charge pump:different time step;different results, why?

[allennlowaton]

Good day guys.

I badly need your ideas about this one.
I'm doing a charge pump. I'm amazed with what happened.
I change the time step on the transient analysis and it
yields different outputs.

I can't understand it.

Please refer to the figures below:

1. This one is the result of this transient analysis
.tran 100us 400ms



2. This one is the result of this transient analysis
.tran 1ms 400ms



This is the circuit of this simulation. It shows only the 4 stage but I simulated 9 stages.



The time step with 100us yields a 3v output
But, the time step with 1ms yields only 2.4v....

Help please............

 

[LvW]

As a general rule, time steps should be as small as possible - of course, with respect to the final simulation time.
I think your second run with a maximum step of 1ms leads to incorrect results.
An indication for that are the "waves" you can observe in comparison with the first run.
More than that, to evaluate and justify simulation results one should have a rough feeling about the results to be obtained. Therefore, which result is more close to your expectations?

 

[keith1200rs]

You don't say how fast you are clocking it, but I would start with the maximum time step being two orders of magnitude smaller than the clock period. Default maximum timesteps are usually far too long for sensible results.

Keith.

 

[allennlowaton]

thank you very LvW...

i'm utilizing a 0.5v input, since I'm using 9 stages of this one, I'm expecting an output of around 4 volts.
The frequency of the clock is 25kHz.

I have tried a time step of 10us and I got 3.7~3.8 volts.

Follow up question, is it suggested to make my time step equal to the period of my clock?

Thanks guys...

 

[AdvaRes]

I can't see figures. Btw, if you are using spectre you can minimimize you simulation step by clicking on Options (for Tran Analysis) and then set step and/or maxstep to 1ps for exemple.

 

[allennlowaton]

good day keith..

correct me, if I'm wrong, I should have a maximum step of 80us? My clock period is 40us.

That is what I have understood from your message.

Thank you.

 

[AdvaRes]

good day keith..

correct me, if I'm wrong, I should have a maximum step of 80us? My clock period is 40us.

That is what I have understood from your message.

Thank you.

No he meant maximum step of 20µs.
Generally it's better to be below the tenth so what I suggest you is to use a maxstep below 4µs.

 

[allennlowaton]

halo AdvaRes..

I'm using Synopsys Hspice...

Guys, what could be the possible conclusion (or reason) that we can obtain in this problem?

Thank you...

 

[keith1200rs]

good day keith..

correct me, if I'm wrong, I should have a maximum step of 80us? My clock period is 40us.

That is what I have understood from your message.

Thank you.

No he meant maximum step of 20µs.
Generally it's better to be below the tenth so what I suggest you is to use a maxstep below 4µs.

I actually suggested two orders of magnitude smaller, so for 40us clock I would suggest 400ns maximum timestep.

Keith.

 

[AdvaRes]

I actually suggested two orders of magnitude smaller, so for 40us clock I would suggest 400ns maximum timestep.

Keith.

Thanks keith1200rs for your explanation.

halo AdvaRes..

I'm using Synopsys Hspice...

Guys, what could be the possible conclusion (or reason) that we can obtain in this problem?

Thank you...

The narrow the step is, the more precise the result will be.

 

[keith1200rs]

The reason for the problem is that Spice will increase the timestep up to the maximum if the results converge within the absolute and relative accuracy settings. However, it is possible for it to produce inaccurate results or miss events completely using this method so you need to force it to keep a short timestep to avoid the problem. I never use the default maximum timestep on any transient analysis. It might depend on the simulator, but it always seems too coarse.

Keith.

 

[allennlowaton]

As I minimized the time step, I now obtain an appreciable results.

Here is the result for the simulation with a time step of 4us.



The voltage at the output node is already 4.1volts.

Added after 3 minutes:


oh no..

sorry for that guys, I had inserted the wrong file.

This is the correct one.



Added after 21 minutes:


keith, LvW and AdvaRes...

thank you very much for the enlightenment.

 

[allennlowaton]

hello guys...

i made a simulations with a different time step.
Please refer the results as shown in the table
Indeed, as I decreased more the time step, the maximum voltage increases. But, the % difference becomes smaller.

 

[keith1200rs]

You can now create your own 'rule of thumb' using your results.

Keith

 

[allennlowaton]

yeah...you're right keith...
slowly, I started to feel and loved the true essence of research.
thanks my EDAboard mentors....