[confused] Torque sensor, gear ratio, pedal assist

[mschwett]

Sorry, I'm confused. My neighbor's Aventon Level.2 is a torque-sensing hub-drive, and so is my ebike.

My ebike can only know crank torque, as it has no reliable way to measure crank speed — in which case, it can't know my input power, as you point out. But it still manages to deliver very natural-feeling assist.

Many reasons to believe that the Aventon doles out assist in much the same way as my bike, but your hub-drive could well be more sophisticated.

Assist terminology drives me nuts! The "cadence sensor" on most "cadence-sensing" ebikes doesn't measure cadence (crank RPM) at all. It just detects the presence or absence of crank rotation. And a "torque-sensing" assist system that also measures and factors in true cadence would be better described as a "power-sensing" system.

My ebike is torque-sensing but not power-sensing.

hmmmm! maybe that’s why some hub drives are accused of having a “jerkier” assist feeling, because they really don’t know how fast you’re pedaling? but given that there is little or no torque on one side of the crank once per cycle (since your feet don’t pull up on the pedals!) i’m certain that very rudimentary programming could also discern rotational speed. why do you say it couldn’t know? a graph of the torque on the “axle” between the cranks would peak and fall at exactly twice the frequency of the rotational speed, because when either pedal is at the top and the other at the bottom, there’s significantly less work being done. maybe they don’t bother and it’s more like an on-off switch, which coupled with the speed sensor is used to determine the assist level? someone here probably knows.

i know for a fact that my hub drive road bike knows rider cadence - it puts it on the screen, along with rider watts, which of course can’t be measured without cadence.

 

[McCorby]

Jeremy, for comparison, you really should try a bike with a Bosch, Shimano, or Brose mid-drive.

 

[Jeremy McCreary]

hmmmm! maybe that’s why some hub drives are accused of having a “jerkier” assist feeling, because they really don’t know how fast you’re pedaling?

My TMM4 rear dropout torque sensor might know how fast my cassette is turning. More likely case, it only knows that the cassette's turning. The latter's a perfect proxy for whether or not I'm pedaling. But no way to turn any of that into a reliable cadence on a 10-speed bike with non-stock gearing like mine. (Search the forum for "TMM4" for details.)

Yet, nothing jerky about my assist under normal circumstances. However, l do get a little surging in one uncommon and easily avoided/remedied situation — very high assist, low gear, and low total resistance.

but given that there is little or no torque on one side of the crank once per cycle (since your feet don’t pull up on the pedals!) i’m certain that very rudimentary programming could also discern rotational speed.

Agree, controller software could in principle read cadence from the ripple in the torque signal — say, with signal processing based on the fast Fourier transform. Since my surging clearly syncs 1:1 with my power strokes, I know that the controller can at least see the ripple.

But is the controller CPU fast enough for that processing? And does it use the ripple to make assist decisions? Hard to know.

why do you say it couldn’t know? a graph of the torque on the “axle” between the cranks would peak and fall at exactly twice the frequency of the rotational speed, because when either pedal is at the top and the other at the bottom, there’s significantly less work being done. maybe they don’t bother and it’s more like an on-off switch, which coupled with the speed sensor is used to determine the assist level? someone here probably knows.

Can't find any other sensor in a position to measure true cadence. I've seen the bottom bracket cartridge. No visible sensor components there.

i know for a fact that my hub drive road bike knows rider cadence - it puts it on the screen, along with rider watts, which of course can’t be measured without cadence.

Given the huge price difference, not at all surprised that your hub-drive's more sophisticated than my lower middle-tier bike. Yours is clearly power-sensing. Mine's only torque-sensing.

I also gather from reports on EBR that some purely torque-sensing schemes are better than others. I seem to have lucked into a good one.

 

[Jeremy McCreary]

Jeremy, for comparison, you really should try a bike with a Bosch, Shimano, or Brose mid-drive.

I'd certainly like to experience the difference first-hand, but no new ebike in my foreseeable future — and short of conning an LBS, none available to try now.

 

[McCorby]

I'd certainly like to experience the difference first-hand, but no new ebike in my foreseeable future — and short of conning an LBS, none available to try now.

Most LBSs that sell the major brands would be more than happy to let you test ride them with no expectation to purchase. Just say’n.

 

[mschwett]

My TMM4 rear dropout torque sensor might know how fast my cassette is turning. More likely case, it only knows that the cassette's turning. The latter's a perfect proxy for whether or not I'm pedaling. But no way to turn any of that into a reliable cadence on a 10-speed bike with non-stock gearing like mine. (Search the forum for "TMM4" for details.)

Yet, nothing jerky about my assist under normal circumstances. However, l do get a little surging in one uncommon and easily avoided/remedied situation — very high assist, low gear, and low total resistance.



Agree, controller software could in principle read cadence from the ripple in the torque signal — say, with signal processing based on the fast Fourier transform. Since my surging clearly syncs 1:1 with my power strokes, I know that the controller can at least see the ripple.

But is the controller CPU fast enough for that processing? And does it use the ripple to make assist decisions? Hard to know.



Can't find any other sensor in a position to measure true cadence. I've seen the bottom bracket cartridge. No visible sensor components there.


Given the huge price difference, not at all surprised that your hub-drive's more sophisticated than my lower middle-tier bike. Yours is clearly power-sensing. Mine's only torque-sensing.

I also gather from reports on EBR that some purely torque-sensing schemes are better than others. I seem to have lucked into a good one.

ahhh ! i see, i thought the sensor you were referring to was in the bottom bracket! now i understand, and agree, no easy way to tell from back there! lots of different ways to do these things, each with pros and cons for sure.

this is the bottom bracket of a mahle x20 system - i have no idea how it really works, since the cool thing about it is that allows the use of a totally standard bike crankset. i can easily see how rotation would measured but i’m somewhat baffled how it could measure torque since the crank axle just spins freely within it... but it does. not super accurately, but it does.

the replacement sells for around $200 retail, so it’s probably a $75 part or something.

 

[Jeremy McCreary]

Out of curiosity, what's the brand? How does that torque sensor look like?

The bike's a 2023 Surface 604 V Rook. The rear dropout torque sensor is a TMM4 by IDbike. Search the forum on "TMM4" for details.

Some hub-drives clearly implement torque-sensing assist better than others. I seem to have lucked into a good one.

 

[Winfried]

How is purely torque-sensing PAS actually implemented?

Curious as to how purely torque-sensing PAS is actually implemented in ebikes with rear hub motors? Few details on the internet, but simple control models along these lines keep popping up: At a given PAS level p, the motor's output torque Mp will be Mp = Xp R where R = rider input torque at...

forums.electricbikereview.com

 

[Kayakguy]

Yes, if I understand your questions correctly. Someone here will probably offer a more detailed and better qualified explanation but if, for example, I want to ride north of 20mph on my Priority Current, I'll likely need to be in the highest gear (5th gear in this case). The amount of pedal assist has less to do with top speed, although it's obviously easier to get there with more assistance.

My Ariel Rider C class claims to employ both cadence and torque sensing. I definitely feel a surge of power when I suddenly increase my leg input. But that surge seems very brief--just one or two rotations of the crank.

 

[mschwett]

The bike's a 2023 Surface 604 V Rook. The rear dropout torque sensor is a TMM4 by IDbike. Search the forum on "TMM4" for details.

Some hub-drives clearly implement torque-sensing assist better than others. I seem to have lucked into a good one.

that’s really cool, there are a bunch of videos and spec sheets describing how it works, it’s basically a strain gauge that measures the deflection of the mounting plate caused by the chain pulling on it.

all it would need to know to determine power would be how fast the wheel is spinning - it actually doesn’t matter what gear you’re in, or how fast the cranks are going around since the relationship between torque, speed, and power is constant.

long before the currently popular crank arm and pedal based power meters, many road cyclists used the “powertap” hub which incorporates a number of the same hall effect strain sensors into the hub itself - and since the hub is spinning, you can then calculate power very precisely.

i am guessing your rear hub motor knows how fast it’s spinning, so the controller actually DOES know how much power is being used in total - whether it could separate your contribution from the battery’s, not sure! but it illustrates how many different ways there are to do engineering

 

[Nvreloader]

I have found a couple of ebikes that has torque sensing (and) PAS on the bottom brackets crank arms???

That don't make sense to me and these are middrive step thru bikes,
which I am looking for the better half so we can ride around here, everything is up hill from here.

Can they combine both together?

 

[Jeremy McCreary]

I have found a couple of ebikes that has torque sensing (and) PAS on the bottom brackets crank arms???

That don't make sense to me and these are middrive step thru bikes,
which I am looking for the better half so we can ride around here, everything is up hill from here.

Can they combine both together?

My hub-drive has both in one unit on the right rear dropout, where it simply detects cassette rotation as a proxy for pedal rotation.

The cadence sensor is used to disable the motor when the pedals aren't turning — e.g., when waiting at a stoplight with your foot on the up pedal. Wouldn’t want pedal pressure turning on the motor in situations like that.

 

[PedalUma]

Here is one that is going out in 45 minutes. It is a mid-drive with a torque sensor and active sensor of cadence, in other words it has a power meter. It is a 90 Nm 500W cargo motor with a premium battery. The owner rides Mt. Tam where mountain biking was invented. It is geared for climbing. I installed a comfort bar and matching grips with a stem riser.

 

[harryS]

I don't doubt that many torque sensor bikes also need cadence sensors. There are probably plenty of power delivery situations where the algorithms want to know if the pedals are spinning too, And that would apply if the torque sensor is in the rear axle. The electric brain also wants to know if the gears, hence the wheel, is moving .

 

[Jeremy McCreary]

that’s really cool, there are a bunch of videos and spec sheets describing how it works, it’s basically a strain gauge that measures the deflection of the mounting plate caused by the chain pulling on it.

all it would need to know to determine power would be how fast the wheel is spinning - it actually doesn’t matter what gear you’re in, or how fast the cranks are going around since the relationship between torque, speed, and power is constant.

long before the currently popular crank arm and pedal based power meters, many road cyclists used the “powertap” hub which incorporates a number of the same hall effect strain sensors into the hub itself - and since the hub is spinning, you can then calculate power very precisely.

i am guessing your rear hub motor knows how fast it’s spinning, so the controller actually DOES know how much power is being used in total - whether it could separate your contribution from the battery’s, not sure! but it illustrates how many different ways there are to do engineering

That PowerTap is an interesting ANT+ device. The G3 rear hub version measures numerical values for wheel speed, the torque delivered to the rear wheel through the chain, and "virtual cadence". (Guessing they get the cadence from the torque ripple, but it's said to be pretty accurate.) Toss in an assumed drivetrain efficiency, and you have all the parameters needed to estimate rider power at the crank on a motorless bike.

On a rear hub-drive ebike like mine with wheel speed reported by the motor and chain-related rear wheel torque from a TMM4 sensor, you might be able to tease out rider power in principle. But you'd need a very sophisticated motor-specific controller with knowledge of motor torque and mechanical power output at all times.

Did the math, and you'd also need either a separate true cadence sensor or some way for the controller to know the current gear ratio.

Doubt such a controller and sensor package exists in the hub-drive realm.

 

[mschwett]

That PowerTap is an interesting ANT+ device. The G3 rear hub version measures numerical values for wheel speed, the torque delivered to the rear wheel through the chain, and "virtual cadence". (Guessing they get the cadence from the torque ripple, but it's said to be pretty accurate.) Toss in an assumed drivetrain efficiency, and you have all the parameters needed to estimate rider power at the crank on a motorless bike.

On a rear hub-drive ebike like mine with wheel speed reported by the motor and chain-related rear wheel torque from a TMM4 sensor, you might be able to tease out rider power in principle. But you'd need a very sophisticated motor-specific controller with knowledge of motor torque and mechanical power output at all times.

Did the math, and you'd also need either a separate true cadence sensor or some way for the controller to know the current gear ratio.

Doubt such a controller and sensor package exists in the hub-drive realm.

i don’t think you need to know the gear ratio - you know how fast the rear wheel is being spun and you know the torque present at the axle. together that’s power, i think? if the motor was on, you’d have to have a way to remove that effect because it increases the wheel speed and the torque on the dropout, right?

the complexity of that is probably why mahle opted to just put it in the bottom bracket rather than the rear wheel - their BB power / cadence sensor looks to be a pretty simple device and uses standard cranks. i don’t actually know how it works but it probably measures the downward “squeeze” on the shell from the axle or bearing, since the BB itself has no rotating parts attached to the axle.

 

[Jeremy McCreary]

i don’t think you need to know the gear ratio - you know how fast the rear wheel is being spun and you know the torque present at the axle. together that’s power, i think? if the motor was on, you’d have to have a way to remove that effect because it increases the wheel speed and the torque on the dropout, right?

the complexity of that is probably why mahle opted to just put it in the bottom bracket rather than the rear wheel - their BB power / cadence sensor looks to be a pretty simple device and uses standard cranks. i don’t actually know how it works but it probably measures the downward “squeeze” on the shell from the axle or bearing, since the BB itself has no rotating parts attached to the axle.

Yes, very familiar with the relationship between power, torque, and angular speed. Unfortunately, not as simple as that when the rider and motor are both acting on the rear wheel at once to produce the measured wheel speed.

To disentangle rider and motor contributions to total rear wheel power and torque, you need to EITHER measure cadence independently OR know the gear ratio to get cadence from wheel speed. Go through the equations, and you'll see that there are too many unknowns otherwise.

Then, of course, you'd need a controller programmed to process this data and display the results.

 

[mschwett]

Yes, very familiar with the relationship between power, torque, and angular speed. Unfortunately, not as simple as that when the rider and motor are both acting on the rear wheel at once to produce the measured wheel speed.

To disentangle rider and motor contributions to total rear wheel power and torque, you need to EITHER measure cadence independently OR know the gear ratio to get cadence from wheel speed. Go through the equations, and you'll see that there are too many unknowns otherwise.

Then, of course, you'd need a controller programmed to process this data and display the results.

interesting, so if you turned the motor off you could measure rider power?

conceptually, would you not be able to measure the total power in use (from torque and wheel rotation) and then just subtract out the amount of power the motor is being fed times 80% or whatever the estimated efficiency of the motor is?

 

[Jeremy McCreary]

interesting, so if you turned the motor off you could measure rider power?

conceptually, would you not be able to measure the total power in use (from torque and wheel rotation) and then just subtract out the amount of power the motor is being fed times 80% or whatever the estimated efficiency of the motor is?

Let me look at the calculations again. As I recall, you'd also need motor torque for the 2nd paragraph. The TMM4 measures only the torque coming through the chain.

 

[mschwett]

Let me look at the calculations again. As I recall, you'd also need motor torque for the 2nd paragraph. The TMM4 measures only the torque coming through the chain.

that’s what i’m wondering about - is the reaction to the motor torque also in the dropout where the TMM4 is located?