The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> RF Simulators >> Harmonic balance no guarantee for start-up?
https://designers-guide.org/forum/YaBB.pl?num=1234428781

Message started by Visjnoe on Feb 12^th, 2009, 12:53am

Title: Harmonic balance no guarantee for start-up?
Post by Visjnoe on Feb 12^th, 2009, 12:53am

Dear all,

I had a discussion with someone who claimed that it is possible that an oscillator exhibits nice, steady-state oscillation as the outcome of an harmonic balance simulation, while in reality it might not start up. This would then, for example, be seen in a transient simulation.

The reasoning is that the gm required for startup is/can be bigger than the one required to sustain oscillation. I do however not see the physical mechanism behind this, since to my opinion, the losses in the circuit are the same (in startup versus steady-state).

Any comments welcome.

Title: Re: Harmonic balance no guarantee for start-up?
Post by pancho_hideboo on Feb 12^th, 2009, 4:07am

I can't understand your question correctly.

Visjnoe wrote on Feb 12^th, 2009, 12:53am:

I do however not see the physical mechanism behind this, since to my opinion, the losses in the circuit are the same (in startup versus steady-state).

There could be two solutions as steady state solutions where "negative impedance"+"loss impedance"=0.
One is an oscillation, the other is non oscillation, that is, static state.

Autonomous HB analysis can find this steady state oscillation solution without any start up for oscillation.
For example, autonomous HB analysis can give oscillation solution for X'tal OSC having very High Q without any startup.
But it requires very accurate oscillation frequency as initial guess.

Old HB analysis which didn't have transient assisted ability required special oscillator probe for Autonomous HB analysis.
For example, Agilent MDS required this special oscillator probe.
Even in Agilent ADS, this special oscillator probe has remained as option.
But there are cases where it can't find oscillation solution for this pure frequency domain technique.

Current HB analyses have transient assisted ability to get good initial guess for oscillation, here start up for oscillation is used.
And it doesn't require special oscillator probe.

Again see the following to your previous post.
http://www.designers-guide.org/Forum/YaBB.pl?num=1216380172/7#7
http://www.designers-guide.org/Forum/YaBB.pl?num=1216380172/13#13

Title: Re: Harmonic balance no guarantee for start-up?
Post by Visjnoe on Feb 12^th, 2009, 5:41am

Just to make my question more clear: is it possible that harmonic balance shows an oscillator to be oscillating (steady-state), while in reality it can not start up?
By steady-state I mean really a sinusoidal behavior. I'm using a recent, commercial simulator (SpectreRF).

In other words: does harmonic balance give confidence that a oscillator will start up?

Regards

Peter

Title: Re: Harmonic balance no guarantee for start-up?
Post by pancho_hideboo on Feb 12^th, 2009, 5:56am

Do you understand HB Analysis ?

Visjnoe wrote on Feb 12^th, 2009, 5:41am:

is it possible that harmonic balance shows an oscillator to be oscillating (steady-state), while in reality it can not start up?

If your circuit doesn't satisfy small signal oscillation condition, there is no oscillation solution in steady state.

Visjnoe wrote on Feb 12^th, 2009, 5:41am:

By steady-state I mean really a sinusoidal behavior.

What do you mean ?

Visjnoe wrote on Feb 12^th, 2009, 5:41am:

In other words: does harmonic balance give confidence that a oscillator will start up?

There is no "Start Up" concept in pure frequency domain Autonomous HB Analysis.

Again see the following to your previous post.
http://www.designers-guide.org/Forum/YaBB.pl?num=1216380172/7#7
http://www.designers-guide.org/Forum/YaBB.pl?num=1216380172/13#13

Title: Re: Harmonic balance no guarantee for start-up?
Post by Visjnoe on Feb 12^th, 2009, 6:02am

I understand HB and also that there is no startup in harmonic balance.

The question is as follows: if you SIMULATE an oscillator using harmonic balance and all looks fine, is that a guarantee that in REALITY/MEASUREMENTS it will start up and oscillate (assuming no model errors or simulator errors)?
If not, why not?

Title: Re: Harmonic balance no guarantee for start-up?
Post by pancho_hideboo on Feb 12^th, 2009, 6:09am

Visjnoe wrote on Feb 12^th, 2009, 6:02am:

The question is as follows: if you SIMULATE an oscillator using harmonic balance and all looks fine, is that a guarantee that in REALITY/MEASUREMENTS it will start up and oscillate (assuming no model errors or simulator errors)?
If not, why not?

You can't know oscillation margin by steady state analysis even though there are no model errors or simulator errors.
You also have to invoke small signal analysis

http://www.designers-guide.org/Forum/YaBB.pl?num=1233667204

Title: Re: Harmonic balance no guarantee for start-up?
Post by Visjnoe on Feb 16^th, 2009, 11:35am

I still haven't got a good answer to my original question.

I understand that HB does not show startup behavior (per definition) and also that you cannot deduce oscillation margin from it.

However, if you SOLELY run HB on a given oscillator, can you be sure he will startup and reach the steady-state condition (in reality)? If not, which possible mechanism would prohibit you from making such a deduction?

The way I see it, the steady-state oscillation verified with HB proves that the gm is enough to satisfy the Barkhausen criterion. Let's assume everything is also OK over corners etc. and you have effectively taken into account all losses.

For the oscillator not to startup in reality, would imply that somehow during the initial startup phase the 'losses' seen by the oscillator are bigger, requiring a bigger gm. I just cannot think of a mechanism which would result in bigger losses during startup (when the oscillator's amplitude is ramping up) versus steady-state.

If the losses do not vary over time, I would think that satisfying Barkhausen in steady-state is a sufficient condition to guarantee startup (without explicitly running a transient simulation to verify it).

Title: Re: Harmonic balance no guarantee for start-up?
Post by pancho_hideboo on Feb 17^th, 2009, 3:32am

Visjnoe wrote on Feb 16^th, 2009, 11:35am:

I just cannot think of a mechanism which would result in bigger losses during startup(when the oscillator's amplitude is ramping up) versus steady-state.

Autonomous HB analysis is no more than numeric solution for homogeneous nonlinear equation.
Even though autonomous HB analysis can find out a solution of stable oscillation,
in transient behavior the circuit states could trace different paths and arrive at different terminal such as oscillation stop depending on starting method of oscillation, that is, initial conditions.
Losses during startup could be both bigger and smaller than steady state.
Also magnitude of negative impedance could be both smaller and bigger than steady state.

Visjnoe wrote on Feb 16^th, 2009, 11:35am:

If the losses do not vary over time, I would think that satisfying Barkhausen in steady-state is a sufficient condition to guarantee startup.

You are misunderstanding.
Barkhausen is Necessary Condition for Oscillation not Sufficient Condition.
And Barkhausen are never conditions for duration of oscillation.
http://en.wikipedia.org/wiki/Barkhausen_stability_criterion

If Barkhausen is Necessary Condition for Oscillation, a expression of "Barkhausen in steady-state" is meaningful.
But if it is Sufficient Condition, a expression of "Barkhausen in steady-state" is very strange.

Title: Re: Harmonic balance no guarantee for start-up?
Post by Visjnoe on Feb 17^th, 2009, 4:25am

Ok, I think we have convergence now.
Barkhausen is necessary but insufficient condition and the losses can vary over time.
This answers my quesions.