temperature in transient sim

Dear Karev11,

Karev11 said:

I expect frequency difference between 0C and 125C to be the same between these two cases, (and then use case2 to reduce number of total simulations), but they are different by 3x to 4x.

is this expected? what could be the reason?

I certainly have no where near enough information to indicate the source of the difference, but I suspect that your dynamic parameter  simulation is not allowing enough settling time between temperature steps. The change in temperature at a simulation time is effectively a "transient pulse" and the time constant of your VCO must be considered.

Information that would be helpful includes your spectre version, the accuracy settings, (i.e., your input.scs tran statement and spectre run line), simulation times, VCO type and if powered by a regulator voltage with its time constant and interval over which you are changing temperature, temperature interval,  how you re measuring frequency and over what interval,  and the actual differences in frequency...

Shawn

Hi Shawn,

I doubled check that frequency has been settled for the measurement.

As to other factors that could affect frequency, I suppose they're common to both cases and therefore should be the reason for this?

I don't know how new and robust is the "dynamic parameter" feature

Kevin

Dear Karev11,

Karev11 said:

As to other factors that could affect frequency, I suppose they're common to both cases and therefore should be the reason for this?

I can't say as, once again, I don't have enough information. For example, what is the interval over which you are doing the frequency measurement in each simulation? 

I also disagree with your comment. The repeatability of the frequency measurement is limited by simulation accuracy. Hence, if you were to measure in simulations that don't behave the same (as in your two simulations), the difference in the frequency between the two simulations can be just due to your accuracy and spectre simulation settings. For example, if the VCO frequency is changing due to temperature, I would expect the timestep to be less than one in which the temperature is not changing. A lower tilmestep can impact the frequency measurement.

So, once again, the added information I suggested that would be helpful to provide more insight on your results would be very useful.

Shawn

Dear Karev11,

Karev11 said:

As to other factors that could affect frequency, I suppose they're common to both cases and therefore should be the reason for this?

I can't say as, once again, I don't have enough information. For example, what is the interval over which you are doing the frequency measurement in each simulation? 

I also disagree with your comment. The repeatability of the frequency measurement is limited by simulation accuracy. Hence, if you were to measure in simulations that don't behave the same (as in your two simulations), the difference in the frequency between the two simulations can be just due to your accuracy and spectre simulation settings. For example, if the VCO frequency is changing due to temperature, I would expect the timestep to be less than one in which the temperature is not changing. A lower tilmestep can impact the frequency measurement.

So, once again, the added information I suggested that would be helpful to provide more insight on your results would be very useful.

Shawn

In addition to the things that Shawn has mentioned, I would also like to point out that using dynamic parameters to make a 125º step in temperature is a bad idea and asking for trouble. When you use dynamic parameters to change something, the change is instantaneous - which means you have a sudden discontinuity in the results which can lead to all sorts of convergence and consistency issues. Those should (assuming it can converge) eventually dissipate as it settles, but given that this a totally non-physical change it could certainly lead to questionable results. It would be better to make a more gradual change in temperature - maybe using the transient param_step option (search on this forum for param_step to find some examples - it's not on the tran form unfortunately).

Dynamic Parameters are not new - it was added in 2006 - and it is robust, provided you don't ask for something that doesn't make sense. I'd venture that this doesn't make sense. Unlike a pulse source where you have control over the edge speeds, and the simulator can take shorter time steps to get across the sharp change, the change here is instantaneous and so you have to try to ensure that any change is small enough that it won't adversely affect the simulator.

It's also possible that your circuit has more than one stable mode of operation and so may be depend on the starting point as to which frequency you get. It's pretty hard to tell all the details, or whether it's accuracy controls, without sight of the example.

Andrew.