High side driver

[MechanoMan]

I'm looking at designing a microcontroller-driven bipolar stepper driver for a 48v supply.

The difficult part seems to be the high side driver for voltages >30v for a high-side NMOS. All I could find on Mouser was the Freescale MCZ33883EG, which looks sort of "ok" except it's $5.10 and 2 would be required per stepper, also not entirely sure about the speed.

The low side drive isn't the big concern here so these combined H-bridge devices aren't as helpful as say a driver with 4 highside outputs.

I browsed some plans for drivers made from discrete components. The problem there is that most don't have provisions for making the Vboost that is actually used for the high side NMOS gate. I can see ways to make that but they're expensive/complicated and these drivers themselves have a lot o speed questions to go through.

Any quick & dirty highside driver solutions in the 50v range that I'm missing?

[kreutz]

Use a half bridge driver, they are widely available up to 600V or even more and include the "bootstrap" feature. Go to International Rectifier's website for more info.

[juantolete]

you can use IRF2112 until 600 vdc.
Iam building a bipolar driver for 160vdc 8A using A3986 and IR2112

[MechanoMan]

Originally Posted by kreutz Use a half bridge driver, they are widely available up to 600V or even more and include the "bootstrap" feature. Go to International Rectifier's website for more info.

Ya well got a part recommendation?

Originally Posted by juantolete you can use IRF2112 until 600 vdc. Iam building a bipolar driver for 160vdc 8A using A3986 and IR2112

I saw IR2112... I wasn't 100% clear on the spec sheet but it doesn't seem to have a bootstrap voltage source. Which would mean all the difficult stuff you'd have to build anyways.

A3986? What? That has an onboard bridge driver and wasn't meant to be used with an external bridge driver, the delays would be really questionable because its logic to prevent shoot-through current was tailored for the on-board driver. Adding the delay of ANOTHER driver in addition to that might not work.

Did you read the thread on the A3986 build? Allegro created a severe design limitation by giving the duty cycle a fixed-% blanking time and making the blanking interval huge, which makes it impossible to meet the intended smaller current needed for microstepping when source voltage is much greater than coil voltage. And that was driving steppers with a 48v source. With 160V source, this problem would be so severe that not only is microstepping gone but it could (depending on your stepper motor voltage) be unable to even regulate the current for the whole steps.

I'm EE and once I saw the problem described it's a real facepalm. It's a consequence of something stated right there in the spec sheet, and realistically there's no way a driver can do anything to compensate for it. Believe me, I was frustrated too- I spent pretty much a week doing an efficient SMD board layout, had all the parts selected, then found the thread that showed the A3986 was basically never going to work as a microstepping CNC driver. Bummer... I started to pull my hair out over how-am-I-gonna-make-this-work and then told myself to chill because the A3986 can NEVER work no matter what you do so just accept it. Not unless you discard the microstepping entirely which would cripple it far below my minimum specifications.

[kreutz]

Originally Posted by MechanoMan .......Ya well got a part recommendation? .......

It depends of how you are controlling the bridge, meaning having separated Low and High driving signals versus just one signal to control both with fixed or adjustable dead-time on the driver instead that on the firmware.

There is another detail, bootstrap high side recharging requires that the low side be On for at least a couple of uSec periodically enough to recharge the bootstrap capacitors, if your circuit does not have that possibility, you need a charge-pump feature and the options are more restricted. selecting the right Mosfet drive depends on the total charge of the Mosfet gates, the driver power supply voltage and the required switching times, that is why I re-directed you to the IR website.

Kreutz.

[MechanoMan]

Yes, all 4 sides of the bridge will gated separately.

OK, lemme make sure I have the terms sorted out. Charge pump is where we generate a (Vsupply+ Vgate-to-source) rail with a capacitive charge pump. Or you could technically use an inductive boost supply to do the same thing. Bootstrap is where the capacitor charges off a low side source when the rail is pulled low. I was using those terms somewhat interchangeably.

Well, I'd originally figured a charge pump WOULD be required, since a CNC will idle leaving one high side active and one inactive. But now that I look at it, the most likely method for regulating the current would involve switching off that high side and switching to the low side on that halfbridge, recirculating the current through both the low side transistors. So OK there should be a guarantee of switching on the high side during idle provided Vsupply>(Icoil * Rcoil) which will always be the case.

If it's turning the duty is less straightforward but since it's turning we have frequent changes in the bridge direction so we should be safe there.

I was looking at IPD640N06L, which is a 1.6Vth logic level NMOS. Qgate=6.4nC @ 5Vgs @ 18A 48V (actually only switching 2.8A in this circuit). Ciss=350pf. That's an SMD TO252 and I'm hoping the switching losses will be low enough to just keep cool with the copper PCB area.

I'd like to use 5v for the gates and avoid making other supplies. I do see a problem with the bootstrapping where the diode drop to the bootstrap leads to a significantly lower Vbootcap when it's only a 5v source, but if it should be tolerable esp if it's a schottky.

And my preferred supplier is Mouser. Not a real biggie but getting parts from elsewhere really bogs down the parts costs with shipping.

[MechanoMan]

Hmmm... well, the IR2112 won't allow for 5v gate drives. Requires Vbb-Vss & Vcc to be 10v-20v and there's an undervoltage lockout to disable it for anything below that. Looks like ALL the IRF drives may be this way, a few are 9v min but no 5v stuff.

It's possible to make a supply for the higher voltage of course, just less desirable. Trying to keep it simple and cheap.

There's the LTC4440-5 which is a half-H which works with 5v gates. But at $2.50 per half-H that's $10 per driver which is pretty darn pricey overall.

[NC Cams]

We learned some VALUABLE lessons about logic level fets, and "High speed" drives" ulling current when we did some R/C car esc's a while back.

Dont think for a minute that you'll drive a fet at 5Vgs to turn it on hard and fast - it don't happen. With IRLS44'. you'd fry your hands at 30 amps with 5Vgs and you could run them sans heat sink with same current with 10Vgs.

The IRFS40 non logic level was about the same. IT would fry you at 10-12Vgs at 30 amp, yet be cool to touch at 15-18 Vgs.

When you study the I vs Vgs curves for any fet, you'll see why as they graph the effective resistance vs gate voltage in this plot.

The "charge pump" is built into the IR and other gate drivers. IT is part of the diode and capacitive network that is part of the "high side" drive. Want more 'holding time?' Use bigger caps and tantalum for the caps. Saw a H/S drive where 0.1's plus 10uf were used in parallal as part of the charge pump.

A trick we learned was NOT to use 100 percent duty cycle.

This lets you run a bit more P/S voltage for better curret/torque at part "throttle" and then you simply run a duty cycle of 96 percent at "full throttle". Motor never knows the difference - runs at rated speed while current limit of the drive prevents low speed over heating.- ditto that for low speed PWM at say 20-30 khz of typical DC servo amp drive.

my take on a similar project; run NON-logic level fets, run half bridges at 15-18vdc on control side, use level shift on logic side to go from 5 to 15-18 on logic input side.

BTW, Farichild has both low side and side drivers as a separate entity. This way, you can run 4 separate drivers and drive them with any form or logic you wish.

Finally, I"d be more inclined to RUN The half driver - especialy if they have dead time built in. THis will prevent you from getting surprised by "shoot thru" and the smoke that it creates. NO need to reinvent the wheel by finding a solution for that. Recall that fets turn on faster than they turn off and a lazy "offing fet" will smoke via shoot thru when the same side of the bridge is coming.on.

ANother reason NOT to have 100 percent on times.

Also keep in mind that surplus grade motors may not to run at the outstandingly fast step rates of today's PC's. Their inductances and sizes may be just fine at 20-30 khz switching speeds. WHen you turn up the wick with swithing speed, new issues develop, especially with steppers.

Keep in mind also that you may not want to run a 600v mosfet on a 100 volt circuit - reason, high internal resistance. Look at the 400 and 200 volt fets, lower on resistance, especially i fyou drive them at the voltages listed above.

finally, DON"T use a logic chip to drive a fet. USE A FET DRIVER. THis is especially true with big die sized fets. Since the gate is essentially a capacitive load, logic IC"s won't/can't do as good a job as true fet drivers. We learned that even an PNP/NPN transistor totem pole worked better than a logic IC at driving a fet hard and quick..

[kreutz]

Originally Posted by NC Cams The "charge pump" is built into the IR and other gate drivers. IT is part of the diode and capacitive network that is part of the "high side" drive. Want more 'holding time?' Use bigger caps and tantalum for the caps. Saw a H/S drive where 0.1's plus 10uf were used in parallal as part of the charge pump. ..

That is "Bootstrap" not "charge pump"..........MechanoMan has it right :"Charge pump is where we generate a (Vsupply+ Vgate-to-source) rail with a capacitive charge pump. Or you could technically use an inductive boost supply to do the same thing. Bootstrap is where the capacitor charges off a low side source when the rail is pulled low".

Kreutz.

[kreutz]

Originally Posted by MechanoMan Hmmm... well, the IR2112 won't allow for 5v gate drives. Requires Vbb-Vss & Vcc to be 10v-20v and there's an undervoltage lockout to disable it for anything below that. Looks like ALL the IRF drives may be this way, a few are 9v min but no 5v stuff. It's possible to make a supply for the higher voltage of course, just less desirable. Trying to keep it simple and cheap. There's the LTC4440-5 which is a half-H which works with 5v gates. But at $2.50 per half-H that's $10 per driver which is pretty darn pricey overall.

Logic level mosfets are driven from a +10Volt gate supply otherwise you will never get short switching times.

Don't worry about the single quantity distributor prices, they will go down with production volume...

What do you think about this one: HIP4081A?http://www.intersil.com/data/fn/fn3659.pdf
Full bridge mosfet driver with a real charge pump.

[kreutz]

Originally Posted by MechanoMan I'm looking at designing a microcontroller-driven bipolar stepper driver .........

Don't worry about a charge pump feature on the Mosfet driver if what you want is to drive steppers in chopper mode bipolar configuration. When idle there will be always current chopping, at high speed there will be always phase switching, calculate Cbootstrap for worst case timing at about 200 RPM (without chopping) and you are covered. Don't let it reach 100% duty cycle when chopping. Limit the minimum Vmotor voltage on your specs in order to make sure there will be current chopping at 200 RPM.

[NC Cams]

Anybody have a link to/for the AN9405 Ap Note listed for the Harris/Intersil fet driver???