**Important** reality check on motor, voltage, current etc.

[Hyena]

I don't know what he meant by TO-220 and TO-247 packages

It refers to the physical appearance/size/characteristics of the fets. The ones used in most controllers we've be familiar with are TO-220s (on the left in the picture below) The TO-247's are more like the ones on the top right (it's not labelled but it looks like a 247 to me)

 

[John in CR]

Then the graph isn't representative. Notice though in the graph that max power falls off above that knee. Falling torque and falling power mean decreasing performance.

Also note that if you mod the shunt, the current settings are incorrect until you recalibrate. If you set the limit to 80A and reduce the shunt resistance by half, then the controller thinks 160A is 80A. That's likely what has lead to your blown controllers. The controller needs to know the shunt resistance in order to calculate current.

John

 

[John in CR]

I don't know what he meant by TO-220 and TO-247 packages

It refers to the physical appearance/size/characteristics of the fets. The ones used in most controllers we've be familiar with are TO-220s (on the left in the picture below) The TO-247's are more like the ones on the top right (it's not labelled but it looks like a 247 to me)

Thanks, then the 75A that LFP threw out there was more of a generalization. Got it.

John

 

[rhitee05]

Sorry, John, didn't mean to toss around too much lingo. Hyena has it correct, almost all the common controllers including Infineons use TO-220 packaged FETs. A few controllers (like my Headline/Cyclone controller) use TO-247 FETs, which are a little wider. TO stands for "Transistor Outline", there are a whole bunch of TO-X standardized packages.

Another picture from Wikipedia:
http://en.wikipedia.org/wiki/File:TO-220_Package_Four_Different_Projections.jpg

I seem to remember looking at the data sheet on the 4110 mosfets and saw something like 115A. Also, MWKeefer mentioned to me once that the common mosfet in controllers that has a voltage limit of around 83v was good for 100A, though I don't remember the part number.

If you were confused so far, this is where it gets worse. There isn't really a standard way of rating MOSFETs. Different manufacturers might use slightly different tests, and there are frequently multiple ratings under different constraints. I'll pull some numbers from the common IRFB4110 FETs as an example.

- The silicon die inside the FET is rated for continuous 180A assuming the case is at 25C. At 100C, it's rated for 130A continuous. That's what the silicon itself is capable of. This is an absolute maximum - any higher risks letting out the magic smoke.

- Connecting the silicon to the package leads are several thin wires called bond wires. According to IRF, these wires are good for 120A continuous. This is also an absolute maximum.

- IRF rates the 4110 for up to 670A (!) in very short pulses. That's based on the thermal limits of the die - smaller pulses can be longer or more frequent, etc.

- Finally, IRF recommends a maximum of 75A continuous for all leaded (non-surface mount) packages. This limit is based on I^2*R heating of the leads (which is what LFP stated a while ago here). Since this is a thermal limit, higher pulses are okay with wide spacing, long enough for the heat to dissipate. The limit is the same for both TO-220 and TO-247 FETs. The limiting factor here is how fast the leads can sink heat to the PCB, so with a few tricks you can do a bit better. That's why they call it a "recommendation," since it's application-specific. I wouldn't count on our controllers to do any better than the baseline 75A, though.

It's pretty common these days for the silicon die to be capable of more current than the package itself. Just like everything else these days, it's a little bit of a marketing gimmick to put the highest number front and center. Look at the top of the 4110 datasheet and you'll notice they emphasize the max 180A current that you'll never be able to achieve.

Stick to 75A per FET and you should be pretty safe.

 

[amberwolf]

FWIW, Hyena, all the top row look like TO-220 size and lead spacing to me. I have a few TO-247 FETs and their leads are much further apart, as well as thicker. I have to go find them and take a pic, since the web fails me for comparison pics.

I don't know what he meant by TO-220 and TO-247 packages, apparently EE lingo, though I do see you guys throw the term "packages" around pretty often.

TO just stands for "Transistor Outline", and the number is which version of that outline is used for a particular physical package that surrounds a transistor's silicon die and gives it wires to hook up to the outside world.
http://en.wikipedia.org/wiki/Transistor#Packaging

The one we see most often for phase FETs:
http://en.wikipedia.org/wiki/TO220
Used to be you'd see this type for power transistors:
http://en.wikipedia.org/wiki/TO3
but they took up a lot of space. I think they could be generally better as the whole surface was metal, but silicon only touched the flange side, not the dome side, which usually had air (or inert gas) underneath. So no direct conductivity of heat to the front of the case. Even plastic is theoretically better than the air gap.

There are a few versions of it, including the TO220FP which is totally plastic encased (except for the leads) for electrical isolation (but not good for thermal conductivity). The TO220AB I think is the one with regular metal tab and metal back, with thicker/wider legs for getting that ~75A capability.

TO-264 is more desirable; larger package with more surface area and much thicker/wider legs. I forget what the other common large package is. Then there are even bigger ones....

In the smaller types:
An old metal-can style:
http://en.wikipedia.org/wiki/TO-18
for what we now typically see in these:
http://en.wikipedia.org/wiki/TO92

I seem to remember looking at the data sheet on the 4110 mosfets and saw something like 115A.

Sure. There are even some FET datasheets I've seen that show 150A continuous for a TO-220 packaged FET, but in reality that's not usually practical. 100A or more continuous out of a TO-247, easy enough. Many datasheets show specs for more than one package of the same silicon, and often enough the specs are for the biggest package, and do not always have any info about the limits for the smaller package, assuming that an engineer looking at the datasheet will already know the package limitations. I blew up more than a few parts because of that in my old quests to design and build brushed and brushless controllers. :lol: I had no idea about that sort of thing, until I ran into datasheets here and there that *did* mention it.

Also, MWKeefer mentioned to me once that the common mosfet in controllers that has a voltage limit of around 83v was good for 100A, though I don't remember the part number.

The silicon is good for that, and you might get 100A thru the TO-220 package it is enclosed in, but it will heat due to the small lead size.

For anyone without much of an electronics background that wants way too much information about transistors, this page has it:
http://101science.com/transistor.htm
Can't vouch for it's accuracy, just skimmed it, but it has a lot of explanations/etc.

 

[Hyena]

If you set the limit to 80A and reduce the shunt resistance by half, then the controller thinks 160A is 80A. That's likely what has lead to your blown controllers. The controller needs to know the shunt resistance in order to calculate current.

Yeah, I may have introduced more badness there.
I added a few strands of copper to the shunt that resulted in peak currents of 120a. I then dialed the rated current back in software to around 40 amps which gave me a watt meter measured battery current of 80 amps. This is what I was after, a controller that had a high hardware limit that I could easily tweek back down with the software while experimenting. In hindsight this may not have been the best of ideas... :lol:


Also re: fet packages, that pic was just the first one I came across, see this thread here, it gives you a better indication of the size differences and also the thermal characteristics of the 247s

http://www.endless-sphere.com/forums/viewtopic.php?f=2&t=17188&p=251463&hilit
Hopefully Lyen makes some of these soon

 

[amberwolf]

Yeah, here's the one I used in my first successful brushed controller rebuild, on the top upside down in the pic, just sitting on the heatsink. It's an OnSemi NTY100N10G, 100V 100A, in a TO−264 package:

It is plastic encased everywhere except the rear face and the leads. Makes bolting it down not require a white spacer top-hat washer, and the edges can touch other FETs without fear of shorting.

The one on the left upside down is a 4110, TO-220AB package, out of a damaged controller.

Below it is something in a TO-220FP form the late 90's. In the middle is a Toshiba TV transistor in one of their own case designs, not done to any standard. it's leads are about like a TO−264 but the package is considerably smaller in volume and surface area. It's also fully enclosed like the ones on either side, so no electrical conductivity to the back, and poorer thermal conductivity too.

So there is quite a variety of packages, and some are definitely better at conducting high currents than others, while still dissipating the heat generated by doing so.

 

[John in CR]

I looked up the datasheet and the IRFB4110 has a max continous rating of 180A @25°C and 130A @ 100°C. I'm glad I looked because the 4115's in my 150V max controllers have a 104A and 74A continuous rating at those respective temps, so I will have have to be more careful in those settings.

Given those ratings for the 4110's, and assuming we've adequately beefed up the traces to handle these phase currents, is it reasonable to expect durability with phase current limit settings of 100A, 200A, and 300A for 6fet, 12fet, and 18fet controllers respectively, or is that still too high given the imprecise application of the phase current limit? Or are there current spikes that swing too wildly requiring a more conservative setting?

John

 

[swbluto]

is it reasonable to expect durability with phase current limit settings of 100A, 200A, and 300A for 6fet, 12fet, and 18fet controllers respectively, or is that still too high given the imprecise application of the phase current limit?

100 amps per mosfet? Uhhh, no. 70-75A is the region where the TO-220 package becomes the limiting factor as liveforphysics has explained. For reliability, I'd probably (Likely conservative) rate it at 70% of the package limitations or 50-55 phase amps per mosfet.

In a 12 mosfet controller, that'd be 100-110 amps. I'm personally running mine at 55.

 

[amberwolf]

Regarding the IRFB4110 specs:

From IRF's datasheet, at the top right corner:

• ID (Silicon Limited) 180A
  ID (Package Limited) 120A

where this note applies to the 180A:

Calculated continuous current based on maximum allowable junction temperature. Bond wire current limit is 120A. Note that current limitations arising from heating of the device leads may occur with some lead mounting arrangements.

Meaning, the case design and lead size is probably not suitable for those currents.

The stuff I discuss below is the Absolute Maximum ratings, and in many cases you might not want to design around being able to use those, but rather go with some percentage lower to give longer life and greater margin of reliability.

• ID @ TC = 25°C Continuous Drain Current, VGS @ 10V (Silicon Limited) 180A
  ID @ TC = 100°C Continuous Drain Current, VGS @ 10V (Silicon Limited) 130A
  ID @ TC = 25°C Continuous Drain Current, VGS @ 10V (Wire Bond Limited) 120A

Notice on that last one, the 25°C. Meaning that when the case temperature is still 25°C it could theoretically deal with a max of 120A at the wire bond. It won't be at that temperature once you start putting current thru it inside a controller.

Then there is the power dissipation the case itself can handle:

• PD @TC = 25°C Maximum Power Dissipation 370W
  Linear Derating Factor 2.5W/°C

Let's say it gets to 75°C inside the controller during use, and that the casings of the FETs normalize at that. That's 50°C above 25°C, so 50 x 2.5W = 125W less that you can run thru it, or only an absolute maximum of 245W thru each 4110.

Then there's the Thermal Resistance of the case.

• RθJC Junction-to-Case 0.402°C/W
  RθCS Case-to-Sink, Flat Greased 0.5°C/W
  RθJA Junction-to-Ambient 62°C/W

For every watt you put thru the case, it'll have that much extra problem dissipating the power, and heat up more because of it. Assuming a perfect cooling solution for the heatsink it's bolted to, then at 245W, it'd rise 245 x (0.402 + 0.5) = 220.99°C of thermal resistance, if I did that right (probably not :lol. Meaning that you can't actually put anywhere near 245W of power thru that case for any length of time, or you'll evaporate the silicon die inside, which generally has an operating temperature of 175°C absolute max.


Another consideration is that the RDSon goes up with temperature, and the low value spec'd for is only at 25°C. There is a Fig.4 of "Normalized On-Resistance vs. Temperature" charted at 75A and a gate voltage of 10V, starting at -60°C and going up to 180°C. The "normalized" means that they don't list the actual resistance on the left side, they list a multiplier, using 1.0 at 25°C. Just to make it complicated for us to read.

Assuming worst case RDSon of 4.5mΩ, then it would range from 2.25mΩ at the low temp to 11.25mΩ at the high temp (probably right before the die starts to fail). So the resistive heating will get worse (probably double or more) as the FET heats up, making the whole problem even worse.

Then we have the reality that almost certainly there will not be a really flat, perfect connection between the back of the FET and the heatsink, and either too much thermal grease or none at all, making for an even higher thermal resistance at that stage. Worse, almost all the controller designs I have seen do not bolt the FETs to the case, they bolt them to a bar that *then* bolts to the controller case, which is yet another thermal junction, increasing the heat problem even further. Even if you lapped the back of the FETs and both sides of the bar, as well as the inside of the controller case, so that all of these things mate so perfectly you don't even need thermal grease, it'd still be less efficient than a solid piece of metal.

And then you have to get rid of the heat, and hardly any controller has much in the way of effective heatsinking fins on the outside. Plus however they're mounted on the bikes, there often isn't enough airflow to keep them really cool (or the environments don't allow it anyway, like here in Phoenix).

I've never been all that good at math, so I may have screwed up calculations above, but the numbers I started with are from the spec sheet, so someone else can check me on the math and correct me if I'm messed up in the head again.


Basically all this means is that whatever the data sheet says is a max, you should derate it a LOT for actual usage conditions. :lol:

 

[Tiberius]

I just did a quick derivation (which I think is correct...) that might help clear up some of the confusion:

Assumptions:
- The PWM frequency is fast enough that the current waveform can be considered a straight line (should be true)
- Neglecting "small" voltage drops - wires, FET on resistance, etc.

A controller is functionally identical to a DC-DC buck converter. We can ignore the three phases and use a simple model:
View attachment 3

When the controller is on, the model is thus:
View attachment 2

Using Kirchoff's Voltage Law, we can derive the equation: Vbatt-Vbemf-Iavg*Rphase=Lphase*dI/dt, where Iavg is considered to be the average current over the PWM period (simpler this way).

So, during the on-state the current slope is given by: dI/dt=(Vbatt-Vbemf-Iavg*Rphase)/Lphase

When the controller is off and the diode is freewheeling:
View attachment 1

Using the same technique we derive the KVL equation: -(Vbemf+Vdiode+Iavg*Rphase)=Lphase*dI/dt

During the off-state the current slope is given by: dI/dt=-(Vbemf+Vdiode+Iavg*Rphase)/Lphase

The waveform looks like this:


Using the slopes found above, we can calculate the current ripple during on- or off-time by the duty cycle:
deltaI(on)=(Vbatt-Vbemf-Iavg*Rphase)/Lphase*D (where D is the duty cycle)
deltaI(off)=-(Vbemf+Vdiode+Iavg*Rphase)/Lphase*(1-D)

If the system is operating in steady-state at a constant average current, the magnitude of both deltaI terms must be equal:
(Vbatt-Vbemf-Iavg*Rphase)*D=(Vbemf+Vdiode+Iavg*Rphase)*(1-D)

Note that the inductance divides out. The actual PWM period is not important, so long as it's short enough that assumption #1 above applies, because the period would also divide out.

We do a little bit of algebra to get:
Iavg=(Vbatt*D-Vbemf-Vdiode*(1-D))/Rphase

Compared to the other terms, Vdiode is pretty small, so we can get rid of it and simplify to:
Iavg=(Vbatt*D-Vbemf)/Rphase

This equation agrees with a few things we know to be true:
- If the motor is at a dead stop (Vbemf=0) and the controller goes to WOT (D=1, no limiting), the current is Iavg=Vbatt/Rphase, the Ohmic limit
- If the throttle setting exactly equals the motor BEMF (Vbatt*D=Vbemf), no average current flows (in this simplified model)
- Similarly, at WOT (D=1), average current also drops to zero when BEMF reaches battery voltage (Vbatt=Vbemf)

I hope you're still with me, because now I'm going to point out how this helps the discussion!
- All other things being equal, decreasing the duty cycle decreases the average phase current flowing. This is why limiting works.
- All other things being equal, higher speeds -> higher BEMF -> lower phase current.
- If we assume limiting is in effect and phase current remains constant, higher speeds -> higher BEMF -> higher duty cycle D. This is why the throttle gradually gains more effective range as the bike speeds up when limiting is applied.
- Holding a constant throttle (D), if you go up a hill and speed starts to drop (Vbemf decreases), Iphase will increase assuming limiting is not in effect.

Eric,

This is good stuff. Why don't you post it in the technical reference area?

I agree with your derivation, but for one minor point. In calculating the deltaI values, you could use the instantaneous i instead of Iavg. Then you take the values as the ripple tends to zero (ie, L is large or PWM freq large). The end result is the same, but it keeps the pedants at bay.

Nick

 

[liveforphysics]

While were're on the subject of Mosfet datasheets, they have another part that is pretty shady, but an industry standard.

We know it's easy enough to calculate a heat estimate for the conduction state of the FET, just RdsOn * current^2. So, you might think, ok, as long as this number is lower than the Pd-watts number listed, everything should be good. Well... this Pd-watts number is the amount of heat the fet package is capable of transferring when it's at it's peak temp before failure, layed up against a miracle heatsink at ambient temperature.

So, if we take the RdsOn of the IRF4115, it's 9.3mOhm nominal. But... That's with a 20degC junction temp. If you look at the figure 4. graph, you can see that if you're going to be pushing the FET to it's operational junction temp, like ~140C, that RdsOn gets bumped up by a factor of about 2.3x.

So, when you first turn on the bike on a cool morning, and the FET junction is at 9.3mOhm, when you pull 100amps through it, you're looking at 93watts of heat at the junction. If you're running the thing up a hill, the same 100amps becomes 213watts of heat being produced.

Now, you gotta looks at the thermal transfer between the junction (silicon) to the fet's back, and then from the fet's back to the sink. For this device under the best of best conditions (definitely not what you get in an infinion with a silicone pad) , you're looking at 0.4w/c + 0.5w/c, or 0.9w/c. This means, if you've got a magic heatsink that stays at ambient temp always, then you're actually only capable of 126w (the max Pd rating of 380w doesn't have a lick of bearing in this circumstance) being continuously created if you don't want to exceed 140degC junction temps, and this is with a magic heatsink that never gets warm. This translates to 76amps being the maximum current the IRFB4115 can handle continuously when connected to a magic heatsink that never gets warm. Curiously close to the 75amp limit on the continuous current the legs of the fet can handle (total coincidence in this case). In real life, when the heatsink gets up to 30-40c, you're looking at ~50amps or so that the fet is capable of handling continously, even though it could handle 420amps per fet for bursts (fractions of a second).

 

[John in CR]

93 Watts of heat seems incredibly high especially times 15fets in my controller. That would mean my case is dissipating something 1400W on a hill, which doesn't make sense. Because I ride around in current limiting mode most of the time, my controller case barely gets any warmer during hill climbs, but anything close to 1000W of heat loss seems way high. Can a controller really be less than 90% efficient under any conditions except very short ones?

This would seem that it's very worthwhile to address better controller cooling as a means to increased performance. I'm putting holes in motors, so why not do the same with controllers?

Also, I'm looking to set the phase current limit as the one I feel is of prime importance, so I'm not talking about battery side any more unless I'm trying to calculate overall consumption. My battery side current limit might as well be set the same as my phase current limit on my bikes.

John

 

[liveforphysics]

That 93watts of heat is just one fet under 100amps of load and its only 93w for the instant you first switch it on and its at room temp. Its more like 200w+ once it been running for a few seconds.

Fortunately, the controler doesnt load just 1 fet at a time and a you parallel them the current is shared between them, which cuts the heating down by 4 on each fet, and each fet is only loaded for about 2/3rds of them time, so you're looking more like 30watts of heat or so per fet when a controller fitted with 12x irfb4115's is supplying 100phase amps. That would mean something like ~360w of controller heating.

Those wicked 85v IXYS fets i recomended as an option for your 24fet controler that use the entire tab for the silicon thermal connection to the back tab and double-thick legs would make for a controller able to around 400 phase amps continous for making the same 300-400w of controller heating.

 

[John in CR]

Even 360W seems high. My 50W halogen light gets far hotter than my controllers. Sure it has maybe 20% the surface area for cooling, but it's also out in front getting full wind instead of tucked under the motorcycle seat with drastically less flow.

That controller sounds like a monster that can take my motors to full task, so they'll really need the ventilated covers for that controller. It should be fun.

John

 

[justin_le]

Thanks man
Yep 18S lipo that comes off the charge at 75v. I used a voltage of 70v as I wasn't sure exactly how the simulator allowed for nominal voltages and sag but I guess now with the battery resistance of 0.04v I can recalculate it more accurately using the nominal voltage of 67v.

So I guess from the graph I can now assume I'm going to be drawing phase currents of ~150a as I move off the line. Is this too much for a 12 fet with 4110s ? In theory should it be able to handle spikes of 200 amps ? Or does the 4x battery current thing come in to play ? I guess at 1/4 throttle it'd only be drawing 30-40 amps max which even at 4x still brings it back to ~150a max phase current.

The problem with this setup is that you are at a power point with the hub motor where all kinds of 2nd order effects start coming into play. I showed before that the saturation of the Nine Continent hub motors starts to happen at about 70-80 N-m:
http://endless-sphere.com/forums/viewtopic.php?f=2&t=14494&start=15#p218312

So if the simulator is showing more than 80 N-m, as it does in the first half of your graph, then your actual torque output may be less, and the actual phase current draw would be more. However, at these current levels the 9C windings heat up extremely fast, and in very short order the phase resistance will be up 60-70% higher than when it started, which has the effect of reducing your phase current draw but greatly compounding heating issues.

With this combo of motor / voltage, it would be very prudent to set a controller phase current limit of 80-90 amps or so if your controller will allow you to do that.

Justin

 

[justin_le]

How do we go about setting the battery side and phase current limits? Is the battery side limit just to protect the battery, and the phase limit is to protect the controller? Assuming a battery that can handle extreme current, is there any reason not to set both limits the same?
...Where does the battery side current limit really come into play?

The only reason for a battery side current limit is indeed to protect the battery and keep the current pulled from the pack within the spec of what the cells and BMS are rated for. For typical ebike packs, PING batteries etc. this is important. But with the high rate RC cells, there is no reason why you wouldn't just set the battery current limit to be the same as the phase current.

This will change the ride behaviour a little bit though. When you go full throttle from a start, instead of being in a constant input power regime to the hub, you'll be in a constant torque regime, with the motor power continuously increasing as the wheel goes faster and faster.

Justin

 

[John in CR]

I think the guys bringing up the fact that reducing the duty cycle also reduces phase current were on the right track, but they were missing the point that during acceleration the controller is trying to increase the duty cycle, not decrease it. WOT does not necessarily mean 100% duty because current limiting will prevent the controller from going to full duty regardless despite our throttle signal asking for full duty. It's the partial duty that makes the controller work harder, and the higher the current the harder it's working, so current limiting mode is the hardest.

I don't know the math and electronics behind what goes on in the controller, but I believe I have a pretty firm grasp of what's taking place. As Luke mentioned, motors and controllers only see that discrete moment in time, and those moments change at some electrical speed.
eg the controller processor runs through this routine:
Is the throttle signal asking for a higher duty cycle than it currently is?
If yes, then has a current limit been reached?
If not, then step the duty/voltage up by one unit (for which the motor will draw more current)

If you follow this logic through every part of an acceleration starting at 0 throttle at a dead stop, then it becomes crystal clear what goes on. Just apply battery current/duty cycle=phase current to see the current multiplication.
I don't know at what speed these decisions take place in the controller, but they occur too fast for us to notice even many consecutive iterations. I also don't no what the units are for each step up in duty/voltage, but I'd think they must be small. If this isn't how the program in the controller works in it's simplest form, please let me know. It sure seems like the right way to do it to me.

What we don't know, or at least hasn't been discussed yet, is whether partial duty can be harder on a controller even if a current limit hasn't been reached, because of the more rapid switching or narrower ON times in the duty cycle. This would be the case where we are accelerating, but not at maximum acceleration, or going up a hill at less than maximum speed for that hill.

John


We should also discuss whether or not an abitrarily low battery current limit could create instances of unnecessary current limiting by the controller and cause more phase current multiplication.

 

[John in CR]

How do we go about setting the battery side and phase current limits? Is the battery side limit just to protect the battery, and the phase limit is to protect the controller? Assuming a battery that can handle extreme current, is there any reason not to set both limits the same?
...Where does the battery side current limit really come into play?

The only reason for a battery side current limit is indeed to protect the battery and keep the current pulled from the pack within the spec of what the cells and BMS are rated for. For typical ebike packs, PING batteries etc. this is important. But with the high rate RC cells, there is no reason why you wouldn't just set the battery current limit to be the same as the phase current.

This will change the ride behaviour a little bit though. When you go full throttle from a start, instead of being in a constant input power regime to the hub, you'll be in a constant torque regime, with the motor power continuously increasing as the wheel goes faster and faster.

Justin

Thanks Justin,
Great, I love the feel of that constant pull right on up to speed. Plus it means potentially a bit quicker acceleration.
John

 

[Hyena]

However, at these current levels the 9C windings heat up extremely fast, and in very short order the phase resistance will be up 60-70% higher than when it started, which has the effect of reducing your phase current draw but greatly compounding heating issues.

Yeah it does get pretty warm, even with side covers drilled for air cooling and 12 ga phase wires. Even generating heaps of heat more power in still = more power out, and whats a guy to do in the quest for performance and bigger EV grins

With this combo of motor / voltage, it would be very prudent to set a controller phase current limit of 80-90 amps or so if your controller will allow you to do that

Will do, thanks for the advice Justin
What are your thoughts with phase current in relation to battery current ? It sounds like the 2.5x battery current might be a bit of a misnomer ?

 

[dbaker]

Hyena,

Would you please run graphs for 24s Lipo as well? That might illustrate the risks and operational issues with going big!

 

[rhitee05]

What we don't know, or at least hasn't been discussed yet, is whether partial duty can be harder on a controller even if a current limit hasn't been reached, because of the more rapid switching or narrower ON times in the duty cycle. This would be the case where we are accelerating, but not at maximum acceleration, or going up a hill at less than maximum speed for that hill.

Yes, partial duty is always harder on a controller for heat generation. In these controllers that just use the FET body diodes to freewheel, usually the diode drop is the largest source of losses. These diodes typically have a max forward drop of about 1.3V, so loss is 1.3*Iphase.

At the max limit of 75A per FET, that's 97.5W each. Worst-case for this is obviously very low duty cycles, such as a hard takeoff from a stop. This is a reasonable worst-case heat estimate, switching losses and I^2*R losses will generally not be higher. The good news here is that only one bank of FETs is conducting at a time and they rotate as the controller commutates. In a 12-FET controller, you might have losses of 195W at 150A. But that averages out to about 16.25W per FET or a worst-case of 32.5W per over just the low-side FETs.

 

[rhitee05]

I agree with your derivation, but for one minor point. In calculating the deltaI values, you could use the instantaneous i instead of Iavg. Then you take the values as the ripple tends to zero (ie, L is large or PWM freq large). The end result is the same, but it keeps the pedants at bay.

Thanks, Nick. I was trying to make it simple and avoid using the actual exponential formulas, thus assumption #1 that PWM period << L/R. If you want to use the instantaneous current, then I*R varies as the current goes up and down and it gets messy fast. If it keeps the pedants at bay, know that the average current is equal to the instantaneous current at the center of the PWM pulse, so deltaI is symmetric about that point. I think the limit you suggest is still valid, that deltaI goes to zero as L increases to infinity or period decreases to zero.

It may be a little presumptuous to talk about limits, though, considering this is a back-of-the-envelope derivation based on a few assumptions and some hand waving.

 

[Tiberius]

Hi Eric,

What I'm suggesting is a neat trick that removes the need for any hand waving. It doesn't even add any extra lines of maths. Sorry, I used to be a mathematician, then a physicist and then an engineer.

Nick

 

[John in CR]

What we don't know, or at least hasn't been discussed yet, is whether partial duty can be harder on a controller even if a current limit hasn't been reached, because of the more rapid switching or narrower ON times in the duty cycle. This would be the case where we are accelerating, but not at maximum acceleration, or going up a hill at less than maximum speed for that hill.

Yes, partial duty is always harder on a controller for heat generation. In these controllers that just use the FET body diodes to freewheel, usually the diode drop is the largest source of losses. These diodes typically have a max forward drop of about 1.3V, so loss is 1.3*Iphase.

At the max limit of 75A per FET, that's 97.5W each. Worst-case for this is obviously very low duty cycles, such as a hard takeoff from a stop. This is a reasonable worst-case heat estimate, switching losses and I^2*R losses will generally not be higher. The good news here is that only one bank of FETs is conducting at a time and they rotate as the controller commutates. In a 12-FET controller, you might have losses of 195W at 150A. But that averages out to about 16.25W per FET or a worst-case of 32.5W per over just the low-side FETs.

Just an hour too late for my controller. Since I modded the shunt a few days ago, I've done numerous repetitive launches at both max acceleration and partial, all with no issue. What killed it was 200m of road that was so bumpy that I had to ride at about 5mph. Then when I pulled out on the smooth road and took off the controller failed almost immediately. That was the 3rd time controllers blew in exactly that manner on 3 different motors, even one brushed. Each time was on normal takeoff after someone road around for a while at near idle, probably including very small short pulses of the throttle from bumps. Each time the low speed usage made the controller scorching hot.

John