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

[Hyena]

Having a switch to change voltage or motor wind would have been advantageous if this is a hub motor.

Thats exactly what I was about to suggest. Especially in your case John where yours is wound for such high speeds.
You almost need a little 5ah 24v pack to switch to!
Riding slow definitely sounds like an issue. God forbid we'd actually pedal at these walking speed :lol:

If you noticed your controller was red hot I guess that should have been a warning bell to give it a rest. Fitting a temperature alarm or better still a thermostat that cuts power to the controller or inhibits the throttle would probably be a good safety mechanism. I've noticed mine gets warm (maybe 50 degrees - warm to the touch but not hot enough to want to take you hand off) but both my controller failures have been within seconds of powering up the bike so the temperature shouldn't be an issue.

 

[swbluto]

Regarding my controller failure, it's definitely something about the very low speed, possibly some unintentional throttle pulsing on such a rough road, but not speeding up and slowing down of significance. It was very low power riding, less than 100w.

Interesting. Do you have a meter of some sort to observe the power going in? Something like a cycle analyst that'll tell you the max amps, even though it may not seem obvious from the instantaneous readings, would be insightful. A data logger would be better.

 

[Hyena]

Hyena,

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

24S brings the pain! Assuming the same 80 amp limit I was using before you're looking at phase currents ~175 amps off the line.

Here's a graph plotted for 24S lipo using a 9x7 9C in a 26" wheel with a 65a controller - which would be the case for alot of people that are running Lyen's 18 fet 4110 or 4310 65a controllers.

You'd definitely want an 18 fet controller but the phase current isnt alot higher than what I was running with a 12 fet. This is dangerous for me but allows a reasonable safety margin for an 18 fet. You can see from the graph that you're current multiplying for alot longer though, until nearly 50km/hr which would be for alot your ride when not WOT and blasting along. I'm not sure what top speeds people are actually documenting on 24S but I suspect wind resistance etc knocks the figure on paper back a hell of a lot.

If you ran a 100 amp limit the peak phase currents would be about 195a, equating to about 65a per fet which is probably as much as you'd want to pump through them given the previously mentioned 75a ceiling for them. The higher battery current does push the current multiplication curve to the left though, with it cutting back about 33km/hr

 

[liveforphysics]

A series/parallel switch on the battery could be pretty handy for times when you want to just roll along at walking speed, or stuck in traffic, or chugging up a steep trail at a walking pace. I think maybe a 4:1 ratio would be about right, so an 80v 10Ah pack turns into a 20v 40Ah pack when you flip the switch. This would let PWM duty% increase by ~75%, making the phase currents decrease to roughly 1/4 what they were before. But... it would mean only about 1/4 the torque as well...

I'm thinking if you don't want to sacrifice torque down low with a low-turn-count motor, but want reliability, the best path to get there is a controller using real fet packages, like TO-264's or at least TO-247's. Maybe for something easy to DIY, even the SOT-227b.

A package like the SOT-227b has a junction to case, and case to sink combined Rth of 0.23C/W... The TO-220 packages are more like 1C/W. That means if a TO-220 fet and an SOT-227b fet both have an identical RdsOn, you're looking at about a 4x increase in the continuous current handling over a TO-220 package with identical silicon rated specs. But, at least for the one I'm looking at here, it's got an RdsOn typical of about half what the TO-220's in the 100v range are packing. You also get legs on the package rated for 200amps RMS continuous. Making a 12-fet controller from these buggers would be something in the ball-park of a 48fet controller made from TO 220 fets.

The reason you don't see this stuff used in ebike controllers is fairly simple. You would be looking at a wholesale price of $110usd just for the FETs alone to make a 12fet controller using SOT-227b packages, and it can't be simply solder-bathed together to assemble, must be hand mounted and hand assembled to a custom bus-bar array. From a manufactures perspective, this seems like a huge parts cost and pain in the ass vs just adding another handful or two of $0.50 TO-220 fets that can just run through a solder bath.

 

[Icewrench]

Infineon at part throttle...



Edit : It is a poor pic 10 miliseconds part throttle 50 volts scale + and-.

 

[swbluto]

Infineon at part throttle...

Mind explaining what you're measuring? What's the magnitude? (What's each vertical interval worth?)

(Although, looking at the graph, it doesn't look like it has vertical interval marks. Eeeks. What's the peak value then?)

 

[TylerDurden]

That's a 9C 36V, no load, probing at two phase leads.

here is the full throttle shot:

http://endless-sphere.com/forums/viewtopic.php?f=2&t=19681&p=287260#p287260

I'm ASSuming that's amps.

 

[John in CR]

The reason you don't see this stuff used in ebike controllers is fairly simple. You would be looking at a wholesale price of $110usd just for the FETs alone to make a 12fet controller using SOT-227b packages, and it can't be simply solder-bathed together to assemble, must be hand mounted and hand assembled to a custom bus-bar array. From a manufactures perspective, this seems like a huge parts cost and pain in the ass vs just adding another handful or two of $0.50 TO-220 fets that can just run through a solder bath.

I wonder if that's what they're doing for the multi-KW scooter controllers. Those hubmotors have an even higher Kv than mine and have to be reliable at low speeds carrying a much heavier load because the scooter weighs 250-300lbs or more. I can get those kinds of motors with matched controller for only double the price of mine.

 

[liveforphysics]

When you see the quick dash of pwm at the start, then the pause, then the middle section, then the pause, then the end section, this is the pattern a sine wave controller switches with to create a current waveform in a sinewave shape. Its really weird to think that jagged pwm, pause, jagged pwm, pause jagged pwm results in current making one half of a smooth current sinewave but it does when that switching pattern is switched across an inductor.

Did anyone know these things are doing sine current?? Are they all doing sine current?? Is it a setting we can program?

Infineon at part throttle...



Edit : It is a poor pic 10 miliseconds part throttle 50 volts scale + and-.

 

[swbluto]

When you see the quick dash of pwm at the start, then the pause, then the middle section, then the pause, then the end section, this is the pattern a sine wave controller switches with to create a current waveform in a sinewave shape. Its really weird to think that jagged pwm, pause, jagged pwm, pause jagged pwm results in current making one half of a smooth current sinewave but it does when that switching pattern is switched across an inductor.

Did anyone know these things are doing sine current?? Are they all doing sine current?? Is it a setting we can program?

Infineon at part throttle...



Edit : It is a poor pic 10 miliseconds part throttle 50 volts scale + and-.

There was a video I did of the battery current at one time. I think the shape of the wave will correspond to the motor current, but the peaks will likely be higher since some of the current contributed to the motor will be from the capacitors and the video didn't see the current "after the capacitors".

[youtube]OzenQU_lcBY[/youtube]

I could get a motor current video...

 

[johnrobholmes]

These aren't sine controllers, they just chop. The sine wave is from the motor bemf.

 

[liveforphysics]

The coils only know the current, they dont feel voltage.

That funny chop, pause, chop, pause, chop waveform IS how a sine generator makes a sinewave with FETs.

These aren't sine controllers, they just chop. The sine wave is from the motor bemf.

 

[swbluto]

These aren't sine controllers, they just chop. The sine wave is from the motor bemf.

Yes, they aren't "sine controllers", but when you apply to a square voltage wave to an inductor, you do get a sine-shaped current curve since the inductance smooths the "current".

"True Sine" controllers probably use an analog approach like an LC tank circuit to create the sine drive.

 

[johnrobholmes]

We still don't have any sine controllers out there that are usable, they are all 6 step block commutation. I don't care what you want to call them, but they don't output anything near an actual sine wave when the motor isn't there to smooth it out.

 

[Icewrench]

The scope leads were attached to 2 of the phase wires. My guess is that we might be looking at a power feed pulse, then a switch to a return pulse. Kinda out on a limb here. But I thought the cycling within the phase pulse was interesting.and real darn fast.

My scope only reads voltage.

 

[liveforphysics]

Remember, only the current form is felt by the motor. A motor doesnt care if the wire has a thousand volts in it, zero volts it, or 1 volt in it, the only thing making any magnetic field is the current, and that pattern oddly makes a pretty as can be sine wave in the current. Its the same pattern a sine wave DC to AC inverter uses. The "true sine" inverters use the same pattern, but switch it through an onboard inductor to smooth it, and a normal inverter lets the load of your device do the work.

We still don't have any sine controllers out there that are usable, they are all 6 step block commutation. I don't care what you want to call them, but they don't output anything near an actual sine wave when the motor isn't there to smooth it out.

 

[Tiberius]

These aren't sine controllers, they just chop. The sine wave is from the motor bemf.

Yes, they aren't "sine controllers", but when you apply to a square voltage wave to an inductor, you do get a sine-shaped current curve since the inductance smooths the "current".

"True Sine" controllers probably use an analog approach like an LC tank circuit to create the sine drive.

Not quite. It depends on the value of the inductance. The PWM frequency is chosen so the inductance is effective, but the effect should not be seen at the commutation frequency. Indeed if the inductance was effective at the commutation frequency, it would be very difficult to drive the motor.

Although these are usually called BLDC or PMBLDC motors, technically they are in fact AC motors.

Nick

 

[liveforphysics]

Here are some pics of the switching waveforms used by sine inverters and sine drives.

This is the waveform used by a "true sine" inverter.

 

[swbluto]

Revisiting the free-wheeling losses, I noticed that wikpedia mentioned that...

http://en.wikipedia.org/wiki/Buck_converter

"By comparing these equations the reader will note that in both cases, power loss is strongly dependent on the duty cycle, D. It stands to reason that the power loss on the freewheeling diode or lower switch will be proportional to its on-time. Therefore, systems designed for low duty cycle operation will suffer from higher losses in the freewheeling diode or lower switch, and for such systems it is advantageous to consider a synchronous buck converter design."

When I calculated switching losses in another thread, it appeared that the amount of heat was comparable to the amount of heat generated by mosfet resistance at high currents. I haven't done calculations for low phase currents, but just for quickie assumptions that I know is probably wrong, let's assume that the losses are comparable at lower phase currents. With a mosfet resistance of 10 mOhm (let's assume it's running hot), the amount of power at 20 amps is 4 Watts. Let's double that to account for switching losses (8 watts) or quadruple that if you prefer for a "high estimate" ( 16 watts). It appears that at 20 amps, you'd also encounter 20*1.3 V = 26 watts of free-wheeling losses, so it appears free-wheeling may account for the majority of the heat at low duty cycles. Let's assume that's for a given mosfet and the mosfet has a total thermal resistance of 2 C/W, with 42 Watts, it looks like you're looking at possibly 84 degrees above ambient or, with it 30 degrees c outside, it looks like the mosfet temperature would converge to 104 degrees C. Compared to a controller that might normally be 50 degrees C during normal usage, I could see how super low duty cycles may increase the risk of failure if you're running at super low throttle for a period of time.

So, in essence, it appears the solution may be a synchronous controller. Anyone know what controllers out there are synchronous?

 

[Tiberius]

I would question whether its worth the bother in many cases of generating a sine wave drive for a motor.

a) The motor itself is not necessarily a sine wave. Ie., the Bemf may not be a sine wave.
b) The "smoothness" is only going to be seen at low speeds.

Nick

 

[rhitee05]

swbluto, that's exactly the point I think I tried to make a little while ago. Diode loss (call it Pdiode) should always be higher than the Rdson loss (call it Pon) - if your FET has more than 1.3V drop while fully on, you have other problems. The total loss is roughly Pon*D+Pdiode*(1-D), so the lower the duty cycle the more diode loss matters.

FWIW, the formula for switching loss is Psw=1/2*Vbatt*Iphase*(Ton+Toff)*Fpwm, where:
Vbatt is full battery voltage
Iphase is the average phase current
Fpwm is the PWM frequency
Ton and Toff are the turn-on and turn-off times respectively.

Consult the figure below:


If you're looking at waveforms, Ton is equal to the time in the regions marked 2 and 3 in that plot. The switching loss occurs because current reaches it's full value before the voltage across the MOSFET starts to drop. So, for a very brief time, the MOSFET is feeling losses equal to Vbatt*Iphase! The turn-off waveform is the same just in reverse.

I once remember someone saying the typical switching time was 1 us for these Infineon controllers. Seems high, but just use it as an example. In a system with 50V battery pulling 100A phase current:
Psw = 1/2*50V*100A*(1us + 1us)*10kHz = 50W.
Ouch. However, only the top FETs are PWMing, so we can assume no switching losses for the lower FETs. And, each bank of FETs is only switching 1/3 of the time due to commutation, so now we're at 16.7W per bank, spread over however many FETs per bank. Roughly 5.5W per high-side FET in an 18-FET controller using these numbers.

There are some other switching losses associated with charging and discharging the gate capacitor, losses in the resistors, etc, but those are all very small and unimportant in the big picture.

 

[liveforphysics]

I could see how super low duty cycles may increase the risk of failure if you're running at super low throttle for a period of time.

O
M
G

Has it been 2 full years of you fighting me on this now? Is it finally over? Do you finally trust Matt when he says his controllers get hot when he takes a hill at light throttle, and stay cool when he climbs up it WOT? You're not just going to tell Matt it's just because he doesn't know how to feel the temperature? (or what ever other 10's of excuses you would try to rationalize why every one telling you empirical evidence light throttle makes their controllers run hotter are all just fools, because you put a GUI on a spreadsheet, making you the God of all things motor related.)

 

[ZapPat]

Since you guys are talking about motor BEMF waveforms, here's a real world example of what a hub motor produces during commutation. Many thanks to Justin for posting this, it has come in handy a number of times and is very informative of the true nature of our "BLDC" hub motors.

This graph is taken from phase to phase (green), so I'm guessing that the ground to phase waveforms must look very close to sines, with maybe some slight flatening of the sine's peaks. What we can be sure of is that it's very far from the square wave (pink) we drive it with. The end result is a non-constant amount of current (blue) being drawn by the motor, or a kind of commutation ripple if you will. This is what causes that vibration we can hear in direct drive hubs, specially in light ones like the 9C. These hubs would otherwise be much more silent during strong current demand periods.

What we want to do I believe is match the motor's BEMF shape but with some added positive offset that will produce a nice smooth current draw through the whole commutation cycle instead of those two spikes we see above.

However the downside to sinewave-like ouput in a controller is higher switching losses since you would never get to that PWM-free state that 6 step controllers get to at WOT and doing no current limiting. I suppose a hybrid low-speed sine / high-speed 6 step solution could work well though?

Pat

 

[swbluto]

Yes, if you have high physical loading like going up a hill, then the freewheeling losses will definitely be significant. Nothing like "still relatively high" phase amps and a 1.3V multiplicative constant of a diode to ruin your day, despite the relatively low resistive losses. I think the assumption behind air planes is that when one reduces the throttle, the physical loading decreases like .... *always* - slower propeller = less loading on the motor. Unlike a bike where your physical loading can actually increase regardless of the throttle amount, like encountering a hill.

Anyways, I charted some example values to get a feel for the losses. This makes some simplifying assumptions, some of which are grossly simplified, but I think the trend is about the same. I was going to post it but it seems like it needs a bit more work to clarify under what situations the given phase currents would actually be encountered.

Anyways, in my example calculations, the free-wheeling losses seem to peak at 40% throttle and the total controller heat peaks at 90% throttle. The full-throttle ESC heat seems to match the heat generated by the ESC at 20% throttle; below 20%, the ESC heat is less than full throttle. This assumes a 404, 100v, .002 ohm controller, and accelerating on flat-land from a stop (And a 200 amp phase limit). Cruising on flat-land calculations would probably be different in that full-throttle phase current probably wouldn't be that high for <40 mph bikes, and so total controller heat might actually be greater at 10-40% throttle than full throttle due to free-wheeling losses.

This doesn't seem to explain why a controller is hotter at "super low throttle"(Like 10%) than full throttle as in John's case, but I'm thinking that airflow may have something to do with it. A controller probably cools a lot better at 50 mph than it does at 5 mph, assuming exposure to the 'wind'.

Btw, my simulator in the current form calculates freewheeling losses and resistive losses but ignores switching losses. So, just need to add in a switching formula and it's good to go as a controller heat simulator.

So, anyways, the important question:

Does anybody know of a synchronous brushless controller?

 

[John in CR]

Is it possible that the super low speed/low load issue creating failures is the result of big thermal imbalance between high and low side fets? I've had 3 of these identical failures. None occurred during the very low throttle use, instead all 3 occurred on the first significant twist of the throttle immediately after the low speed use? eg My thought is that half of the fets are extremely hot just from being switched on/off so much at extremely low duty, and then when a load of significant current is requested half of that phase's fet bank allows more current than the other half can handle because they're much hotter already.

John