Max amps of a 18 FET 4115 controller, check my math.

[zombiess]

I got a 18 FET 4115 controller and have just been looking through data sheets and playing around with numbers and trying to figure out realistic safe battery amp and max phase amp settings to program the controller to if I want to get totally crazy. I am running 30S lipo with a peak voltage of 126 volts on a 9C 2806 in a 20" wheel with vented side covers and 12 AWG phase wires.

IRF4115 says highest RDSon of 0.011 Ohm Max so for this example I'll say they are each 0.015 mOhm.
Max power dissipation at 25C is 380W with linear 2.5W/C derating so I'll say a 125 degree operating temp which gives 380-2.5*100=130W with an awesome heat sink. I'll derate that to 90W using the case as the heat sink.

So there are 6 per phase, 3 per high/3 per low leg of a phase. That means in parallel each three have a 0.005 ohm resistance total with 3*90W dissipation of heat for 270W dissipation per 3 total.

Since the FETs only care about the motor I'll use phase amps in this example.

With the above consumption each set of 3 FETs should be able to pass 180 Amps of current per phase wire before going pop. 100% DC * current squared * RDSon

Working backwards that gives us combined power dissipation for 3 FETs @ 125C of 270W / .005 RDSon for a total of 54000. Then take the square root of 54000 and you end up with each 3 FET bank supposedly able to pass a maximum 232 amps per half of a phase. If we go WOT we will get 232 amps of current per phase wire being delivered at 0 rpm. If we stick to the generally rule of running 2.5 to 3.0 phase to battery current that lets us with max battery current of 77 to 93A battery current. I typically aim for the middle and go use 2.66 time my battery current, so 232 phase amps / 2.66 = 87 battery amps maximum and you are now on the verge of making FET popcorn if the ones in your controller are 100% perfect to spec.

Since I never like to run things that close to spec if possible I'll allow 25% head room to allow for imperfect simultaneous switching times, losses/etc creating more heat or what ever other issues occur.

This I believe this leaves me with a general rule of them for my Lyen 18 IRFB4115 MOSFET controller of a max battery current setting of 65 amps and max setting of the phase amps between 162A and 195A for my 30S 126V setup.

These are just numbers that I picked from data sheets and hopefully got the math mostly correct. I would seriously appreciate anyone who knows more about this than me to double check my work. I tried to work with what I felt were somewhat decent safety margins by assuming really bad operating conditions such as 125C FET temps, not being able to fully drive the FETs into 100% saturation thus keeping RDSon higher than it's spec'd max of .011 ohm, etc.

These number also assume staring from a dead stop and just opening up the throttle 100% so everything gets maximum current. Phase current and battery amps drop from that point of course

These numbers all seem to match up very nicely with the ebikes.ca simulator. What do you guys think? Is my math correct? Is my derating good enough to find the theoretical maximum operating settings for my combo? Please point me in the right direction if I've messed this up royally.

Thanks,

ZombieSS

 

[zombiess]

Anyone care to guess if I'm even on the right track with this?

 

[icecube57]

i would use those settings on a 4110 controller but I would derate it about 1/3 for the 4115

Those data sheets get their values from low voltage tests with magical heat sinks... real world is much different...

Ive sat down and done the math like you did an realized that a bone stock controller is running at around 40% capeability.. I would push it to 70 or 80% but not much more unless you are cooling the hell out of it... current handling goes down with heat. So if the 4115 has a higher rsdon than the 4110 and is less capeable you need to derate accordingly or prepare to buy another controller.

You already know the controller ball parks the current why flirt with the maximum data sheet specs just so the controller can ball park guess the current to low and blow a bank of fets.

 

[icecube57]

The lyen 4115 mosfet controller was never intended to run 65A. its stock config is 45A but can be pushed to 65A if you must.. Just do the usual upgrades you better have a low inductance high resistance motor... or shes going to give u the finger.

 

[Alan B]

You may want to review a design note for power FETs. Paralleling them doesn't divide current evenly, so need to derate for that. I think the rule of thumb is that you only consider the additional FETs will carry half of the balanced current they should. So take credit for 1.5 FETs for 2 in parallel, or 2 FETS for 3 in parallel.

Consider the switching time and the power generated during this transition as well. Worst case is phase current times half the supply voltage times the switching time times the frequency times two. It can be substantial if the switching time is slow.

Another source of heat is the freewheeling diode that conducts the phase current when the PWM FETs are open. So this will be (1-pwm) * phase current * forward diode drop voltage.

Not sure if you compensated for three phases and the commutation duty cycle. There is also some kickback due to commutation that causes some extra current at the moment it changes.

Not sure if you took the worst case RDS, have to look at the highest die temperature and component variation to get that.

 

[icecube57]

Clears throat... well damn

 

[gwhy!1]

You also need to take into account of the layout of the fets as regards the needed heatsink. i.e one of my 6 fet controllers ( in a taller box with a 5mm higher spreader bar will work all day long pushing 3.5kw and the controller only gets warm, one of my 6fets in a standard box it gets hot but nothing to worry about, Now one of my 12fet controllers in the standard box gets to hot @ 3.5kw if pushed for to long ( but when fitted with a 10mm higher spreader bar its fine upto 6kw before I need to start worring about the heat ), but on paper it should be good for around 7kw ( if we are going by your maths ) and a 18fet should be good for around 10.5kw ( I dont have a 18fet controller) but In real life the 18fet will probably get just as hot as one of my 6 or 12 fet only pushing 3.5kw ( this is assuming we use all the same fets in all the controllers ), I belive that one of the causes for this ( apart for the usual problems with paralleling fets ) is because of the size of the spreader bar is unable to shift the heat from the inner most fets so this in turn causes a imbalance in fet temps and the more fets you have the more of a problem it is, I work it out using a basic 6fet setup first then de-rate all the figures by 10% then add this to another bunch of 6fet (de-rated)figures to become a 12fet then de-rate those figures by 10% and so on. This has worked well for me (upto now).

 

[gwhy!1]

You may want to review a design note for power FETs. Paralleling them doesn't divide current evenly, so need to derate for that. I think the rule of thumb is that you only consider the additional FETs will carry half of the balanced current they should. So take credit for 1.5 FETs for 2 in parallel, or 2 FETS for 3 in parallel.

This fits in very well to my experiences.

 

[zombiess]

I'm trying to get my a better understanding of how the power stages of these controllers work, but I'm trying to take baby steps. I went through the data sheets and understand many of the specs, but I don't have all the knowledge needed to fully utilize them.

From the responses I got a lot of info, but I'm not sure if you guys read what I originally posted where I showed my math and the assumptions I made and the derating I did with some SWAG. I know that paralleling FETs never = perfect current sharing and I also assumed a 125C operating temp of the FET's and derated the power dissipation to even lower than what the spec sheet said because most heat sinks aren't as good as what they test on for specs. I went with 90W per FET. I was trying to come up with something that is more real world vs ideal data sheet info.

I did not account for the fact that these are 3 phase motors and each 3 FET bank of an 18 FET controller is only on for 1/3rd the time. I think I also just found the mistake that I didn't series the resistances of each phase. I used a RDSon of 0.015mOhm for each FET which 3 in parallel is 0.005 Ohm but did not take into account the other 3 FET bank which would bring the total resistance up to 0.010 Ohms so I need to rework some of my numbers based on that flaw.

A little later tonight I'm going to try and rework the numbers with some of the info posted here. Please pick what I'm doing apart. I really want to understand some of this better and try to come up with some estimated safe operating values for controllers without using the empirical method of try it till you fry it.

I'm basing my assumptions on higher inductance motors such as the 9c 2806 since I own one and not something like a xlyte 5302. I'd love to be able to have a motor choice be made and give estimated safe values based on it's inductance, but know this will be tricky.

 

[zombiess]

So I just re ran some numbers, tell me if these look any better.

The values I used for this were as follows
6 FETs per phase, 3 high side 3 low side.
Went worst case scenario and used a 125C operating temp vs spec sheet 25C values which took the 380W per FET power handling down to 80W each and a total of 480W for the entire controller which would be one toasty chunk of aluminum.
I used a thermal derating of 3 W/C per watt in stead of the spec sheets 2.5 W/C
I derated the FET's by a factor of 1.5, 3 FETs in parallel is more like 2 in parallel
I took the RDSon Max of 11mOhm and made it 20mOhm for each FET to account for the increased resistance at high temp (couldn't find the spec so I basically doubled it)

These values gave me a theoretical max phase amp handling of 155Amps (which 3 parallel IRFB4115 FETs should handle) with a 126V input. Dividing that by 2.66 gave me 58 battery amps (which means it would have to be set for 53A to account for over shoot).

I only calculated this for one 6 FET bank because I figured the others would be the same, but each bank would only operate at 1/3 duty cycle because there are 3 total phases. I realize burst operation should allow this to be a bit higher, but I'm just trying to guesstimate the safe settings for the controller with my limited knowledge. I tried to assume worst case scenarios I could think of. I know my ebike is never run at these levels because I don't have a place to do it. I believe my current settings are 38A battery 110 phase 126V and in the 6 lap race I only managed 65 watt hours/mile with a top speed of 50mph over the 4.8 mile course and that is by far the hardest I've ever ridden it. Motor itself was hot but no stench smoke or need to do any active cooling, controller was only a bit warm.

Do these numbers look any more realistic? Lots of guess work involved but I tried my best with my current level of knowledge.

 

[Electroglide]

So I just re ran some numbers, tell me if these look any better.

The values I used for this were as follows
6 FETs per phase, 3 high side 3 low side.
Went worst case scenario and used a 125C operating temp vs spec sheet 25C values which took the 380W per FET power handling down to 80W each and a total of 480W for the entire controller which would be one toasty chunk of aluminum.
I used a thermal derating of 3 W/C per watt in stead of the spec sheets 2.5 W/C
I derated the FET's by a factor of 1.5, 3 FETs in parallel is more like 2 in parallel

I took the RDSon Max of 11mOhm and made it 20mOhm for each FET to account for the increased resistance at high temp (couldn't find the spec so I basically doubled it)

The datasheet has a curve that gives you the multipler verses Tj...so 125C ~= 2.1 which gives an Rdson of 23.1mohm @ 125C. Good guess on the Rdson.

The factors that will influence the power dissipation as mentioned previously are the turn on and turn off time of the mosfet banks which for higher voltage operation tends to dominate the losses more so, the conduction losses, the reverse recovery losses, the diode conduction losses, and any shoot through or avalanche losses due to bad design.

Knowing the worst case total power dissipated per fet then allows you to figure out if you have the right thermal path from the mosfet junction to wherever the heat will finally end up...which is the ambient air for ebikes. This is what determines how much long term current you can safely pass through your controller. The mosfet datasheets are only a starting point for figuring this out. Headline specifications mean nothing because they are given for the device only and not for real world applications. So when they say that the device can dissipate 380Watts, that is really an unusable number because it is assuming that you can keep the back of the case at 25C. The calculation is based on the thermal resistance from die to case which is listed as 0.4C/w working backwards you can get the wattage listed (175C-25C)/0.4(C/W) = 375Watts Not sure why is disagrees with their number other than the Rjc is listed as max and they may have used a nominal value instead. What you need to figure out is what the thermal interface material that is used between the fets and the heatspreader and calculate its thermal resistance. Typical values are in the 2C/W to 1C/W ranges [EDIT] for electrically insulating types. If that thermal interface is located further down stream in the thermal path it will have a different thermal resistance. Once you know that drop, you need to figure out what the thermal resistance of your heatspreader to ambient is or to another heat sink. It gets a little bit difficult to figure that out if you don't have good geometries to work with or known heat sinks with published specifications. For short term currents, you need to know what that thermal interface drop is along with your heat spreader heat capacity. Heat capacity is like a capacitor in electronics. It stores heat energy rather than coulombs of charge. Like a capacitor, the heat capacity and the thermal resistance give a similar response as a low pass filter does in electronics. This means that you can dissipate a lot of wattage for a short term until the heat sink temperature rises to its steady state maximum temperature dictated by the thermal resistance across all interfaces, the maximum die temperature, and the wattage flowing through those interfaces.

So lets take an example...make it a worst case
Commutation frequency = 200Hz ....you can use this to estimate an average dissipation for a controller operating in BLDC mode and the expected temperature rise on each mosfet bank.

duty cycle mosfets = 90%
fsw = 20khz
ton = 1usec
toff = 1usec
Rdson = 23milliohm
Assume synchronous rectification
# of parallel devices = 3
Assume equal sharing of current
Assume phase current = 100A
Assume battery voltage = 127V
deadtime = 500ns
maximum desired Tj = 125C

losses due to hard switching mosfet bank
Psw ~= 127V * 100A * (1e-6+1e-6) * 20kHz/6 = 84.7Watts ... this is total switching losses for one sector of operation

Diode losses...ignore reverse recovery losses. (freewheeling mosfet bank)
Pdeadtime ~= 0.65V * 100A * 0.5/50 = 0.65Watt...more or less neglible.

Conduction losses on the hard switching mosfet bank
Pconduction = (100)^2 * 0.023 * 0.9/3 = 69Watts

Total watts on the hard switching bank is 69+ 84.7 = 153.7W or 51.2Watts per fet.

The synchronous rectification mosfet bank will always be on during the sector so you will have only the conduction losses
Pconduction Sync = (100)^2 * 0.023 /3 = 76.7Watts or 25.6Watts per fet

The freewheel mosfet bank will have mainly the diode losses, and a small conduction loss, ignoring the reverse recover losses. Pfreewheel ~= Pdeadtime+Pconduction = 0.65W + (100)^2 * 0.023 * 0.1/3 = 8.32Watts or 2.77Watts per fet

Maximum Total losses for controller ~= 153.7 + 76.7 + 8.32 = 238.7Watts

This would be the case for a stalled bldc motor operating only in one sector. The bank dissipating the highest [EDIT]power is in the hard switching bank...ie the bank that interrupts the current by the mosfets switching on and off.

It would have an approximate maximum dissipation of 51.2Watts per fet. If you have 0.4C/W Rjc thermal resistance and you add the drop from the interface material to you heatsink which could be 2C/W for a TO-220 case, you get 2.4C/W from die junction to heatsink.

That means that your heatsink could not rise above T-Heatsink <= 125C - 51.2W * 2.4C/W = 2.12C Since that is not very practical unless you are running a chiller on your bike, you have to back off the maximum phase current that you want to run. You may have notice that the thermal drop through the interface dominates the thermal resistance from junction to heatsink, so improving that drop will allow you to run a higher phase current. You have to also note that there is a thermal resistance associated with copper and aluminum so going direct from the mosfet case to metal improves the thermal resistance but does not make it go away. Anyways, the numbers above are made up for the most part and it assumes operating at the highest die temperature allowed, which would not occur except under certain conditions. The calculations for switching are approximations since the switching losses are linearized versions of the non linear voltage and current curves during a switching event. They however are close enough to use for first pass calculations which allows you to figure out what you need to design for to get what you want in performance.

So if I made a few mistakes I am sure others will point them out, but the main take away from this is that mosfet datasheets should not be taken at face value. The electrical and thermal implementation will drive the real maximum current that is achievable.

 

[zombiess]

electroglide, you juts gave me so much information that I've been trying to figure out by example. Thank you so very much.

I'm just trying to come up with a very simplified way to know when one is getting really close to the danger zone of their settings. I have started a new thread with a spreadsheet I created here http://endless-sphere.com/forums/viewtopic.php?f=2&t=33181. It's not even close to anything you posted and makes some huge assumptions, but I'd appreciate if you look at it. I'm going to try and incorporate some of the information you just posted here into it and see if I can get it closer. Seems like it's all about the temperature and some fans on the controller could make a big difference when doing lot's of power bursts like wheelies, showing off, etc.

I do not know what the thermal resistance of our controller cases are, but I suspect most of the FET failures we are experiencing are from high burst current at very low commutation speeds.

Without going back through all your calcs and trying to reverse engineer the data I need, what do you think of the settings I arrived at by taking a SWAG of having a battery amp setting of 50A, 135A phase @ 126V on a hot controller that is 105C? From a dead stop the block time burst would have to stay under 80A batt and 216A phase if the FETs are hot at 105C already.

Of course if I just started the bike up and the controller is sitting at 50C I shouldn't have to worry unless my block time amp burst went over 105a battery/284A phase since I like to use a 2.7 bat/phase multiplier because it rides smoother that way for me.

 

[liveforphysics]

I just started typing up an answer to this, telling you to ignore everything but temp corrected RdsOn (because the 25degC spec NEVER matters), and then work backwards from the combined Rth data, add in your silicon pad Rth (generally at least +1w/degC), call your heatsink temp ~80degC (because locally on the back of the FET, even when you're feeling 60degC, it's 80degC at the surface touching the pad to touch the FET tab. And now you can estimate something.

But then electriglide had a pretty solid post explaining it well.

So, I will just give my calculation for what each 4115 is good for continuously.

0.0275 ohms at 150degC Tj.

0.9w/degC+1C/W (for the pad) = 1.9C/W.

Sink contact surface 80degC, peak junction temp 150degC

70degC deltaT. 1.9C/W = 36.8w thermal budget on the device.

About half is going to be conduction losses, half will be transitional conduction, flyback clamping on the body diode, switching losses, etc.

That leaves us with 18.4watts for our 0.0275ohm resistance. (I^2)*0.0275 = 18.4. I = 25.8amps per device continuously, leaving roughly no margin of safety. For burst loading, you should be able to double that for a few second impulses at a time.

 

[zombiess]

OK Luke, since you worked the math and came out to 25.8A continuous per device what is the proper way to rate this 18 FET controller for continuous duty since there are 3 phases with a high and low side and each having paralleled FETs?

1. Every FET, 18 FETs * 25.8
2. 9 FETs (combined high and low side) * 25.8
3. 18 FETs / 1.5 (due to unequal current sharing) * 25.8
4. 9 FETs / 1.5 (due to unequal current sharing) * 25.8

I'm confused because there are 3 phases and no more than one phase is on at a time.

I just really want to understand how these things are operating and BLDC is new to me, but I feel I'm learning pretty quick.

Thanks.

 

[zombiess]

Now that you guys have gone through several examples I'm going to go back to playing with my spread sheet since I understand way more of what's going on. Your examples helped me make a lot more sense of the graphs on the data sheet. For some reason I just couldn't find the darn RDSon VS Tj even though it was right there, didn't realize it was a multiplier... derp

 

[liveforphysics]

It means, 77.4amps is the perfect world continuous phase current.

If you're trying to relate that number to where you set battery amps, it's simply impossible to determine without looking at the motor as well. On a colossus motor, 77.4amps of phase current would happen at below 10battery amps... On a 5305 or something, you could likely set battery amps pretty close to phase current and never see a problem as long as you avoided spending time at low RPMs. If you wanted it to be bullet-proof on a colossus, you simply cant, just the hystersis of trying to switch and see the current rise in the shunt would be enough to explode the controller. If you want it to be bullet proof on a high-turn-count hubmotor, I imagine setting it around 40amps battery current and 70amps phase current would be a good starting place, but if you did a lot of extended very low speed operation, it bet you could still blow it.

 

[Electroglide]

I concur with Luke(?) on this. Controllers that do battery current regulation need to be matched to a particular motor if you want to make sure that you don't pop any fets. Back when I was working at Accelerated Systems, they went over to a chinese controller factory for a tour on behalf of Bafang, and what they came back with was that every controller was set up to be matched to a particular motor. They had a large number of variations that they were producing that had as small of a difference as not populating one resistor and they new their cost down to the fractional cent as to what each component added to the unit cost. So when buying controllers by themselves, you run the risk of blowing them up because they have likely been set up for a motor other then yours. If you had phase current regulation, a lot of they issues would go away because the manufacturer would be limiting the output to what the controller could safely handle.

 

[zombiess]

Hey guys,

I took a lot of what was said in this thread and incorporated it into a new spreadsheet which will hopefully provide some tighter tolerances. I'd appreciate if you looked at it. It's located in this thread.

http://endless-sphere.com/forums/viewtopic.php?f=2&t=33181&p=482024#p482024

I had some fun calculating the polynomial function needed to model the RDSon temp curve, first time I've ever done that but I think I got the multiplier pretty darn close. Now I finally know how to mathematically represent curves... yay! Polynomial functions FTW! Wish I had known about these before today but didn't even know what they were called until googling how to come up with a formula to input x and get y based on a curve.

I haven't gotten into incorporating the FET switching and diode losses yet, but that might I might not need to go that far, we'll see how bored I am

 

[zombiess]

A quick stupid question. How indicative is the outside temp of the controller to the FET temp on the inside? Anyone put a thermal probe to see how they correlate such as a 60C controller case = an FET temp of 80C? Does a split like that even sound reasonable or would it be more like a 60C case = 120C FETs?

I have no idea how good the case is at conducting heat away from the FETs, but I do know that mine has never been so hot I couldn't hold my very girly sensitive hands on it for as long as I like... aka just warm. Had my motor hot enough to not be able to hold on for more than a few seconds, probably what most of you consider starting to get hot. I've just got super temp sensitive fingers, no joke my tongue and my finger tips are about equal with sensing temperature, hot or cold. I can't even hold onto a cold can of soda for more than 30 seconds without it starting to hurt

For this controller simulation I set my RDSon @ 15mOhm (I'm trying to over compensate to find the safe envelope) and derated the parallel FET's by 1.5 so that the 6 on a phase only act as 4 do on paper. If by some chance the FET drivers manage to fully saturate the FETs then I've got a lot more margin with 11mOhm RDSon, but I doubt that's the case.

Back on topic, playing with my spreadsheet and comparing it to the ebikes.ca simulator for my setup if I can keep the FET's temp under 88C I might be able to run 126V@50A battery with a 2.7 multiplier giving me 135 phase amps with a decent safety margin.

The ebikes.ca simulator says I'd need 175 phase amps at WOT until I got to 11mph where they would drop down to 135A and fall pretty fast from there. I'm guessing that with block time set at 1.0 sec I'd see a burst of 75A battery which would give me around 200 phase amps for the one sec which is the max before failure on my spreadsheet with a FET case temp of 88C. I know the FET's can handle higher bursts as well, I'm just trying to find the most reliable power zone. I'm probably going to install fans on the controller to help keep the temps down since that effects the power handing the most.

Am I getting any closer to a realistic prediction? Why don't we have a way to download data directly to our brain in seconds dammit. This has taken me at least 15 hours of racking my brain to learn this crap from scratch because I knew almost NOTHING about FETs and controllers. I very much appreciate the help guys. Make me unstupid.

 

[liveforphysics]

A quick stupid question. How indicative is the outside temp of the controller to the FET temp on the inside? Anyone put a thermal probe to see how they correlate such as a 60C controller case = an FET temp of 80C? Does a split like that even sound reasonable or would it be more like a 60C case = 120C FETs?

If the temp you can measure on the outside of the case is 60C, then it's safe to consider the surface the thermal pad is contacting to be 80C (lots of variable here, but our setups are about as thermally poor as things come). Then you have to know the C/W to know the FETs junction temp. In this example, the pad is adding ~1C/W, the fet case to tab + tab to junction was 0.9C/W. This means you're looking at roughly 1.9C/W. So, if your FET is dealing with 30 watts of heat, it's going to have 30 *1.9 = 57degC hotter than the 80degC case surface you're contacting. Or 137degC junction temp. The junction temp is the temp that matters for finding the right RdsOn multiplier.

If you want to make your model really easy, just call your switching/body-diode clamping/transitional conduction losses 50% of the RdsOn losses. Sometimes it will be more than 50%, sometimes it will be less than 50%, and if you ever get any gate ringing in your driver, it's going to be like 95% of your losses lol, but 50% is a fair back-of-the-napkin number for modeling.

 

[zombiess]

A quick stupid question. How indicative is the outside temp of the controller to the FET temp on the inside? Anyone put a thermal probe to see how they correlate such as a 60C controller case = an FET temp of 80C? Does a split like that even sound reasonable or would it be more like a 60C case = 120C FETs?

If the temp you can measure on the outside of the case is 60C, then it's safe to consider the surface the thermal pad is contacting to be 80C (lots of variable here, but our setups are about as thermally poor as things come). Then you have to know the C/W to know the FETs junction temp. In this example, the pad is adding ~1C/W, the fet case to tab + tab to junction was 0.9C/W. This means you're looking at roughly 1.9C/W. So, if your FET is dealing with 30 watts of heat, it's going to have 30 *1.9 = 57degC hotter than the 80degC case surface you're contacting. Or 137degC junction temp. The junction temp is the temp that matters for finding the right RdsOn multiplier.

If you want to make your model really easy, just call your switching/body-diode clamping/transitional conduction losses 50% of the RdsOn losses. Sometimes it will be more than 50%, sometimes it will be less than 50%, and if you ever get any gate ringing in your driver, it's going to be like 95% of your losses lol, but 50% is a fair back-of-the-napkin number for modeling.

FWIW, 60C is about as hot as I can hold my hand on for maybe 2 seconds before pulling away, none of my controllers have ever gotten this hot... yet.

If you look at my spreadsheet you will see that I modeled for this, even used your 1.9 C/W model based off the .50 C/W tab to case and .40 C/W die to tab junction and 1 C/W pad Went through all this including finding the polynomial function to get to the correct RDSon correction factor based on die temp. But once again, you just supplied me with more info about estimating the switching losses which I believe I've taken into account for by starting out with a 15 mOhm RDSon instead of calling it 10 mOhm. I figured 10 mOhm was close enough to the min/max RDSon divided by 2, then added 50% and called it 15 mOhm.

I hope I did that right. Is what I'm saying making sense?

 

[CamLight]

A quick stupid question. How indicative is the outside temp of the controller to the FET temp on the inside? Anyone put a thermal probe to see how they correlate such as a 60C controller case = an FET temp of 80C? Does a split like that even sound reasonable or would it be more like a 60C case = 120C FETs?

I have no idea how good the case is at conducting heat away from the FETs, but I do know that mine has never been so hot I couldn't hold my very girly sensitive hands on it for as long as I like... aka just warm. Had my motor hot enough to not be able to hold on for more than a few seconds, probably what most of you consider starting to get hot. I've just got super temp sensitive fingers, no joke my tongue and my finger tips are about equal with sensing temperature, hot or cold. I can't even hold onto a cold can of soda for more than 30 seconds without it starting to hurt

One of the biggest problems with using the controller case temp as an indicator of the FET junction temp (the important value) is that both the worst possible performing thermal setup and the best performing setup give you the same results...a cold controller case.

Imagine a system without any heat sinking at all, just FETs standing up inside the case. The outside of the case would be cool to the touch and would give no indications that those FETs are frying. Now imagine a system with huge amazing heat sinks, monster fans and perfect contact between FETs and other cooling components. The case would be cool to the touch in this case too.

Bottom line...in my opinion, the case temperature can't be used for even a remotely wild guess of FET temperature. Even with the narrow range of case styles and construction methods, you don't know if the manufacturer is using pads or insulators between the FETs and the case. You don't know what kind of heat spreader, if any is being used, etc.

Best is to measure the case temperature of the FET. For TO-220 FETs, this is the point where the metal tab enters the epoxy body of the FET. This isn't the "case temperature" referred to in FET specs though!!! That temperature is measured directly against the rear of the FET, through a hole in the heat sink. But, it come within a few degrees of the "real" value for a controller setup and you can just add, let's say, 5-degrees C to the temperature you measure at the tab/epoxy case junction temperature to come up with a value you can plug into your equations. Just calculate the power each FET is dissipating and that case temp value will give you the all-important junction temp.

When comparing against like-constructed controllers though (heat spreader/no spreader, pads/no pads, etc.), you can often use the temps measured by others to get an idea of what your max external case temp should be to be safe. Consider 80% of the max junction temp rating to be the highest junction temp you should ever reach for max reliability. That includes considering the ambient temp INSIDE the case! The outdoor temp doesn't matter much if the inside-the-case temp is 50C higher. You can go all the way to the max junction temp rating most of the time without giving up too much since this is an intermittent-power application.

 

[zombiess]

Ok, getting a lot further but need some questions answered.

Parameters
Total number of FETs = 18
FETs per Phase =6
RDSon = .015 ohms
parallel Derating factor = 1.5, means I'm treating 6 FETs as 4
Thermal interface to case = 1.9 C/W (based of LiveforPhysics example)
FET Case temp = 60C

Controller settings:
Voltage = 127
battery amps=48 (yes I know FETs only see phase current)
multiplier=2.66
Phase Current = 127A

1. My spread sheet is basing all of the calculations on a single phase, so is that 127A phase split between 18 FETs on 3 phases or just 6 FETs on a single phase, should I use my 1.5 derating factor?
2. Up until now I've been doing everything by single phase. My 127 phase amps/493 phase watts is spread across 6 FETs I calculated 493W/6 = 82W per FET. If that is the case and the thermal resistance is 1.9 then 60C + 82*1.9c/w =216C junction temp. Should the phase amps, phase watts etc be divided by the total number of FETs or just the ones on that phase?

Main question is what numbers do I divide by the total number of FETs and which ones by the number of FETs per phase?

 

[liveforphysics]

Whoa whoa. In an 18fet controller, all phase current passes through groups of 3 parallel fets, not 6.

 

[zombiess]

Whoa whoa. In an 18fet controller, all phase current passes through groups of 3 parallel fets, not 6.

I thought on an 18 FET controller each phase had 6 FETs, 3 on the high side of the coil and 3 on the low side of the coil.
Is it 3 because only one back is PWM and the other three are just turned on to complete the circuit for the phase?

I'm really confused now. Got any links to a quick explanation (not 100 posts of discussion) of 3 phase controllers. Going through the motions I am right now might make me look kinda dumb initially, but I'll get there eventually.