Sensorless LYEN Controller+Turnigy RC Motor Maximum RPM Test

[Lyen]

Hello ALL,

Since there are many people are asking me and wanted to find out the actual RC motor rotation speed with the sensorless controllers, I have purchased a couple Turnigy 80-100-B 130Kv Brushless Outrunner motors from HobbyKing for testing and future implementation. I have mounted the motor in an iron bracket for testing purpose.

Below is the physical demonstration using a 6x4110 Sensorless Mini Monster Controller LYEN Edition that has the sensorless circuitry embedded on the controller circuit board and the Turnigy 80-100 130Kv motor and a 48v LifePO4 10Ah battery at 53v. The result is also applicable with the 12 FET sensorless version controller. The test tool being used is a RPM tester. It requires me to put a reflective tape on the motor shaft for accuracy so the reflection of the laser light can shoot back to the tester's sensor for rotation counting. You can see it from the video.

[youtube]TLWgoT4A9EY[/youtube]

The average speed out of multiple tests is around 3700-3800 RPM. The speed is actually more than what I have expected from the previous 9C direct drive hub motor speed which only go about 2500 RPM. The controller started trigger the overload protection and cutoff after that. The workaround is to prevent the speed from spinning above 3700 RPM by reduce the throttle speed to 43%.

The Turnigy motor is stated that it rotates at up to 130 RPM per volt, so 28v (3700 RPM divide by 130 RPM = 28.46v) should be the maximum recommended voltage to operate with the sensorless controllers.

Regards,
Lyen

 

[GGoodrum]

Interesting test. These outrunners have more poles than an AstroFlight 32xx motor, however, which only have 8 poles. In theory, the AF should be able to spin faster. I think I have an extra 3210, which I think has a kV of 135, if you'd like to try the same test on. PM me your address, if you do, and I'll send it up to you,

-- Gary

 

[jag]

Lyen: Thanks for providing real experimental results. How many poles was that motor?

Like Gary, I'm also curious about the Astro case, so it would be great if you could try that also.

I'm wondering if there is some easy modification to the sensorless signal path that could increase max bandwith and thereby RPM? I had an analog clyte sensorless that I intended to put the scope to and see if I can find something out. However now it is broken (Shorted itself and smoked when I forgot the bike it was on out in the rain for a couple of hours. -- motor was not even running). Might have to get a new sensorless unless this one is salvagable.

 

[johnrobholmes]

I have an astro here that I could test with the SL module as well. I also should have the astro with sensors that I will tach if that helps.

 

[Jeremy Harris]

If that controller is really a Xie Chang, with the 116 chip, then it will run a 12 slot, 7 pole pair, 130Kv outrunner a lot faster than that. Something else, other than over-speed, is causing the controller to trip out, probably a high current pulse tripping the secondary current limit interrupt (we found the fix for this earlier this year - there's a thread on it here somewhere), it isn't running out of commutation frequency capability.

I've run a 12 slot 120 Kv motor at over 6000 rpm from a baby 6 FET Xie Chang and I believe that these controllers can run much faster than that.

The high peak current trip circuit in these controllers is crude, just a transistor directly sensing the voltage across the shunt (it's a separate circuit, not associated with the primary, software programmable, current limit). If the voltage across the shunt exceeds about 0.6V for a few (which I think equates to about 120A for a 6 FET controller, 240A for a 12 FET) this transistor turns on, shutting down the controller. It's possible to either disable this feature, increase the duration of the current peak that triggers it, or to change the peak limit current by adjusting some component values, but both these measures introduce some element of risk to the health of the FETs. Having said that, if the FETs have been uprated this may not be a problem.

Jeremy

 

[johnrobholmes]

It is the sensorless module that is limiting the RPMs, Jeremy.

 

[rebelpilot]

I'm wondering if there is some easy modification to the sensorless signal path that could increase max bandwith and thereby RPM? I had an analog clyte sensorless that I intended to put the scope to and see if I can find something out. However now it is broken (Shorted itself and smoked when I forgot the bike it was on out in the rain for a couple of hours. -- motor was not even running). Might have to get a new sensorless unless this one is salvagable.

jag,
I'm not sure if this is what you would consider an easy modification, but STMicroelectronics application note AN1946 outlines the use of a simple circuit utilizing a TS274 high speed Op-amp.
I placed a copy of this app note in the 12 kw rc motor thread.
http://endless-sphere.com/forums/viewtopic.php?f=30&t=16728&start=1065

 

[Jeremy Harris]

It is the sensorless module that is limiting the RPMs, Jeremy.

That's interesting. I have a couple of those modules here, I'll have a look and see why they're frequency limiting.

Looking at the sensorless board it just seems to use a couple of LM339 quad comparators as zero crossing detectors for back EMF sensing. As far as I can see, these should be OK up to a switching frequency of around 300kHz with a small (around 100mV) input signal, and maybe around 1MHz with a large signal at the input, so it seems unlikely to be these causing the limiting. There are some time constants in the input filtering though, maybe that's what slugs the frequency response down. The puzzle is that, looking at the components, I'd have said that none of the time constants looks long enough to limit commutation frequency to this degree, although it's not easy to tell as the capacitor values aren't readable - I'm just guessing based on the size of the capacitor chips.

Jeremy

 

[Jeremy Harris]

OK, I've just done a bench test on a sensorless module to see just how high it will go and if it's the likely cause of the speed restriction reported here.

First off, I hooked up a sensorless module (which is electrically the same as the built in option on the board in question, with the 339 quad comparators) to a spare small RC motor, one with a Kv of 1020. Unfortunately, the only ESC I had spare to run it was limited to about 12V, which should have run this motor to about 12,000 rpm. In practice I couldn't get it to go above about 8600 rpm, but still fast enough for this test. It's a 12 slot, 7 pole pair motor, so electrically similar to the bigger motor above - if the sensorless module works at this rpm on this motor it should work at the same rpm on any similar configuration motor.

The sensorless board gives perfect, clean Hall outputs when checked on my old scope, from the lowest speed that the motor will run right up to 8,600 rpm, no problem at all. I was getting a very nice 1000Hz square wave from the module at full rpm, (1000 x 60) / 7 = 8,571 rpm, so whatever is causing the limitation reported here it isn't the sensorless circuit, that's for sure. I would suggest that the sensorless board will go a fair bit higher than I tested it too - if I can find a higher voltage ESC around I'll see how high I can push it.

This may help shed some light on just why the rpm was lower than expected and why full throttle couldn't be achieved. At least it should remove when suspect cause.

Edited to add:

While having some lunch I realised I could just put a function generator right on the phase connection and inject a signal to see just how fast these modules would work, so I've just done another test. The Hall output starts to get a bit flaky at around 5kHz, equivalent to around 43,000 rpm for a 12 slot, 7 pole pair motor. The combination of these two tests pretty much proves that the sensorless circuitry isn't causing the problem.

Jeremy

 

[Lyen]

Hi Jag,

How do I determine the number of poles for the motor without disassemble it? The specification from Hobby King do not have this info listed.

The Crystalyte analog sensorless controller may not have the same circuitry as the one I have mentioned so it may be even more challenging to deal with.

Lyen: Thanks for providing real experimental results. How many poles was that motor?

Like Gary, I'm also curious about the Astro case, so it would be great if you could try that also.

I'm wondering if there is some easy modification to the sensorless signal path that could increase max bandwith and thereby RPM? I had an analog clyte sensorless that I intended to put the scope to and see if I can find something out. However now it is broken (Shorted itself and smoked when I forgot the bike it was on out in the rain for a couple of hours. -- motor was not even running). Might have to get a new sensorless unless this one is salvagable.

 

[Lyen]

I really appreciate your input Jeremy. It is very informative. I have to clarify that there are two kind of sensorless for the existing digital controllers. The one you are looking at has the sensorless circuitry embedded on the controller circuit board. The other kind of the external add-on sensorless module that you solder wires onto the existing sensored controller.

Below is the video I made for the external add-on sensorless module soldered onto an existing 18 FET 4110 controller. In this video, you will notice the differences in speed and characteristic. The first different is the maximum RPM is around 2200-2300 RPM. The second is the motor started the stutter after 2300 rpm but continue to go higher if you force it. You will see the motor tries to jump out of the very heavy iron bracket after 2300rpm whereas the 6 FET sensorless controller will simply shut itself down once it reaches the limit.

[youtube]g-MrA9CJui8[/youtube]

So in reality, the embedded module does processes faster than the external add-on module. The RPM differences is between 1500 rpm.

Regards,
Lyen

 

[Jeremy Harris]

Lyen,

The tests I've just done were on the add-on module, but the circuit is the same for both - it just uses LM339 quad comparators to derive the Hall signals from the phase wires. I believe the actual chip on the built in version is the AS339, but that is functionally identical, just from another manufacturer. Here's a 'scope shot image showing the sensorless board running at about 12,600 rpm on an outrunner of the same configuration as yours (I found a higher voltage ESc to drive the motor):



As you can see from this image, the period of the Hall signal (the yellow trace) is around 680uS, which equates to about 1470Hz. The motor phase trace is the purple one (the motor is at full throttle, so there's no PWM). 1470Hz for a 12 slot, 7 pole pair motor is about 12,600 rpm (1470Hz x 60 / 7 = 12,600).

I don't know what your problem is, unfortunately, but I can say categorically that the sensorless board produces clean Hall signals at a frequency much, much higher than the rpm where you're seeing a problem, so I doubt that it's the cause.

Jeremy

 

[liveforphysics]

If letting off on the throttle a bit lets the RPM creep up higher, he may be running into that instantanious over-current limit thing. My motors would do something similar when I was first playing with the sensored setup until I did that over-current analog circuit retard/delay mod.

If you're going to ask why it did it at a few thousand RPMs and not just starting out, I don't have a clue, but going to partial throtle would let the RPMs creep up higher, and that's a symptom I'm seeing in common with Lyen's test here.

 

[Jeremy Harris]

That's my view, too, Luke. One thing's clear, the sensorless bit works as advertised up to silly rpm territory, so I doubt that it's the cause.

Quite what would cause a current spike at this high rpm I don't know, but the video shows exactly the same shut-down symptoms as we were seeing earlier this year when we first discovered this problem.

Jeremy

 

[rebelpilot]

One thing's clear, the sensorless bit works as advertised up to silly rpm territory, so I doubt that it's the cause.
Jeremy

Is the Hall signal symmetrical? The displayed waveform does not show this.

 

[GGoodrum]

I sent a 3210 to Lyen today, so hopefully the circuit will be able to spin it up around 6K. If it does,I'll be happy.

-- Gary

 

[Jeremy Harris]

Is the Hall signal symmetrical? The displayed waveform does not show this.

It's very slightly out, but as the controller is edge-triggered, using just the Hall rising edge I believe, this doesn't make any difference.

The error is small, anyway. From that trace it looks like the Hall high time is about 320uS and the Hall low time is around 360uS, which is probably as good (or better) than the symmetry you get from Halls inside a motor.

Jeremy

 

[liveforphysics]

I sent a 3210 to Lyen today, so hopefully the circuit will be able to spin it up around 6K. If it does,I'll be happy.

-- Gary

I don't know if even a 2-turn Astro is as low of winding resistance as the HXT motor he is using, so it should likely be fine.

If he does the current-interupt mod to this controller, the big HXT motor will spin up past 6k as well.

Lyen- The current interupt on these things is a totally analog circuit, so software changes have no effect on it. You gotta swap out a couple resistors to fix it. There is a big thread on how to do it somewhere on here, I think it was a smart Aussie or Brit that was the one who figured it out.

 

[Jeremy Harris]

Lyen- The current interupt on these things is a totally analog circuit, so software changes have no effect on it. You gotta swap out a couple resistors to fix it. There is a big thread on how to do it somewhere on here, I think it was a smart Aussie or Brit that was the one who figured it out.

Very true, it's just a transistor with it's base connected to the shunt via a simple RC network, so that if the shunt voltage peaks above about 0.6V for more than a very short period (probably uS) the controller gets an interrupt to shut down.

It was figured out by a multinational team, cooperating via this forum, but a couple of us were Brits............

Jeremy

 

[Thud]

In case any one is looking for that info:
http://endless-sphere.com/forums/viewtopic.php?f=2&t=16910
I perfer the "adding a resistor" mod as it doesn't tottaly disable the circut...I have also run into the cut out issues on every infineon up to the 18fet units with my outrunners. so 3 cheers for these guys who showed me how to fix it.

 

[John in CR]

I don't know if this might help diagnose the issue.

I've got one of the sensorless controllers with the add-on module. I ran it on my big hubmotor, which has 48 magnets and 51 slots in the stator. Start up is great with only odd noise in rare rotor position from a dead stop. In fact the start is smoother and far quieter than with the sensored. Starting about 500rpm under load power is diminished, and it won't go above about 600rpm under load. The motor sounds different, and there must be higher currents too because both motor and controller get hotter than normal at those speeds. There's no cutout, just a rough sounding operation and lack of power. The controller without the sensorless module has no problems right up past 1krpm under load.

Since Jeremy has verified a clean signal and far higher frequency from the sensorless modules, it indicates the issue must be one of timing. Is it maybe that these sensorless modules are set for a timing that is somewhat retarded? That makes sense to me, since they are designed for hubmotor use and good start-up is a key issue for sensorless.

John

 

[Jeremy Harris]

I don't know if this might help diagnose the issue.

I've got one of the sensorless controllers with the add-on module. I ran it on my big hubmotor, which has 48 magnets and 51 slots in the stator. Start up is great with only odd noise in rare rotor position from a dead stop. In fact the start is smoother and far quieter than with the sensored. Starting about 500rpm under load power is diminished, and it won't go above about 600rpm under load. The motor sounds different, and there must be higher currents too because both motor and controller get hotter than normal at those speeds. There's no cutout, just a rough sounding operation and lack of power. The controller without the sensorless module has no problems right up past 1krpm under load.

Since Jeremy has verified a clean signal and far higher frequency from the sensorless modules, it indicates the issue must be one of timing. Is it maybe that these sensorless modules are set for a timing that is somewhat retarded? That makes sense to me, since they are designed for hubmotor use and good start-up is a key issue for sensorless.

John

I think you may be right, John. I've hooked up the sensorless module to a couple of motors this afternoon and found that it runs the low Kv (around 90) 5330 hack motor I have fine at the highest voltage I can get on my bench supply (~38V). This doesn't really test things though, as the combination of low motor Kv and limited supply voltage only let me get to about a bit over 3000 rpm. This motor is a 12 slot, 7 pole pair motor, just like the HXT that Lyen is using. The 5330 started reasonably well and ran perfectly right up to full throttle, with no sign of any hesitation. The Hall signals coming out of the module looked nice and clean, with no sign that they were on the limit of dropping out.

Next I tried a small, 1020Kv motor, again a 12 slot, 7 pole pair one. This didn't work at all well. The motor refused to start most of the time, with the controller trying to start then shutting down. Every once in a while it would spin up, but would only get to around 4000 to 4500 rpm before suddenly cutting out, pretty much just like Lyen has been seeing.

As this small motor can't possibly have been causing the current trip to go (it's a small park flyer type, roughly equivalent to a 400 brush motor) it has to be something else that's making it stop. The signals coming from the sensorless board still look perfect right up to the time that the motor stops, with no indication of them going flaky from over speed, as I saw during the function generator test at around 5KHz input.

The timing does look a bit awry. Here's a capture from my crappy USB 'scope (limited bandwidth, so not good for the motor phase signal):


This picture is the TowerPro 5330 running at part throttle, and shows that there is a fair bit of timing skew between the sensorless pseudo-Hall output and the phase voltage. Note that the purple phase trace probe was switched to 10X so is 10V per division, whereas the yellow Hall trace is at 2V per division.

For comparison, here is the same 'scope capture for the motor running on it's internal, neutrally timed Hall sensors (it is connected in wye):


The timing shift is pretty clear to see from these pictures and probably explains the speed restriction problem. One thing it does tell us is that timing ain't too critical in order to get these motors working, though!

Jeremy

 

[MitchJi]

Hi Gary,

I sent a 3210 to Lyen today, so hopefully the circuit will be able to spin it up around 6K. If it does,I'll be happy.

-- Gary

I thought you were waiting for Hall Sensors in the Astro's, due mostly to poor start-up. What problem(s) does running a Sensorless LYEN Controller with the Astro solve for you?

Maybe the owner of tppacks should build and sell Astro Hall Sensor kits.

Very true, it's just a transistor with it's base connected to the shunt via a simple RC network, so that if the shunt voltage peaks above about 0.6V for more than a very short period (probably uS) the controller gets an interrupt to shut down.

It was figured out by a multinational team, cooperating via this forum, but a couple of us were Brits............

Lyen should probably sell his Sensorless Controllers with the fix.

 

[GGoodrum]

I thought you were waiting for Hall Sensors in the Astro's, due mostly to poor start-up. What problem(s) does running a Sensorless LYEN Controller with the Astro solve for you?

I am mounting a multi-prong attack on getting SOME sort of solution that will let me run a 3220 at a voltage higher than the ridiculous 50V limit on pretty much all RC controllers. I love the AF motors, but not with overpriced problematic controllers that limit the voltage. Lyen's controllers are extremely reliable, and I suspect that if the sensorless module works at a high enough rpm, we won't need things like slipper clutches to keep the controller from killing itself. If the module doesn't work, I will be forced to continue to wait for a sensor solution to appear.

 

[Jeremy Harris]

Lyen's controllers are extremely reliable, and I suspect that if the sensorless module works at a high enough rpm, we won't need things like slipper clutches to keep the controller from killing itself. If the module doesn't work, I will be forced to continue to wait for a sensor solution to appear.

It looks (from the 'scope traces above) as if the sensorless circuit has a significant timing issue. The skew on the pseudo-Hall signals that come from this module is almost certainly what's causing the rpm limit. Quite why the sensorless circuit does this I'm not sure and I don't have time to do a detailed series of tests to find out how the timing skew changes with frequency. However, having found some pretty strong clues as to where the problem lies it should be easy enough for Lyen to dig around and get a fix, maybe by getting some of the RC values that are causing the unwanted phase shift changed. There is a potential problem with this, as the filtering has to be good enough to keep out the 15kHz PWM frequency noise that's present, even on the unpowered phases, and allow reliable detection of the underlying commutation frequency.

My guess is that the sensorless circuit was designed for use at relatively low commutation frequencies, such as might be found at normal speeds with a hub motor. This would explain why the filter time constants are as low as they are.

Jeremy