I am designing an H-bridge and I am wondering about the diodes I should use in my circuit. The first one is the one at the input of the MOSFET driver supply, I guess it must be a schottky that can resist to 12V and a low current, then I would say the same thing for the diode in parallel of the gate resistor, tell me if I'm wrong. On the other hand it is for the diodes at the terminals of MOSFETS where I have a little difficulty to define the characteristics, should I use a schottky? Knowing that it is for a DC motor 24V and 4A I think I will choose a diode 50V and minimum 8A.
Almost none of the diodes are useful:
D5, D6, D10, D11 are in parallel with the MOSFET intrinsic body diodes, which very likely have lower voltage drop, therefore the diodes never carry current. They are also shown as a slow-recovery type (1N4007), which will interfere with the operation of a PWM system, if they should carry significant current (greatly increasing switching loss, and perhaps skewing timing). Since they won't carry current here, that doesn't matter, and they're just dead weight -- a couple pF extra capacitance loading the circuit.
Curiously, there is some rationale for placing them, in this case. The IRLL3303PbF specifies its body diode with maximum 1.3Vf at 4.6A. Now, this is a higher current than the 1N4007 is tested at (1A), but it's highly unlikely that going from 4.6A down to 1A would lower Vf by more than 130mV. In any case, the 1N4007 is rated 1.0V maximum (at 1A), which seems to be an improvement.
However, I suspect the 1.3V figure is just wildly pessimistic, not so much a worst-case statistical value as a guard band to avoid having to, say, test any parts to it, or to have to discard any parts that fail a more meaningful limit. Case in point, notice the typical data in Fig. 7 gives 1.0V at a whopping 14A, and less than 0.8V at 1A. Now, I'm not saying it's impossible for the Vf to be as high as 1.3V (at rated current), I can't know that from these data alone -- but it seems pretty damn unlikely that it ever would be. Typically, diode Vf (for general diodes, I mean) varies within maybe 100mV of the typical value, not several hundred. I would be inclined to disregard this particular parameter.
D3, D4, D8, D9 are in parallel with gate resistors that are already quite small. The IR2104 has a fairly weak output: +130/-270mA. When the driver is working as hard as possible, it only develops a 4.05V drop across the 15 ohm gate resistor. This is more than a diode Vf, so the diodes will conduct -- but this has only very minor impact on turn-off time. Effectively, the gate driver has an internal resistance around 60 ohms, so connecting an additional 15 in series with it -- or a diode (of about 3 ohms at peak current) in parallel with that resistor -- has a tiny effect. Basically the diode acts to reduce turn-off time from 255 to 214ns.
The gate resistors here are more likely serving to dampen possible oscillations as the transistors turn on/off, and for that, some tens of ohms is fine (which we have here).
If different rise/fall speed is desired, a larger gate resistor should be used (more than 50 ohms would have noticeable effect with this driver, or a more powerful driver can be used), and then the parallel diode will have useful effect. At the low load currents in this circuit (at most an ampere or two?), it seems doubtful that any special consideration is needed.
Diodes that are not optional, are D2 and D7! And these are chosen appropriately: an ultrafast diode to supply bootstrap power to the high side gate drivers.
Unrelated circuit critique: the loop between
LO, MOSFET, and
COM / GND should be made very short. Apparently there is a shunt resistor connection in this path, possibly a wired connection even. If this is the case, I would strongly recommend adding bypass capacitors between low-side source and GND, so that high-frequency switching currents don't have to flow through that wired path. Conversely, the supply/drain side should be -- and is (R22 in series, C22 providing bypass after it).
Huh, the 10k pulldowns are also specified as "0.1%". Even 10% would suffice, of course.
I suppose the inductors are quite small for the current and switching speed as well, but maybe
V_MOT is actually 5V or something, and relatively high dI/dt is acceptable, I don't know. It's not that the inverter is very slow or anything anyway, it should be able to run about 100kHz well enough. Larger inductors can be used to lower the switching frequency, improving efficiency.
 Neat fact: these are shown in two places already. The arrow in the middle of the symbol means the substrate connection, which is tied to source, thus hard-wiring a PN diode from source to drain. (The three thick lines in a row, in the middle of the symbol, represent drain, channel/substrate, and source elements. They would be shown joined (a single line) if depletion mode; this is an enhancement mode MOSFET.) For some reason, this fact was rarely taught even 2-3 decades ago, leading many to draw an explicit diode from source to drain as reminder. Curiously, this symbology has also been forgotten, leading users still to place extra diodes in parallel...
D3, D4, D8, D9 are not really useful, the gate resistor is to reduce the rush current caused by the MOSFET gate into the IR2104 and these diodes allows to have a faster gate capacitance discharge when the gate is pulled down, probably for faster switching on the negative edge.
The type is just normal diodes, with a reverse voltage that is higher than the gate drive voltage. They can be quite small and handle 1-2A and being fast-switching type.
D5, D6, D10, D11 are flywheel diode that avoid the MOSFETs to be destroyed by inductive loads.
These diodes are also normal diodes, but needs to be able to handle higher current. You want fast switching diodes. For the current, it really depends on the inductance of the load, but if you want to be safe, ~20A range should be safe as long as you are not driving a massive motor. Their reverse voltage should be higher than your rail voltage (V_Mot).
D2, D7 are for the IR2104 to generate the high side gate drive voltage, seems like a capacitor charge-pump design.
The resistor R25, R16, R18, R19 are used to "allow" the fast switching of power MOS-FET. But the "doing" is not the same when switching ON or switching OFF. This is the reason of these diodes D3/4/8/9, some times with another serial resistance. If time switching is not important, these can be omitted. For some reasons of symetry, there are also used when you are paralleling devices. You can also add a capacitor between gate and source if you want "identical MosFet parts". This is ok when they are "part" of the same "lot" of components if ... packaged together.