Thermal for >1 FET connected in parallel

General
,
1

Hi,

I need to connect 3 FET in parallel in the same Heat Sink and I would like to know if there is any way to easily configure this through a parameter or I have to connect three equal FET in the schematic?

Thanks in advance.

2

Hi @alefarri

Paralleling three discrete MOSFETs is generally not the best approach. In particular, if all three devices, or even just one of them, use an ideal On-resistance Ron, the entire current will flow through a single device.

The demo model “Three-Phase T-Type Inverter” demonstrates a recommended way to handle parallel devices. In the mask of the corresponding subsystem, there is a parameter called Number of parallel devices.

In the initialization section of this subsystem (Ctrl + M → Initialization), you can find the code that scales the thermal model according to this parameter:

if ~isnan(num_par_device) && ((floor(num_par_device) ~= num_par_device) || (num_par_device < 1))
  error('"Number of parallel devices" must be an integer and greater than zero.');
end


if(~isempty(therm_mosfet.Von))
	therm_mosfet.Von.i = therm_mosfet.Von.i*num_par_device;
end
if(~isempty(therm_mosfet.Eon))
	therm_mosfet.Eon.i = therm_mosfet.Eon.i*num_par_device;
	therm_mosfet.Eon.E = therm_mosfet.Eon.E*num_par_device;
end
if(~isempty(therm_mosfet.Eoff))
	therm_mosfet.Eoff.i = therm_mosfet.Eoff.i*num_par_device;
	therm_mosfet.Eoff.E = therm_mosfet.Eoff.E*num_par_device;
end
if(~isempty(therm_body_diode.Von))
	therm_body_diode.Von.i = therm_body_diode.Von.i*num_par_device;
end
if(~isempty(therm_mosfet.CauerChain)) 
	therm_mosfet.CauerChain.R = therm_mosfet.CauerChain.R/(num_par_device);
	therm_mosfet.CauerChain.C = therm_mosfet.CauerChain.C*num_par_device;
end
if(~isempty(therm_body_diode.CauerChain)) 
	therm_body_diode.CauerChain.R = therm_body_diode.CauerChain.R/(num_par_device);
	therm_body_diode.CauerChain.C = therm_body_diode.CauerChain.C*num_par_device;
end

if( ~isnan(num_par_device))
	ron_mosfet = ron_mosfet/(num_par_device);
	ron_body_diode = ron_mosfet/(num_par_device);
end

if( ~isnan(rth_ch) && ~isnan(num_par_device))
	rth_ch = rth_ch/(num_par_device);
end

From a simulation performance point of view, this approach is quite efficient: regardless of the selected number of parallel devices, only a single scaled device is simulated. The underlying assumption, however, is ideal current sharing between the parallel devices.

If you are using an XML file provided by a semiconductor manufacturer, it is also worth checking whether they offer a dedicated PLECS library component, as this functionality may already be implemented there.

Do you already have a specific manufacturer in mind?

3

Hi again and thanks for you response.

I see, but I just tried to simulate two mosfet in parallel with Ron=0 and the current is splitted in two.

On the other hand, I studied a bit the model you mention, but my path is to have a more simplified design. I am running the model from a python script setting all the parameters with .set method. Currently studying three Infineon mosfet with its own Thermal Description saved in the Tehrmal Description Search Parth, I though it would be possible to modify the Thermal Description in a way such as:

plecs.set(model_name, ‘ThermalDescription’, ‘FF1000UXTR23T2M1’).

Then, with Manual Switches, select how many mosfet in parallel are used.

Thanks in advance.

4

This can be observed in simple circuits. Depending on the surrounding network and the numerical integration, the resulting current sharing may however differ. If you choose this modeling approach, it is therefore important to verify that the current sharing is represented as intended.

From your described application, using a single device and adjusting the thermal description together with the number of parallel devices appears to be the most suitable approach. Alternatively, you can build a Configurable Subsystem that models two, three, or more devices in parallel and enables the required configuration.

If you want to automate the analysis and keep it more general with respect to the number of parallel switches, the proposed workflow is recommended.