Unexpected results when using a MOSFET to pulse a capacitor and measure the current

I am attempting to charge a capacitor to 25 Volts and then release a pulse of energy every 1 second by sending a gate voltage signal of 12 Volts to a MOSFET. The expected testbed current is 25 Volts / 1 Ohm = 25 Amps. However, I only measure 25 Amps when initializing the capacitor's startup voltage to 25 Volts. All other times the measured current is much lower than expected (~ 3.6 Amps). I am confused where the current is going because the capacitor voltage is fully recharged to 25 Volts between each pulse. I also notice that the drain-source current being measured starts to rise before the pulse signal begins, which may or may not be part of the problem. Any help is greatly appreciated!

Answer 1

[Joel Van Sickel]

Hello Braydon,

thank you for the updated model. That has made it easier to see what is going on.

1st: the default tolerance settings for this type of model are unfortunately too loose. The faster time constant requires a tighter tolerance. If you go to your model configuration settings under solver, if you change it from 1e-3 to 1e-4, you will get consistent results per pulse. This use to be the default if you used an electrical template, but I think they backed off it again to 1e-3. For many electrical simulations you will find that you need 1e-4 still.

2: the other issue is that simscape doesn't handle infinite derivatives well. In this case, because the gate junction is being modelled instead of using a piecewise linear mosfet, it is messing something up in the solution. If you apply filtering at that specific simulink to ps converter. This will stop the current from flowing before the pulse hits. (I am guessing using a fixed step solver might also fix this, but this is the better solution) I chose a pretty small time constant so that it barely slows down the pulse, but you can make this filter even smaller, just know that the smaller this number, the smaller time steps will be taken which can slow down a larger simulation. You could alternately try utilizing the prebuilt gate driver block that ships in part of simscape electrical. However, this appraoch with a filter will work just fine.

I have attatched a model where I made the changes so that you can see the results. You can see I was also trying different solvers. You can probably leave yours as auto, but ode23t or the 15s are good options if you pick a specific one.

Regards,

Joel

Answer 2

[Joel Van Sickel]

Hello Braydon,

There are a few things to consider. the first is if the pulse is actually as long as you want it to be. Your Rg value limits the dv/dt. If you lower your resistance, or lower your resistance and increase your pulse width, you can get a higher peak current. The device will not instantly turn on to 25 amps as you are doing a detailed simulation of the switch.

You may also want to consider if you should really be using this specific part number (switch) for 25 amp pulsing. I didn't do an indepth look at the spec sheet but it didn't seem like a device designed for pulsing 25 amps at the pulse width you have specified. Again, I didn't look at it long so I may be wrong.

Regards,

Joel

[Braydon Westmoreland]

Joel, thank you for the reply.

You are correct that the selected switch is not designed for 25 Amp pulsing which was an oversight on my part when throwing this simulation together. This simulation is a simplified version of the actual simulation I'm currently working on which contains 4 of these switches in parallel. I’m experiencing these same issues with the more complex simulation, so I made this simple one to illustrate my problem for this post. I have since adjusted the testbed load resistor from 1 Ohm to 5 Ohms in order to reduce the current down to 5 Amps which is more appropriate for a single switch design.

After reducing the current to 5 Amps I am still experiencing the same problems as before. The tb_drain current is the expected 5 Amps for the initial pulse but drops to 3.6 Amps for all remaining pulses, just like it did when pulsing at 25 Amps. Additionally, the gate1 voltage begins to rise before the gate voltage pulse (gvp) begins which causes current1 to rise before it’s supposed to. This effects the sampled current measurement that is supposed to be taken halfway through the pulse.

In response to some of your other suggestions:

• The pulse width of 100 microseconds is intentional for this simulation in order to model how we plan to use this device in our lab.
• I tried different values for Rg but none of these eliminated the problems mentioned above.

Updated files are attached.

Thank you for your time,

Braydon