DC Motor Control: H-Bridge +5A, 48v

Mike... how about the h bridge board from Procyon Engineering Pascal Stang? 50V 5A as I recall...

On 3 May 2005 04:30:29 -0700, Winfield Hill wrote in Msg.

I've used some of their older stepper-motor control modules (not chips). They work fine. The new stuff (the TMC3xx modules) looks even neater. About a year ago I pestered them about servo motor stuff and they said they'd been working on it for a long time, but it kept been pushed away by more pressing issues -- they seem to be a VERY normal company ;-)

--Daniel

I'm a big fan of the HIP4081 family, as you'll know if you've seen my dozen or so posts here on s.e.d. suggesting them as the solution for various problems. But in the TMC249 we have an IC that provides an important functionality for motors, stuff that an HIP4081 can't do.

I have often used an HIP4081 with a fancy controller to drive MOSFETs, because the controller was a wimp. For example, the UCC3895 is a fine IC, a resonant phase-shift PWM controller that "implements control of a full-bridge power stage by phase shifting the switching of one half-bridge with respect to the other. This allows constant frequency pulse-width modulation in conjunction with resonant zero-voltage switching to provide high efficiency at high frequencies," as they say in the datasheet. But it has rather wimpy 100mA gate-current output drive (despite its 1MHz PWM frequency), and it expects one to use a transformer to solve the flying high-side n-channel MOSFET drive. Whew! Two serious strikes against it. But paired with an HIP4081A it became a truly elegant solution to a tough problem (e.g., my 500W 10kV 600kHz resonant tank-circuit driver). So to me the HIP4081 family is well used in conjunction with other powerful ICs. Trinamic's TMC249 and other similar powerful chips may be a good examples.

--
 Thanks,
    - Win

Are referring to any particular chip?

This thread is rather hard to follow, it seems as if nobody is providing any context.

Robert

No. TMC239 or 249 needs a lot of peripheral for voltages up to 65V so that the HIP4081 seems to be the better choice, even that I need an additional uC for SPI. The area saved beacause of fewer components allowed more area for pcb-heatsink...

--

Michael Wieser

thats the german Farnell, I don`t know about "your" Farnell..

hth

--

Michael Wieser

Tell you about what?

Some people don't even bother to respond to googlers that are too lazy or stupid to learn to cut and paste - or maybe google has a setting to quote some context, but then the google kiddies would have to bother to find the setting.

Thanks, Rich

continuous,

capabilities

needs

products.

so

my

MOSFETs,

fine

of a

half-bridge

pulse-width

provide

But

1MHz PWM

flying

against it.

tough

So to

powerful ICs.

examples.

Several years ago I was using the HIP4082 in a couple motor controllers and had a lot of trouble with EMC radiated susceptability. When the controller was exposed to a big enough field it would start switching incorrectly and turn on upper and lower MOSFETs at the same time, resulting in a spectacular failure. This was several years ago so I don't really remember the field strengths to get it to do this, it was large but not too unreasonable, maybe 10V/m at around 100-300MHz, maybe. Eventually with carefull board layout we were able to get it to pass our internal testing requirements. It is not easy to improve EMC performace on a high power device, the usual band aids such as ferrites just saturate all the time, and don't really work, so you have to be much more clever about how you filter noise. The product was never particularly reliable, we had a lot of failures in the field, and the HIP4082 was the scape goat. Eventually we replaced it with a couple IR2110's and some extra dead time circuitry, which was not nearly as clean of a design as the HIP4082, however the reliability was infinitly better.

I always suspected part of the problem was the single ground pin on the device for both the high power side and the logic inputs. Keeping the input logic signals clean is always a chalange when they are referenced to a high power ground plane.

Has anyone else experienced similar problems? Does anyone know if Intersil has done anything to address EMC susceptability? If this problem has been fixed in the IC, I would consider it again. It really was a clean design.

Regards,

Ethan Petersen

This may be a possibility. But I have always used the HIP4081A instead, which has separated logic-ground and MOSFET-source pins, with an allowed +/-2V transient voltage difference. End of the problem, or at least well on the way to solving it. But if I was forced to live with an HIP4082 instead, I'd add to or increase the gate series resistors to slow down the FET and thereby reduce the FET's source-lead bounce, and isolate the driver at the same time.

--
 Thanks,
    - Win

controllers

the

switching

I

was

to

EMC

ferrites

be

the

couple

as

infinitly

the

the

referenced

really

My problem was more due to the EMC susceptability that the ground bounce from switching transients. The HIP4082 has logic internal that is supposed to make it impossible for the top and bottom FETs from turning on at the same time. The example circuit in the data sheet shows the input logic holding the high side FETs on all the time and relying on the guts of the IC to take care of everything. Unfortunately the internal logic would screw up when exposed to just the right (wrong) RF signal, and the fireworks begin.

Keeping the input logic signals clean during the switching transients is difficult with only one ground pin, but not insurmountable. Gate resistors are always a good idea for driving power MOSFETs. I would want a pretty good justification before not using one. You can get some ugly oscillations on the gate without it. Not to mention radiated emmisions from uncontrolled switching speeds.

Another good trick for controlling switching speeds is to use a gate resistor and add a cap from gate to drain to increase the Miller capacitance. This makes the FET look like an integrator and you get a very clean slope on the drain voltage.

This works for each leg of a bridge if the current is always flowing through the load the same direction. Then it is always one FET controlling the switching and the other is just there for synchronous rectification. If current starts to flow in the other direction the wrong FET will start to control the switching, and the miller cap will cause some ugly transients on the gate of the FET that is suppose to be off.

Ethan Petersen

In a previous life I worked for a electric forklift manufacturer. We designed our own motor controllers. It could PWM about 700A at 36V into each of the lift and traction motors. Fun stuff, where interesting things happen when things fail. The first version used the HIP4081, but it was later redesigned to use the HIP4082 to save some money.

Now it's been a while so you might want to check the datasheets to make sure I know what I'm talking about (I am on a slow dialup connection at the moment or I'd check myself), but I seem to remember one huge difference between the two chips.

The 4081 has built in oscillator for the charge pump circuit. That means you can turn on the upper FETs and just leave them on. The 4082 doesn't have the oscillator. It just uses a cap and diode to power the upper FETs. So that means that it needs constant PWM of the upper FETs to keep the cap charged. We just changed the software to put a maximum PWM of 0xFE out of 0xFF, and it worked fine.

With the 4082, if you just leave the upper FETs on at 100%, the cap will quickly discharge and the gate voltage will start to drop and bad things will happen. :-)

Carl Smith

What kind of FETs were you using? PWM Frequency?

Now that I think about it, the HIP4082 had the easy job. :-) The traction motors were separately excited motors. The field was reversible with an H bridge that the 4081, and later the 4082, controlled, and it took only 50A or so tops, and usually more like 30A. The FETs were Harris RFG70N06's in TO-247 packages. This thing was designed in 1994 or so. There are much better FETs available now.

The interesting part was the armature drive. Since the field was reversible, the armature didn't have to be reversible to be able to change the motor direction. The armature drive was a bank of six IXYS IXFN200N07 in the SOT-227 "minibloc" package (a block with 4 screws on top). The gate drive for the bank of FETs was a

4424 (both drivers in the package paralleled for higher current) connected to the gates through 10 ohm resistors. This controlled up to 700 amps maximum, and 350 or so in average usage.

PWM frequencies weren't very high. The system couldn't handle the switching losses at really high frequencies. I think the field was PWMed at around 700 hertz, and the armature less than that. Maybe like 400 Hz or so. The motors did whine with these PWM frequencies. At one point we implemented a varying PWM frequency that started higher as you began to move and lowered as the speed went up. That gave more torque from the motors when it was needed to get the forklift rolling, and cut down on heating from switching losses by lowering the frequency when the torque wasn't needed as much.

One of the things I always wanted to do just as a joke, but never did, due to having real work to do, was re-program that frequency changing routine to change the PWM frequency such that the motor whine would play the theme to Jeopardy. :-)

Maybe later when I have more time I'll post some stories of the interesting failures that can happen in such a system, if anyone is interested.

Carl Smith

I've worked on the same sort of drive (IR MGDs though) 1000A peak, nominal battery voltage of 24-48V, switching frequencies of 10kHz dropping to 5kHz when 'plugging'. Originally used IRFZ44s

You can get some interesting pyrotechnics when these fail under load.

Robert

I remember an incident where I leaned down to take a close look at a board under test, right when all the armature mosfets decided to blow like popcorn. I jerked back just quick enough that the foot long flames flying out the side just missed burning off my eyebrows. :-)

Another memorable one wasn't a failure, but more of a software bug. The controller stored the configuration in the flash memory. Early in the prototype development the software guys were more worried about getting something working than saving write cycles on the flash memory where the configuration data was stored, so they erased one of the flash banks and wrote the configuration to the beginning of that bank on each power down. The process was to open the main power contactor, then erase the flash bank, save the configuration, then power off the controller. Later when they had the time they implemented a wear leveling procedure that wrote the configurations consecutively through that flash bank, and the software used the last one on power up. This made it so that you only had to do the erase on one out of several hundred power off cycles, since the size of that flash bank was much larger than the size of the configuration data.

The problem was that this eliminated the 1 second delay, caused by the wait for the flash to erase, that had been there on every power down. This is where I should point out that due to the inductance and mechanical design of the main power contactor, it stays closed for about 0.1 seconds after power is removed from it's coil. The result was that the controller de-energized the main power contactor, wrote the configuration to flash, and powered off the controller. When an erase was not needed, the controller powered off before the contactor had actually opened. This resulted in all the mosfets on both traction controllers and the lift controller turning on, and full power being applied to all the motors momentarily. It caused a huge arc as the main power contactor tried to open under load, and more impressive to engineer types like me, the battery cables that dangled down the side of the forklift to the battery would jump apart by about SIX INCHES momentarily. Addition of a short delay to the power down routine fixed the problem. :-)

Carl Smith

The hardware design defaulted to all the MOSFETs turned on?

--
 Thanks,
    - Win

Yes. Yes, it did.

That one is kind of a sore spot with me. I suggested many times over several years, to the engineer that was in charge of that design, that he add a resistor from the armature FET gates to ground to keep them from floating on when the power to the gate drive circuitry went away. Something like a 10k resistor to ground would have cured the problem and presented no significant extra load to the driver chip. He never wanted to rev the board to add one resistor, because he didn't think it was a problem.

Here's another story. Maybe it should be titled "Ground is relative." Working at this forklift manufacturer was my first real engineering job just out of college. At that point I don't think I had designed anything that handled more than about 10 amps of current, and I end up at a place that designs motor controllers that control 700A. I learned a lot quickly. Like when ground isn't. :-) The forklift had a capacitor bank connected across the battery, and the negative bus bar was used as a central ground point for all the high current stuff. Each of the three motor controllers were connected to this point by a cable about 18 inches long. I could put a scope probe ground on one end of the cable and the tip on the other end of the cable. This cable is about half an inch in diameter. But there would still be a spike of voltage as much as 7 volts across this length of cable at each PWM pulse turn on, when the controller was running at full current. That always amazed me, that there could be 7V of drop on an 18 inch cable half an inch in diameter. I learned a lot quickly about how to handle such problems...

I managed to avoid being that close. :) Although I have fried a trace or two.

I once did the SW for a variant that used a pair of controllers to control a series wound traction motor in an H-bridge configuration. We had the current oscillating at about +-700A IIRC with the motor reversing direction every few seconds. I could swear I could feel the magnetic field off of that sucker. Twitching 00 cable does make for an intriguing sight.

Robert

Of course not ;)

Why were the capacitor banks so far from the FETs? That's an order of magnitude further (or more) apart than I've seen on any non-SCR based design.

It would seem to me that you end up nullifying a lot of the benefit of the caps by placing them so far away and using such a low switching frequency.

Robert