Sensorless LYEN Controller+Turnigy RC Motor Maximum RPM Test

[GGoodrum]

Yes, I agree. This is probably a "fixable" issue, but my thought is that it might not be needed with the AstroFlight motors. These have four pole pairs, vs. the 7 pairs on the big Turnigys. If the latter can get to 3700 rpm, I'm hopeful the 32xx will get to 5500-6000. For my immediate needs, I have a 3220-7T that has a kV of 97 in wye mode. I want to use this for a friction drive, and I want it to be powered by an 18s2p LiPo pack, so somewhere around 67V nominal. The rpm would be somewhere between 5500-5800 rpm, max.

-- Gary

 

[John in CR]

From my perspective solving the sensorless issue will be the cats meow for 3 reasons:
1. No hall sensors- Then the only motor problem can be worn out bearings.
2. Smoother startup- With sensorless my motor doesn't make the startup growl that also comes with a slight vibration using a sensored controller. To me that noise must come from imprecise commutation, so no noise should mean more power, less heat, and better efficiency. I'd take silence even if it cost some efficiency and power.
3. Assuming it's a timing issue, then solving it will also give us more top end too.

 

[johnrobholmes]

Proper sensored systems will have better startup, hands down. What I would like is a controller than can run sensored for startup, and then switch to sensorless on the fly for best efficiency or power. It looks like something needs to be changed with the way these 116 chip controllers are running with so much sensorless timing retard.

 

[rebelpilot]

Could someone please read appendix 10 and comment.

Thanks

 

[olaf-lampe]

I followed the thread with big interest.
Has anyone checked the 'deadtime' between high and low Fets? Maybe its too short and the current control trips because of a shoot thru?
-Olaf

 

[Jeremy Harris]

I followed the thread with big interest.
Has anyone checked the 'deadtime' between high and low Fets? Maybe its too short and the current control trips because of a shoot thru?
-Olaf

Good point. I don't have a fast enough 'scope to make a meaningful measurement, but if one of you guys with nice, fast 'scopes wants to have a look it would help a great deal, not just with this problem either.

Jeremy

 

[John in CR]

Proper sensored systems will have better startup, hands down.{/quote]

The Lyen sensorless I have starts smoother and quieter than any sensored controller I've run on my big hubbies, "hands down". Only from a dead stop does it do one little reverse bump, maybe 1 in 10 times, which must be just the wrong rotor position. Otherwise it's smooth as silk like a good brushed motor, and no motor growl under load like with sensored controllers.

 

[johnrobholmes]

From a dead stop is what I am referring to, thus the term STARTup. Claim that sensorless startups are better all you want, at zero RPM there is no feedback and the motor must be forcibly rotated with a startup routine. Noise is not an indicator of reliability of startup. Only one out of 10 are bad starts? That is a 10% chance, and an indicator that the startup routine is not as good as sensored startups. Roll a few MPH and you get out of that dead zone and have reliable starts all day, we have known this for years on sensorless ebikes.


Large CNC equipment uses sensors at low speed and sensorless at high speeds. You would think that they would get rid of the sensors if it was better to do so. Less parts to fail, right?


Full disclosure, I use both types of controllers frequently. The 116 controller with sensorless commutation is quite smooth, and they do the job well. On vehicles where I need reliable startup I will choose sensors first, however. Even a Castle Creations startup routine (arguably the best in the world) isn't good enough to rival sensors at zero RPM. I want the best of both worlds, but right now this thread is just focusing on getting around the limits of the little sensorless add on.

 

[Lyen]

Hello Gary and all respectful fellows,

I have received an Astro 3210 RC motor from Gary for testing purpose yesterday. I have just drilled four additional holes on the iron bracket and mounted the motor. After that, I have performed a series of tests. Here is the setup:

Battery: 48v 10Ah LifePO4 ducktape type battery at 53v
WattsUp meter
Lyen 6 FET IRFB4110 MOSFET equipped sensorless controller with 116A MCU
Program settings: 30A rated current, 80A phase current, LVC = 45v, throttle = 100%

The test result on the Astro 3210 is surprisingly satisfactory than using the Turnigy 130Kv motor. The controller is totally the same way it was when testing the Turnigy motor "stock" with no additional modification.

The maximum RPM out of 5 video recorded tests are from 6510 to 6515 RPM which is about 40-45% more than the Turnigy 130Kv motor. The other beauty of it is I was able to turn the throttle all the way up without cutting out. The maximum current draw is about 7.12amps at wide open throttle at 53v.

Here is the size comparison between the two motors:


Here is the video for the 5 tests performed on the Astro 3210 with the 6 FET 4110 sensorless Lyen controller:

Warning: Please cover your ears when watching the video due to high pitching sound coming from the motor.
[youtube]2noo3GEyX8g[/youtube]

Thanks everyone for bringing in ideas to get this improved. I will continue to find out what would happen at even higher voltages and remedies as suggested.

Regards,
Lyen

 

[Jeremy Harris]

Good news, Lyen, as it pretty much confirms that the problem with the higher pole pair motors is probably timing skew.

I'd guess that it might be difficult to resolve this with the present comparator-based sensorless circuit, as the filtering needed to keep out PWM artefacts will inevitably cause enough phase shift at frequencies approaching the filter notional cut off to cause problems.

The best way around this would be to do the sensorless detection digitally. A digital filter could have a much sharper cut off, with a relatively flat phase response right up to cut off. This would then allow operation at higher commutation frequencies.

Jeremy

 

[GGoodrum]

Awesome! That is pretty much what I was expecting to see, but with a higher voltage, you might get even more rpm. That motor has a kV of 135, so if you are running it at about 53V, that 6515 rpm number would be just about right, if you factor is some efficiency loss. At 6515 the efficiency would be about 91%, which is just about right for these motors. The efficiency maxes out at 93% at 7500 rpm.

Anyway, I'm a very happy camper, because I now have a solution to spin my 3220-7T, which has a kV of 97, at around 5500, which is all the speed I need. This is great news indeed.

-- Gary

 

[spinningmagnets]

I am still new to learning electrical theory, so I apologize in advance...

There will always be hot-rodders who push the edge of the envelope, and I love their work (100V, 40-MPH, etc). I have recently been polling a lot of college students, who I think are a prime market to sell E-bikes to. They have a hundred newbie questions. "If this, then why not that?" I have been forced to learn more so I can explain better why I have made the choices I have, and why I recommend the things I do.

If we use higher volts, we can use fewer amps to do the same job, so how high is "reasonably" cost-effective and safe, plus uses fairly available parts (not rare) with a variety of choices. I think the manufacturers sell kits with a winding to provide 20-MPH at 36V on purpose. They know upping the voltage to 48V is very easy for the customer to do, but once over-volted, the manufacturer is not responsible for any problems that happen.

Above 55V I am told that some people may be electrocuted. I am not afraid of 60V and 72V, but if we want to convince NON E-bike folks to embrace this, I have to ask myself "what makes sense to the average non-technical customer?"

Several people are developing an easy-to-install and affordable Hall-sensor kit, and Lyen is improving his controllers to work better with non-Hall RC-motors, so wonderful things are about to happen. What I most want is a trouble-free plug-and-play controller for 20V-55V, which will work with un-sensored RC motors. I believe most of the different styles of recent friction-drive builds will prove to only need the 6-FET, but once these are evolved to reliably run sensorless-RC, there is a 9-FET and 12-FET (if needed).

I am just one customer, but I think a LOT of people will end up wanting the high-performance and small size of a low-kV outrunner on 44V-48V

 

[John in CR]

From a dead stop is what I am referring to, thus the term STARTup. Claim that sensorless startups are better all you want, at zero RPM there is no feedback and the motor must be forcibly rotated with a startup routine. Noise is not an indicator of reliability of startup. Only one out of 10 are bad starts? That is a 10% chance, and an indicator that the startup routine is not as good as sensored startups. Roll a few MPH and you get out of that dead zone and have reliable starts all day, we have known this for years on sensorless ebikes.

Large CNC equipment uses sensors at low speed and sensorless at high speeds. You would think that they would get rid of the sensors if it was better to do so. Less parts to fail, right?

Full disclosure, I use both types of controllers frequently. The 116 controller with sensorless commutation is quite smooth, and they do the job well. On vehicles where I need reliable startup I will choose sensors first, however. Even a Castle Creations startup routine (arguably the best in the world) isn't good enough to rival sensors at zero RPM. I want the best of both worlds, but right now this thread is just focusing on getting around the limits of the little sensorless add on.

I didn't intend to make an all encompassing claim, unlike your hands down comment. I just reported what it did with my motor, not any other motor or any other type of machine. None were bad starts by your definition, just there where some occasions where it made a single knock before taking off forward normally. I made a bunch of starts from a full stop with no pedaling, because it was my first use of sensorless controller and I wanted to see how well it started. Some kind of non-start would be a show-stopper for me, but It started with 100% reliably for me when I was running it. I only took it off because of problems above 25-30mph.

The difference in my results is probably due to the much higher pole count. I have a new motor with even more poles, so I'll give the controller a try with that one too before I mod it for sensored operation only. The speed/timing issue with my motor won't work for me.

FWIW, a sensored motor has to have a startup routine too, because those hall sensors aren't reporting anything from a dead stop, so the controller knows no more about the rotor position than a sensorless controller does. Though much more rare I have seen the similar kind of knock to start rotation with several different sensored controllers too.

 

[Hillhater]

Good news, Lyen, as it pretty much confirms that the problem with the higher pole pair motors is probably timing skew.

I'd guess that it might be difficult to resolve this with the present comparator-based sensorless circuit, as the filtering needed to keep out PWM artefacts will inevitably cause enough phase shift at frequencies approaching the filter notional cut off to cause problems.

The best way around this would be to do the sensorless detection digitally. A digital filter could have a much sharper cut off, with a relatively flat phase response right up to cut off. This would then allow operation at higher commutation frequencies.

Jeremy

does this mean that this controller is not compatible with the Turnigy motors ..as it stands ??

 

[johnrobholmes]

FWIW, a sensored motor has to have a startup routine too, because those hall sensors aren't reporting anything from a dead stop, so the controller knows no more about the rotor position than a sensorless controller does. Though much more rare I have seen the similar kind of knock to start rotation with several different sensored controllers too.

Hall sensors show rotor position at zero rpm. Either one or two sensors are in a high state at any given time, with the combination telling the controller exactly which phases to excite for locomotion.

 

[John in CR]

FWIW, a sensored motor has to have a startup routine too, because those hall sensors aren't reporting anything from a dead stop, so the controller knows no more about the rotor position than a sensorless controller does. Though much more rare I have seen the similar kind of knock to start rotation with several different sensored controllers too.

Hall sensors show rotor position at zero rpm. Either one or two sensors are in a high state at any given time, with the combination telling the controller exactly which phases to excite for locomotion.

I thought they were just on/off, so position isn't known with great precision until the edge of a magnet passes the sensor. Isn't it only the strong signal to report timing as the magnets pass the sensors that gives sensored the advantage at very low speed over sensorless, since the electrical feedback from the motor is still so low? Do I have this wrong?

If sensored really knows exact position, what would account for the notably smoother operation of the sensorless that I saw under load from start to maybe 10mph with my motor? Why do sensored make any noise at all, because it's always sounded and felt to me like imprecise commutation?

 

[spinningmagnets]

I am still new (again, apologise in advance) it is my understanding that in a 3-phase motor, at least one phase is energised at any given time due to wave-form overlap? I have heard the only on/off motor is a one-phase with a dead spot in-between each phase (Axels DIY plywood axial-motor)

 

[johnrobholmes]

I can't comment on the smoothness or sound difference between your SL and sensored controllers, John. It may be timing related, or the positions of the sensors that are causing uneven commutation. It shouldn't be a PWM difference, although it is possible I suppose. I notice the same effect on my Lyen type controllers, so you are not alone.

The three sensors have six permutations that are possible, matching with the 6 step commutation cycle of the controller. In this type of controller it is just a digital on/ off instead of an analogue signal. While you are correct in that it isn't great precision, it is enough precision to always generate torque in the intended direction.

100
110
010
011
001
101

With this signal the controller knows what phase to trigger with what polarity.



Now, why do you perceive smoother operation with the sensorless module? Just a guess, but it may be less vibration and your setup isn't on the edge of stalling at low speeds. Higher pole counts do help for sure. The higher the commutation speed for a given wheel RPM, the easier it is to startup sensorless. I will be interested to see how the astro does under load with the SL module.

 

[MitchJi]

Hi,

Thanks Gary!

Even a Castle Creations startup routine (arguably the best in the world) isn't good enough to rival sensors at zero RPM.

So is the following a correct summary (assuming you have an Astro and you can live with the potential [unmodified] RPM limitations)?:
Advantages LYEN Sensorless:

• Less expensive.
  Higher voltage.

Advantages Castle Creations HV160:

• Better startup.
  More compact.
  Higher amperage (compared with an unmodified LYEN Controller).

 

[Jeremy Harris]

I'll lay money on it being the timing differences between sensored and sensorless that are giving John in CR the apparent change in smoothness. My guess is that his Hall sensors aren't optimally placed for smooth running and the timing skew that the sensorless unit chucks in just happens, probably more by luck than design, to make his motor run more smoothly, under the conditions he's operating it at.

My limited experience with testing on the bench shows that in sensored mode start up is better, low speed torque is massively improved and mid to high speed running seems to be identical to sensorless. The differences are almost certainly down to Hall placement, I'm sure - mine are all accurately set at neutral timing with the winding configuration I'm using, which may not be the case on some hub motors.

Jeremy

 

[Hillhater]

Hi,

Thanks Gary!

Even a Castle Creations startup routine (arguably the best in the world) isn't good enough to rival sensors at zero RPM.

So is the following a correct summary (assuming you have an Astro and you can live with the potential [unmodified] RPM limitations)?:
Advantages LYEN Sensorless:

• Less expensive.
  Higher voltage.

Advantages Castle Creations HV160:

• Better startup.
  More compact.
  Higher amperage (compared with an unmodified LYEN Controller).

I could be mistaken (probably am ! :? ) , but isnt this comparing a 1.5-2kW controller (Lyen)..... to a 4-5 kW ESC (Castle 160) ?
You would even out the price differential by picking one of the smaller (80-100a) CC ESC's.

 

[John in CR]

I'll lay money on it being the timing differences between sensored and sensorless that are giving John in CR the apparent change in smoothness. My guess is that his Hall sensors aren't optimally placed for smooth running and the timing skew that the sensorless unit chucks in just happens, probably more by luck than design, to make his motor run more smoothly, under the conditions he's operating it at.

My limited experience with testing on the bench shows that in sensored mode start up is better, low speed torque is massively improved and mid to high speed running seems to be identical to sensorless. The differences are almost certainly down to Hall placement, I'm sure - mine are all accurately set at neutral timing with the winding configuration I'm using, which may not be the case on some hub motors.

Jeremy

Torque during take off was definitely lower with the sensorless, but I've got an apples/oranges comparison due to significant current limit differences. An imperfection in sensor placement makes sense, but wouldn't that manifest as a problem at higher speeds or significant inefficiency? I do run these motors at higher rpm than almost anyone with a hubbie, due to high speeds and a small 19.5"dia wheel. Just yesterday I was at over 1krpm on the highway. If the physical timing was off, then I have to believe I'd have significant problems with 48 magnets passing that out of whack sensor with each rotation of the wheel.

Every hub motor I've run has a similar kind of noise in the first few mph of take off with moderate to full throttle and a load on the wheel. I just always attributed it to my thinking that perfect commutation is impossible due to the multiple coils on each phase and mismatch in the number of magnets vs coils, so each coil on a phase is in a different relative position to it's passing magnet at any given time. The sound isn't loud or bothersome. It's just more sound than the hum of an electric motor under load, and even AJ commented that he thought hubbies were noisy when he heard that sound in someone's video. Take off with the sensorless was essentially silent, so of course I want that if it's possible.

Once I have another bike ready, I'll give the sensorless another go and try to be more scientific in my observations. I'll get some video with sound too.

John

 

[Lyen]

Here is another test performed today using the Astro 3210 from Gary at 86.2v. 5 tests were executed with the following results:

test 1: 6665 RPM
test 2: 7077 RPM
test 3: 6999 RPM
test 4: 7015 RPM
test 5: 7065 RPM

Observation: I was able to get high RPM if accelerate the throttle slowly. The controller did cutoff this time with high voltage at 86.2v.

[youtube]LMxHfTrNTC0[/youtube]

stay tuned...

 

[GGoodrum]

Okay, so now it looks like we are running up against a processing speed limit. At 86V, this motor should spin at around 10,000 rpm. Its most efficient rpm, however is 7500 (93%...). In any case, My 3220-7t has a kV of 97, and I want to run it on 18s, or about 66-70V, so the rpm will be just under 6k, which this module seems to handle quite well. Put me down for a 4110-based 12-FET version, just as soon as you can build one.

-- Gary

 

[olaf-lampe]

Okay, so now it looks like we are running up against a processing speed limit. At 86V, this motor should spin at around 10,000 rpm. Its most efficient rpm, however is 7500 (93%...). In any case, My 3220-7t has a kV of 97, and I want to run it on 18s, or about 66-70V, so the rpm will be just under 6k, which this module seems to handle quite well. Put me down for a 4110-based 12-FET version, just as soon as you can build one.

-- Gary

If you won't run above 75V, why don't you try the irfb3077 instead? IMHO it's even better than the 4110.
-Olaf