LTspice SiC MOSFET Simulation Convergence Problem

Thread Starter

beleg

Joined May 14, 2016
16
Hi everyone,

I am trying to simulate a SiC MOSFET from ON Semiconductor on LTspice. The one that I am using is:
NVHL020N090SC1. I've downloaded the model from manufacturer's website and it is a hex file which means this is a black box for me.
I've used SiC MOSFET models from Wolfspeed and I had no problems whatsoever but for this one I cannot simulate even a basic switching cell
let alone a complex converter. I am always getting "Time step too small error" .I am by no means a SPICE pro but I've tried some convergence problem solutions that I've found online but nothing worked for me so far. Can you please point my error? I am stuck. I've added the LTspice simulation and the MOSFET model that I am using.
 

Attachments

ericgibbs

Joined Jan 29, 2010
14,131
hi beleg,
The semi file is encrypted.
* LTspice Encrypted File
*
* This encrypted file has been supplied by a 3rd
* party vendor that does not wish to publicize
* the technology used to implement this library.


E

Update:
Is there any equivalent MOSFET that you could use.?
 

Thread Starter

beleg

Joined May 14, 2016
16
Yeah exactly it is encrypted. That is what I meant by "hex file" and "black box". Unfortunately, there are not so many options when it comes to the SiC MOSFETs. This one has a relatively low price and I would love to use this one in my designs. I already used similar MOSFET models from Wolfspeed and they work fine with similar configurations.
 

Papabravo

Joined Feb 24, 2006
17,242
Is this happening during the simulation or is it happening during the determination of the DC operating point?
In LTspice group on groups.io there are numerous threads dealing with various aspects of this problem. Setting a small maximum timestep can help avoid this error because it happens when they use nonlinear devices to determine the "appropriate" timestep. Changing the integration method may also produce results. You need to be careful with this approach because it trades accuracy for convergence. This is especially true when poles of a transfer function are on the verge of slipping into the right half plane, causing the response to diverge exponentially.
 

Thread Starter

beleg

Joined May 14, 2016
16
For this configuration it is happening during the simulation. I've tried to simulate with 1e-10 maximum timestep, it goes further in the simulation but only for few cycles. I've used "alternate" for the solver instead of normal but that did not do the trick. I've tried to add 1Meg ohm resistors from every node to ground, still no luck. I am quite annoyed because this is as basic as it can get. Either there is some problem with the model (I doubt it but you never know) or I am doing something incredibly stupid.
 

Papabravo

Joined Feb 24, 2006
17,242
For this configuration it is happening during the simulation. I've tried to simulate with 1e-10 maximum timestep, it goes further in the simulation but only for few cycles. I've used "alternate" for the solver instead of normal but that did not do the trick. I've tried to add 1Meg ohm resistors from every node to ground, still no luck. I am quite annoyed because this is as basic as it can get. Either there is some problem with the model (I doubt it but you never know) or I am doing something incredibly stupid.
Box it all up in a .zip file and post it to groups.io Those folks have what I consider a high probability of success in figuring out what the problem is. At the very least they may have insight on what the problem might be, especially if it is with the simulator and not with the model. I'm making an assumption that the model was not designed for LTspice, but for some other SPICE flavor.
 

Thread Starter

beleg

Joined May 14, 2016
16
Box it all up in a .zip file and post it to groups.io Those folks have what I consider a high probability of success in figuring out what the problem is. At the very least they may have insight on what the problem might be, especially if it is with the simulator and not with the model. I'm making an assumption that the model was not designed for LTspice, but for some other SPICE flavor.
As soon as you mentioned the group I created an account. I will check previous post and post my problem if nothing comes up.
In the manufacturer's website where I downloaded these models from it says LTspice models.
MOSFET Models by Manufacturer

Thanks for the help.
 

Papabravo

Joined Feb 24, 2006
17,242
This seems to be pointing the finger at the model, but it is often the case that the model may be the victim and not the problem. I'm not that familiar with this technology and I am puzzled by the text of the error message which seems to suggest that the device is being modeled as a "jfet". Is that correct or am I missing something? Also what is the specification for source V2? In particular the rise and fall times.
 

Papabravo

Joined Feb 24, 2006
17,242
Thanks for the details on V2. Are those rise and fall times reasonable for this part? Like I said I'm not really familiar with these parts, but it seems like small rise and fall times with big derivatives could be problematical.
 

Thread Starter

beleg

Joined May 14, 2016
16
This seems to be pointing the finger at the model, but it is often the case that the model may be the victim and not the problem. I'm not that familiar with this technology and I am puzzled by the text of the error message which seems to suggest that the device is being modeled as a "jfet". Is that correct or am I missing something? Also what is the specification for source V2? In particular the rise and fall times.
You are not missing anything. It should say MOSFET. I do not know why it says JFET. I have used different rise and fall times from 5ns to 100ns, did not work. In principle it should work with 5ns.
 

Papabravo

Joined Feb 24, 2006
17,242
You are not missing anything. It should say MOSFET. I do not know why it says JFET. I have used different rise and fall times from 5ns to 100ns, did not work. In principle it should work with 5ns.
I'm interested in what is going on here and that is why I am asking questions. I took a look at the onsemi site and noticed that the "ONSEMI_SiCMOSFET_900_ltspice.txt" file was 584K or just over half a megabyte and 4 times bigger than the other two files. Could that be significant in that the LTspice file is the only one which is encrypted?
 

ericgibbs

Joined Jan 29, 2010
14,131
hi beleg,
Note it says myjft1 which suggests a self made version of a jfet which is required to get the Model to function a SiC device.

E
 

Papabravo

Joined Feb 24, 2006
17,242
hi beleg,
Note it says myjft1 which suggests a self made version of a jfet which is required to get the Model to function a SiC device.

E
After the word "instance" in the error message string, is the string "j:u1:1:rd1" which also suggests that the simulator thinks it is dealing with a JFET model. It is not necessarily an error to do this, but it does look "strange". I'm pretty sure there is a reasonable explanation for doing this, but I just don't know what it is at the moment.
 
Top