I am trying to make a simple H-bridge controller for a brushed DC motor and I'm not having a lot of luck getting it to work.

I am using N-Channel Mosfets (IRF540S). Datasheet Here

I am driving them (or trying to) with MCP-14700's. Which are high/low side combination drivers with built in bootstrap for the high side. Datasheet Here

My PWM is supplied with a PIC18F.

The entire contraption can be see here.
And my schematic is here.

The problem:
When I view the waveform going INTO the MCP-14700, everything appears to be correct. I get a nice square wave on Q1 and Q4, and a flat nothing on Q2 and Q3 (for going forward).
(Waveforms I've tried are anything from 500hz to 10khz with 50% DC)

However, the output pins of the MCP-14700's behave in odd ways. (Probe reading across Vgate-source). Typically the low side FET will show the correct waveform. The complimentary high-side FET will show a bastardized waveform, but that is still similar enough. The two FETS that should be *completely turned off* show all kinds of bizarre spikes and waves, of up to 5V. Enough that they are also trying to conduct (obviously very bad). Eventually the MCP drivers go into thermal shutdown (handy feature, that).

A variety of things I've tried:
Replaced motor load with a simple resistive load. No change.
Add explicit 10k pulldown resistors to the inputs of the MCP-14700. This had some marginal improvement, even though the datasheet shows the MCP as having internal pulldowns.
I have metered for shorts and such for a good while and it appears to be put together alright. I've ordered parts to make another, though, just to be sure.

My question: am I misunderstanding how the MCP-14700 is supposed to work? Should a single chip control opposite corners of the bridge, instead of one side like I've shown it? Can the high-side bootstrap function without the low-side switching? Intuitively there must be a maximum duty cycle of the high side, or else the bootstrap cannot recharge, but I don't see it in the datasheet. Is there an obvious blunder on the schematic?

I think this particular example is beyond debugging with my 1 channel of scope, so I'd just like to not repeat mistakes when I make attempt number two...

Thanks.

Did you read section 5.4? Layout is critical, probably needs ground plane board.

They do not show a full bridge application, although certainly possible.

Using N fets for high drivers, with bootstrap gate supply, can be tricky.

Using P fets on top makes the H bridge a trivial circuit. If low power, even normal

bipolar transistors will work. And it can be run statically, unlike the bootstrap chip.

Hi,

Problem with n-ch on the high side from what I can see is that to turn the FET on would require the gate voltage to be 4 or 5v higher than both drain & source thus higher than the supply to the motors. I guess the MCP would look after this but only if the supply to the MCP was high enough.

p-ch on the high side is easy. I've used an Arduino to drive directly as well as TC-4431 / TC-4432 controllers.

Ian.

It helps to follow a chip/fet makers applic diagram EXACTLY.

Often they have little eval boards. Makes it easy. Then you copy their layout files.

Originally Posted by jglenn

Did you read section 5.4? Layout is critical, probably needs ground plane board.
Using N fets for high drivers, with bootstrap gate supply, can be tricky.

Hrm, alright. I didn't think this would be *that* picky if I kept my switching frequencies low. I've kept my traces short as possible on the MCP's and the bottom of the board IS a solid-fill ground plane. I don't think I'm that far astray of their layout recommendations. Some of these datasheets will give an actual PCB layout. I don't see it on this particular one though.

I'm going to build another one, just to reduce the possibility that I've damaged something in soldering or have a bridged trace somewhere that I haven't spotted. Going to use through-hole FETs this time, as I had a hell of a time soldering the D2PAK package.

Would it be more reasonable to make a single driver chip handle "forward" and one "reverse" FETs? I'm guessing it doesn't actually matter, as the datasheet says they are designed for "unique drive schemes".

Problem with n-ch on the high side from what I can see is that to turn the FET on would require the gate voltage to be 4 or 5v higher than both drain & source thus higher than the supply to the motors. I guess the MCP would look after this but only if the supply to the MCP was high enough.

MCP supply voltage is 4.5-5.0V. It's supposed to handle the job of getting Vgs above threshold. Does it matter that may load is supplied higher? (7.2V)

You might try just using the setup to drive a singled ended driver, one high and one
low Nfet. With just a resistor load. If you are doing something wrong, that will
reveal it.

Originally Posted by jglenn

You might try just using the setup to drive a singled ended driver, one high and one
low Nfet. With just a resistor load. If you are doing something wrong, that will
reveal it.

In that vein, I just tried sending PWM to a corner of the bridge at a time, holding the other 3 inputs low.

And all the outputs would work correctly in turn.

The moment I open two corners of the bridge to conduct, it all goes haywire.

Appreciate the help so far.

Is there a popular high-side driver for this time of application? Or is there nothing inherently wrong with my selection of MCP-14700?

The high side chip may need a path to ground to work properly, if broken, the whole
thing crashes. You can try these:

http://www.irf.com/product-info/cic/...driverics.html

I have a bunch of TC4420 and TC4804CPA, but I don't think are for high side.

Digikey and Mouser have P ch fets you can use for high side, or must you

go thru the complex way? That is for companies making zillions of units,
so they get the cost benefit of buying a million N ch fets. Also, P ch cost more,

and can be hard to get because laptops, Iphones, etc drain the supply (pun!).

Originally Posted by jglenn

http://www.irf.com/product-info/cic/...driverics.html

There's some good application notes on there I will read through, at the least.

or must you go thru the complex way?

I'm giving it one more go, because I'm stubborn that way. I usually find there was some bizarre workmanship defect in my circuits and they rarely work the first time...

Appreciate the feedback. Just wanted to know that I don't have something obviously wrong and that it's probably something more obscure that is causing my grief. Definitely on to P-channels if the second go-around fails.

The chip you are using should work, but would be nice if they had an app diagram
for the full bridge, all the waveforms, etc. Layout really can be critical, weird parasitics
can wreak havoc. Forget about the white breadboards for HF work. You may even
have ground loop problems. Learn about "loop area" and transmission line techniques.

A line pair carrying HF, with sharp edges, power or signal, if.....

they are separated, NOT run parallel, or twisted pair, or coax, or ground plane protected,

CAN

become a loop antenna, and send or receive EMI.

Know the "source, path, receptor" model for interference.

You can get output circuits talking to the input circuits through drain to gate
capacitance, or transformer leakage capacitance. Layout is everything.

Hi,
I think you're on the right track but I don't think you can statically switch through the various states of the 4 FETs because your gate drive to the two upper FETs is AC bootstrapped and relies on continuous switching of the output drive. Without this the top FETs will briefly turn on then start to fall back to the lower voltage of the driver chip itself.
Normally the diagonally opposite pairs of FETs will conduct to give one direction and the other pair for the reverse direction. Upper FETs can be PWMed while the corresponding lower devices can be held on. I think you need to maintain switching to preserve your upper FET gate drive so I guess 100% duty cycle is out.

Cheers,
David

Here's a design of mine........it's 2ch but you'll get the jist. Also, it's a lot higher power than you'll need probably so choose your FET's accordingly if you decide to use it. I use it to drive a couple of wheelchair motors.

There's a bit of logic up front so that all you need is a single PWM input and a simple direction input. This also helps protect the circuit from turning on hi & lo on the same side accidentally.

The resistors on the gate of the FETs are very important:-
The 2K2ohm resistors protect the 4431/4432 drivers from over-current and also help protect Vgs.
The 50N06 & 47P06 Mosfets in my circuit have a max. Vgs of +/-25vdc and the 15K/2k2 resistors will give approx 21vdc Vgs. So, if you happen to choose Mosfets with a max. Vgs of 20vdc for example then you can drop the value of the 15k resistors to 8k2 which will give you approx. 19vdc approx for a 24vdc supply.

The 15k resistors also protect the gate from floating in the event of a 4431/4432 driver failure (open circuit).

I use an Arduino to provide the PWM and direction control signals.

http://www.ianjohnston.com/content/i...ge/PCB/Mcb.pdf

Ian.

FET bridge drive circuits can be trickier than you think.

As already mentioned most high side drivers require the output to be continually switching in order to pump up the high side PSU.

You need the right combination of gate source impedance. Too low and it tends to ring causing high FET dissipation, too high and it can't charge the gate capacitance quickly enough causing high dissipation. Also if say the top right FET switches on very fast, the Miller capacitance of the lower right will try to turn it on too, causing ... high dissipation!

Usually the drive circuitry has a small amount of cross conduction delay to make sure devices are off before swapping diagonal paths. If you're monitoring the output signal without resistive load the two lines go all over the place. I usually place temporary resistors from each side to gnd to clean things up while testing.

Not sure on your load but you show no anti-parallel snubber diodes. Low power bridges may get away without, but for higher power the intrinsic diodes in the FETs aren't very good.

Steve

And don't forget about DEADTIME!

If you turn on hi and lo fets at the same time you go BOOM.

Hi,,

May be ( may be ... ) you should have a look to something like the LTC 1154 in the High side part... and try to use logic Mosfets !

just my two cents ...

Alain