How to simulate Crystal Oscillator in HSPICE?

[ysz]

initial conditions crystal oscillator spice

changing 1u to 100n in .tran 1u 200000u
will not help - it is "initial" sep size. Simulator will then change step size to anything.

Maybe you are right. But I have ever heard someone encountered the same problem -- the output swing is too small. He finally solved the problem by changing the transient step size to 0.1ns.
I don't know why.

Factors that determine the computed DELMAX value are:
.OPTIONS RMAX and FS
Breakpoint locations for a PWL source
Breakpoint locations for a PULSE source
Smallest period for a SIN source
Smallest delay for a transmission line component
Smallest ideal delay for a transmission line component
TSTEP value in a .TRAN analysis***********
Number of points in an FFT analysis

the tsetp in .tran sentance will change the delmax, the delmax will change the accurcy of the simulation.
so I through it is useful.

 

[fashion]

simulating a crystal oscillator

Maybe you can increase the resolution of your .tran statement

 

[haoshs]

oscillator transient crystal

Thanks for the promt reply!

I just found in HSPICE manual the following:
For circuits with very high resonance factors(such as crystal tank circuits,and active filters), .OPTION DELMAX should be set to less than period/100.

Share with you!

(Other conservative simulation options also should be used in crystal simulation as mentioned by Fom's post.)

regards,
jordan76

Seting delmax is great idea. it really works. TX

 

[promethean_spark]

I also found that the simulation would start oscillating in the microvolts, and run over-night ramped up to less than 1mV. I was able to jolt it into life instantly by adding a current source from ground to the node between the inductor and capacitor in the crystal model, and turning it on at 1mA for 100ns. We're trying to get that crystal to vibrate, and this is like hitting it with a hammer.

istart n_1 0 pwl 0 0 20u 0 20.01u 1m 20.02u 0

Power was applied at 10us.

This should work easier than trying to start it with a .ic statement because it's easier to converge oscillator circuits when you start the transient analysis with everything off.

Care should be taken to verify that this oscillation doesn't damp out eventually though, and an AC analysis to verify gain/phase at the oscillation frequency is also needed to demonstrate how much gain margin you have.