Compact Field Oriented Controller, ASI + Grin, limited run

Alan B said:
Thanks Justin,
I was wondering about the top speed loss, not the efficiency difference. I switched from a trapezoidal to a sine wave controller (Xie Chang to Sabvoton) on my Cromotor bike and lost about 10% of the top speed (with no field weakening). Just curious if this is similar. The only bike I have that might work for this has a fairly low top speed, wondering how much I might lose by this.
Thanks!

Hi Alan, great question and observation. I setup a test using a small 6" 10 pole motor with a CA and ran it unloaded from 20V to 55V with both the ASI field oriented controller and a sensored infineon trapezoidal controller, recording the RPM and current draw with a CA. Here's the result
RPM vs Voltage.jpg

For a given motor RPM, both the trapezoidal drive and FOC drive have the same current draw, but the FOC needs about 8% higher voltage. Plotting the actual kV in RPM/V you can see it's a perfectly flat line across the tested motor speed with both controllers, but the trapezoidal drive is 48.73 RPM/V, while the FOC spins it at 45.07 RPM/V, so without using any of the field weakening features of the ASI you loose about 7.5% of your max unloaded speed.
kV Constant vs RPM.jpg

How this affects the loaded performance is something of another question, and I'll try to get a dyno test in with an identical motor running the ASI versus Infineon controllers.
 
riba2233 said:
Thanks for the answer! :D Do you have the info on the max erpm?

Still haven't hit the "max" yet, but I did just get it up to 1200 Hz or 72,000 eRPM, running a 7 pole-pair KV 210 turnigy motor with a 50V power supply. I didn't feel comfortable going any faster than this without a blast shield between me and the motor because it sounded like the turnigy was going to fly apart at any point.

Turnigy Max RPM Test.jpg

But yeah, 72K eRPM is no problem, and I suspect it can go a fair bit further. You can see in the picture above I put a current sense resistor on one of the phase leads to look at the phase current waveform profile on my scope as a possible indicator of it getting close to some limit, but it was hard to make much out as the signal to noise ratio is very poor when the motor is spinning unloaded like this.
 
thats the first really meaningful FOC vs. trapez comparison i have seen. Thanks!

justin, are you able to try max voltage (75V)? what about installing a propeller on the motor to simulate some load? :)
im looking forward for the dyno tests - that will make it clear what benefits each type of controller has.
 
Justin, do you dyno only for performance or also system efficency under same conditions ( watt in + same max.RPM)?
 
madin88 said:
thats the first really meaningful FOC vs. trapez comparison i have seen. Thanks!
justin, are you able to try max voltage (75V)?

Well I realized we used a 63V rated high ripple bus capacitor in the enclosure and so I'd need to take that out and replace it to run up to the max of 75V. So probably not. But 60V with this motor would still be 86K eRPM and I'll make sure it can handle that OK tomorrow once I have the motor better secured.

what about installing a propeller on the motor to simulate some load? :)

That would probably wiser than slowing it down with my hands, which is what I did for at least a couple seconds before the heat reflex reaction kicked in :shock:

im looking forward for the dyno tests - that will make it clear what benefits each type of controller has.

Me too. I've also got a decibel meter so that we could in principle quantify the sound difference while I am at it, but I think that the noise from the dyno load motor might overpower things.

\/ampa said:
Justin, do you dyno only for performance or also system efficency under same conditions ( watt in + same max.RPM)?

When I do the dyno test I have the controller set to maximum power and then sweep the RPM from the unloaded speed down to the point where the dyno motor can't spin smoothly anymore, usually about 80-100 rpm. While doing this I'm recording the motor torque, motor rpm, controller current, and controller voltage. So it would be the system efficiency and system power plotted vs. RPM with each controller. This is what most people would be interested in (ie, how does a given motor/battery compare with the FOC vs a Trapezoidal controller). To do the same condition test that you describe would require running the FOC test at a 7.5% higher supply voltage with a 7.5% lower battery current limit. I'll try to squeeze that in too.

For these comparison tests It would be great if I could also measure the torque ripple too, and while the load cell measuring torque has a very fast dynamic response, the amplifier circuit I've got scaling it to a 0-5V range for the CA3 does quite a bit of low pass filtering so I'd need to rework that before being able to hook it up to the scope.
 
justin_le said:
To do the same condition test that you describe would require running the FOC test at a 7.5% higher supply voltage with a 7.5% lower battery current limit. I'll try to squeeze that in too.

do you think it would be possible to compare the FOC controller in flux weakening mode so that RPM is similar to the trap controller? I think its not that easy, because if no load RPMs set similar they still can differ under load i guess.
maybe 93,5% throttle on the trap controller also is a good option instead of changing the battery voltage.
 
madin88 said:
do you think it would be possible to compare the FOC controller in flux weakening mode so that RPM is similar to the trap controller? I think its not that easy, because if no load RPMs set similar they still can differ under load i guess.
maybe 93,5% throttle on the trap controller also is a good option instead of changing the battery voltage.

Flux weakening would distort the result.
Reducing throttle if it have not the power throttle like a Kelly trapezodial controller is also distorting.

I would do it like Justin mentioned. Adjust the input voltage that both make same max. RPM under same load.
Than meassure under same conditions '' watts in @ xRPM vs watts (torque) out '' and compair results.
 
justin_le said:
For these comparison tests It would be great if I could also measure the torque ripple too, and while the load cell measuring torque has a very fast dynamic response, the amplifier circuit I've got scaling it to a 0-5V range for the CA3 does quite a bit of low pass filtering so I'd need to rework that before being able to hook it up to the scope.

First off, I want to thank you for the absolutely wonderful information you are gathering and presenting.


for the torque ripple, can you have a second load cell amp so you don't have to fuss with your main setup? I imagine it would have to be passively watching the load cell and not applying voltage to it so the two amps didn't fight each other.



Absolutely great work Justin. I'm still a bit behind programming my controller, got food poisoning last Friday and it just pushed everything out.
 
madin88 said:
im looking forward for the dyno tests - that will make it clear what benefits each type of controller has.

OK. Well here is the first of them. For my test setup I was running the dyno bench with a Crysatlyte TC3080 hub motor and a lab power supply set at 48.0V for the supply source. The trapezoidal controller I was using is one of our 25A infineon based devices, the actual measured current at the current limit was 24.5A and then decreaseing to ~23.5-24A as the load increased. For the Field Oriented Controller, I had the BAC500+ set to run pretty much unlimited and then used a CA to effect a 24.5A current limit, so that it had the same max power input as the infineon. Of course it took a second for the current limit to kick in as I was sweeping down the RPM so that is why you see a little overshoot at about 370 rpm.

Dyno Comparisons with TC3080 Hub at 48V.jpg

The graph comparison is very interesting and I'm glad you guys encouraged me to do this, because I hadn't actually given too much consideration on the actual change in behavior you'd experience when switching to a nominally spec'd FOC from a trapezoidal drive.

You can see that the max RPM is clearly on the order of 7-8% higher with the trapezoidal drive as expected.

The peak efficiency is pretty much the same 88-89% for both systems, although with the FOC controller it happens at the correspondingly lower RPM.

But then under loads at lower RPM's the efficiency of the FOC is about 3% better than the trapezoidal drive. As well, the output power and torque are higher for the FOC even with the same battery current draw. So say you are riding the bike at 300 RPM (~38 kph in a 26" wheel). Both the infineon and the BAC500+ controllers would be drawing 24.5A, but the BAC500 would be putting ~33Nm of torque on the wheel, compared to ~31 Nm for the infineon.

The effect is a little bit exaggerated in the graphs since the actual battery current draw on the infineon was more like 24.0A at this point, versus 24.5A on the BAC, but it's still quite evident that under loads all else being the same you get measurably better torque, power, and efficiency with the FOC.

I was not actually expecting this result to be honest, but there it is. In a nutshell:

Trapezoidal = higher top speed (7-8%), more power when operating at these higher speeds
Field Oriented Controller = Lower top speed (without field weakening), but definitely better torque (~5%) and efficiency (~3-4%) under load
 
Thank you Justin.
That are results which are expectable for compairing slow and fast wind motors on same controller.
Could you also make a dyno test with both controllers have same max. RPM?
 
\/ampa said:
Thank you Justin.
That are results which are expectable for compairing slow and fast wind motors on same controller.

No that's not quite true. I was thinking along those lines for a minute but we really shouldn't see the improved efficiency and torque when the same power input is being delivered at the same motor RPM. If we have the same controller and two slightly different motor winds, the torque and efficiency when the controllers are both power limited is nearly identical. The slower motor does NOT show more torque and better efficiency like we have here.

Could you also make a dyno test with both controllers have same max. RPM?

Up next, but I'm going to call it for a day and go home for the night rather than sleeping at work, again.. So stay tuned for this test in the morning.

Alan B said:
Fascinating!

I'm seeing slightly better efficiency on my commute with the FOC controller compared to the trap. The difference is not huge (in the neighborhood of 10%), but I wasn't expecting it. Looks like there is more to this story. Nice!

Hey neat, while data collected 'on a commute' is a bit rough and anecdotal, it's nice when things do line up with measured values. I suppose though that you are also commuting a bit slower with the FOC drive and that would be some part of it too?
 
Maybe a bit slower with FOC, but not much as I was using a 3 speed switch on Trap that was close speedwise to what I'm using on sine, however the sine control is smoother so I suspect the smoothness is part of the efficiency improvement. It certainly is not a controlled enough experiment to place any real precision on it however. In any case a real torque throttle is wonderful, for some reason even better than what I was able to get with CAv3.

It is so great to see some controlled experiments and data, Thanks again!
 
Wow! Can you give us an overlay of just the efficiency difference, but scaled better to see how much better it is? If I'm seeing things correctly, it might be worth 4-5%!!! :shock:
 
The FOC controller definitely gets noticeably less hot during a normal commute that the Infineon did, especially regen seemed to heat the Infineon up a lot but doesn't seem to as much on the Foc.
 
justin_le said:
madin88 said:
im looking forward for the dyno tests - that will make it clear what benefits each type of controller has.

OK. Well here is the first of them. For my test setup I was running the dyno bench with a Crysatlyte TC3080 hub motor and a lab power supply set at 48.0V for the supply source. The trapezoidal controller I was using is one of our 25A infineon based devices, the actual measured current at the current limit was 24.5A and then decreaseing to ~23.5-24A as the load increased. For the Field Oriented Controller, I had the BAC500+ set to run pretty much unlimited and then used a CA to effect a 24.5A current limit, so that it had the same max power input as the infineon. Of course it took a second for the current limit to kick in as I was sweeping down the RPM so that is why you see a little overshoot at about 370 rpm.



The graph comparison is very interesting and I'm glad you guys encouraged me to do this, because I hadn't actually given too much consideration on the actual change in behavior you'd experience when switching to a nominally spec'd FOC from a trapezoidal drive.

You can see that the max RPM is clearly on the order of 10% higher with the trapezoidal drive as expected.

The peak efficiency is pretty much the same 88-89% for both systems, although with the FOC controller it happens at the correspondingly lower RPM.

But then under loads at lower RPM's the efficiency of the FOC is about 3% better than the trapezoidal drive. As well, the output power and torque are higher for the FOC even with the same battery current draw. So say you are riding the bike at 300 RPM (~38 kph in a 26" wheel). Both the infineon and the BAC500+ controllers would be drawing 24.5A, but the BAC500 would be putting ~33Nm of torque on the wheel, compared to ~31 Nm for the infineon.

The effect is a little bit exaggerated in the graphs since the actual battery current draw on the infineon was more like 24.0A at this point, versus 24.5A on the BAC, but it's still quite evident that under loads all else being the same you get measurably better torque, power, and efficiency with the FOC.

I was not actually expecting this result to be honest, but there it is. In a nutshell:

Trapezoidal = higher top speed (7-10%), more power when operating at these higher speeds
Field Oriented Controller = Lower top speed, but definitely better torque (~5%) and efficiency (~3-4%) under load

Thanks for running this test.
PLEASE try to run this test again at different throttle... 75%, 50%, 25% ... vary the amps or whatever your setup needs to simulate this . :shock: :shock: :shock:
 
liveforphysics said:
Wow! Can you give us an overlay of just the efficiency difference, but scaled better to see how much better it is? If I'm seeing things correctly, it might be worth 4-5%!!! :shock:

Here's a better scaled closeup. It seems to average a steady 3% improvement in efficiency right across the board once both controllers are in their current limited regime. And 3% is still no small matter when it comes to overall system efficiency. Trying to eek that much improvement via copper fill, thinner laminations, higher grade steel etc. would be a lot of work compared to just changing the controller's drive scheme.
View attachment 1

If you compare at equivalent torque outputs the difference is even more significant. Say the motor needs to produce 40 Nm to climb a given grade, the FOC would do that at 210 rpm running at 76% system efficiency, while the infineon trapezoidal controller would need to slow down to 180 rpm at which point it is running at a 69% system efficiency. So the real world case at this load would result in a 7% improvement in system efficiency with the FOC setup.


Again, I was not expecting this kind of result, because my gut instinct would have been that the trapezoidal controller would generally have more power and speed rather than less. But that only holds true when you are up closer to the unloaded RPM.

okashira said:
PLEASE try to run this test again at different throttle... 75%, 50%, 25% ... vary the amps or whatever your setup needs to simulate this . :shock: :shock: :shock:

The test I did already contains a full swath of throttle levels. Full throttle only applies above ~400 rpm when both systems are drawing less than the current limit. To the left of this when they are battery current limited, then the %throttle is automatically decreasing downwards as the pwm output adjusts to keep the battery current limit in check.
To actually vary the throttle input would be somewhat fruitless for comparison since the BAC500 has a torque throttle while the infineon controller has a voltage throttle, so "50% throttle" has a totally different meaning in each case.
 
\/ampa said:
Could you also make a dyno test with both controllers have same max. RPM?

Here it is. I repeated the test with the FOC controller using a 51.4V power supply, rather than 48.0, so that the motor had the same unloaded RPM. In order to normalize the input power, I then set a 22.5A current limit with the CA, so that it matched approximately the 24A current limit of the 25A infineon controller.



So in this case, in every single metric the FOC does better. It has a stiffer rollback in RPM as you add torque close to the unloaded speed, it has higher peak power output, torque output, and better efficiency, at every single point on the graph. So if you account for the reduced V/RPM and choose your motor winding and battery voltage accordingly, it's a no brainer.

In light of this data, I now have to largely retract a statement I made to izeman earlier in this thread
justin_le said:
In general there is no measurable gain at all in efficiency with a FOC versus a regular trapezoidal controller, that is not a selling point at all. The motor efficiency may be a tad higher because of fewer harmonics, while the controller efficiency is definitely worse because of switching losses on all the channels.

Scrap that, the gains in motor and drive efficiency are easily measurable and well overshadow the additional controller losses. Improved efficiency is indeed a selling point too :eek:
 
thanks for the tests justin. i very appreciate your work.
further it would be interesting what happens to efficiency if the FOC controller works with flux weakening at same RPM as the max rpm of trapez controller.
when i swapped my infineon for adaptto i noticed very similar consumption in this situation.
 
justin_le said:
\/ampa said:
Could you also make a dyno test with both controllers have same max. RPM?

Here it is. I repeated the test with the FOC controller using a 51.4V power supply, rather than 48.0, so that the motor had the same unloaded RPM. In order to normalize the input power, I then set a 22.5A current limit with the CA, so that it matched approximately the 24A current limit of the 25A infineon controller.



So in this case, in every single metric the FOC does better. It has a stiffer rollback in RPM as you add torque close to the unloaded speed, it has higher peak power output, torque output, and better efficiency, at every single point on the graph. So if you account for the reduced V/RPM and choose your motor winding and battery voltage accordingly, it's a no brainer.

In light of this data, I now have to largely retract a statement I made to izeman earlier in this thread
justin_le said:
In general there is no measurable gain at all in efficiency with a FOC versus a regular trapezoidal controller, that is not a selling point at all. The motor efficiency may be a tad higher because of fewer harmonics, while the controller efficiency is definitely worse because of switching losses on all the channels.

Scrap that, the gains in motor and drive efficiency are easily measurable and well overshadow the additional controller losses. Improved efficiency is indeed a selling point too :eek:

Can you do one more at ~13 amps? I would LOVE to see how things look at a partial relative load.
 
okashira said:
Can you do one more at ~13 amps? I would LOVE to see how things look at a partial relative load.
Since they run different voltage now. Its better to say another dyno test with 500W power draw from both.
 
\/ampa said:
okashira said:
Can you do one more at ~13 amps? I would LOVE to see how things look at a partial relative load.
Since they run different voltage now. Its better to say another dyno test with 500W power draw from both.

That would be fine also.
 
Back
Top