Regen on a Zero: how it works, what it gives back, how to use it

The Gen2 dash has a REGEN meter which, in the manual’s words, “indicates the amount of TORQUE on the rear wheel and the POWER (energy) feeding back into the battery” during regenerative braking [1, p. 29]. The SR/F splits it in two. Its torque meter shows the torque at the rear wheel “while accelerating (output) or decelerating (regen)” [3, p. 36], and its power meter shows the rate at which energy is “feeding back into the power pack” [3, p. 37].

On the SR/F a traction function also acts on regen. When the wheel speed sensors report the rear wheel slipping as the bike slows, drag torque control raises or lowers the regenerated power [3, p. 95].

A regen event lasts 3 to 10 s, so a reading once a minute lands on about one in ten. To see what that does to a total, the regen energy of the same rides was added up twice: from every Bluetooth reading, and from one reading a minute, with the minute starting at each of its 60 seconds in turn [4].

Read once a minute, a bike’s regen total came out anywhere from about a quarter to about twice the real figure. Over many rides the log can give a rough total. It cannot say which slowdown the regen came from, and with no brake bit it cannot split the lever from the throttle.

The lines below are the first 11 minutes of the descent in the worked example, from both logs. What to look for:

On the four bikes with coast regen at 0%, 83 to 97% of the regen energy came with the lever on. On the three with coast regen at 50 to 66%, 19 to 52% did. Without the lever, the bikes with coast regen set above 0 showed regen in 11 to 14% of their readings; those at 0 in 0.4 to 5.9%.

The current is not always negative with the lever on, because regen fades at low speed. No lever reading under 15 km/h showed regen (680 readings). Between 15 and 30 km/h, 29% of the lever readings showed none, and above 30 km/h, 2%.

Regen power at the battery grows with speed (table). Divided by the speed, it gives a force at the wheel that stays nearly the same at every speed: 200 to 350 N with the lever. On a bike with its rider that slows you at about 0.8 to 1.6 m/s², 0.08 to 0.16 g. The power is measured; the force and the deceleration are worked out from it [4].

The distance to stop at a steady deceleration is speed² ÷ (2 × deceleration). From 50 km/h on regen alone that is about 95 m at 1 m/s², 120 m at 0.8 m/s² and 60 m at 1.6 m/s². Any firmer stop needs the friction brakes for the difference. Below 15 km/h there is no regen at all, so the last metres of every stop are friction: from 50 km/h that part holds 9% of the motion energy, from 30 km/h 25%.

A slowdown here is the speed falling by 8 km/h or more with the throttle off; 652 were recorded. The energy it has to lose is its motion energy plus the height it drops (from the phone’s barometer, or GPS altitude where the phone has none). That energy goes three ways. The air and the tyres take some while the bike still covers ground, which is the best use of it. Regen puts some into the pack, read at the battery and divided by 0.85 to count it at the wheel. The friction brakes take the rest [4].

Two of the three parts are worked out. Air and tyres use a drag area of 0.40 m² (0.55 m² in brackets), a rolling coefficient of 0.012 and an 85 kg rider. The brakes are what is left, so they move with the mass, the drag area and the 0.85 efficiency.

Slowdowns that do not end in a stop were done almost entirely with regen. Stops left about a fifth of their energy in the brakes. Descents left the most: riders used the lever more downhill, and regen could not hold the bike on its own.

Above 60 km/h the same arithmetic gives the brakes a negative share (−11% at 0.40 m², −26% at 0.55 m²). There the air takes most of the energy and the guess at the drag decides the answer, so this page gives no split for faster slowdowns until each bike’s drag is measured.

Each stop and slowdown was matched to OpenStreetMap at the point where it ended [6]. 99 of the 164 stops, 60%, were at traffic lights, with a median wait of 17 s.

Stops at a traffic light left 17% of their energy in the brakes. Stops where nothing was mapped (queues, unsignalled junctions, parking) left 43%. The data has no other vehicles in it, so traffic itself is not visible; the stops with nothing mapped are the closest it gets. OpenStreetMap coverage also varies: 45 of the 164 stops matched nothing.

A long descent with regen at its maximum suggests heat. The log does not show it: no temperature warning, and every reading far from a limit. It does not say why the main board restarted either. Nothing in it rules in or out the board’s 12 V supply, a connector, or the board itself. The cause is not established.

Across all the bikes whose logs we hold, unexpected main-board restarts while moving are not tied to regen. There were 50 on 9 bikes (no Key Off in the 120 s before, last Riding record above 10 km/h): 6 in regen, 12%, and 34 with the pack driving the motor. Regen is 14.7% of moving Riding records, so a restart was no more likely during regen than at any other moment of riding [4].

To read a log like this one, treat a restart as one sequence across both logs and take its times from the BMS. On a descent, add up the current: the SoC can rise far faster than the current explains, and it is worked out again after each restart.