New "Lingbei MM28.250 torque sensor mid drive"

stinkycats said:
casainho said:
For what was shown before, I really think torque measure can be measured on the timing difference between pulses of the hall sensors.
Yes, the phase difference of two pulse trains can indicate the amount of torque. However, as I pointed out, a pulse train is nearly useless as a measure of torque if you don't have enough pulses-per-revolution. One pulse every 360° is useless. One pulse every 180° (2 pulses per revolution) is almost as useless. One pulse every 90° means you have to rotate the pedals 1/4 of the way around before you get a torque measurement. This is also fairly useless. In my opinion, you would need at least 32 magnetic poles (every 11.25°) to get a smooth torque reading without moving the pedals too far.

There is no way to see if the axle has that many magnetic poles from the pictures posted here, or even if it is pulse train based or magnitude based (analog output).
Additionally, the minuscule angular deflection in a short length of solid shaft resulting from human torque makes this approach unlikely to succeed. A new arrangement would have to radically different from what @kisazul posted.
I’m not the sharpest knife in the draw when it comes to electronics & I don’t have a proper scope but I do have a Fluke Scopemeter https://www.inlec.com/fluke-105b-oscilloscope, which might work.
Need help in identifying where on this wiring layout I might find a torque sensor signal.
IMG_0295.jpg
I suspect I rather need to be in the controller, right? If so, does anyone know how to remove this custom 3-indent locknut? @kisazul – you must have managed it?
IMG_0294.jpg
Also see lack of grease on my secondary gears, if done properly it would have been quieter.
 
Hi. I do not have a special key for loosening this nut. I took a thick flat screwdriver and unscrew it with a hammer counter-clockwise.

P.s. But after it turns out to unscrew this tricky nut with the bearing, the shaft will not be able to remove. To extract it, it is necessary to remove the oil seal on the other side of the shaft. then unscrew the 2 retaining rings with a small screwdriver, turning it counterclockwise. Next, the shaft is removed in the direction of the nut.
 

Attachments

  • IMG_20181005_134447.jpg
    IMG_20181005_134447.jpg
    77.3 KB · Views: 4,879
wheelyphased said:
Need help in identifying where on this wiring layout I might find a torque sensor signal.
You need to tap the signal from these wires:
View attachment 1

which are connected to the controller here:
lb2.png

We can probably assume the red and black wires are V+ and GND, but you should test it first with your meter (check continuity between the pcboard V+ and GND planes and the wires). The green and blue wires are probably your hall sensor outputs. Hook those up to your A and B scope inputs. When powered up, spin the shaft and look for a signal. If you can spin the shaft at a steady speed (maybe chuck it up on a drill), set your scope to auto-trigger, then select X-Y mode. If the signals are coherent (steady, triggering at the same time, roughly sinusoidal/rounded-rectangle), you should see a straight diagonal line. As you delay the chainring mount of the shaft, probably by simply putting pressure on it with your hand to simulate a torque differential, you should start to see that diagonal line on your scope change to roughly an oval (or somewhat close to it) shape.

If you see a continuous oval shape as you apply pressure, that means we do have a viable torque signal and a modified firmware that would properly use the torque signal, like in the TSDZ2 firmware mod, is possible.
 
Great instructions @stinkycats I can't directly access the ground plane or the V+ on the controller board because they are buried under soft silicone potting compound which has been poured after assembly into its housing but the black wire from the PAS hall sensor has continuity with the black wire of the adjacent 5-pin connector so I'm going to assume it's ground. I am building a probe harness so I can reliably access the 4 hall wires whilst the motor is (a) powered up & (b) running. Here's a better view of the 9 MOSFET controller.
IMG_0296 copy.jpg
What is the 5-pin connector likely for?
 
If I remember correctly, the 5 pin is for the HAL sensors of the motor(VCC, GND, 3x output) and the 4 pin is for the torque/PAS sensor.

So you want the 4-pin connector wires to the outside for measurements.
 
Re: The TSDZ2 torque sensor:
I'm not sure exactly what the communications protocols etc are, or the details of how they do it, but:
The coils of wire are large diameter (maybe 3 inch / 80mm diameter?? - I don't have one here to measure) each attached to the rim of a disc perpendicular to the pedal axle. One disc spins with the pedals, and the other is stationary and is pressed against the face of the rotating one by 3 light springs.
Obviously both power and communications are passed from one coil to the other by induction, implying AC.. clever stuff.
Unfortunately I didn't take many photos while I had the unit stripped, but here's a photo of the pedal spindle / torque sensor / coil assembly:
View attachment 1
TS Tq sensor 1 (2).jpg

I agree that we need to know whether the Lingbei motor has a good torque sensor or not. Poor programming could account for the poor performance even if there is a good torque sensor. On the TSDZ2, Casainho was able to identify the torque sensor signal and display it on an oscilloscope, proving that it varied with the force applied to the pedal. If someone could do the same with the Lingbei motor, we would all know what the situation was.
 
Drum said:
One disc spins with the pedals, and the other is stationary and is pressed against the face of the rotating one by 3 light springs.
Obviously both power and communications are passed from one coil to the other by induction, implying AC.. clever stuff.

That's not clever at all, as it's not contactless and very fragile. There are much better systems on the market, like the Sempu or the Bafang torquesensor.

But I agree, it would be interesting to know, if the Lingbei system works properly.

regards
stancecoke
 
Hi Stancecoke, thanks for the link to that article.. it contains more information than I had.
However, I disagree when you say "it is not contactless": the coils of wire are not physically transmitting current by direct metallic contact: they are electrically insulated from each other, and transmitting information by induction of an AC signal. By my understanding of the word, that makes them "contactless".

Fragile, well yes, I have seen photos of those coils that have worn / mechanically failed, so I wouldn't argue that one.

When I said I thought it was clever, I meant that both power and signal are communicated via only two wires from the controller to the stationary coil, AND through a contactless inductive coupling, something I hadn't seen before. Whoever thought up the system and made it work knows at least a bit about electronic components and can think beyond what has been done before. Of course, this particular system may have been copied from a similar setup from a different manufacturer, but if so, I was not aware of it.
The specific design of the components may not be great, as proven by the failures.
 
stancecoke said:
That's not clever at all, as it's not contactless and very fragile.
But I agree, it would be interesting to know, if the Lingbei system works properly.
As this seems to be a parenthesis on the TSDZ's torque sensor, the Lingbei could be equipped with a better make (but with faulty software, bad production control, defective intems sole). So far, we've read reports from only one user of the v2 Lingbei. That's not much.
 
Drum said:
When I said I thought it was clever, I meant that both power and signal are communicated via only two wires from the controller to the stationary coil, AND through a contactless inductive coupling, something I hadn't seen before.

You can build a "two-wire" system that is contactless in electrical and mechanical manner :wink:


wheelyphased said:
Piggyback wiring complete.

I'm looking forward to see your results!

regards
stancecoke
 
wheelyphased said:
Almost there. Just mad busy right now.
Piggyback wiring complete.
IMG_0298 copy.jpg
Test bench
IMG_0300.jpg

Nice! Can that fluke meter measure 2 channels? I expect a VCC 5V, GND, and 2 pulse trains.
 
Here are some scope measurements of the V2 motor ‘hall’ sensor output at the 4-pin connector in the controller. There are 4 wires: red (assumed VCC), black (assumed GND), green and blue. I apologise if this reads as if I don’t know how to use a scope or know ‘lectroniks. That’s because I don’t. Bear with me.

This is the voltage between V+ (red) and GND (black) with display ON.

Now we change to scope mode. You may read the gain/settings at the top of each screen. The Fluke has an auto settings function for idiots like me which makes seeing consistent screens in images a nuisance but this is best I could do. Hereafter channel A is connected to the green sensor wire, channel B is connected to the blue sensor wire.
a) This is at rest. It’s the same whether the display is booted or not.
A.jpg
b) This is with the display powered up & crank driven by a drill (hex bolt head in chuck driving the hex of a crank fixing setscrew). The drill causes a strange distortion of both channels either because it’s a cheap Chinese drill that leaks current or because the drill speed control is rubbish and vibrates.
B.jpg
c) Now with a third channel enabled. The grey shape is A vs. B (or X-Y for scope people). This is at rest, powered up.
C.jpg
d) This is drill-driven, just at the point where the motor kicks in (more about that later).
View attachment 4
e) This is manually cranked, with one crank arm attached, just as the motor kicks in. This is hard to do & needs 3 hands so you are seeing the screen holding an image in the seconds it takes to pick up a phone.
E.jpg
f) Once more, as (e).
F.jpg
(g) And again.
G.jpg

I’m puzzled as to why channel A appears to be the only signal channel. I was expecting to see a signal in 2 channels. Equally puzzled as to why channel A voltage drops as the motor kicks in, is this sensor phase shift?

It is instructive to comment on what is physically happening here. The motor is sitting on a bench under it’s own weight and driven from the chainring side either by drill or by hand crank against no load. This is a rather unstable environment and I assumed would be pretty difficult to apply enough dragging torque to obtain a signal. Not so.
Cranking the motor gently up to a cadence of about 60 achieves nothing, 80, the same - the motor doesn’t initiate assist. Simply applying a drag to the chainwheel outer cover with hand pressure whilst the crank is rotating is enough to get the motor to kick in. To be clear, without that torque, the motor won’t run. Similarly, applying hand torque without rotation does not start it. Having fired it up it will run happily for as long as the crank is rotating. Sometimes the required torque is high, needing a firm grip (and a fast reaction to withdraw fingers) sometimes it is almost nothing. Sometimes it then runs fast, sometimes slow. It is inconsistent and unpredictable in the extreme. I tried this at least 40 times.
I wondered if I could get the motor to start at all with torque alone. Certainly not with hand pressure but stupidly jamming a screwdriver against the casing did. It also bent & threw the screwdriver 10m. Now there is a rough spot once per crank revolution. This motor is a torque monster.
Measurements aside, it appears there is an element of torque sensing but it is set with a very low threshold if the crank is turning or it acts as a foot switch and doesn’t have modulation i.e. crude. Without manipulating the motor gently on the bench this way the torque-sense/switch would be undetectable because loads on a real bike are way higher. 2 ways to look at this, either we have an incredibly sensitive torque-sensor that is a reprogram dream, or, it’s a switch.
Either way, color me surprised. What can you tell from the scope?
 
I'm seeing a 57Hz sine wave(are you in the US?). Are you sure you grounded the signal? The signal shouldn't be so sensitive to external influences.

Trying to see a torque sensor signal on the bench like this will be very difficult as you can't apply much torque.
 
wheelyphased said:
Here are some scope measurements of the V2 motor ‘hall’ sensor output at the 4-pin connector in the controller. There are 4 wires: red (assumed VCC), black (assumed GND), green and blue. I apologise if this reads as if I don’t know how to use a scope or know ‘lectroniks. That’s because I don’t. Bear with me.
This is fantastic wheelyphased. Great job!

The 60Hz interference is unfortunate. You can buy a pair of cheap probes if you need to do something like this in the future. We can ignore the in-phase X-Y 60Hz signals though, as shown here:
View attachment 3


Just as the motor kicks in, we can see two 180° out of phase signals:
View attachment 2


As torque is applied with the motor powered, we can see those signals become more in phase:
View attachment 1


And then we hit payday:
scope4.png


There's that oval I was hoping we'd see! It looks like we truly do have two signals that vary phase based on applied torque.

I'm sending off an email to Ling Bei now for a motor . . .
 
Surprise! A torque sensor has been found. Is this a contactless wire free sensor type?
Do Linbgbei sell motors direct?
 
hey sickle where are these on your website I can find it?
 
devo1223 said:
hey sickle where are these on your website I can find it?

http://recycles-ebike.com/search?controller=search&orderby=position&orderway=desc&search_query=EBBS02&submit_search=

That's the partner website I provide support for. FYI though, they are not the 12 mosfet, we have a few of the older models left... I like them but (theroyal?)we definitely need to do something about the programming, which clearly leaves much to be desired and doesn't take advantage of the variable/incremental torque signal/reading.

It's not on my own eco-ebike website... I only have a couple new models, and working out my next move concerning this motor...
 
I'm hanging my head in shame about posting images with ungrounded probes. I didn't even know this was necessary, now I've found the little sleeve on the probe that facilitates this. Here's another set.
Channels A & B (green & blue wires), at rest, powered
View attachment 2
Channel B, cranking
Channel B, cranking.jpg
Channels A & B, cranking
View attachment 4
Channels A, B, A vs. B, cranking free. The A vs. B plot is the fainter grey box with diagonal lines
Channels A, B, A vs B cranking free.jpg
Channels A, B, A vs. B, cranking with hand torque applied.
Channels A, B, A vs. B cranking with torque.jpg
Although it's a slight parallelogram shape, it is most likely that I'm unable to generate enough restraining torque. Additionally, this time I can start the motor using crank movement alone (no restraining torque). I was starting with too high cadence, staring with a low cadence fires the assist up without torque (i.e. like a BBS).
 
We can very clearly see between the second-to-last and last image that there's a variation between pulse train differentials. The free-cranked picture shows approximately 106ms between pulses while the hand-torque-applied picture shows roughly 42ms.

Interestingly enough, assuming the bottom picture is rotating at between 60-80RPM, we can see roughly 120ms width pulses. I'd make a guess that there's 6 magnetic poles per sensor, or one pole every 60°. This isn't close enough together for a single sensor to make an effective torque sensor. However there are two sensors offset from each other by 50%. That means the combined pulse train is every 30°. Very clever. This corresponds with your previous observation that approximately 30° of rotation is needed to engage the motor.

Having to move the pedals 30° or more before torque is determined is, in my opinion, too little. Imagine you've stopped on a steep hill, accidentally leaving your bike in high gear and there's a truck barreling down on you. To get out of the way you'd have to really crank down on the pedals, and since you're cranking down so hard you're moving the two pulse trains closer together, you're increasing the detection distance. So now you have to rotate 45° or more just to get a motor assist. Splat! You're road kill!

Maybe we can come up with a magnet mod where we adhere a bunch of thin, magnetic strips to the rotor to increase the pulse train density. That's something to consider in the future.
 
Now it seems this motor has possibilities (enhancenments in future firmwares, either official or unofficial, at some point in the future), and has been tested for some reasonable time, would be great if any could give a review in some detail about what it has to give at the present time: noise, perceived durability, performance and feels, differences between old and new firmware, with old BBS02, how much the torque sensing effect is present, how it alllows gear changes, some photos and/or a video, etc.
 
This is a follow up on the scope measurements. A sanity check if you like. I took a look inside my V2 motor. It’s not easy to work on and without kisazul’s disassembly tips I may have failed. Here is the sensor.
Sensors.jpg
You can see a small PCB with 1 hall? sensor head that isn’t covered by epoxy potting and another one further inboard (toward the center of the motor) that is, partially. The axial position of these heads align perfectly with 2 magnetic rings that are mounted on the crank shaft. Now the shaft.
Shaft.jpg
There are 2 rings, not magnetically strong & from the surface finish I would guess they are sintered so could have multiple poles.
Rings.jpg
The outboard magnetic ring mounts on an inner carrier & both are glued to the crank shaft. You can see overspilled glue. That locates the ring and provides a datum. The inboard magnetic ring is held between 2 circlips and is also glued in position. It’s easy to see that this arrangement provides for rotating the inboard ring between its circlips before the glue sets to give a desired phased signal between it and its twin during sub-assembly of the shaft.

That’s all. There is no other sensor either on, or adjacent the clutched gear. Compared to kisazul’s images, there is no difference between his earlier motor and my latest V2 motor. So, if the speculation in this thread that a torque signal is derived from the change in phase between these 2 sensors under shaft torsion, it must all happen in the shaft between these 2 rings, right?

The diameter of the crankshaft at the ring mounts is 17mm, the centre distance of the rings and sensors is 9mm and the OD of the magnetic ring is 24mm. Assuming this is a solid shaft, machined from one piece of carbon steel (the crank securing bolt holes are blind, so this is a safe assumption), then the angular deflection between the rings when the cyclist is a track professional is given by
α = 584 L T / G D4 = 0.163º
where L is 9mm, T = 200Nm, G= Modulus of Rigidity for carbon steel, D = 17mm
That’s a maximum peripheral deflection between the rings of 0.024mm or just over 1 thou.

This is a basic arrangement unlike the clever stuff in http://lup.lub.lu.se/student-papers/rec ... 044800.pdf
and it doesn’t take a genius to work out that if it was the intention to measure torque, you might at least start by putting some distance between the rings and/or making the shaft hollow/partially tubular to reduce its modulus. They didn’t do any of this, which suggests that torque measurement was not the design intent.
I attribute the ovality in my ungrounded X-Y scope plots to noise (and ignorance of taking proper electronic measurements) and differences in my grounded measurements to inconsistent crank speed. It is inconceivable that I could exert sufficient torque reaction in the shaft by hand of 10Nm, leave alone 200Nm. These are real difficulties with taking measurements under uncontrolled conditions as fantasy2 suggested.
The conclusion is blindingly obvious.
 
Interested as well,

If I got this little ( low power mid-drive ) I would pushing the limits of it's Volt and Amp usage, so does this Lingbei motor have a thermal shut down feature for when it starts to get to hot ?

mgtroyas said:
Now it seems this motor has possibilities (enhancenments in future firmwares, either official or unofficial, at some point in the future), and has been tested for some reasonable time, would be great if any could give a review in some detail about what it has to give at the present time: noise, perceived durability, performance and feels, differences between old and new firmware, with old BBS02, how much the torque sensing effect is present, how it alllows gear changes, some photos and/or a video, etc.
 
ScooterMan101 said:
Interested as well,

If I got this little ( low power mid-drive ) I would pushing the limits of it's Volt and Amp usage, so does this Lingbei motor have a thermal shut down feature for when it starts to get to hot ?
I am curious for this on this motor. This is very important to avoid demagnetization and so permanent damage/loss of torque of the motor!!!

On TSDZ2, took me sometime to believe that are the hall sensors the first ICs that protect themself (and turn off their signals) when the motor get very hot probably at around 125 degrees (hall sensors are installed inside the motor so they are the first ICs to get hot). The original firmware, in absent of the hall sensors signals, simple turn off the motor coils.

I have to install motor temperature sensor on my TSDZ2 motors: https://github.com/OpenSource-EBike-firmware/TSDZ2_wiki/wiki/How-to-install-motor-temperature-sensor
 
Back
Top