Tahmid's H Bridge - Does it work?

[lfreddi]

Hi,
In the following days i'm planning to build an H bridge to help a friend that needs to control speed on a DC motor of 12V and around 6A nominal.
I have been searching information on the past weeks on how these bridges operate with mosfets and IGBT's, but I want to ask for an expert if this circuit will work for my particular case of study.

I took the Tahmid's blog H bridge using the IR2110 mosfet driver (Tahmid's blog: Debugging the Bridge: Tips for Successfully Designing (Full / Half) Bridge Circuits), and I put fast optoisolators to protect the Atmega328p (in the image there is an atmega168, but I will use and Atmega328p), and two different LM7805 in order to give full isolation between control circuit (the microcontroller) and "command circuit" (the IR2110).

I have the following questions:

1. Is it correct to operate the H bridge using only 2 pins? I mean having the following operation:
   pin 3 OFF, and pin 5 OFF, then the motor has no supply on its terminals.
   pin 3 ON, and pin 5 OFF, then the motor moves clockwise (for example).
   pin 3 OFF, and pin 5 ON, then the motor moves anti-clockwise.
   pin 3 ON, and pin 5 ON is Forbidden state.

2. The IR2110 manages the timing of the bridge mosfets in its own circuit or should I worry about the possibility of cross-conduction in the bridge if I make this circuit that is controlled only by two pins?

3. Is it correct to operate this circuit with PWM on both mosfets at a time? I mean having the same PWM applied on Q1 and Q4 for example (having Q2 and Q3 off). Or is it better to have Q1 always high and apply PWM only on pin Q4?

Thank you for your time
Lucas

[MarkT]

Firstly that circuit has 1k resistors from gate to source on the MOSFETs - lose them, they severely load
the gate bootstrap circuitry. The 22uF bolstering the bootstrap circuit can also go, its a consequence of
the erroneous resistors.

Secondly note this kind of circuit cannot drive permanently high, it relies on the PWM waveform for
the bootstrapped high side gate supply, meaning you must use PWM with a duty cycle no
higher than about 95%, otherwise the high side gate supply will droop.

Basically follow the circuit in the manufacturer's datasheet, use PWM of 1kHz or higher and duty
cycle from 0% to 95%.

You'd be far better off using a high-low gate driver chip that has shoot-through prevention and a
single input logic signal, such as the IRS2004. Shoot-through prevention in the driver is a real boon,
insist on it. (In that chip's datasheet its called "cross conduction preventtion")

There are chips that support an entire H-bridge (HIP4081A for instance), and a 3-phase bridge
(which is 1.5 H-bridges) like the FAN7388, which can reduce parts count. The HIP4081A has
an independent charge pump so can drive the high MOSFET permanently high (so long as you
don't put a resistor between gate and source)

And decouple the gate driver chip well with ceramic caps (10uF is good, 1uF is minimum), from
a stable spike-free 12V supply. Put spikes on that 12V supply and BANG! bye bye circuit. This
means the 12V gate driver chip supply cannot be the same supply as the H-bridge itself - been
there, done that, never again.

The nice thing about this style of H-bridge is that the H-bridge supply is completely independent
of the gate driver chip - either can be powered without the other with no issues.

[lfreddi]

Thank you for your answer MarkT.

Firstly that circuit has 1k resistors from gate to source on the MOSFETs - lose them, they severely load
the gate bootstrap circuitry. The 22uF bolstering the bootstrap circuit can also go, its a consequence of
the erroneous resistors.

Ok, good thing to know, I will do it that way without the 1K resistors and the 22uF capacitors

Basically follow the circuit in the manufacturer's datasheet, use PWM of 1kHz or higher and duty
cycle from 0% to 95%.

I was planning to implement a 32kHz PWM from the atmega328p (running on internal 8Mhz oscillator), this should be okay, right?

You'd be far better off using a high-low gate driver chip that has shoot-through prevention and a
single input logic signal, such as the IRS2004. Shoot-through prevention in the driver is a real boon,
insist on it. (In that chip's datasheet its called "cross conduction preventtion")

In the IR2110 datasheet it says the following:
"The output drivers feature a high pulse current buffer stage designed for minimum driver cross-conduction".
I looked in the IRS2004 datasheet and it says that exact words so I suppose they both have the same logic circuit inside.

And decouple the gate driver chip well with ceramic caps (10uF is good, 1uF is minimum), from
a stable spike-free 12V supply. Put spikes on that 12V supply and BANG! bye bye circuit. This
means the 12V gate driver chip supply cannot be the same supply as the H-bridge itself - been
there, done that, never again.

Do you mean that I can't use an LM7805 that is fed from the 12V supply from the H bridge to feed the gate drivers? And if I put a really big capacitor on the input of the LM7805, something like 2200uF and a 100nF to decouple the high frequency, that should be okay?

The nice thing about this style of H-bridge is that the H-bridge supply is completely independent
of the gate driver chip - either can be powered without the other with no issues.

My idea was to isolate in the most extreme way I can the different "sectors" of the circuit, I mean the microcontroller, gate drivers, and bridge. So that a fault or short circuit won't affect everything.

I attach the circuit updated with your comments

[MarkT]

lfreddi:
Thank you for your answer MarkT.

Ok, good thing to know, I will do it that way without the 1K resistors and the 22uF capacitors

I was planning to implement a 32kHz PWM from the atmega328p (running on internal 8Mhz oscillator), this should be okay, right?

Very fast, 16 or 20kHz might be less demanding of switching losses, and allow a higher max
duty-cycle percentage.

In the IR2110 datasheet it says the following:
"The output drivers feature a high pulse current buffer stage designed for minimum driver cross-conduction".
I looked in the IRS2004 datasheet and it says that exact words so I suppose they both have the same logic circuit inside.

No, the IRS2004 has shoot through prevention, the IRF2110 does not. Shoot-through prevention is nothing to do with the current buffer output stages of the chip, its about coordinating between high and low drivers to
avoid them both being on simultaneously.

Do you mean that I can't use an LM7805 that is fed from the 12V supply from the H bridge to feed the gate drivers? And if I put a really big capacitor on the input of the LM7805, something like 2200uF and a 100nF to decouple the high frequency, that should be okay?

High low gate drivers require a clean, decoupled 12V supply. 5V will engage undervoltage lock-out and
nothing will happen.

The decoupling needs to be something like 1 to 10uF ceramic per driver chip, electrolytics do nothing useful at these speeds. 2200uF is pointless for fast logic decoupling anyway, everything's over in a few
tens of nanoseconds. That's the sort of value for the bridge power rail itself.

The bootstrap caps should be about 100nF ceramic for those MOSFETs, can be a bit more, don't go
overboard with them. The 12V decoupling must be at least 10 times as large as the bootstrap caps.

My idea was to isolate in the most extreme way I can the different "sectors" of the circuit, I mean the microcontroller, gate drivers, and bridge. So that a fault or short circuit won't affect everything.

I attach the circuit updated with your comments

Isolation is a very good idea. I see you have fast optocouplers (a requirement), but they will pull high
during the reset time which is very bad news, you are going to have to rethink those devices, all MOSFETs
should be OFF during powerup/reset time.

I strongly suggest you change to the IRS2004 or similar, shoot-through can be instant destruction in a
powerful bridge like this.

One way to protect the logic side from the gate driver is a series resistor on the gate driver pin, and
on the Arduino output add a schottky diode to the Arduino 5V rail. Or find active high opto couplers.

You can lose the resistors and diodes inline to the gates, they are pointless in this kind of circuit, the
gate driver's output resistance is plenty.

Much better to add a 15V zener across gate/source of every MOSFET (right at the device's leads),
such zeners are very useful to protect against spikes on the gate electrode. And yes, 15V, not 12V, they
must be very high impedance at 12V or they will droop the bootstrap cap too fast (less of an
issue with faster PWM though)

[lfreddi]

Very fast, 16 or 20kHz might be less demanding of switching losses, and allow a higher max
duty-cycle percentage.

I looked the timing possibilities of the atmega328p and it can provide a PWM of: 62.5kHz, 31.3kHz, 7.8kHz or 3.9kHz. I assume the better from these and according to your comment is 7.8kHz.

High low gate drivers require a clean, decoupled 12V supply. 5V will engage undervoltage lock-out and
nothing will happen.

The decoupling needs to be something like 1 to 10uF ceramic per driver chip, electrolytics do nothing useful at these speeds. 2200uF is pointless for fast logic decoupling anyway, everything's over in a few
tens of nanoseconds. That's the sort of value for the bridge power rail itself.

The bootstrap caps should be about 100nF ceramic for those MOSFETs, can be a bit more, don't go
overboard with them. The 12V decoupling must be at least 10 times as large as the bootstrap caps.

[...]

I strongly suggest you change to the IRS2004 or similar, shoot-through can be instant destruction in a
powerful bridge like this.

[...]

You can lose the resistors and diodes inline to the gates, they are pointless in this kind of circuit, the
gate driver's output resistance is plenty.

Much better to add a 15V zener across gate/source of every MOSFET (right at the device's leads),
such zeners are very useful to protect against spikes on the gate electrode. And yes, 15V, not 12V, they
must be very high impedance at 12V or they will droop the bootstrap cap too fast (less of an
issue with faster PWM though)

I changed the schematic following all these advices, I put a 4N25 optocoupler that enables the +12V for the logic side of the gate drivers. The IR2104 (the only driver similar to the IRS2004 that I can obtain in my country) has pull down resistors on the input IN, so it shouldn't be a problem to have all the 12V logic circuitry "floating" during the reset, because it should be pulled down by these resistors inside the IR2104.

I also put connections going from the motor 12V to a step up to boost it to around 15V, in that way I can use a LM7812 to obtain the regulated 12V for the drivers, my idea was to use a MT3608 module (ICStation MT3608 DC to DC Step Up Converter Module | Boost Power Supply Module).

Please correct me if I'm wrong, but from what I have read about the bootstrap the way this circuit should operate is:

• When the circuit is in strandby the bottom mosfets (Q2 and Q4) are ON, this enables the charge of both bootstrap capacitors C10 and C12.
• When the motor should be moved clockwise (for example) a PWM signal is applied on the IN of the upper IR2104, turning ON the Q1 mosfet during the duty of the PWM signal, and during the OFF part of the PWM signal the bootstrap capacitor C10 is charged again. In this way, during the duty of the PWM a current path is going through the motor from Q1 to Q4.

[lfreddi]

Can anyone confirm if this way of operation is correct for an H bridge? Will I have a problem if I do it that way?

Please correct me if I'm wrong, but from what I have read about the bootstrap the way this circuit should operate is:

• When the circuit is in strandby the bottom mosfets (Q2 and Q4) are ON, this enables the charge of both bootstrap capacitors C10 and C12.
• When the motor should be moved clockwise (for example) a PWM signal is applied on the IN of the upper IR2104, turning ON the Q1 mosfet during the duty of the PWM signal, and during the OFF part of the PWM signal the bootstrap capacitor C10 is charged again. In this way, during the duty of the PWM a current path is going through the motor from Q1 to Q4.

Thank you

[MarkT]

You cannot power the output section of the 6N137 with 12V, it will fry, its a 5V part.

I think I understand the new opto coupler - disables the pullups for the 6N137's during reset?

Doesn the 4N35 have enough gain to do this (its trying to drive 1k1 at 12V (12mA) on the output with
only 10mA to its input...)

I guess it will work though, the gate drivers logic inputs don't need the full 12V or anything like
to operate.

[lfreddi]

You cannot power the output section of the 6N137 with 12V, it will fry, its a 5V part.

You mean I have to feed 5V on the 6N137 pin 7? Will the IR2104 work properly with 5V logic?
If I put 5V on the output of the 6N137, then it shouldn't be neccesary to add the 4N25. With a BC327 with a diode on its base and commanded by an atmega328 pin I can do this habilitation from the 5V supply into the pin 7 of both 6N137, what do you think?

I think I understand the new opto coupler - disables the pullups for the 6N137's during reset?

Thats right, in that way there isn't a High output in the power stage during reset