Putting half bridge is in parallel the Infinion BT N8982

[Simon]

Evening all. I have a requirement to drive a 33 amp motor. I have come across the Infinion half bridge driver chips for example the BT N8982 which headlines an amp rating of 55 A. Obviously it is worth looking firstly at the RSD on and this could be as high as 10 milli ohms on either high or low MOSFET. This resistance at 33 A would produce 10 W of heat which may be a little excessive not to mention the fact that these small pins have to take the full power. I'm loathe to move away from this chip as it is short circuit protected over temperature protected and has other nice features. I am considering the possibility of putting two in parallel. Now I'm going to guess this will come at least with some caveats, for example I assume the traces leading to the input pins would need to be pretty much the same length to ensure the timing is identical between the two half bridges. It would course be beneficial that they are mounted not too far apart so that they are thermally identical and share the load equally. Apart from these small obstacles is there any other reason I should not do this?

The alternative is to find some sort of driver chip to drive a discrete high and low side MOSFET which could be of the through-hole type and heatsinked.

[Zero999]

I've think you've got the decimal point a little off:

Obviously it is worth looking firstly at the RSD on and this could be as high as 10 million ohms on either high or low MOSFET. This resistance at 33 A would produce 10 W of heat

I assume you meant 10 milli Ohms. LOL

MOSFETs tend to be fairly easy to connect in parallel, as the on resistance has a positive temperature coefficient, so if one device gets hotter, than the other, it will pass less current.

I wouldn't worry too much about synchronising the switching because one IC should be able to handle the full current for a short length of time.

[Simon]

Yes Long live speech recognition there is a vast difference between Milli and million even in my garbled speak.

The problem is that it's 1/2 bridge. When the signal goes low it will switch off one side wait and then turn on the other side. So if the timings go wrong I could have a massive shoot through.

[Andy Watson]

The devices are rated for 50A continuous - I would be tempted to go down the route of providing sufficient heatsinking to make one device work.
Have a good look at section 5.3.3 "Current Limitation" on the datasheet. It seems to imply that if the current limit is reached the output will automatically flip to the opposite state - this might cause some problems if the output is connected to a second device that hasn't flipped.

[Simon]

there is also the switching losses, at 33/2 amps (16.5A) is well below the "static" current rating.

[MagicSmoker]

Is that 33A peak, or full load current (assuming this is an AC motor and we are talking about per phase rating)?

If this is a "full load" current rating then be aware you might need 3x that rating for "thermally meaningful" periods of time to deal with sudden increases in load.

Discrete switches can be paralleled fairly easily (within reason), but so-called "intelligent" power modules - or power stages which have gate driver and other assorted functions built in - are a whole different story; you'd best inquire with the manufacturer about that.

[JamesAus]

What are the motor specs, at what voltage do you intend to drive it, and under what load? I have designed these into a product and have thousands in service. They are very well protected as you are obviously aware and basically never fail. Typically I'm driving loads <10A @ 12-14V with a soft start / stop implementation and braking. The maximum current I've run them at repeatedly is ~25A for a few seconds, with a very low duty cycle. During these on times there is no huge temp increase with just heatsinking on PCB. If you like I can carry out some tests for you over the weekend and see how they behave at >30A continuous (if that's really necessary). 33A seems like a lot to run continuously though.
Regards
James

[Simon]

Hello JamesAus

Well if you can test that would be very nice of you but don't feel obliged. I am going to check with infinion as to whether putting them in parallel is a good idea. The motor is a brushed DC motor at 12V and the maximum current is 33.6 A (no they can't use 24V). I have run one of these at 10A and they were not bad heating to 50C (I think or was it 70) on a 2oz copper board. I did fill the heasink pad with via's and have more copper on the back. Obviously 33A and Isqr losses are going to make this a very different prospect but the device will be in air flow so any heatsinking could make the world of difference.

[JamesAus]

Hi Simon

Since you have already worked with these I guess you can run the tests yourself anyway. One thing I would do is just try it with the motor / application in question and see what happens. You already know you can't kill these things. When you say the maximum current draw by the motor is 33.6A would this not be the stall current? or the maximum current at maximum power? If this is the case I would suspect you'll rarely if ever hit that figure and could be managed with the way you drive it. What is the rated wattage of the motor?
James

[Simon]

Unfortunately I don't have the actual motor with me. I've had a customer enquiry and they need a quote and obviously pretty soon a device that works. So it's a tossup between careful thermal engineering on the intelligent driver or it's a case of throwing the towel in and using a driver chip and discreet MOSFETs that could possibly be through-hole and have a proper heatsink on them.

The motor's maximum load current is 33.6 A at 12 V although this is an automotive applications so maybe the motor would go a bit quicker at charging battery voltages although at 33 A I'm sure there will be significant voltage drop across supply cables. So motor power -wise we are talking 400 W. My experiment so far on a 2 ounce copper board have produced a temperature rise on ambient (let's say of 20°C) of 35°C. This was driving 10 A worth of load and finding the most efficient switching frequency for that motor setup. Extrapolating to 33 A and bearing in mind that resistive losses are to the square of the current increase we are talking of in fact a 10 fold increase in power dissipation. That would take my temperature rise to over 300 degrees and is obviously unfeasible.

I can heatsink the driver through the PCB using a 0.8 mm PCB to reduce the length of the via and solder filling that will carry the heat to a insulated thermal pad with a heatsink bolted down to it. This will be in airflow but I don't know if I'm going to get enough heat conduction through the board and into the heatsink regardless. So this is where I start to consider putting two of these devices in parallel as this would also make it easier to get the current into the power pins as believe it or not despite the 55 amp rating of these devices the pins are very close together and there is one ground pin and one power supply pin so you can't surround it in a reasonable sized PCB trace all you can do is have as much trace off each side as you can. With two devices in parallel we would only be halving the on resistance and other pseudo-resistive losses and still be facing a 10 fold increase in the I square term so the heating would go down to just half which is still a heck of a temperature but then just maybe I would get enough heat through the circuit board quickly enough to the heatsink.

As this is starting to sound all a bit mad as a colleague suggested I should perhaps consider using a discrete driver and discreet MOSFETs as this way I can get the pick of the MOSFET and even put some in parallel or use a part with a heatsink tab that can be screwed down to a big chunky heatsink. The only problem I'm having here is that most drivers I am finding only seem to extol their virtues for MOSFETs with a gate capacitance of up to 10,000 pF. The problem I am having is that any decent MOSFET to do the job is at least 3000 pF and if I were to put anything in parallel that would soon increase and I have found MOSFETs with as much as over 5000 pF input capacitance so putting two in parallel which sometimes would still be necessary get to be a huge amount of input capacitance that the chip may or may not be able to deal with as they do not talk about anything over 10,000 pF. Obviously increasing the gate capacitance will slow down the turn on and turn off times possibly negating the gains I have made in lower on resistance. The only real advantage could be that in fact by reducing the switch on and switch off times I get a lower slew rate and less EMC problems which I will have to pay for with the increased heating but if I'm using a part that can be screwed to a heatsink this becomes a negligible problem and being able to fan out the PCB traces the carry the total current is also a great advantage.

Unfortunately you can't search Farnell for MOSFETs based on gate capacitance, some websites allow you to search on gate charge. I always assumed gate charge would be proportional to gate capacitance as I thought this was their physics. I think I might try Digi Key and see what their parametric search is like.

[Simon]

What can I say? Long live digikey the following seems to be a reasonable compromise: http://media.digikey.com/pdf/Data%20Sheets/ST%20Microelectronics%20PDFS/STP100N6F7.pdf

[JamesAus]

Given the unknown quantity with the motor / application it's probably a wiser solution to use separate drivers and paralleled FETs. I imagine when Infineon get back to you they would be pointing you in the direction of some of their higher power FETs and gate drivers. Since you were looking at the BTN8982 to begin with I guess you were planning on monitoring output current? Are you going to use a current shunt now or another method for this?

[Simon]

Yes I have emailed Infinion but I don't know how long it will take them to respond. They will probably suggest discrete MOSFETs and another driver chip. I have been recommended the LT 1158 this does permit current sensing which would be nice to do as it allows for short circuit protection.

[max_torque]

I wouldn't get too hung up on the gate capacitance of your chosen MOSFETs, because in this application, the easiest way to avoid excessive switching losses is just to not switch very often.  A 33A 12V brushed motor, will, almost certainly have, what is technically know as a "sh*t lot" of inductance, and as it'd driving a fan, current ripple is also not really an issue (within sensible limits for resistive loss minimisation).  As such, i bet you can run a low fundamental pwm frequency at say 100Hz or something similarly low, so a relatively low current gate driver, which takes say 10us to cross the miller plateau, will not give rise to a large total switch loss. Given the low voltage requirement (say 30Vdc max) you can pick a mofset with a very low RDSon (probably around 1mO), plenty of choice, in TO-220 or even TO-247 packages if you really want to go beefy (247 package allows for much larger pcb traces due to the wider pin spacing). Alternatively for SMC, IR do their DIRECTFET range, which are very nice to heatsink, being fully metalised packages.

[Simon]

Searching on Digi Key (most other suppliers are pretty pointless unless you are checking the actual stock Digi Key seems to have an excellent parametric search and the widest stock available)  for the lowest drain to source resistance MOSFETs did not bring up any T0247 packages as you suggested the other day when we spoke about this they don't seem to have been used for new designs of MOSFET. I did see the IR DIRECTFET packages and thought they might be easy to heatsink as they have a full metal body that I've not used them before so was a bit iffy. I suppose the heatsink will be screwed down across the top of them with a sill pad in between which would be fairly simple to implement mechanically. One big concern I do have is that this could become a monstrosity to physically assemble once the pick and place has been done as it people's labour that cost them money not the pick and place assembly that generally ends up costing not much more than the price of the PCB and the components as assembly houses get very good pricing on their components so if you're building enough to reduce the setup charge there is actually no PCB assembly cost to worry about.

I'm not too worried about the gate capacitance providing the chip will not have a problem with it I think I will go with your suggestion of the LT 1158 as looking around there is not anything any better in terms of specifications and detailed datasheet information. I looked at a part from Fairchild that can drive more gate current and the first thing the datasheet talked about was the drawbacks of high current switching with the chip which I thought was a bit silly. What's the point of designing a chip and then starting your datasheet with "Ah yes but". The LT 1158 seems to be designed to go up to 100 kHz so yes at the low frequencies we are talking about I suppose gate capacitance wouldn't be a problem.

Ideally I should probably get one of the actual fan and run it at different frequencies. When I tried out the Infinion chip I used a range of frequencies to drive a pair of fans in parallel (I don't know if having two fans complicated anything but that was the only way of getting higher current for testing purposes) at low frequencies I got not a lot of heating as would be expected, I started at around 100 Hz and moved up, going past 500 Hz the chip started to get noticeably warmer and at 2500 Hz it was getting quite hot I think it was around the 70° mark. But as I came through 2700 Hz to 3000 Hz chip suddenly cooled right back down to the same temperatures it was at when I was using 100 Hz at about 45 or 50°C which is quite good when you're looking to extrapolate to hire currents. The other consideration I will have here is EMC given the amount of current I don't know if switching very slowly is going to be better than switching at what seems to be a higher frequency that has the same thermal efficiency but perhaps has less currents moving up and down as the fan is switched on and off as that would be the only explanation as discussed there is lower ripple at this particular frequency so perhaps it would be better for EMC I don't know. Of course there might be so much inductance in a motor of this size that it will be a low-frequency anyway that has the least ripple.

Of course having a higher gate capacitance would mean longer rise and fall times which would in effect carry out the same function as the slew rate control on the Infinion part and help reduce sharp current rises and EMC problems. I had thought of using T0220 parts as at least I can just screw a heatsink to them but perhaps the metal caseed parts you suggest are worth a look at. If anything if they need just a little bit of cooling considering it will be forced air cooling a piece of aluminium could be screwed straight across the top of them or you could get a single hole into any heatsink.

[Simon]

Bingo: http://www.farnell.com/datasheets/1911991.pdf?_ga=1.147776306.40049004.1474651687

Using two will make gate capacitance over 12'000 pF but the on state resistance will be minimal

[max_torque]

EMR doesn't really care about the fundamental frequency (within normal, sensible and practicable pwm frequency limits for such things, ie 100 to 100kHz) it cares about the RISE and FALL time of the signals edges.  Hard switching your fets WILL produce a "squarer" output wave form and hence the harmonics of those edges will produce more EMR.  IN all cases, as usual, it's a trade off between EMR and efficiency.  The harder your switch, the less switching losses you get, but the more EMR you get.

For something like an ultrahigh efficiency DC:DC or similar, where you want to be able to drive the smallest inductor possible (for cost and space reasons) then you would choose a fundamental pwm frequency as fast as possible, (these days, up above 200KHz probably), and they would need to have hard switching, with times to fully cross the miller plateau below 1us to prevent the FETs going over temp.

But in your case your load is fixed already and it IS a huge inductor (probably, most high current DC brushed motors are massively inductive due to design constraints) so you know you can switch very slowly, so slowly as to almost make the switch losses irrelevant, because they happen so infrequently. Also with just 12v, your dV/dT isn't going to be terrible either, again, helping matters.  I'd pick a decent FET driver (like the LT1158) and carry out a gate drive impedance exercise with the real fan motor you are driving to work out the lowest fundamental PWM you can use (before current ripple becomes excessive (need to do that at 50% duty - max ripple point) perhaps aim for <10% ripple current) and how slowly you can then drive the gates.  I'm going to guess that probably even a very soft drive of say 5us to switch will probably be ok, because it's only happening once every 10ms!

[max_torque]

PS, for directFET info, look here:

http://scolton.blogspot.co.uk/2011/06/directdrive-well-it-certainly-looks.html

[Simon]

Yes i guess that EMC won't be a problem with the low frequency and rise/fall times and with the low on resistance (0.7mR at most) any ripple current losses will be minimal. Just laying out the board now taking great care over the symetry of the MOSFET's with respect to wire pad and sense resistor positions etc. even at 4oz copper a 3.5mm wide trace resistance can rival that of the MOSFET's, I suppose if I am forced air cooled I might get away with 2oz but that will need strict testing.