A question about transient analysis

[listen]

Hi,

I encontered a strange phenominan on my transisient analysis. The circuit is an oscillator, I want to investigate the oscillation behaivior when it is injected/not injected a signal source.

First, I inject the signal source, the oscillation is OK.

Second, I remove(delete) the signal source, the circuot did not oscillate as expected. It is OK.

But, if I did not remove the signal source, I just disconnected to the circuit. As my understanding, the circuit should not oscillate and the signal source should not have any impact to the circuit. But, I found the oscillate work just as it is connected to the osciilator.

Anyone pls. shed some light on it?

Thanks

 

[tgootee]

Re: A question about transiient analysis

If it was a real circuit, then even just sharing the same power supply, or even just sharing a small section of a ground conductor, could couple the signal source frequency into the oscillator. Also, if there is any significant "enclosed loop area" in the oscillator circuit, it would act as a receiving antenna, and any enclosed loop area in the signal source could act as a transmitting antenna. And if two conductors, from the two circuits, were physically close-enough to each other, the signal source's frequency could inductively or capacitively couple to the oscillator circuit.

 

[LvW]

Re: A question about transiient analysis

Listen, I don`t understand your problem.
The oscillator is a circuzit that is able to produce a signal WITHOUT any signal source.
Why do you "inject" a signal? And where? Does your oscillator circuit has any input?

 

[KerimF]

Re: A question about transiient analysis

If you leave the signal source even disconnected, it may provide the simulator an indication of the maximum timestep.
To verify the validity of this point, you may remove the source completely and set the maximum timestep manually (according to the source frequency).

 

[listen]

Re: A question about transiient analysis

Listen, I don`t understand your problem.
The oscillator is a circuzit that is able to produce a signal WITHOUT any signal source.
Why do you "inject" a signal? And where? Does your oscillator circuit has any input?

Hi,

The oscillatotr is the injection oscillator. In addition, the oscillator will not oscillate in transient analyais if without an initial seting.

For my circuit, I wonder if it is a simulation error?

Thanks

 

[KerimF]

Re: A question about transiient analysis

Does your simulator have a setting for the maximum timestep?

 

[LvW]

Re: A question about transiient analysis

The oscillatotr is the injection oscillator.

Why don`t you give us such an important information not from the beginning?

For my circuit, I wonder if it is a simulation error?

Yes - I understand. However, do you expect that we are able to localize a probable error without any circuit diagram and without any further information about your simulation profile?

 

[Kevin Aylward]

Re: A question about transiient analysis

A “oscillatable” oscillator, when simulated in Spice, may or may not start without giving the oscillator a kick start. It can be quite common for an oscillator to not require a kick, but there is no guarantee of this. So, usually one turns on the power supply at T=0, or feed a pulse through a very small capacitor to a suitable node.

 

[listen]

Re: A question about transiient analysis

Hi KerimF,

Thanks for the helpful response!

You mean setting the max timestep in simulation, does that mean the max timestep will introduce the error of calculation for the simulator and this calculation error then acts a noise source to kick off the oscillator?

Thanks!

If you leave the signal source even disconnected, it may provide the simulator an indication of the maximum timestep.
To verify the validity of this point, you may remove the source completely and set the maximum timestep manually (according to the source frequency).

- - - Updated - - -

- - - Updated - - -

- - - Updated - - -

A “oscillatable” oscillator, when simulated in Spice, may or may not start without giving the oscillator a kick start. It can be quite common for an oscillator to not require a kick, but there is no guarantee of this. So, usually one turns on the power supply at T=0, or feed a pulse through a very small capacitor to a suitable node.

Hi Kevin，

would u please clarify further why the oscillator may need not a kick start in simulation? Is it for the calculation error?

Thanks

 

[dick_freebird]

Re: A question about transiient analysis

There is always finite numerical error in the initial condition
and this may or may not suffice to unbalance the oscillator
core enough to start it going. The simulator stops at "good
enough", not "perfect". But sometimes you get close enough
to perfect balance that the oscillations never start without
additional stimulus. This comes down to cases and conditions.
Changing the netlist in any way can change the error residue.

If you're doing things like nested loops trying to get PVT
coverage, you don't want to give it the option of not starting,
so you kick it in some way that doesn't corrupt the operation
afterward (I prefer a brief current pulse starting from and
returning to zero).

 

[listen]

Re: A question about transiient analysis

Thanks dick_freebird，

It is indeed helpful to me!

Regards
listen

There is always finite numerical error in the initial condition
and this may or may not suffice to unbalance the oscillator
core enough to start it going. The simulator stops at "good
enough", not "perfect". But sometimes you get close enough
to perfect balance that the oscillations never start without
additional stimulus. This comes down to cases and conditions.
Changing the netlist in any way can change the error residue.

If you're doing things like nested loops trying to get PVT
coverage, you don't want to give it the option of not starting,
so you kick it in some way that doesn't corrupt the operation
afterward (I prefer a brief current pulse starting from and
returning to zero).

 

[crutschow]

Re: A question about transiient analysis

In real life circuits have noise that will start an oscillation. Simulations are ideal and do not necessarily have such noise so the oscillation may not start. Thus they may need an added perturbation to initiate the oscillations. A simple way is to add a small pulse from a current source, since a current source has infinite simulation impedance and thus has no affect on the circuit values or circuit operation after the oscillations start.

 

[LvW]

Re: A question about transiient analysis

In real life circuits have noise that will start an oscillation. Simulations are ideal and do not necessarily have such noise so the oscillation may not start. Thus they may need an added perturbation to initiate the oscillations. A simple way is to add a small pulse from a current source, since a current source has infinite simulation impedance and thus has no affect on the circuit values or circuit operation after the oscillations start.

I think, most probably, it is not noise that starts oscillation but the power switch-on transients which cause something like an impulse response within the circuit. But this is - more or less - a theoretical question.
However, also in circuit simulations such a supply voltage switch-on at t=0 will start the oscillation safely.