Design recomendations for Mosfet H-Bridge

Question

I'm trying to build my first Mosfet H Bridge, the goal is to build a bridge with Vin = 5 - 12 V, Imax = 30 A, managed by microcontroller.

I have read about functional principle and Mosfet parameters, but I have some doubts about the implementation.

I have seen that many people use Half bridge drivers like in this answer but I have not seen that galvanic isolation is implemented in any design, furthermore in most of these chips that I have seen just have one Ground or Reference Point, with exception of IRS21844, so I supposed that these IC's don't have isolation.

I know in advance of the IGBT Gate Drive Optocoupler existence. So my questions are:

Is enough safe to use only chips like IRS21844 to drive the bridge and if not which optocoupler you recommend?.

What security precautions should I take into account in addition to TVS diodes, Pull Down resistors and Zener diodes in the gates?

Which is the best method to control the bridge, Optocoupler + Anti shoot through circuit + float power supply, Half bridge driver + optocouplers?

Which is the best way to get a float power supply?

Excuse me for so many questions, but is the first time I work with Mosfets and my intention is to learn the right way to build H bridge with these :).

Answer 1

I would suggest starting with a smaller H-bridge project first. The control of MOSFETs seems simple, but there are some subtleties which can easily bite you if you are not careful. When designing a smaller H-bridge project, it's not too hard to protect a circuit in such fashion that even if something goes horribly wrong it won't do much damage (and may very well not do any, though whatever caused the problem will need to be fixed in any case). This is useful not only from the standpoint of reducing the amount of time spent replacing blown components, but also from the standpoint of being able to figure out what happened. If something goes wrong in a 12V 30A H-bridge controller, it's likely to cause enough destruction quickly enough that it will be hard to analyze the pile of cinders and determine what failed first.

Beyond that, the reason companies make H-bridge controllers is that other companies buy them, and the reason that other companies buy them is that a surprising number of discrete components are required to safely and optimally control a MOSFET. If one is trying to control a MOSFET whose Absolute Maximum Vgs is 15 volts, and an inductive transient causes Vgs to reach 30 volts even for a microsecond, the part may easily be destroyed (the voltage required to destroy the part would most likely be around 20 volts, but one shouldn't rely upon it being above 15.01). Keeping Vgs below 15 volts may not sound hard, but factors like parasitic gate-drain capacitance can lead to some unpleasant surprises.

Answer 2

Half bridge drivers like the IRS21844 or IR2111 or IR2104 were originally intended for use in off-line switchers. In that application bridge voltages of ~300V are typical. So, that's why you see max voltage ratings for those parts of 600V. These parts are safe to use in low voltage applications, like 15V to 30V without isolation. But, they also have under voltage lockouts so that they will not work for Vcc voltages of less than about 10V (I've not seen any that work with Vcc under about 8V). So, would not be useful if you really need to operate as low as 5V.

For the low voltage you are talking about, there is no safety reason to isolate.

The advantages of integrated half bridge drivers is that they take care of things like phased switching of the two FETs (to prevent shoot through or cross conduction), and pull down resistors (so they don't float high). They are meant to be driven from a low current source like a micro-controller.

Your biggest problem in addition to the low voltage (5V) that you are going to face is the high current (30A). One way to simplify an H bridge is to use P channel FETs for the top switches ... then you don't need to drive the gate above the supply rail. But P channel FETs of the same voltage and die size of an N channel will have about 3 times the \$R_{\text{ds}-\text{on}}\$ as the N channel, which makes conduction loss a problem.