The drive belt: tension, alignment and what changed

Zero has used a Gates toothed carbon belt on every motorcycle it has built. That continuity is exactly what makes belt tension dangerous to look up: the belt looks like one component with one specification, and it is not. Across the model years in this reference Zero publishes 4 different frequency windows, and three of them apply to belts of the same 11 mm pitch. The window is a property of the motorcycle — its swingarm, its span length, its torque — not of the belt.

The table below is every window, with how many resolved model years each one governs. It is computed from 35 owner manuals rather than asserted: the extractor reads the "Recommended Tension Range" table out of each manual and the join then attaches it to every model year that manual covers, so a window cannot be attributed to a machine whose own manual says something else.

Two things about the shape of this table matter more than the numbers in it. First, the windows overlap in time, not in sequence: in model year 2023 a rider could buy a DS specified at 42.5–73.6 Hz and an SR/F specified at 62–82 Hz from the same dealer on the same day. A figure keyed only to a year is therefore ambiguous even when the year is right. Second, the 62 Hz lower bound arrives twice, on two different platforms, with two different upper bounds two hertz apart — which is a strong hint that the upper bound is doing real work and is not a rounding artefact.

Pitch is the distance between two adjacent belt teeth, and it is the one belt dimension Zero states in the owner manual — printed in the tension table, in the same cell group as the frequency window, in every manual from MY2017 on. There are only two values in 35 owner manuals: 8 mm on the narrow belt, and 11 mm on everything that came after it. Two model years are an exception worth flagging: the MY2015 and MY2016 manuals state no pitch anywhere at all, so the 8 mm (inferred) figure this page shows for them — 10 rows of the resolved table — is carried from the MY2017 manual, whose surviving narrow-belt row is the same 220-tooth belt. It is labelled as an inference everywhere it appears, because an inference must not read as a printed specification.

The archive's specifications table gives the 2017-and-later belt a 14mm pitch. It also gives the SR/F 11 mm, two rows below, for a belt of the same profile — so the table disagrees with itself, and the row that agrees with Zero is the later one. The sibling sprocket page in the same archive prints 11mm for the 2017 wide belt too. Where two archive pages disagree, this is one of the cases where one of them is simply right.

Width is the opposite situation: Zero almost never states it. The 14 mm and 17 mm and 20 mm figures in circulation are community measurements attached to part numbers, not manual specifications, and this page reproduces them only in the archive table below with that provenance attached. The single exception is the newest belt — the Moto X9 on the DSR/X and the 2024-and-later DS and DSR, which Zero describes in the specification table as "Gates Carbon Drive Moto X9 belt, 25 mm wide". That is the only belt width on this page with a manual behind it.

The practical consequence is narrow but sharp. Pitch and tooth count together define the belt; width does not, and a wider belt on the same pitch will not fit a narrower sprocket. If you are ordering, the two numbers to be certain about are the pitch from your own manual and the tooth counts you have physically counted.

Zero has specified belt tension three different ways, and a reader who mixes the methods gets a number that looks like a specification and is not comparable to one. All three appear in manuals that are current somewhere in this reference.

The Tension Tester — a Krikit-style deflection gauge — is the original method and the only one the MY2015 and MY2016 manuals give: press the gauge into the non-cogged side of the belt at the midpoint of the span until it clicks, and read the load off the arm. Those manuals specify 20 kg to 30 kg and print no frequency at all. The frequency method arrives at MY2017, alongside the Gates Carbon Drive smartphone application, and from then on every manual prints both a load window and a frequency window for the same belt.

The 2026 light line abandons both. The LS1 and XB are measured with a Gates 508C ultrasonic tension meter, in newtons, with the meter's mass, width and span parameters given in the manual: the LS1 wants 500 to 800 N with the meter set to 4.6 g/m, 23 mm and 356 mm, and the XB wants 550 ± 50 N with its own set. Neither figure is a frequency and neither converts to one without the belt’s mass per metre, which is why the meter parameters are part of the specification rather than a footnote to it.

The steps below are the frequency procedure as the Gen3 service manuals give it, because the two instructions in it that riders most often skip are the two the manual is most emphatic about: rotate forwards, and rotate twice.

Belt alignment is the part of this job that most rewards being done properly and is least well documented in the community record. A belt running against a sprocket flange wears asymmetrically and fails early, and the failure looks like a belt problem rather than an alignment problem, which is how it gets misdiagnosed.

The archived procedure tells you to set alignment by making the marks on either side of the swingarm agree. On the Gen3 bikes that is explicitly the wrong method, and Zero says so in the same paragraph of three separate service manuals: the marks stamped on the swingarm are for initial alignment, they are not equal on the two sides of the swingarm, and they must not be used as final adjustment marks. The archived page also gets to the right answer two sentences later, in its own prose, when it says to roll the wheel forward and track the belt flush against the inside of the rear sprocket — that tracking check is the real one.

The manual figure for tracking is a 1–2 mm gap between the drive belt and the belt-side face of the rear sprocket, confirmed after the axle nut is torqued and after two more full belt revolutions. That is a one-sided gap against a named face, not a belt centred with a millimetre of clearance either side, and the difference matters because it tells you which way to move the wheel when you are out of tolerance. The archive’s own note that torquing the axle nut nudges the belt about a millimetre outward is consistent with Zero putting the confirmation step after the torque step rather than before it.

The hardware changed with the platform too, which is why the archived tool list reads oddly on a modern bike. On the S/DS and FX platforms the adjusters and their jam nuts are both 13 mm, the axle nut is 27 mm at 102 Nm, and the service manuals specify Loctite 242 on the axle threads and a jam nut torque the archived procedure omits entirely — 26 Nm on the S/DS platform and 20 Nm on the FX platform, which are not the same figure. On the Gen3 bikes there are no jam nuts: the belt tension adjustment bolts are T45 and are torqued to 5 Nm at the end of the job.

The rear sprocket has been 90 teeth on every 11 mm-belt Zero and either 130 or 132 on the narrow-belt machines. The front pulley is where the variation lives, and it is the part a rider is most likely to order wrongly, because the number moved several times without the belt profile changing at all.

Three moves are worth naming. The FX family went from 90T/20T in MY2017 to 90T/18T in MY2018 and has stayed there, changing the final drive ratio from 4.50:1 to 5.00:1. The DSR/X arrived in MY2023 with 90T/22T, and the DS and DSR joined it on the ADV platform in MY2024 — the DSR at 90T/22T and the DS, on the same belt, at 90T/20T. And on the S and DS the count is pack-dependent rather than year-dependent: 6 specification sheets in this index print it as a pair rather than a single figure.

That pack-dependence is the trap a year-range table cannot express. A 2019 Zero S is not one specification: the ZF7.2 has an 18-tooth front pulley and the ZF14.4 has a 20-tooth one, in the same manual, on facing pages. Neither is wrong and neither is the default. The only reliable way to know which one you have is to count.

The archive’s section on riding with stripped teeth reasons from a 28-tooth front pulley of which about half the teeth are in mesh, and concludes that a continuous gap of perhaps a dozen missing teeth "might" be survivable at low torque. The arithmetic describes a 2012–2016 machine and nothing built since: from 2017 the front pulley is 20 teeth, 18 on the FX family and on the small-pack S and DS, and 22 on the DSR and DSR/X. Fewer teeth carry the load, each one carries more of it, and a bare stretch of belt is proportionally more likely to slip. Zero’s own guidance is not a tolerance at all: replace a belt showing exposed tensile cords or cracking at the base of the teeth, and replace a belt that has ratcheted before the next time you ride.

The archived tables stop at 2020. Everything Zero has done to the final drive since then is therefore absent, and the gap is not cosmetic: it is a different belt, a different pulley and a different tension window. The DSR/X from MY2023 and the DS and DSR from MY2024 run a Gates Carbon Drive Moto X9 belt, which Zero's specification tables describe as 25 mm wide — the only belt width Zero states anywhere in an owner manual.

Their tension window is 62 Hz to 84 Hz, two hertz wider at the top than the SR/F's, and their tester specification is given as a centre and a tolerance rather than a range: 77 ± 20 kg, printed in the manual as 170 ± 44 lb. That format is worth noticing, because a centre-and-tolerance specification tells you where Zero wants the belt, and a window does not.

The pulley pairing is 90T/22T on the DSR/X and the DSR, and 90T/20T on the DS built on the same platform — so even inside one manual, on one belt, the front pulley is not a constant. A rider who takes "2017+" from the archive and orders for a 2026 DSR gets a belt of the wrong width and a pulley of the wrong count, and would tension it 19.5 Hz slack at the low end.

This page corrects a community archive, which obliges it to be straight about our own record first. Until this week the maintenance sheet this site extracted from the service manuals carried one belt tension figure per model with no year dimension at all, and for four models that single figure was the newest one applied retroactively to every machine that ever wore the name.

The DS row was the clearest case. It read 62–84 Hz, which is the MY2024-and-later ADV figure, and it was attached to every DS in our data including the machines through MY2023 whose own manuals specify 42.5–73.6 Hz. The S, SR and DSR rows had the same shape. All four have now been corrected, as a verified overlay in `data/manuals/spec-corrections.json` re-applied by `backend/scripts/apply_spec_corrections.py`, with the CI job failing if a regeneration of the LLM-extracted sheet loses one — because a hand edit to that file would have been silently undone by the next extraction.

The schema is still the root cause, and the correction works around it rather than fixing it. Our maintenance record is keyed by model alone, so the year split now lives inside the string as prose: "MY2014–2023 (SDS platform): 42.5–73.6 Hz. MY2024 onwards (FST ADV platform): 62–84 Hz." That is readable and it is not queryable, and nothing stops the next figure from being written without its years.

The correction also still overreaches at the early end, which is worth stating plainly because this page is the reason anyone would look. The corrected strings begin their frequency range at MY2014. The MY2015 and MY2016 manuals publish no frequency at all — they specify a Tension Tester load of 20 kg to 30 kg on the narrow belt, and the frequency method does not appear in a Zero owner manual until MY2017. So for three model years the corrected figure is still a specification those manuals do not contain.

What this page does is refuse to read from that sheet either way. Every tension figure here comes from the manual text directly, keyed to the manual’s own model years, through the generator named at the foot of the page.

A Zero belt makes noises, and the community record about them is better than it looks while being attached to the wrong explanation. The archive documents a pronounced groan in the wet that goes away when the bike dries, and recommends a light application of silicone spray — with a sound warning about petroleum distillates attacking the belt — while attributing the noise to a hypothesis about the motor’s position encoder losing accuracy in high humidity.

Zero’s FX-platform service manual explains the same noise mechanically and without the hypothesis: the belt makes noise when it is forced to run hard against the flanges on the sprockets, which happens because of sprocket misalignment, wheel misalignment or runout from a damaged component, and the permanent solution is to correct the alignment or replace the damaged part. It also says that hosing the belt and sprockets off after a ride usually ends the noise on its own, and that if it does not, the remedy is to spray the belt with water, let it dry, and apply a thin coating of dry silicone spray to the tooth side of the belt — dry silicone dries instantly, sheds grit, and does not affect belt longevity.

So the archive’s remedy is endorsed by the manufacturer, with one correction and one caveat. The correction: apply it to the tooth side of the belt, not to the motor shaft behind the front pulley. The caveat: the manual treats it as a symptom treatment, and puts alignment first. A belt that needs silicone to be quiet is telling you something about its tracking.

The awkward part is that the same service manual, two pages earlier, states that "drive belt and sprockets do not need lubrication of any kind" and that cleaning should use water and a soft brush with no detergent. Both sentences are Zero’s, in one document, about one belt. This page does not pick a winner: the honest reading is that no lubricant is required for the belt to work, and that dry silicone is sanctioned specifically as a noise remedy. Anything wet, oily or petroleum-based is ruled out by both sentences at once.

Belt tension is on Zero’s maintenance schedule four times over, and the belt itself has a hard replacement interval rather than a condition-based one. The schedule below is from the Gen3 service manual; the FX-platform manual states the same replacement mileage in prose, as "the belt must be replaced at 24,000 mi (37,000 km)".

The two early checks are the ones riders skip and the ones that matter most. A new belt beds in, and the first hundred miles is when the tension a dealer set will have moved most. After that the interval is long enough that a rider who checks only at service will go a year between measurements — which is fine, and is also why the every-ride visual inspection is on the list separately.

The scope is the drive belt as Zero specifies it in the owner and service manuals for model years 2015 and later: pitch, tension, tooth counts, alignment method and the maintenance interval. Five things are deliberately outside it.