Greetings,

Simulating crystal oscillators got a lot easier in MMSIM12.1...We have made enhancements to both Harmonic Balance and transient analyses.

In Part 1, I discussed **Improvements to the Harmonic Balance** use model. With the new streamlined Choosing Analyses form, you now can focus on getting your simulation results, rather than the setup of the analysis.

Now, in Part 2, I'll cover **Improvements to transient analysis.** We’ve added a new feature to transient analysis that allows you to reach steady state more quickly should you decide to simulate your crystal oscillator using transient analysis.

*Important*: *Be sure to use IC615 ISR14 or later to see the new MMSIM12.1 features in the GUI.*

**Part 2: Spectre Transient Analysis Enhancement **

**Here is the new method that you can use to speed up transient analyses of crystal oscillators. **

**First, we’ll simulate without using the new parameters…then we’ll add them in and show the difference.**

· Set up a transient analysis in the Choosing Analyses form:

· Set the** stop time** to 10K periods of oscillation - in this case, 1m (1 millisecond).

· For crystal oscillators, I always use *errpreset = conservative.*

· Click on the *Apply* button. Then click on the *Options* button.

Set the following** transient analysis options:**

· Set the* **Time Step *parameter *maxstep *to about 20 timepoints per period. Note that Spectre accepts expressions (see the form below), which simplifies things for you.

Set the following *Algorithm Parameters:*

· For* Integration Method*, select *traponly. *By default, when errpreset=conservative, the integration method is gear2only. However, gear2only can dampen oscillations and shouldn’t be used – you want your oscillator to oscillate! So, I recommend using either trap or traponly when simulating crystal oscillators.

· Set *Accuracy Parameter relref *to *alllocal*. Note that this is one of the options that are set by default when you select *errpreset=conservative*.

· Leave all of the other parameters at their default values.

For more information on the transient parameters, type* spectre -h transient* in an xterm window.

· *OK* the Transient Options form. *OK* the Transient analysis form.

· Run the analysis (*Simulation->Run* or green arrow ).

The transient analysis output is shown below. Access the results by selecting *Results – Direct Plot – Main Form, *select* tran *in the *Analysis *section*, voltage *in the* Function *section, and* *select the resonator net in the Schematic).

**Note that in 10K cycles, this oscillator is still ramping up – it will take quite a while to start. The peak voltage is approximately 1.47V (yes, that is volts, not KV).**

**As a comparison:** Since we know that harmonic balance solves for the periodic steady state solution, we can use this as a check. How far away from steady state are the above transient analysis results? We will look at the transient assisted (tstab) and steady state waveforms from a harmonic balance simulation. Below is the tstab waveform (Transient assisted waveform) from the harmonic balance analysis:

Note that the peak voltage is about 900V. Compare this to the transient output voltage from previous simulation where the voltage was on the order of 1.47V. This is quite a significant difference.

Harmonic balance simulates to convergence and the ifft (inverse fast Fourier transform) is plotted below.

Notice that the actual peak voltage at steady state is about 1.3KV. This is significantly larger in amplitude compared to the transient analysis. You can see that we have a ways to go before reaching steady state in the transient analysis.

Now, we’ll use the two new transient parameters ** linearic** and

**.**

*oscfreq*· In the transient options form, select the misc tab. At the bottom of the form in the Additional Parameters section, you see a type-in field called *additionalParams*.

· In that field, type* linearic=yes* *oscfreq=<actual_oscillator_freq>. *Here, the actual_osc_freq is 10M.

· Note the space between the two keyword=value pairs. This is needed to separate the statements.

· Now re-run the simulation for 10ms, and plot the transient waveform.

· The output of the transient analysis **without specifying linearic and oscfreq **is above in green. The voltage is at roughly 1.5 volts and is nowhere near steady state at 10ms.

· The output of the transient analysis **using linearic=yes**

*oscfreq=10M**is in orange. Steady state was reached in approximately 3ms. Note that the voltage starts at about 900 Volts peak and at steady state is approximately 1.3kV.*

**Summarizing:**

We have enhanced transient analysis to improve convergence. When using the new transient options *linearic* and *oscfreq*, transient analysis starts with a solution closer to the steady state value, and the transient simulation converges much faster. Although you do need to do some typing (filling in of fields), in the end the time saving is well worth it.

Have fun simulating!

Best regards,

Tawna

e have enhanced transient analysis to improve convergence. When using the new transient options linearic and oscfreq, transient analysis starts with a solution closer to the steady

Hi Tawna, This feature is exciting . One question I have is if we use linearic = yes and give approximate oscfreq is there any chance that even the oscillator closed loop gain is less than 0dB(but close to 0dB) and circuit shows that it will oscillate.Bharath

It doesn't work for me :-(

This is what I see:Warning from spectre during transient analysis `tran'.

WARNING: Linear IC: Fail to find out initial frequency. [ Early Reject ]

Can't find linear initial condition. Running transient from DC with the specified initial conditions. Can you contact me privately about it?