The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> amplitude of xtal osc
https://designers-guide.org/forum/YaBB.pl?num=1251157408

Message started by uu@uk on Aug 24^th, 2009, 4:43pm

Title: amplitude of xtal osc
Post by uu@uk on Aug 24^th, 2009, 4:43pm

Hi all,

I have a xtal osc and its amplitude looks good when it settles in the simulation. But the lab measurement shows only 1/3 of the swing comparing to simulation result. Anybody has any idea about that? Is there any parasitic resistance hurts the oscillator? BTW, the xtal osc runs at 25MHz.

Thanks a lot

Title: Re: amplitude of xtal osc
Post by buddypoor on Aug 25^th, 2009, 1:08am

The amplitude of a free running oscillator is sometimes a problem.
In theory: If you exactly meet the oscillation criterion, the amplitude can have any value (within the power supply limits), which meens: it is not determined.
In practice: Normally, it is determined by an amplitude limiting mechanism.
Question: Did you incorporate such a limiting circuit into the oscillator ?

Title: Re: amplitude of xtal osc
Post by uu@uk on Aug 25^th, 2009, 9:19am

Dear buddypoor,

Thanks for your reply.

if the amplitude can have any value when the osc barely meet the oscillation criterion, can the simulator have different amplitude value for different runs? since I run the transient simulation, the results always give me the same amplitude.

usually the xtal osc has a automation gain control circuit to manage the current after the xtal osc settles. I am not sure whether this is the limiting circuit you are talking about.

Thanks

Title: Re: amplitude of xtal osc
Post by wave on Aug 25^th, 2009, 12:09pm

Crystals are a specialty of themselves. Often poorly documented in literature that is very piece-meal.

You are likely ignoring parasitics of the XTAL element itself, the board, pad/ESD structure, the ampl detection loop, and/or the scope probe monitoring the amplitude.

A complete model has many many passive elements and the interactions can be surprising! I'd start with your ESR and Co values.

Wave ;)

Title: Re: amplitude of xtal osc
Post by buddypoor on Aug 25^th, 2009, 1:28pm

Hi uu@uk,

of course, I agree with all the arguments WAVE has listed. But I´m not sure if they are the reason of your amplitude problem. Mostly, these parasitic effects cause frequency shifts.
But there remains one important question: Is your output sinusoidal or rectangular ?
Quote:
...can the simulator have different amplitude value for different runs? since I run the transient simulation, the results always give me the same amplitude.
That´s normal, because there is no reason for the simlator to calculate in a different way. It uses some initial conditions to arrive at a certain amplitude,

usually the xtal osc has a automation gain control circuit to manage the current after the xtal osc settles. I am not sure whether this is the limiting circuit you are talking about.

What do you mean with "manage the current" ?
When I mention a limiting mechanism I mean, of course, a circuitry which is non-linear and brings the poles from the right hand s-plane
back to the Im axis (or at least in the vicinity) when the amplitude increases. Such a amplitude or gain control device is necessary in each harmonic oscillator circuit. In some cases this is established by hard limiting at the power supply rails. But then, the signal is distorted.

Title: Re: amplitude of xtal osc
Post by Forum Administrator on Aug 25^th, 2009, 6:43pm

Are you sure you simulated to steady-state? Often times crystal oscillators settle so slowly that you can fool yourself into thinking you have settled, but still have a very long way to go.

-Ken

Title: Re: amplitude of xtal osc
Post by uu@uk on Aug 25^th, 2009, 9:58pm

Dear Ken,

Yes, I am also worrying about it. So I set the Q in the simulation lower than the real one in the lab. And then I let the xtal ocs start and run 7500 periods. eventually I obeserved the voltage of AGC, it is really flat after some long time(about 2500 period). I think it settles in this case. Is that correct?

Thanks

Title: Re: amplitude of xtal osc
Post by ywguo on Aug 26^th, 2009, 3:27am

Hi UU,

A few points jump into my mind.

1. The parameters for your crystal, is it the same as those in the simulation?
2. The parasitics at the PCB board.
3. Is the current equal to what you expect. That is important. It seems that circuitry has Auto Level Control. I guess it may be implemented with something tuning the current.
4. Ken has mentioned that simulation amplitude is sensitive to algorithm for transient analysis. So do you know the amplitude theoretially?

Best Regards,
Yawei

Title: Re: amplitude of xtal osc
Post by uu@uk on Aug 26^th, 2009, 11:57pm

Dear Yawei,

I do not think the xtal model is exactly same as that in the lab since I lower the Q(by increasing the resistance) purposely to make the osc startup time shorter.

The parasitic resistance from PCB should not be significant here. it is about 1 ohm.

Thanks.

Title: Re: amplitude of xtal osc
Post by wave on Aug 27^th, 2009, 3:18pm

uu@uk wrote on Aug 26^th, 2009, 11:57pm:

Dear Yawei,

I do not think the xtal model is exactly same as that in the lab since I lower the Q(by increasing the resistance) purposely to make the osc startup time shorter.

You should lower the Q by changing the series L-C of the L-R-C.
This will affect your frequency accuracy a bit, but not amplitude.
Changing the ESR will definitely change your amplitude.

Wave

Title: Re: amplitude of xtal osc
Post by ywguo on Aug 27^th, 2009, 10:57pm

Hi uu,

What Wave said is correct. By the way, what is the simulation set up? Have you tried another alogrithm?

wave wrote on Aug 27^th, 2009, 3:18pm:

uu@uk wrote on Aug 26^th, 2009, 11:57pm:

Dear Yawei,

I do not think the xtal model is exactly same as that in the lab since I lower the Q(by increasing the resistance) purposely to make the osc startup time shorter.

You should lower the Q by changing the series L-C of the L-R-C.
This will affect your frequency accuracy a bit, but not amplitude.
Changing the ESR will definitely change your amplitude.

Wave

Title: Re: amplitude of xtal osc
Post by loose-electron on Aug 30^th, 2009, 7:22pm

Crystal simulations are a compromise at best -

Q of a crystal is about 10E5, and this can mess a simulator up. (Run an AC sweep of a crystal and you will see the errors due to the resolution of the frequency steps) Add to that the parasitic crossover capacitance, which introduces a zero in very close proximity of the pole due to the crystal resonance. Both mess up the gain response and transient accuracy of simulations.

Also, if you are trying to get a correct crystal model, you need to to include the resonances of the overtones. Also, remember this is an electrical model of a electro-mechanical system.

Common strategy here:

Simplify the model by getting rid of the crossover parasitic.
Lower the Q of the circuit while keeping the ESR the same, so the insertion losses are the same.
Quick start the model by using a current as an initial condition on the inductor in the crystal model.

my 2 cents worth....
Jerry