MosFet Selection and protection

[pikstart]

Hi!

I am searching for the appropriate MosFet to build an H-Bridge for 40Volts at 15Amps max current. Will drive a coil of about 18mH (a very big 60cm coil of 2mm wire). The frequency that will feed the coil is from 1 to 50Hz.

Unfortunately all MosFets I found have not the SOA that is needed for such a slow signal and such a high current.

Also what is the best protection of the MosFets from the coil kick? TVS, diode or zeners? And how is this calculated? Have not found calculations anywhere, just the need to add such protection. For a small project any power schotky might be OK, but at 40Volts/15Amps?

Thank you,
Ioannis

 

[danadak]

Not sure if this might be of interest -

https://www.st.com/resource/en/application_note/an4892-vipower-m05-hbridges--hardware-design-guide-stmicroelectronics.pdf

Regards, Dana.

 

[pikstart]

Hi Dana.
Thanks for the link. Lots of good info but not on the protection of discrete MosFets. Rather a presentation of the ICs that ST carries as a compete chip solution.

Ioannis

 

[danadak]

Some snubber design info, attached.

TVS you can google that as well.

 

[Diver300]

At the bottom end of the 1 - 50 Hz frequency range, 15 A will be achieved with a much smaller voltage than 40 V.

At the top end of the frequency range, 40 V won't give as much as 15 A into an 18 mH coil.

How are you intending to control the current?

 

[pikstart]

Hi Diver300,

you are right about that. The current cannot be the same in every voltage set (1-40volts) or any frequency (1-50Hz). Also I do know that the coil will be saturated in many combinations, but this is what the requirements need.

To protect from overcurrent I will add a shunt resistor to shut off the MosFet drivers.

My concern is if there is a fullproof way to select the proper components for such working conditions.

In a smaller projects I just selected the diodes in the H-Bridge as Vf> V/R and Vrev-blocking>2*Vsupply. And power dissipated to the diode (Vf*i) < diodes rated power.

But here I have a beast coil! Want to be sure.

Ioannis

 

[virgiox]

you are right about that. The current cannot be the same in every voltage set (1-40volts) or any frequency (1-50Hz). Also I do know that the coil will be saturated in many combinations, but this is what the requirements need.

To protect from overcurrent I will add a shunt resistor to shut off the MosFet drivers.

Hi Pik,

I guess you need to drive it as PWM, with a frequency much higher than 1-50 Hz. Even with 1 kHz like below, you can create a 1 Hz drive signal with a ripple of +- 500 mA @ 15 A. It would have been 50 mA @ 10 kHz.
By doing so you can use any stock mosfet as they are working in a very safe area.

 

[pikstart]

virgiox
thanks for the idea. Never thought of such an approach.

I am very newbie on LTSpice. Could you explain your sources setup of V1 to V4 please?

Regarding the MosFets protection, for such conditions a SS36 would be OK?

Ioannis

 

[danadak]

The drive has to include dead band timing to insure you do not turn on both fets
in a vertical leg, eg. shorting power supply.

Modern micros have this capability in them.

Example of that timing :

Here you can see PH1 and PH2 of the PWM in this processor never overlap, and
dead band timing user controlled in its setup.

Of course one can generate this with CPLD or logic, But timing complications to
deal with why re-invent the wheel.

 

[virgiox]

Could you explain your sources setup of V1 to V4 please?

Regarding the MosFets protection, for such conditions a SS36 would be OK?

Hi Pik,
V3 and V4 are voltage generators emulating a square (rectangular) wave with a 70% duty cycle. It stays on at 10 V for 0.7 ms over a 1 ms period. After 500 periods ( 500 ms) V3 and V4 go to 0 V while V1 and V2 start, like V3 and V4 did before. So you have a pulsed current flowing from left to right into L2 for the first half second and from right to left for the second one.

About mosfet protection, it depends on the switching frequency you are going to use. If you stick to few kHz I guess body diodes should be enough.

 

[virgiox]

pikstart I attached the ltspice schematic, should you have time and will to play with it.

 

[pikstart]

Thank you very much! I will experiment with that.

Ioannis

 

[pikstart]

Tried the simulation file. Why after 1 sec the frequency is increased? The 4 voltage sources do not seem to do anything past 1 sec.

Ioannis

 

[virgiox]

Tried the simulation file. Why after 1 sec the frequency is increased? The 4 voltage sources do not seem to do anything past 1 sec.

Ioannis

After 1 second all 4 generators go to zero. The idea was to test 2 half waves, one period with some easy to build generators. You can use BV (behavioural generator) to create more complex waveforms, i.e. a 1 s period square wave multiplied by a 1 kHz 70% duty cycle square wave, but usually simulation become awfully slow and I'm too lazy to play with simulation parameters to speed it up.

 

[pikstart]

The four generators indeed go to zero, but after one second theres is a high freq sinusoidal signal on the output for the rest of 0,1 second.

I put a 100ohms-100nF snubber across one of the Mosfet and that signal stopped. Seems like a ringing or something.

Well, the Simulator seems like a magic tool!

 

[virgiox]

I understand what you mean. When both mosfets are turned off, the voltage you read with a probe it's the voltage you read on a voltage divider made with high impedance components: not so meaningful. I guess that even a 1 MOhm resistor to gnd would fit.