I'm guessing you're trying to drive some AC motor that was originally designed to run on 50 Hz AC.
I'm assuming you're using two IRS2113 chips, the "left" chip controlling the two FETs connected to the "left" wire going to the motor, and the "right" chip controlling the two FETs connected to the other wire going to the motor. You probably want to check out the IRS2113 datasheet and the FET datasheet(s).
It is your responsibility to never, ever drive both HIN and LIN of the left IRS2113 chip HI at the same time -- the "self-destruct" state.
As JustJeff points out, if both HIN and LIN of the left FET driver chip are ever high at the same time, the FETs connected to it will, in effect, short the rails ("shoot through", "fuse test", etc.), and much unwanted excitement will follow.
(Never, ever drive both HIN and LIN of the right IRS2113 chip HI at the same time, either).
A few people don't use a translation circuit at all; they directly connect the left HIN and LIN and the right HIN and LIN directly to 4 independent output pins of the microcontroller, and hope that the software doesn't have any bugs that would set the FETs to the self-destruct state.
I prefer using a digital logic gate, such as the external NOT gate you mention, to enforce this "avoid the self-destruct state" condition in the motor driver hardware, rather than software.
It looks to me that the IRS2113 (like many other "high-side drivers") uses a boost circuit that can't really produce a 100% "ON" in the high-side transistor; it assumes that software will periodically turn that transistor OFF, so your software might need to limit the maximum PWM duty cycle to a maximum of something like 255 ticks HI + 1 tick LO.
bipolar
The simplest (in hardware) approach is to drive the entire H bridge/motor system in one of two basic states of the H bridge: either
State 0:
The microcontroller PWM output is LOW.
This drives the left LIN and the right HIN, turning the corresponding FETs off.
That output also drives a NOT gate which drives HI the left HIN and the right LIN,
turning those corresponding FETs ON.
This in effect connects the motor's left wire to HI and the motor's right wire to LO.
If things stay in this state long enough, the motor turns in the direction I call "forward".
State 1:
The microcontroller PWM output is HI.
This drives the left LIN and the right HIN, turning the corresponding FETs ON.
...
This in effect connects the motor's left wire to LO and the motor's right wire to HI.
If things stay in this state long enough, the motor turns in the direction I call "reverse".
With a PWM running at reasonable speeds, setting the duty cycle to 50% HI, 50% LO, you end up with the motor stationary (possibly humming a little).
This 2-state (bipolar) system is sometimes called "Locked-Antiphase". a b
trilevel
Trilevel, aka "Sign-Magnitude Drive"
Adding a third state makes things a little more complicated and difficult to debug, but it typically improves (reduces) harmonics and makes the system more power efficient.
State 3:
Some people use a third state turns off all the FETs, letting the motor freewheel.
(If you manage to turn off any 3 of the 4 transistors and leave the other one ON, that also works just as well as a freewheel state).
If things stay in this state long enough,
the motor "stops".
(In some cases, external forces spin the motor in one direction; the direction is not under the control of the electronics).
Alternate State 3:
Some people use a third state that turns on both the lower FETs, turning off both upper FETs, which slows down the motor ("brake").
(Others turn on both the upper FETs, turning off both lower FETs; that also works just as well as a "brake" state).
If things stay in this state long enough,
the motor stops.
(Even when external forces push the motor in one direction,
this "brake" generally causes the motor to stop).
There are a variety of ways to implement digital logic between the microcontroller and the FET driver that supports trilevel drive, but also avoids the self-destructive states.
(Some motor driver chips such as the IXYS IXDN404 include this anti-shoot-through direction/PWM translation circuit, but the IRS2113 does not).
Such translation circuits generally require a "direction" and a "PWM" line from the microcontroller.
The software sets the "direction" to "forward", and then adjusts the duty cycle of the PWM to control the speed from "idle" to "full speed forward".
(The two states of the PWM, in effect, are translated to the two states "State 0 forward" and "State 3 idle").
Much later, the software sets the "direction" to "reverse", and then adjusts the duty cycle of the PWM to control the speed from "idle" to "full speed reverse".
(The two states of the PWM, in effect, are translated to the two states "State 0 forward" and "State 3 idle").
Translation circuits that support tristate motor control:
a b c d e
You might think that these circuits would have 100% LO, 0% HI PWM duty cycle give "idle" (in either direction), and 0% LO, 100% HI give "full speed" (in whatever direction the "direction" line indicates).
But lots of people use a direction/PWM translation circuit that does something more confusing, and then fix it up in software.
Perhaps the simplest such circuit translation uses the PWM to drive the left LIN and a NOT gate that in turn drives the left HIN.
Another general-purpose output on the microcontroller drives the "direction" signal, which drives right LIN and a NOT gate that in turn drives the right HIN.
PWM frequency is 32 kHz so that's a period of about 32 us and half that time the transistor will be dumping energy into charging up those capacitor. It'll be doing it totally inefficiently too so maybe half that energy again is wasted in transistor heat.
Capacitor acquired energy is \$\dfrac{CV^2}{2} = \dfrac{2.2 \times 10^{-6}\times 144 }{2}\$ = 0.158mJ
BUT because charging the cap is done totally inefficiently, at a guess, the transistor will throw away just as much energy. It does this 32,000 times per second so that's a power of about 5 watts per transistor.
Now I may have over-egged this a bit by assuming that each 2.2uF would become totally discharged in one cycle of PWM but I'd wager that each transistor is doing over a watt just driving these capacitors back to full charge (32,000 times per second). Get rid of the caps and if you still have problems with noise then come back and tell us.
The lower the duty cycle the worse this will be because there will be a longer time for these caps to discharge and hence inefficiently re-acquire more of their 0.158mJ when the transistor switches back on again.
Best Answer
But the motor has to change from full speed in one to another direction or form zero to full speed, that's why we use motors. You can limit the current and consequently the acceleration. the H bridge shall output such voltage that the current remains always in the margins. The best you can do is a PI regulator, the setpoint is rated motor current and the feedback is measured current. When you will start, the regulator will output a certain duty cycle setpoint and it will be increased by regulator output when the motor will gain speed (back EMF compensation). $$M=J\alpha + M_{load}$$ $$M \propto I_a$$ $$U_a=I_a*R_{winding}+U_{backEMF}$$ $$U_{backEMF}=k[V/krpm]*n[rpm]$$