I think the thing with cadence sensors and rear-drive vs. torque sensors and mid-drive bikes in general is that more than the speed of response -- which is a totally solvable problem for cadence sensors -- is the nature of the response. The problem people have with cadence sensor bikes is that they simply put out power that is controlled by the PAS level and exactly nothing else. When you are actually riding a normal bike, you are going to go slower some parts, faster some parts, want to lay down more power here, less power there, and everything in between. How hard you push and how fast you turn the cranks are both variables that play into that. Both of which could be results of things like hill grade, what gear you're in, etc., but also things like not wanting to suddenly jolt up to 12 mph when you're trying to cross the street as the light turns green. Torque sensors get a lot of the envy because the controllers that use them at least generally scale the power appropriately based on the input, which by its nature will feel more natural and more like a regular bike.
In theory, you could make some sort of all-inclusive array of sensors to make something that feels as natural as possible and provides power as you desire based on your settings or some other controls. In practice, of course, they do try to make things simple for themselves as manufacturers. Most all manufacturers are just assembling parts from parts makers and just defining specs -- they're not doing any controller programming or setting up power ramp curves or anything like that. It's all just plug-and-pray for them. The simplest setup I can think of that would probably solve everybody's problems would be a torque sensor on the crank and a wheel speed sensor. Wheel speed is a better measurement than crank cadence because it is affected by gear ratio along with a throttle that can act as a "boost" beyond that limits of the standard power when you are actively pedaling.
Kind of quick back of the napkin calculations make me think of a power response curve based on wheel speed (s) and crank torque (t) that looks something like this. For the sake of easy calculations, I just normalized the variables to a 0..1 representative range, but you can just scale this according to a given PAS level or units of measurement. In fairness, I haven't taken into account the act of coasting, which would basically be this times a min-torque cutoff. This took me 5 minutes to whip out, so I seriously doubt it's beyond the limits of engineering capabilities to come up with something more thoughtful.
Now in fairness, this is an idealized hypothetical, and having written software for electric cars in the past, I know how noisy and unreliable sensor feeds can get, so you ultimately need to do a bit of noise reduction and smoothing of the input time-series data. Also, the relative ratios of how to weight speed-related and torque-related power response are highly dependent on the characteristics of the motor and drivetrain, so you'll need to adjust based on things like the cassette and wheel/tire size, so I'm greatly oversimplifying here. Someone who actually works more closely with ebikes and ebike controllers can think about all those variables a lot better than I am in this simple posting.
In theory, you could make some sort of all-inclusive array of sensors to make something that feels as natural as possible and provides power as you desire based on your settings or some other controls. In practice, of course, they do try to make things simple for themselves as manufacturers. Most all manufacturers are just assembling parts from parts makers and just defining specs -- they're not doing any controller programming or setting up power ramp curves or anything like that. It's all just plug-and-pray for them. The simplest setup I can think of that would probably solve everybody's problems would be a torque sensor on the crank and a wheel speed sensor. Wheel speed is a better measurement than crank cadence because it is affected by gear ratio along with a throttle that can act as a "boost" beyond that limits of the standard power when you are actively pedaling.
Kind of quick back of the napkin calculations make me think of a power response curve based on wheel speed (s) and crank torque (t) that looks something like this. For the sake of easy calculations, I just normalized the variables to a 0..1 representative range, but you can just scale this according to a given PAS level or units of measurement. In fairness, I haven't taken into account the act of coasting, which would basically be this times a min-torque cutoff. This took me 5 minutes to whip out, so I seriously doubt it's beyond the limits of engineering capabilities to come up with something more thoughtful.
Now in fairness, this is an idealized hypothetical, and having written software for electric cars in the past, I know how noisy and unreliable sensor feeds can get, so you ultimately need to do a bit of noise reduction and smoothing of the input time-series data. Also, the relative ratios of how to weight speed-related and torque-related power response are highly dependent on the characteristics of the motor and drivetrain, so you'll need to adjust based on things like the cassette and wheel/tire size, so I'm greatly oversimplifying here. Someone who actually works more closely with ebikes and ebike controllers can think about all those variables a lot better than I am in this simple posting.
