Power / Torque Sensor Roundup

Power and torque sensing is an interest for a subset of us who like the feel of a bike without a hand throttle that responds to how much effort we put into pedaling. I would like to make a bike that uses a torque sensor to drive a mid drive motor without the speed limitations that manufacturers (Bosch, Bronx, ect...) always build in.

Here is a link to the current set of power meters from Power Tap, which, while still expensive, are more capable and cheaper than they were a few years ago. If anyone else has an interest in this and would like to post a link to another system, it would be nice to be able to compare them and see what the state of the art is in 2015.

http://www.powertap.com/page/compare

The best solution would be to get Bafang to add a BB torque sensor to their BBS systems. They have the technology (in the OEM Max drive system, and the separate torque sensor which they currently sell), but they're sitting on it. Maybe they're afraid that if they put too many features into their aftermarket products, it will detract from the Max sales?

Yeah there is a rather big lack of information about torque sensors and setups on this site. Most people seem to be happy with there throttle setup or standard rotational pas. This sort of makes sense if you want a high power setup or don't want to work yourself. But if you have ever ridden a good torque sensor setup you really do not want to have anything else.

Maybe we can make this thread the place for information about torque sensors and setups.

BB Sensors
The Thun torque sensor X-cell RT is expensive but from a German brand and is known to work with CA3 and gives you info on the ammount of work you are doing
there is also a cheaper Thun BB sensor that doesn't give info on how mutch you do http://em3ev.com/store/index.php?route=product/product&product_id=206

There are some Chinese BB torque sensors, no idea how well they work and if you can set them up with CA3
http://www.alibaba.com/product-detail/BB-Torque-Sensor-for-E-bike_1345262360.html
http://www.alibaba.com/product-detail/ISIS-Torque-Sensor-bottom-bracket-for_60205374673.html

Crank Sensor
http://www.alibaba.com/product-detail/OR05D1-OutRider-3-12-magnets-ebike_2002223729.html (looks rather crappy)

There is a Chain tension sensor, Called Beam TS sensor
Haven't seen them sold anywhere and it seems rather fragile

Torque Simulation controllers are sold by BMSbattery, no idea how they work and if it really is as good as a real one. I do not think it is possible as it would still need a rotation sensor and they need a bit of rotation of the crank to work and therefore will have more delay then a BB sensor.
https://bmsbattery.com/ebike-kit/667-s12sh-800w-35a-torque-simulation-sine-wave-controller-ebike-kit.html


I also want to do a build with a torque sensor and mostly i am really wondering if CA3 or controllers can be programmed with a custom curve for the torque sensor, for example program it to help very little at lower speeds and incrementally increase the support to 40kmh and then gradually build support off to no support at 48kmh.
This to make sure i pull up like a regular bike and can drive around 40kmh with my own power still mattering.

I do not have a CA3 so cannot test or see that myself

The V3 can do linear rpm assist scaling with a hard power (assist) limit at a preset speed.
RTM

teklektik said:

The V3 can do linear rpm assist scaling with a hard power (assist) limit at a preset speed.
RTM

Click to expand...

The problem is that this provides the opposite of what you'd want during takeoffs, when you need the most power at the lowest rpm.

Malloot said:

There is a Chain tension sensor, Called Beam TS sensor
Haven't seen them sold anywhere and it seems rather fragile

Torque Simulation controllers are sold by BMSbattery, no idea how they work and if it really is as good as a real one. I do not think it is possible as it would still need a rotation sensor and they need a bit of rotation of the crank to work and therefore will have more delay then a BB sensor.
https://bmsbattery.com/ebike-kit/667-s12sh-800w-35a-torque-simulation-sine-wave-controller-ebike-kit.html

Click to expand...

I'm using a combination of the BeamTS, and an S12S Torque Simulation controller. I've documented my experience here, along with some thoughts on why I feel that torque and cadence sensors provide the best experience (for me, anyway):

http://endless-sphere.com/forums/viewtopic.php?f=3&t=67086

This is a "budget" option, because it does not require a CA. But it's not applicable to a mid-drive system.

Avitt said:

The problem is that this provides the opposite of what you'd want during takeoffs, when you need the most power at the lowest rpm.

Click to expand...

We are talking about V3 support for torque sensors. The CA uses both rpm+torque to determine assist so high rider getaway torque = more assist. That said, for a very broad speed range, it may be necessary to dial up the assist level manually. A conveniently placed pot (thumb or bar-end) makes this fairly easy. Again, there are adjustments to the assist level ramp curve, start assist level, etc. PAS is not simple to adjust for a good feel - particularly if the usage spans different riding/speed situations.

Avitt said:

Malloot said:

There is a Chain tension sensor, Called Beam TS sensor
Haven't seen them sold anywhere and it seems rather fragile

Click to expand...

I'm using a combination of the BeamTS, and an S12S Torque Simulation controller. I've documented my experience here, along with some thoughts on why I feel that torque and cadence sensors provide the best experience (for me, anyway):

http://endless-sphere.com/forums/viewtopic.php?f=3&t=67086

Click to expand...

Since the V3 uses a 0-5V torque sensor input and has the ability to set the zero-torque baseline voltage for a particular unit, you can quite likely use this sensor with the V3. The V3 has adjustable averaging to take the pedal stroke variations out of the equation. Here you would calibrate the V3 for the desired Nm/V that the sensor happened to generate and it should give torque+rpm scaling. I mention this since you have a Beam unit in hand - the TDCM BB sensors from Grin are actually a bit cheaper and already available as a tested V3 option.

Yeah, the TDCM is within $10 of the BeamTS. I would have tried it, if my setup allowed a BB sensor.

Pedal stroke averaging is necessary to deal with the Thun sensor, which only detects torque on the left pedal. But it's a fine line between linking the strokes, and making the system feel detached and insensitive.

Avitt said:

Malloot said:

There is a Chain tension sensor, Called Beam TS sensor
Haven't seen them sold anywhere and it seems rather fragile

Torque Simulation controllers are sold by BMSbattery, no idea how they work and if it really is as good as a real one. I do not think it is possible as it would still need a rotation sensor and they need a bit of rotation of the crank to work and therefore will have more delay then a BB sensor.
https://bmsbattery.com/ebike-kit/667-s12sh-800w-35a-torque-simulation-sine-wave-controller-ebike-kit.html

Click to expand...

I'm using a combination of the BeamTS, and an S12S Torque Simulation controller. I've documented my experience here, along with some thoughts on why I feel that torque and cadence sensors provide the best experience (for me, anyway):

http://endless-sphere.com/forums/viewtopic.php?f=3&t=67086

This is a "budget" option, because it does not require a CA. But it's not applicable to a mid-drive system.

Click to expand...

Thank you for the writeup, makes it clear there is still quite some work to get a torque sensor setup. The Beamts does have an advantidge over the bb setups because it does mnt need CA3 making it allot cheaper if you do not have that. The Beamts seems to have allot of settings by itself.

Any posts detailing how to setup a bb sensor with CA 3?

This does left and right measurements and may have 32 magnets:

http://www.alibaba.com/product-detail/BB-Torque-Sensor-for-E-bike_1345262360.html

Has anyone tried it with a V3?

Torque sensor + PAS sensor provides the ultimate ebike experience, for me.

If you have things adjusted correctly, the experience becomes "just like riding a bike", but with super powerful legs. It's completely intuitive, and natural feeling. The only thing I have to do is shift (to adjust my cadence) and occasionally edjust the PAS level (which modifies my maximum speed).

If you haven't tried it, you just don't know...

Malloot said:

Any posts detailing how to setup a bb sensor with CA 3?

Click to expand...

Just go to Grin Tech's Cycle Analyst V3 Info page - that should get you started. SW setup only took me about 15 using the guide - then tweaking to dial it in. I just installed a TDCM a couple weeks ago and I'm loving it. The hardest part was drilling a 10mm hole in the BB and getting the measurements and alignment right. Justin is really responsive with questions.

I'm using the Thun on my bike with a CAV3.

It works well with the downside of there being a lot of lag (and the lag appears to gets worse when you turn down the PAS-level on the CAV3, which is annoying). It would be nice if the CA figured out an average starting torque ramp from previous start from a dead stop accelerations and ramped up at that average level until it was getting actual feedback from the BB (that takes 1-2 seconds). It does know pedaling quite quickly (well under 1 second).

There is no throttle on my bike and I like it that way. It encourages you to keep working because you can't get more power out of the system than you are putting in yourself.

I've also ridden a pure PAS system (where any pedal movement was enough to start the motor) using a Bafang bottom bracket motor. It felt terrible in comparison because you always needed to adjust the power levels. The biggest surprise was having the bike accelerate hard when I was just lining up my cranks to jump off of a curb.

I wish that there was a CA version specifically for pedal assist bikes with fewer wires, a much smaller display and thus easier to integrate. The CAV3 is the biggest and ugliest part on my ebike.

Alex W said:

It would be nice if the CA figured out an average starting torque ramp from previous start from a dead stop accelerations and ramped up at that average level until it was getting actual feedback from the BB (that takes 1-2 seconds).

Click to expand...

There are two components parameters I think that cause the drag: min watts and min RPM. Bypassing real time requirements of these two parameters would be a safety issue if, for example, a child stepped on a pedal of a parked bike and it took off. A better way to minimize lag I would think is to add magnets to the wheel and set the CA3 accordingly. I assume that for min RPM, it looks for at least two consecutive magnet triggers. Put 4 magnets in and set in watts to a low number - you reduce your lag possibly down to 1/4 revolution.

In torque sensing mode the bottom bracket is sampling over a full revolution of the cranks, I don't think it cares about the wheel. It would be great to get some confirmation though, if it does care about the wheel rpm then I'll find some more magnets and try it out with 4.

I think that there is also a bug where a low aux-level seems to also cause more lag. I sometimes have to "trick" the motor into turning on by setting my assist level to 1, get the motor going, then drop it to 0.5. I've ridden blocks without any aux power at 0.5. That is my normal preferred setting and I'd like to have the bike work well running at half assist.

It would be great if someone with the source code to the CAv3 would ride their primary bike without a throttle for a month. The system is a lot less responsive than some other torque-only bikes that I've ridden. A friend gave up and installed a throttle on his torque assist bike. I don't want to do that, but do want less lag.

I'm also surprised that the bike doesn't cut power to the motor as soon as it sees that my pedal RPMs have stopped. That is easy to measure and the Thun BB has 8 PAS poles. So if I'm pedaling at 90rpm (my normal) there are 12 pole pulses per second. Every time a pole signal happens a timer should start that is approximately 2 times longer than the recent time decayed pole period. So that would kill the motor in one sixth of a second. Today it seems to wait for what would have been one pedal revolution, which is 2/3rds of a second at 90rpm.

Alex W said:

In torque sensing mode the bottom bracket is sampling over a full revolution of the cranks, I don't think it cares about the wheel. It would be great to get some confirmation though, if it does care about the wheel rpm then I'll find some more magnets and try it out with 4.

Click to expand...

Hey Alex, the wheel diameter has nothing to do with it. You do need to make sure that the # Pas Poles is set to 8, and that the Trq Averaging is also set to 8 (or 16 or 24, but those changes would make it less rather than more responsive by averaging over multiple pedal rotations).

What firmware version are you running right now? At some point the starting behavior was updated so if you just start to apply torque on the crank after you haven't been pedaling, then it only does an immediate averaging of that force up to the point where you have the full revolution in place, so it's acting to a pretty immediate signal and not waiting for you to do a full pedal turn.

If there seems to be a lag at startup, then the usual culprits would be either your Start RPM threshold being too high (set it to a low value like 8-10rpm), your WGain setting being too low (the CA is operating in a power limited feedback mode, so a tightly tuned watts feedback gain is important), and/or your throttle ramp rates being too slow. It's really useful to look at both the 3rd screen (showing human watts/rpm and output motor watts) as well as the diagnostics screen (Showing output throttle voltage level) when you are trying to understand what's happening to dial in the behavior.

I think that there is also a bug where a low aux-level seems to also cause more lag. I sometimes have to "trick" the motor into turning on by setting my assist level to 1, get the motor going, then drop it to 0.5. I've ridden blocks without any aux power at 0.5.

Click to expand...

There isn't any bug observed here but it's also possible that the torque sensor's zero torque offset has drifted, or that your Assist Start is a bit high, 20-30 watts is usually sufficient for that.

It would be great if someone with the source code to the CAv3 would ride their primary bike without a throttle for a month. The system is a lot less responsive than some other torque-only bikes that I've ridden. A friend gave up and installed a throttle on his torque assist bike. I don't want to do that, but do want less lag.

Click to expand...

The THUN sensor will always have the limitation that an application of torque on the left crank will never be noted, but with sensors like the TDCM units that respond to left and right pedal forces and which also have a higher PAS pole count the behavior can be made quite instant.

I'm also surprised that the bike doesn't cut power to the motor as soon as it sees that my pedal RPMs have stopped. That is easy to measure and the Thun BB has 8 PAS poles. So if I'm pedaling at 90rpm (my normal) there are 12 pole pulses per second. Every time a pole signal happens a timer should start that is approximately 2 times longer than the recent time decayed pole period. So that would kill the motor in one sixth of a second. Today it seems to wait for what would have been one pedal revolution, which is 2/3rds of a second at 90rpm.

Click to expand...

First, if you haven't yet be sure to install the latest P10 or P11 firmware and make sure that the quadrature mode is set to "2-wire", since this will help in the cutoff time and will also make it absolutely instant with the slightest back pedal. Secondly, you have full control over the cutoff lag time in the PAS setup menu via the Stop Thrsh RPM parameter. Say you set it to 20 rpm = 3 seconds/rev. With the 8 poles of a THUN sensor, then that means if the CA doesn't see a pulse in 3sec/8 = 0.375 seconds, it will shut off. If you'd like it to cut out in 1/6th of a second, then you'd set the Stop Thrsh to 45 rpm, but of course then whenever your pedal cadence falls below 45rpm the CA will keep thinking that you've stopped pedaling. Personally I wouldn't go much above 20-25 rpm for this.

Let me know if any of those things help in getting the behavior response you are after.

This is all a bit off topic form the original post, but I have some stuff to add to that as well too. First, we've been getting some samples of the BeamTS device without any of the internal throttle circuitry so that it's less expensive and easier to hook up to a CA3 for those interested in a lower cost pedal force that doesn't require a BB replacement, with the limitation that the calibration goes out of wack if you have derailleur gears. The raw signal goes form about 3.3V to 3.8V, so not as much of a range than we are used to seeing for best resolution but it still works great and Robbie has it installed on the rear of an IGH bike to good effect.


And secondly, there are many many (at least a dozen I'd say) companies advertizing torque sensing bottom brackets now, like Dapu:
http://www.dapumotors.com/torque-sensors/
Sempu
http://sempumotor.en.alibaba.com/productgrouplist-219984346/Dual_Side_Torque_Sensor.html
and countless others, so hopefully with more options on the market the prices will come down. Do keep in mind though that these are all WAY cheaper than a commercial/industrial torque sensor with similar specs.

-Justin

Thanks for the advice Justin. I'm running the P9 firmware, I always read the release notes on later versions and hadn't seen any noted updates in the torque. It's good to hear that there have been changes. I'll do an upgrade and check my settings and let you know how it goes.

I know that the Thun has the downside of only sampling on the left crank, but the play in the TDCM and it's heavier weight turned me off of it. It would be great to see a higher quality torque sensing BB option. I'd have paid an extra $100 for something with the quality of an old Shimano BB-UN72 (hollow spindle, very well sealed bearings, very smooth bearings) vs the lower quality of the Thun.

there is a new torque sensor from bafang that measures torque from both the left and right pedal:

http://www.szbaf.com/en/components/component/sensor/sen-st01325.html

It has an insane width and is designed for their cranks. The Q factor is going to be huge, like the bbs motors.

I'm looking for e-bike parts that don't get in the way of riding the bike as a normal high performance bike. There isn't a lot offered.

justin_le said:

First, we've been getting some samples of the BeamTS device without any of the internal throttle circuitry so that it's less expensive and easier to hook up to a CA3 for those interested in a lower cost pedal force that doesn't require a BB replacement, with the limitation that the calibration goes out of wack if you have derailleur gears. The raw signal goes form about 3.3V to 3.8V, so not as much of a range than we are used to seeing for best resolution but it still works great and Robbie has it installed on the rear of an IGH bike to good effect.

Click to expand...

Hey Justin, how likely will you do offer the BeamTS? with this, I can incorporate torque sensing on a Bafang BBS02 with the CA3 "throttling" the controller, right?

I've just got back from holiday and whilst there I hired a BH Emotion Neo Xtrem which has a torque sensing derailleur bracket. I must say it worked really well. The controller was a bit digital in that you could feel the step changes in torque as you pedalled harder but the sensor seemed to work very well.
I am a control systems engineer specialising in vfd's but I'm new to electric bikes.
I just wish I had time to develop a good analog controller!
Surely, it can't be too difficult to have a torque setpoint and feedback and feed them into a PID controller to give an output for the motor?
That way, at any speed, the pedalling force required would always be the same.

Justin: Are there known good values for the AGain and WGain with the 15A controller and a 250W geared hub motor? Or does it vary too much bike by bike to have known good values?

I've tuned PID loops before (for servos and VFDs on CNC machines), is AGain what would normally be considered proportional gain, and WGain is the integral gain? What is the derivative gain? The unofficial guide doesn't explain these very well.

I already had the rest of my system tuned as you recommended, except that the stop RPM was set very low (16rpm). I've tried raising this to 30rpm (one third of my normal riding cadence, and half of my lowest riding cadence) to see how that goes. I also raised my ThrottleOut ramp voltage from 2 to 4 volts.

I didn't know that back pedaling was used as a signal to kill output.

I'm confused as to why a hybrid PAS/TRQ mode wouldn't be safe if a PAS-only mode is considered safe. It seems like all of the starting lag could be removed by starting to ramp up power as soon as their is any pedaling input, measured using the PAS portion of the sensor. Once the user has been pedaling long enough (that doesn't take long) to get a torque input then that could take over. This seems like it would be just as safe as PAS mode and eliminate any perceived lag. Since I'm often riding at or above my hub's max speed the main times that the motor helps are during acceleration and when climbing hills. The lag during acceleration kills most of the benefit that I'd get on starts.

flyinbrick said:

Hey Justin, how likely will you do offer the BeamTS? with this, I can incorporate torque sensing on a Bafang BBS02 with the CA3 "throttling" the controller, right?

Click to expand...

No, you can't use the chain tension to sense your pedal torque on a mid-drive system since it has no way of discerning between you pedaling or the motor powering. Both will have exactly the same effect on the BeamTS sensor. It's only really usable for mid-drives if you run the electric motor via a jackshaft, so that only the pedal force is transmitted along the primary chainline with the beam sensor installed.

Alex W said:

Justin: Are there known good values for the AGain and WGain with the 15A controller and a 250W geared hub motor? Or does it vary too much bike by bike to have known good values?

Click to expand...

It would depend more than anything on the voltage pack you are running at. But it's easy to find out, in the case of AGain set your throttle mode to an Amp throttle (with appropriately set Max Amps), and then adjust the behavior until you like it. Then for the WGain, set your throttle mode to be a Power throttle, and tune WGain to have the response that you like.

I've tuned PID loops before (for servos and VFDs on CNC machines), is AGain what would normally be considered proportional gain, and WGain is the integral gain?

Click to expand...

Both AGain and WGain are integral gains, the former for your current limit feedback, and the latter for your watts limit feedback. In a PAS system, only the watts limit is in regular use, AGain won't come into play unless you also have a separate current limit programmed (or are using a throttle too, with the throttle configured as a current throttle).

What is the derivative gain? The unofficial guide doesn't explain these very well.

Click to expand...

The only control feedback system in the CA3 that has a full PID controller with derivative gain is the speed limit function. None of the others (amps, watts, or low voltage rollback) need that kind of tuning term.

I'm confused as to why a hybrid PAS/TRQ mode wouldn't be safe if a PAS-only mode is considered safe. It seems like all of the starting lag could be removed by starting to ramp up power as soon as their is any pedaling input, measured using the PAS portion of the sensor.

Click to expand...

You can easily remove starting lag, just set the Start RPM threshold really low, like 3rpm and it will kick in pretty much right after you do a 1/4 turn of the cranks. The downside is that if you move the pedals a bit and then stop, then the system will stay powered for quite a while before it shuts off which can be disconcerting. You have one pedal rotation before it switches from the Start to the Stop RPM threshold.

Once the user has been pedaling long enough (that doesn't take long) to get a torque input then that could take over. This seems like it would be just as safe as PAS mode and eliminate any perceived lag.

Click to expand...

I don't think any perceived lag on start has to do with torque sensing at all. If you apply torque and pedal at the same time the response will be just the same as a PAS mode configured for the same power output. Most likely it was your slow up-ramp more than anything.

-Justin

justin_le said:

No, you can't use the chain tension to sense your pedal torque on a mid-drive system since it has no way of discerning between you pedaling or the motor powering. Both will have exactly the same effect on the BeamTS sensor. It's only really usable for mid-drives if you run the electric motor via a jackshaft, so that only the pedal force is transmitted along the primary chainline with the beam sensor installed.

Click to expand...

Not all mid drive builds use a single chain for both motor and pedals. A common and simple example is connecting the motor to the left side of the rear wheel with a dedicated chain.

Avner.