It would be best to use a scope on your original circuit and find out what is going on, if you have access to one.

 

If you post your asc file we can have a play too.

 

I'll try to scope the actual circuit tomorrow at some point. In the meantime, here is the .asc file for my simulation. The actual board has been built using identical components to those shown. I have tried both 1uF electrolytic and 0.1uF ceramics for the bootstrap capacitors.

To be honest, my main motivation for using this particular driver/FET combination was that they have built in LT Spice models

 

I have tried both 1uF electrolytic and 0.1uF ceramics for the bootstrap capacitors.

I think I ended up using 0.2 uF polyester for the caps... also, I ran into some trouble because I had miswired the thing when I drew the pcb ... so I recommend you double and triple check everything you've done.

 

The way the simulation looks it should work jsut fine with or without turning the bottom fet on.

 

The way the simulation looks it should work jsut fine with or without turning the bottom fet on.

It worked fine on LT Spice which is why I pressed ahead with the PCB. I'll investigate the actual board with the scope today at some point.

 

The way the simulation looks it should work jsut fine with or without turning the bottom fet on.

Agreed, but simulation confirms it should be more efficient if the bottom FET is turned on (when the top FET is off, of course) because the motor's circulating inductive current then causes a smaller voltage drop across the drain-source path than it would if the only path were the body diode of the bottom FET.
This can be seen if R1 in the asc file posted is replaced by an inductor (~10mH).

 

Agreed, but simulation confirms it should be more efficient if the bottom FET is turned on (when the top FET is off, of course) because the motor's circulating inductive current then causes a smaller voltage drop across the drain-source path than it would if the only path were the body diode of the bottom FET.
This can be seen if R1 in the asc file posted is replaced by an inductor (~10mH).

Yes that is true and why it is allways done like that.

 

I set up my test circuit again this evening. Two series connected lead acid batteries to provide 12 and 24V and an MSP430 development board to generate my PWM signals.

I began by generating a 20kHz signal from one of the PWM channels of the dev board, around 40% duty cycle as shown below. The top trace is the PWM signal applied to the top left MOSFET of the bridge with the bottom right MOSFET gate drive connected to 3.3V, whilst the bottom trace is the load voltage. As can be seen, the circuit works as intended.

Also shown below is the voltage at the top of the bootstrap capacitor with respect to ground.


Finally, the voltage at the gate of the top side MOSFET wrt ground is shown below, indicating that the bootstrap circuit does indeed provide circa 36 V to drive the high side switch.



I'm at a loss to explain why it didn't work before, perhaps as the Arduino was not doing what I thought it was doing as I did not have a scope to verify its output. The bootstrap capacitor would appear to be recharging even without switching on the MOSFET directly below it. I presume this is because even with the back EMF of the motor there is still a volt drop from V+ to the high side of the motor allowing the capacitor to charge?

 

As can be seen, the circuit works as intended.

Where are the images?

 

Where are the images?

Apologies, the images were too large for the post. I'm putting them into a PDF as we speak....

 

The images referred to in my previous post are included in the attached PDF. Please excuse the shaky camera work!

 

The images referred to in my previous post are included in the attached PDF. Please excuse the shaky camera work!

You've got quite an overshoot there, I recommend using TVS diodes, or at least back paralleled zeners.

 

I'm at a loss to explain why it didn't work before, perhaps as the Arduino was not doing what I thought it was doing as I did not have a scope to verify its output. The bootstrap capacitor would appear to be recharging even without switching on the MOSFET directly below it. I presume this is because even with the back EMF of the motor there is still a volt drop from V+ to the high side of the motor allowing the capacitor to charge?

My post #11 about the back EMF was wrong. The clarification in post #13 is more appropriate, the back EMF doesn't come into play until the energy of the inductance and that circulating current is depleted, which will not happen with PWM at normal frequencies, and the diode in the mosfet will recharge the bootstrap cap before that will start to happen.
(actually the back EMF will influence the di/dt of the circulating current and cause the eneregy to be depleted sooner than with just simple inductance, but that is beside the point)

 

You've got quite an overshoot there, I recommend using TVS diodes, or at least back paralleled zeners.

That is a measurement artifact, not a real overshoot. Look at those kilometers of inductance the scope ground lead goes through before it gets to the actual circuit. For large currents it is good to use very short ground lead connected right next to the measured point, or better still just the ground spring contact just a few millimeters apart from the tip of the probe if possible.

 

That is a measurement artifact, not a real overshoot. Look at those kilometers of inductance the scope ground lead goes through before it gets to the actual circuit. For large currents it is good to use very short ground lead connected right next to the measured point, or better still just the ground spring contact just a few millimeters apart from the tip of the probe if possible.

When I started testing the circuit I suddenly became acutely aware that I hadn't included any test points for my scope probes or even a spare ground connection, hence the wire connecting the scope probe to the battery negative.

I ran the circuit for quite some time and there was no apparent build up of heat in the MOSFETs despite their relatively high Ron compared to other similar MOSFETs. I suppose this is partly due to my low switching frequency and the fact that the gate driver is charging the gates very rapidly.

I'm a little bit concerned about how much charge the bypass capacitor on the 24V rail can store even long after powering down the circuit. Gave me quite a fright when one of my croc clips accidentally shorted it! I think I might solder a bleed resistor across it to make sure it discharges when switched off.

 

I ran the circuit for quite some time and there was no apparent build up of heat in the MOSFETs despite their relatively high Ron compared to other similar MOSFETs. I suppose this is partly due to my low switching frequency and the fact that the gate driver is charging the gates very rapidly.

Heating through Rdson is independent of frequency, so each mosfet should not get more dissipation than if it was running full on. Heating through slow drivers on the other hand depends mainly on the frequency, as the fets spend more and more time in the transition and less in conduction as frequency increases.

 

Add a temporary ground point on the PCB for the CRO test point. The PWM drive into the board shows overshoot too and that really should be clean. Is that a ground plane on top? If so, connect a test point to that.
It is worth while to get clean real signals.
The board looks pretty neat.