What's the temperature limit of a controller?

[John in CR]

My low turn count hubmotors have blown lots of controllers, and typical controllers always run very warm even at moderate power. I started putting small blowers to pull fresh air through my higher power controllers, and haven't popped any that were actively vented other than one I forgot to turn the fan on and it blew on the first moderate hill.

I picked up a high power 36fet with irfb4110's from SteveO. It's ventilated, but still gets hot after just 3 or 4 short launches. It's surprising just how quick the heat gets up to outside of the case through the 2 heat sink bars on the fets. It has a thermometer attached to one of the fet bars, and I want to start using it for running even higher power, but I have no idea at what kind of temperature I should use for the alarm or what to look for on the display. I've heard people talking about scorching hot controllers that survive, but even my popped controllers weren't near 100°C at the case near the fets, maybe 70-80C.

I don't know if it makes any difference or has anything to do with how high I can run it, but the fet legs are soldered down near the tip instead of pushing the legs deeply through the board like I've seen on every other controller.

John

 

[dickselectrictoys]

i am pretty sure never mor then 62, however different applications can occur, but never with extended degree lengths. too hot, not right. (venting needed)

 

[amberwolf]

70C is often seen as a max temp spec for the die on various non-power semiconductors, usually called Junction Temperature.

IRFB4110 spec sheet shows 175C as absolute max Junction temp, as another example.

When the outside environment is hotter than that, its harder (or impossible) for the die to shed heat fast enough, and depending on it's design and on present operating conditions relative to it's other max specs, could either reduce it's operating life, damage it in some way that makes it's operation unreliable or cause it to always emit even greater heat, or could actually just cause it to fail (dramatically or otherwise) right then.


I'd stay well below that if possible, though, because you have to account for the thermal resistnace of several layers:
--die-to-case
--case-to-mounting-bar-insulators
--mounting-bar-insulators to mounting-bar
--mounting-bar-to-heatsink
--heatsink-to-air
in most ebike controllers, plus the air temperature itself (since the hotter it is, the less able it is to take on more heat from the heatsink).

Plus if it's an unvented controller, you ahve the heatbuildup inside that, too, making everything even hotter than regular open-air numbers would suggest.


Since the hotter things like FETs are, the worse some of their other specs get, and the more sensitive they then become to transients near their max specs, the more likely they are to pop, then the cooler you can keep them the better, down to ambient air temperatures at least.

Exactly how hot is too hot will depend on the weakest link in any particular controller. I fully expect that capacitors (85C, usually, 105C at best) could be that limit in an unventilated controller, over time, as their electrolyte boils off, and FETs (70C to 125C probably) for instantaneous heat generation inside them that can't get out fast enough.

I don't know the math to show what that lowest rating should be, because of all the stages of thermal resistance to getting it out of the controller, but I would guess that a FET-case/tab temperature of less than half of the max spec for worst-case lowest-rated part would be a good goal.

 

[alfantastic]

Interesting post. I did start a similar post a while back, with reference to Cellman's 30A Infineon 9 fet (EB209) fitted with IRFB3077 FET's.
I intend to install it into a bag with no air flow, and an inch space around the top and sides of the controller casing.
I'm still 50/50 on whether I should do this, as I don't want it failing quickly.
It'll be drawing an average of 20A @ 48v, with short one minute runs of 30A @ 48v, running a MAC 10T.
Would mine be considered as a high power controller, or are the controllers John in CR talk about real monsters?

The question again for me, is would my controller really generate enough heat, to worry about its performance and reliability stored in a bag?

 

[amberwolf]

My 12FET is mounted on the bike frame where it gets lots of airflow on CrazyBike2, and it still gets noticeably warm. That's at "48V", really almost 60V, short bursts of 40A at startup and 7-20A sustained depending on conditions. I suspect in a bag it'd cook pretty quick.

FWIW, if I leave the endcap open that faces forward, and have the rear cover held on only by one screw so it is pivoted partly out of the way for airflow thru the case, it stays at what feels like ambient temperature. With the rear cover fully installed, it's not that much warmer. I've never measured the actual temperature of it that I recall.

 

[alfantastic]

Without resorting to mounting something like an axial fan on the controller to help with cooling, is there any heat dissipating material I could attach to the case to help remove excess heat?

Would attaching a large heat sink, with a thermal adhesive pad to the top of the casing, provide a viable passive solution to dissipating heat?

Heatsink: http://www.ebay.co.uk/itm/Large-Alu...642?pt=LH_DefaultDomain_3&hash=item27bff43302

Thermal pad: http://www.ebay.co.uk/itm/Akasa-Hi-...hermal_Compounds_Supplies&hash=item27c423b115

Click on the above two links to see an example of what could be used.

 

[Philistine]

Would mine be considered as a high power controller, or are the controllers John in CR talk about real monsters?

I am reminded of the joke where the white guy gets off the plane in Jaimaca, and he is in the airport taking a pee, he looks over at the tall Jaimacan standing next to him and can't help but notice he has "Wendy" tattooed on his dick, which is a coincidence because so does he, he taps the Jamaican on the shoulder and says "Hey buddy, your dick has Wendy tattooed on it, and so does mine, isn't that a coincidence we both have girlfriends called Wendy!", and the Jamaican says "No mon, it is flaccid, when it is erect it says "Welcome to Jamaica, have a nice day".

John's motor is so powerful with such high speed wind that he has heat problems with a 36 Fet controller, he basically has three of your controllers and it is still getting hot.

Re your issue Alfantastic, I have found at the power levels you are talking about it depends on the wind of the motor combined with the size of the controller. So, for example, I have a mate who runs a 12 Fet Lyen controller on 48V at 40amps, with zero airflow (ie totally bagged) on a 9x7 9C and he has no problems at all, but I recently tried that with a BMCV3 (which is basically a Mac, but I don't know what wind it is equivalent to in a Mac) on a 12 Fet Lyen at 48V at 40 amps, and I found it got way too hot bagged, so I put it out on the frame with airflow and it is cool as a cucumber. But I run a 24 Fet on a x5304 on 48v @ 50 amps, completely sealed up in an ammo can and it doesn't even get warm. In my opinion with a 12 Fet, and anything over 25 amps, I would want that to have some airflow myself, but having said that, as I said, my mate runs his sealed on a 9x7 fine, and has done so for over a year everyday on a long commute without issue.

EDIT: Sorry BTW John, I didn't mean to derail your thread, because I am really interested in this answer to. I have told this story countless times on ES, but when I first ran my 5303 I tried to do so with an 18 Fet unventilated, and after my first commute, it got so hot that it actually physically exploded, so violently it blew my entire battery box into pieces and caused bystanders to call the police. I can only presume it got so hot that the solder melted and shorted it, or perhaps the caps exploded? Anyway, I also have a temp probe installed in my 24 Fet in my Gborg, because it gets no airflow at all, and I wanted to see how hot it got in case I needed to vent it (but because I was only running at 20S/50amps it never rises much), but I too have wondered what the safe limit is. Really interested in the answer.

 

[zombiess]

From those I've talked to who are in the know, it is suggested to keep the MOSFET case temperature below 80C if you want to extend their life. It is also best to monitor the case temperature on the high side MOSFETs used in the Xie Chang controllers since they are subjected to switching and resistive losses, not just resistive losses like the low side has. At 80C case temp the internal junction temperature of the FET is about 112C for IRFB4110 and IRFB4115 FETs. At a case temp of 100C the internal junction temp is about 140C. To get to 150C you only need a 107C case temp. It's suggested to never exceed a 150C junction temp even if the max is 175C. In short, the cooler the better.

I have a thermocouple wedged between two of my FETs and I've only seen the case temp get over 80C one time and that was during a 15-30% hill climb done at 40% throttle on a 125V pack. I've noticed that the case temps on the FETs cool off much faster on my controller than the internal ambient temp. Both often run in the 65C range but when I stop the FETs quick drop down into the 45C range after 1-2 mins and the internal case temp can take 5-10 minutes to come down into the high 45C range depending on conditions.

 

[BatteryMooch]

It's critical to know the power that the FET is dissipating in order to determine a safe case temperature. If the FET is dissipating 5W, then the case could be 125C without problems. If the FET is dissipating 100W, the case temperature better be a lot lower. It also depends heavily on what the case-to-junction thermal resistance of the FET is and that differs for the various FETs and case styles.

For similar setups, from the same manufacturer, at a limited, i.e., known, range of power levels that each FET has to dissipate, you might be able to establish safe case temperatures. Tough, but it can be done.

Responding to earlier posts....
Using the temperature of the controller's case as a guide to how hot the components are just won't work. A cool case exterior is present in both the worst performing cooling setups and the best. For the worst setups, the heat never gets to the case exterior. For the best setups, it's so efficient that the case temperature doesn't go up. Both setups feel the same when touching the case.

Adding additional cooling to the case of the controller will only remove the little bit of heat that's reached all the way to that part of the case's outer surface. It won't do much for the heat that's being generated at the FET. It certainly won't hurt to add additional cooling though.

Cooling the case interior, via a fan perhaps, could drop the FET temperature 10C-30C. Whether that prevents the components from being overheated, there's no way to say without measuring the component temperatures. Again, definitely wouldn't hurt though.

 

[zombiess]

It's critical to know the power that the FET is dissipating in order to determine a safe case temperature. If the FET is dissipating 5W, then the case could be 125C without problems. If the FET is dissipating 100W, the case temperature better be a lot lower. It also depends heavily on what the case-to-junction thermal resistance of the FET is and that differs for the various FETs and case styles.

This doesn't make any sense to me, could you please elaborate? How do you get a FET case temp to 125C dissipating 5W unless you decided to skip using a heat sink? At 125C case temp you are running a 175C junction temp on 4110 and 4115 FETs, that's a great way to nuke them in short order no matter what your power dissipation is.

In my examples above I was referring to the MOSFET case (Tc) not the controller case.

 

[ZOMGVTEK]

It's critical to know the power that the FET is dissipating in order to determine a safe case temperature. If the FET is dissipating 5W, then the case could be 125C without problems. If the FET is dissipating 100W, the case temperature better be a lot lower. It also depends heavily on what the case-to-junction thermal resistance of the FET is and that differs for the various FETs and case styles.

This doesn't make any sense to me, could you please elaborate? How do you get a FET case temp to 125C dissipating 5W unless you decided to skip using a heat sink? At 125C case temp you are running a 175C junction temp on 4110 and 4115 FETs, that's a great way to nuke them in short order no matter what your power dissipation is.

In my examples above I was referring to the MOSFET case (Tc) not the controller case.

Since its essentially impossible to measure the FET's die temp, you need to assume it based off the spec sheets rated thermal resistance. He's simply saying, that if the FET is just putting along, it can be very close to the max rated case temp, since the thermal path is quite good. As soon as you start asking a ton more out of the FET, and it needs to dissipate much more heat, the die temp is going to rapidly increase, meaning a 'safe' case temp is going to be much lower when the FET is asked to dump more heat.

Assuming the thermal path from the FET case to the heat sink is fairly good, you can assume a negligible delta between the FET case temp, and the heat sink temp. However, as the heat being dissipated on both ends goes up, this delta is going to dramatically increase as well. Putting a fan on the case will provide some minor help, but it will mostly just mask the FET's temperature by mostly cooling the case.

In my experience, when the controller is burning hot to the touch, its too hot. Warm, or even hot, is OK. Just try and avoid beating on a very hot controller, unless its not pushing the FET's too close to design limits. On a 6 FET 500W controller, you can probably run it at 100ºC all day case temp. If your 12 FET 8kW controller has a 100ºC case temp, its going to implode when you hit the throttle.

 

[zombiess]

You can calc the die junction temp with the right formulas. I've got a spread sheet that does a fairly decent job at it and it follows the spec sheets nicely after you enter the parts specs.

I recommend installing a thermocouple on one of your switching FET tabs if you want to push the controller hard, it helps to know when to back off the throttle and let it cool. It's been very helpful to me. I use a 4 channel thermocouple logger made by UEI.

 

[John in CR]

Thanks guys, I was afraid the answer was no real answer other than cooler is better.

Philistine, no need to apologize at all. As long as everyone is having fun any thread I start is open to wander any direction.

Now to other details, and please point out where I might have an inaccurate conclusions.

First, I've used these same little blowers to vent 4 or 5 controllers, and on all of the others the air blowing out was noticeably warmer than with this controller, but this controller's case gets hotter. The flow volume feels the same by hand. I used the same approach, which is to block air flow from passing through the unobstructed space under the board where there's nothing heat sensitive and no heat sources other than resistance in the traces, since I want the air to flow where it's needed. I can only assume that most of the air flow is bypassing the hottest surfaces...at least the caps should be cool. Placement of the controllers has been the same, behind and below the seat, so open to some air flow though my legs and shorts make it less than maximum.

The big difference is the fet bars. The other controllers were the typical single row of fets with the bar bolted to the side of the case with thermal paste. This one has 2 rows of fets and 2 far larger bars attached to the top of the case. Since the case get quite warm so quickly, that tells me SteveO did a really good job on the thermal pathway from the fets to the case, so I don't want to disturb that at all. It looks a bit difficult to get back into place anyway, so the board stays in place and at most I deal with the end caps.

I didn't do the intake holes. SteveO just swiss cheesed the wire side end cap. So first I'll take some steps to better direct the flow intake to go where I want it to go. There's so much more hot surface area in there that the exhaust air needs to be warmer to prove its doing the ventilation is effective. Once I get the air warmer then I can consider adding another fan.

I guess the next step to maximize cooling, is to move the controller from behind me and the seat, which act like a windscreen, to down in front of the down tube. There's plenty of room and this is a fair weather only street machine, and getting that few pounds forward and lower is a good thing.

Hopefully with some changes I can crank the current up by 30-50%, putting it well above 250A and still below the 300A he said he pumped into an X5303. That was with 2 of the lowest flow rate fans I've ever seen, which I had to immediately replace. OTOH, I'm sure that motor just sucked up the peak current as a spike on takeoff. Hubmonster might as well be 3 copper bus bars as the windings, because I have to restrict the phase current so much that extreme current is continuous during acceleration. Current actually increases to peak somewhere above 40mph....as you can imagine I can only steal glances at the CA. It wouldn't hang with LFP's bike, but even with my 260lb self it's a blast for me, and I don't have to worry about flipping it with an accidental throttle twist except when stopped. E-life is good.

John

 

[BatteryMooch]

It's critical to know the power that the FET is dissipating in order to determine a safe case temperature. If the FET is dissipating 5W, then the case could be 125C without problems. If the FET is dissipating 100W, the case temperature better be a lot lower. It also depends heavily on what the case-to-junction thermal resistance of the FET is and that differs for the various FETs and case styles.

This doesn't make any sense to me, could you please elaborate? How do you get a FET case temp to 125C dissipating 5W unless you decided to skip using a heat sink? At 125C case temp you are running a 175C junction temp on 4110 and 4115 FETs, that's a great way to nuke them in short order no matter what your power dissipation is.

In my examples above I was referring to the MOSFET case (Tc) not the controller case.

Yup, I was referring to the MOSFET case temp.

You mentioned getting a FET case, dissipating 5W, to 125C only if there's no heat sink. That's exactly my point. A bad cooling setup can severely affect a FET's operating temperature. So, you need to know the thermal resistances (and power levels) involved in order to say at what temp a FET case can operate safely at. If the cooling is lousy, 5W can kill a FET. If the cooling is good, it can operate without problems at over 100W (assuming TO220 case, blah, blah, blah).

And running a 4110 or 4115 at 125C case temp doesn't mean that the junction temp is 175C. In fact, you cannot possibly tell what the junction temp is knowing only the case temp. You must know the power being dissipated by the FET and any case to sink thermal resistances. Assuming a 0.5C/W case-to-sink thermal resistance (spec for the IRFB4110) and a 1.0C/W thermal resistance for an insulating FET insulator, that results in a 1.5C increase in junction temp over case temp for each watt of power that the FET dissipates.

If the FET is dissipating 5W (and case temp = 125C), then the junction temp is 125C + (5 x 1.5C/W) = 132.5C (this FET is in no trouble)
If the FET is dissipating 50W (and case temp = 125C), then the junction temp is 125C + (50 x 1.5C/W) = 200C (this FET is in big trouble).

If I insulate the entire heat sink and not use an insulating pad between the FET and the heat sink, you get completely different (lower) junction temperatures for the same power levels. And, of course, if I change to a better cooling setup, I get much lower junction temperatures.

I completely agree though that running IRFB4110/4115 FETs at a junction temperature of 175C (their rated max) is a bad idea!!

 

[John in CR]

Camlight and others, am I correct or at least reasonable to assume the SteveO did a good job in minimizing heat path resistance based on the fact that the case near the screws gets pretty warm so quickly. eg My usual route leaving the house is 300yds turn 300yds turn another 300yds to stoplight, with acceleration during less than half of those distances pulling maybe 150A since max current doesn't build to peak in the 100yds or so I'm on the throttle due to near parity in battery and phase currents limits to keep heat down. It's a 36fet with irfb4110's and the only thing I see I don't like is the long exposed fet legs soldered to the board near the ends of the legs.

Good news-
All it took was tape over half of the air intake holes all on the the side away from the fet bars to force significantly more fresh air down the interior channels formed by the 2 heat sink bars, and now the exhaust air is warm. I ran it as hard as possible on 2 runs this morning and the case was a bit cooler than before due to more effective ventilation (just like I'm so stubborn about for hubbie ventilation). I'll do one more test tomorrow the surpasses any real life use by me, which is 10 full stop launches up past 50mph with regen braking back to stop and immediately launch again. That even surpasses race torture, since there's only 1 full stop launch in a heat race. It also surpasses mountain climbing as far as the controller goes, unless I did a 1/3 or so throttle long hill climb with throttle modulations on the way up, something I know never to do after popping several otherwise very stable in use controllers.

If it does 10 back-to-back launches with 70-80A regen back to a stop, which i think it will do since 4 or 5 good fun launches is my typical errand route (shoot me, I like to have fun with my ebikes), then I'll feel comfortable advancing the current limit about 25% and the phase current limit a bit extra for stronger launches. I'm a bit pissed that a moto beat me the other day . Granted she had a huge advantage carrying enough speed into the the red light from behind me, so I didn't get my huge advantage while they shift to 2nd while my acceleration is continuous and building.

Sure 99 and 1 or so is great, but I've take it upon myself as the best thing I can do for electrics in my part of the world is to blow the socks off anything that I get along side of. It doesn't matter that against sportier looking vehicles I take them only by surprise, because my goal is to impress by forcing them to look at my fat ass on a little electric bike with absolutely nothing they can do about it despite hitting full throttle or stomping the accelerator. By the time they've been able to change the encounter I've already made an exaggerated motion with my right arm that I'm letting off the throttle. Nothing puts an LFP type smile on my face than my relatively silent ebike enabling me to hear a sports car or bike trying to wind their shit up after they've already been embarrassed by an electric bike that I built.

Just wait till I go 2wd with a Cargo Bike 2.0, and I can embarrass them with a bike I carry the boys to school with or pick up supplies at the lumber yard that will push me well past 70mph top speed. Note that this typically won't work for most, but in my case going from 1 motor in a 20" to two motors in 24's will give me both a significant increase in top speed and enable me to greatly increase current limits since the load is cut by almost 40% per motor despite the larger wheel.

John

 

[BatteryMooch]

Camlight and others, am I correct or at least reasonable to assume the SteveO did a good job in minimizing heat path resistance based on the fact that the case near the screws gets pretty warm so quickly.

John

The warm screws can mean two things...
1) The cooling setup is efficient (low resistance).
2) The cooling setup is not efficient at all and the FETs are heating up wayyyyy too hot. Some of that extreme heat is leaking out a bit through the high resistance paths and warming the screws.

I'm very inclined to think SteveO did a good job. But there's no way to know how far you can push the setup without taking some temperatures. Is there any way you can borrow a temp meter that has a thermocouple you can put next to one of the FETs during your test rides?

 

[John in CR]

Thanks John,

I'll hook up the thermometer and get actual temps, and in the meantime I'm going to step up the current limits a bit now that I have at least somewhat better cooling with the change in how air flow is directed inside the controller.