easier and artificially quicker startup sims.

I haven't used PSS enough to comment on Q3,
and for Q4, I'd set your expectations lower on Frequency accuracy sims. However I doubt that is a critical sim.
Freq. can be pulled externally by the Crystal and components.

Wave

Title: Re: Some puzzles on Crystal oscillator design
Post by jlee on Nov 23^rd, 2010, 6:22pm

Thank you, Sheldon and Wave. ;D

Title: Re: Some puzzles on Crystal oscillator design
Post by pancho_hideboo on Nov 26^th, 2010, 10:53pm

jlee wrote on Nov 22^nd, 2010, 11:01pm:

1. I have got the following element parameters in the crystal equivalent model:
Cs(motional cap), Cp(parellel cap) and Rs(ESR).
Are they enough to calculate the remaining parameter,
say Ls with a given oscillation frequency?

Yes.
See http://www.yoketant.com.tw/yoketant/terms.htm

jlee wrote on Nov 22^nd, 2010, 11:01pm:

2. I have tried to run simulation by PSS and tran,

What simulator do you use as "PSS" ?

jlee wrote on Nov 22^nd, 2010, 11:01pm:

but the oscillator can normally work without any initial condition under PSS

It is possible if you use HB-PSS.
What analysis engine do you use as PSS ?
See the followings.
http://www.designers-guide.org/Forum/YaBB.pl?num=1234428781/1#1
http://www.designers-guide.org/Forum/YaBB.pl?num=1216380172/7#7
http://www.designers-guide.org/Forum/YaBB.pl?num=1216380172/13#13

jlee wrote on Nov 22^nd, 2010, 11:01pm:

3. Why the simulated oscillation frequencies are different by changing tstab in PSS analysis,
although oscillation has become a steady state?

I can't say nothing unless you show name of vendor, name of simulator and analysis engine which you use.

jlee wrote on Nov 22^nd, 2010, 11:01pm:

4. For example, a frequency of 11.1111MHz is obtained from PSS.
If an accuracy of 11.111111MHz is preferred, how to set up in simulation?

In Agilent ADSsim, there are frequency absolute and relative error setting.