As far as I’m concerned, torque-sensing’s existence is primarily owed not to the fact that it is a better system, but instead is a tool to help persuade an existing customer base (recreational, leg-powered cyclists) not to hate the product (ebikes) quite as much as they already do (either that or to sell ebikes while not cannibalizing sales of analog bicycle brethren).
Say what?
What is Torque-Sensing Pedal Assist?
On an ebike, when a torque sensor is used, it applies a strain gauge to the drivetrain (located either inside the bottom bracket, or in the back of the bike near the gear cluster). This measures the amount of force you apply to your pedaling stroke. If you pedal (work) harder, the assist you receive is dialed up. If you pedal more softly – regardless of cadence – the assist level is reduced… or eliminated.
I have heard it said that torque-sensing “rewards pedal effort” and this statement is both correct and indicative of the root problem with its advocacy. Old school cyclists hate the fact ebikes allow someone to make forward progress, without using their muscles in the first place. By restricting/keying the assist to physical exertion levels, the fact that a motor exists at all is less difficult to accept – and more easily sold to the existing cyclist population.
It also allows an ebike to be sold without denigrating the old-school unassisted version. Zillions of which are still manufactured for sale worldwide. If torque sensing just makes it seem “more like a regular bicycle” then that helps preserve the perception that a normal bicycle is still every bit the desirable, viable product that manufacturers still need to sell millions of.
It is unfair to say torque-sensing is ONLY about these things. Its not. You will also hear people say torque-sensing results in the most ‘natural’ bicycle riding experience for them, since you still have to work hard on the pedals. And the assist increases in proportion to your effort, just like a real bicycle. An ebike goes faster of course, but a physical work ethic is still demanded. So to be fair, torque-sensing does indeed give cyclists who want this a familiar and desirable experience. There is nothing wrong with that.
What is Cadence-Based Pedal Assist?
In its simplest form, its nothing more than this: Your assist level goes up or down based on how fast the crankarms are turning. The amount of effort you expend could even be irrelevant if your gearing is low enough. The only thing that matters is the rotational speed of the pedals/crankarms (strictly speaking it is the spindle’s rotational speed that is measured, but thinking ‘pedal rotation’ is easier to visualize).
So if you want more assist, you just turn your legs faster – not harder. Again in simple circumstances this means you can get into a low gear and easily ‘ghost pedal’ your ebike, without expending any effort. So you are breezing right along right up to either the speed limit of the ebike or the road/path you are riding on (please note I am not saying this always happens… only that it is possible with this type of system).
Such a thing is utterly anathema; deeply, personally hated in the cycling community. There, your progress and ability is hard earned through what can only be described as prolonged, personal, stoic suffering whose level outsiders neither understand nor hope to match. Despite the spandex, funny hats and silly shoes, cyclists know they are endurance badasses (they really are).
Except, fate has dictated these solitary warriors suddenly have to share the road with the Griswolds, blowing past them in their two-wheel Trucksters. Ebikes democratize cycling so that now… anyone can do it? WTF!?!
… not a shock the response of cyclists to ebikes has been negative.
Far From Perfect
Finally, lets make the very important point that criticism of cadence-based systems is often entirely justified. Especially on low cost direct-to-consumer ebikes. Unfortunately, rather a lot of these systems have cadence-based pedal assist that is merely an on/off switch. It pays no attention to how fast the crankarms are turning, or how fast the bike is going. It justfires up and produces one of maybe five different power output levels, steady and regular.
Faced with that kind of behavior, its no wonder so many people think torque-sensing is the only way to go.
Its not so simple
I said above the description of cadence-based pedal assist was in “its simplest form”. There are some big exceptions to this that most ridersare unaware of. Criticism of cadence-based assist systems can be entirely justified… but it should be recognized that the concept of cadence-based assist is not the problem, but rather to crappy, cheap-ass implementations of it.
There are some cadence-sensing ebike motors that have settings both complex and rather profound in how they impact the riding experience. Notice I did not say ‘cycling experience’ because a central tenet of my rejection of torque-sensing is that ebikes are bicycle-shaped-objects, but not bicycles. It is a mistake to treat them as if they should behave the same (unless that is something you expressly want).
The Cadence-Sensing Mid-Drive
If there is such a thing as a ubiquitous aftermarket mid drive on the market, its the Bafang BBS02 and its heavy-duty big brother, the BBSHD. There are zillions of them out in the world. If you want to find yourself a drive that has plentiful aftermarket support, countless users in discussion groups with experience to share, and myriad how-to’s out there on Youtube and the internet in general… well the go-to for those things are these DIY kit Bafang mid drives.
From the factory, quite frankly the performance settings on these drives is awful, and this is well-known. Pedal assist is overpowering and easily lets the bike run away from you. That powerful pedal assist also saps the range of the bike. But the bad news gets worse when you factor in the clunky, laggy motor engagement that bashes on the gears, and generally does its best to wear the bike’s drivetrain down sooner, not later.
Fortunately this is all very well known, and there is a rich settings interface hiding under the hood for these drives, with readily-available free software that will let you dig into the motors and completely change (i.e. tame) their character. With that in mind, here are some screen shots showing those settings.

Some cadence assist options are on the left side above. This isn’t even the Pedal Assist screen for this particular motor. These settings determine assist strength and when, based on both speed and motor rpms, the assist gets cut back.

Here you can determine how much the cut-back is when pedal assist is reduced, and more. The graph explains how the various settings on this screen affect output.
This is not at all the simple on/off concept most people think of when they complain about what they think of as cadence-based pedal assist
UPDATE:
Since this article was written in 2019, I have documented a cycling-oriented approach to this motor in BBSHD Programming for the Pedaling Cyclist.
The KT Controller And A Hub Motor
What everyone commonly calls the “KT” controller is in fact a product of the Suzhou Kunteng Electronics Co., Ltd. KT controllers can be found as original equipment inside of manufactured ebikes, and sold individually as aftermarket upgrades and components for builders who are building their own ebike. This is a company that has its products inside of many ebikes, but unless the owner reads the label on the controller, they would have no idea whether its a product made by this company.
I bring them up because KT controllers are very popular in the DIY ebike building community. They provide a quality product that offers features not readily available with other products in their class.
KT controllers use cadence-based sensing, but apply a proprietary algorithm that someone nerd in the basement was allowed to label as ‘imitation torque control‘. I don’t know where that comes from, but as their (poorly translated) web site says,
“Use imitate torque control mode to help achieve a smooth start-up of the vehicle, the power-assist drive will be smooth and natural. Electric power assist is with fast response and always consistent with the rider’s pedal action to achieve the effect of the torque boost.”
Thats fairly typical Chinglish marketingspeak, but they really are onto something.
- When you start off from a stop, power ramps on via a recognizable curve. It doesn’t just flick from ‘off’ to ‘full blast’
- The startup curve can be adjusted in three separate steps so if that on-curve is too strong, you can set one of three increasing levels of slow-start to give a more shallow slope to that startup curve.
- When starting off and crankarm rotation and speed are both slow, power applied is high. This corresponds to what you need when starting from a stop.
- As speed increases from ‘stopped’, power is smoothly dialed back, so the bike doesn’t run away from you. When cruising at high cadence, very little power is output.
- If during that cruising speed you start going up a hill, the controller senses the combination of reduced speed and cadence and dials in more power to help you get up the hill.
The result is of course not ideal, but it is smooth and easy to adapt to, especially coupled to the other settings options that are available.
So What?
Once again, we see that we are not dealing with a clunky kind of on/off switch that cadence-assist detractors point to. There’s grey area to be had here when it comes to performance.
Torque-Sensing Can Be A Disaster
If you have a physical limitation, torque-sensing doesn’t help you get past it. It does help you go faster while working hard. Studies have shown ebikers in fact can work nearly as hard as, or even harder than bicycle riders… they just don’t realize it. Possibly this is due in part to the exhilaration of being able to go faster, and stay in the saddle for longer periods.
Myself, I am a lifelong cyclist. Or rather, I was. I commuted daily for decades. For many years I eschewed the use of an auto. I commuted and even shopped for groceries by bike (being poor and single had nothing to do with this). But after a couple of heart attacks, my cycling life was over. To stay alive, I gave up the intensely personal activity I most valued. Bummer.
A few years ago, I discovered ebikes, and the one I bought had cadence-based assist. I had no idea there was any other way to do it at the time. I did something many do not: I treated the ebike – which looks like a bicycle but is not one – as a new animal. I threw out much of the knowledge on cycling I had acquired, and started fresh on riding technique.
At the start, pedal effort very quickly led to chest pain and an immediate need to stop doing that. But I could go on if I incremented up the assist and incrementally lowered – but did not erase – pedal effort. This allowed me to keep going (maintain forward progress).
I learned to treat the ebike like an exercise machine. One that went places and was practical transportation. Instead of directly coordinating effort with forward motion, I separated the two. Effort was always maintained, and so was clicking off the needed mileage to my destination. But the two no longer had a 1:1 relationship. This decoupling of effort and speed while maintaining cadence solved everything. The procedure in a nutshell is as follows:
- Set a preferred cadence
- As heart pain occurs (heart pain is not the same thing as getting tired) click up the assist level so effort is reduced – and keep moving. This relieves the heart pain gradually
- On recovery – I’m good after maybe a half block – reduce the assist level back down a click at a time and start working harder again
- All the while maintain the same cadence
- Rinse and repeat as the miles click off to destination arrival.
You are never just sitting in the saddle doing nothing but exercising your thumb on the throttle, unless you’ve really overdone it and have to enforce a complete break. But that should be very rare and only happen in the earliest stages of getting back on the bike and working out as a daily routine.
Again, to belabor the point: I’m using the bike for utility and transportation. My bike has somewhere to go, so the point of cycling is to reach a destination. If I was a recreational cyclist then maybe its fine to slow to a crawl, or stop and sit on a bench for awhile. But a bike as transportation is a different. The point of riding is to get somewhere . So I must maintain forward progress, while managing a constant – but changing – exertion level.
Only cadence sensing is going to let you do that (and I know this from experience. See Afterword below). Its a different riding experience described most simply as a moving exercise machine. Again… not a bicycle.
Different But Still Good For You
Over time and thru repetition, I scaled back the point where pain occurs to where I was able to manage it with gear changes (upshifts) and not changes to the assist level. Now I’m running at top assist speed while maintaining pedal pressure and exertion at all times during the ride. On my Class 3 daily driver I cruise right at 28-30 mph (legal in my jurisdiction) and I get to those higher speeds above the assist limit by myself. All along doing so by maintaining a set, preferred cadence.
And if I overdo it, since I am now running at full power, I can just downshift (maintaining cadence on the easier gear) to take a break. I’ll go a little slower and dip down to the level where the bike starts providing assist again once its speed gets down into Class 3 territory. This is a different way to use cadence-assist. I am not dialing back power: I’m always running at full blast and I’m working to get the bike up to where I am going fast enough so the motor pulls power back on its own due to its going over the Class 3 limit, or adds some in thanks to a downshift and reduced speed.
Note from The Future:
The above describes my experience using an awd geared hub bike with twin KT controllers and a big battery. This one in particular. That was quite some time ago, and its a platform I moved on from long ago. The kind of variations possible in a geared hub system are different than what they would be if I was, say, doing the same thing with a mid-drive-motor solution like a BBSHD. So, two entirely different cadence-based systems, and two different ways to achieve a good ride… but you have to go into it without thinking you know the answers already..
Broadening The Use-Case
Cadence sensing isn’t just for recovering invalids. For the healthy rider, successful use of cadence-based assist as a hard-exercise tool is easily possible, and rooted in that rider not coming into the experience with pre-conceived ideas. Don’t treat it like a bicycle (yes I am repeating this over and over on purpose).
Using the ebike as an exercise machine as you roll down the road, you’ll be getting fit during time otherwise spent sitting in your car and exercising nothing. A torque-sensing ebike can do this too… but if the ebike is meant to also be practical transportation, your physical condition of the moment will have a direct impact on whether you make it to your destination. Not so with cadence assist.
The Future
It took 34 years for the Tour de France to allow bicycles with derailleurs — because not grinding up a slope in the Alps on a single-speed was cheating.
… Isn’t it better to triumph by the strength of your muscles than by the artifice of a derailleur? We are getting soft.
-Henri DesGrange, world-renowned cyclist and original TDF organizer
If someone tried to make that same case today, their opinion would be regarded as fringe idiocy.
So lets take that same interval: 34 years from now, when ebikes have long-since become the accepted norm (just look at the sales figures) as derailleurs did a century ago… will we be espousing technology or methods rooted to the norms of the past? Will a couple of generations of riders who have known nothing else continue to think of torque-sensing assist as giving a bike a ‘normal’ feel?
My money is on ‘no’. Or more accurately… sorta-kinda-no. I think for higher end bikes a dual system could become commonplace, letting riders choose one or the other as they see fit in the moment. One mode for recreation. One for transportation.
If it has to be a this-or-that binary choice, I think torque-sensing won’t survive the test of time. Why? Sheer weight of numbers, and the growth of the automobile replacement market. Look at global ebike sales. Only a small fraction of ebikes are sold in the European and North American markets, where recreational cycling is a thing. Look at the Far East, where bicycles are simply utilitarian transportation and there is no stigma attached to effortless travel. Whats the norm there?
Cadence-based assist.
UPDATE (February 2021):
Its already happening through a vector I hadn’t considered. Recreational ebike riders are starting to upgrade from their cadence-based budget bikes to what the industry tells them was the ideal product: a better bike with higher end components and… torque sensing. I’m seeing reviews from riders not inculcated in traditional cycling ethos, saying the bikes are no longer fun. They can’t just get on a bike and zip around and enjoy the outdoors for as long as the battery holds out… now their bike is making them work at it. What was once an unconsciously-achieved benefit (exercise) is now an enforced requirement. Riders like this, new to the fold, don’t always appreciate the new rules. With the pandemic rushing literally tens of millions of new riders into the fold, the spread of this effect could manifest itself far more quickly than the slow evolution I originally anticipated.
Afterword
Lest I give the wrong impression… I have an ebike that uses torque sensing, and frankly I love it. But its a recreational bike, not suited for a bike that has a job. Going for a fun ride, where I don’t have a problem stopping and sitting down on a bench or a rock for awhile and enjoying my surroundings… Its almost perfect for that. I wish I had time to ride it more.
But by its nature it can’t be a serious transportation tool.