Transient & frequency response



Dear supporter,
For VPD transient simulation, we can get the impulse response, then get the frequency response. In the vpd_electrical.ldev, the 3dB reached 100+GHz, which beyond the theoretical value too much. I want to know if there are problems in transient set?190GHz
vpd_electrical.ldev (5.6 MB)

The error in vpd transient simulation
The transient simulation of vpd

Dear @mapengcheng,

Sorry for the delayed reply. I was working on your simulation to figure out where the problem is.
The reason you are getting such a high bandwidth is because high field mobility (velocity saturation) effects for Germanium are not considered in your simulation. This is especially important when you increase the bias voltage of your simulation (as you did in your file) which can significantly affect the mobility due to the generated field by that voltage. Attached is a modified version of your file with high field mobility model enabled in Ge material properties. Also the gradient mixing in advanced CHARGE solver settings has been enabled which helps in simulation convergence when using field dependent models. This simulation will give you a more reasonable bandwidth.
vpd_electrical (6).ldev (5.7 MB)


Dear @mmahpeykar,
Thanks for your help, I got the reasonable simulation results. Futhermore, When I chaged the value of shutter ton, it will chaged in the range of 30-35GHz, I wonder konw how to set up the Global Source Shutter and other parameters such as Poisson Solver Controls, Mesh to improve the simulation result accuracy.
Looking forward to your reply.


The time at which the shutter turns on shouldn’t affect the bandwidth as long as the system has reach steady state before the end of simulation. It is possible that if you delay the turn on time, the steady state hasn’t been reached and the results won’t be valid.

closed #5

This topic was automatically closed 3 days after the last reply. New replies are no longer allowed.

opened #6


dear mmahpeykar,
There is a job error in vpd transient simulation as follows, which parameter in this simulation leads to mistakes? How to avoid this happen?
1w=2.ldev (5.6 MB)


This version of the file seems to have three times more generation than the one in your previous post so simulation is more difficult to converge. To help with convergence, switch the ‘gradient mixing’ in advanced tab of CHARGE solver from ‘fast’ to ‘conservative’.


When I switch the ‘fast’ to ‘conservative’, this simulation works properly. But when I study the length of Germanium, for different length, this 3dB bandwidth is the same, which is not inconsistent with the reality. Reference:The 3dB bandwidth extracted from the lateral PD.
I add the RL to redo simulation, Whether I set the’ gradient mixing ’ is ‘fast’ or ‘conservative’, the simulation diverged and the result is incomplete. I am very confused.

5.ldev (5.6 MB)


In your case, for a more stable simulation, you should apply the external resistance to the anode contact rather than cathode as shown below:

This won’t affect the results of course since it does not matter at which contact the external resistor is located from device operation point of view.
Also, you need to switch the ‘time integration’ under the 'transient tab of the CHARGE solver from ‘BDF2’ to ‘TR-BDF2’ since BDF2 is not compatible with external circuit inclusion in the simulation. Making these modifications should make the simulation converge.
Regarding the change in bandwidth when you change the length, as it is also mentioned in the reference post you provided, if the transit time is the dominant limiting factor for bandwidth, you won’t see any change in bandwidth as you increase the length.
Another way of calculating the RC limited bandwidth is demonstrated in the example below:
This way you can obtain the RC limited bandwidth separately and verify which one (transit time or RC) is the dominant limiting factor in bandwidth.

closed #11

This topic was automatically closed 3 days after the last reply. New replies are no longer allowed.