Joining Gear Racks for Long Travel: Why More Sections Means Worse Accuracy

Long-section gear rack, 1400 mm, for multi-section travel

Target keyword: joining gear racks long travel

Secondary: gear rack splicing, rack joint pitch error, long travel rack and pinion

Meta description: Real numbers on cumulative error across spliced gear racks. Three racks of 2 m beat twelve of 0.5 m by more than 3×. Includes joint tolerances and mounting method.

Someone specifying a 6 metre gantry has a choice to make that looks like a logistics decision and is actually an accuracy decision: three 2-metre racks, six 1-metre racks, or twelve 500 mm racks.

Short racks are easier to ship, easier to handle, and often cheaper per metre. They are also, for the same quality grade, roughly three times worse on total positioning error.

The arithmetic is worth walking through, because the conclusion is not intuitive.

Error does not scale linearly with rack length

The first surprise is in the pitch error figures themselves:

Double the length and the total error grows by only about 30%. On a per-metre basis, a 2-metre DIN 6 rack is 0.022 mm/m against 0.034 mm/m for a 1-metre rack - the longer rack is better per unit length.

This is a property of how cumulative pitch error accumulates on a ground rack. The grinding machine indexes along the bar with its own accuracy budget; errors partially cancel rather than adding in a straight line. A rack ground in one setup does not accumulate error proportionally to its length.

Every joint adds its own error

Where you butt two racks together, tooth spacing across the joint is set by how precisely you positioned the second rack, not by how precisely either rack was ground.

Figure 1. Total cumulative error over 6 m of travel. Joint count dominates the result.

Two joining methods are in common use:

Companion rack / joining gauge method: error at rack joints < 0.025 mm

Precision rack assembly fixture: error at rack joints < 0.005 mm

Note the size of that joint error relative to the racks themselves. A 0.025 mm joint error is comparable to the entire pitch error of a 1-metre DIN 5 rack (0.026 mm). One sloppy joint can throw away everything you paid for in rack grade.

The worked example

Worked through for DIN 6 module 4 hardened-and-ground helical racks over 6 metres of travel, with companion-rack joints at < 0.025 mm:

Three 2,000 mm racks:

3 × 0.044 + 2 × 0.025 = 0.182 mm

48 mounting screws, 2 joints, zero dowel pins

Six 1,000 mm racks:

6 × 0.034 + 5 × 0.025 = 0.329 mm

48 mounting screws, 5 joints, zero dowel pins

Twelve 500 mm racks:

12 × 0.026 + 11 × 0.025 = 0.587 mm

48 mounting screws, 11 joints, 24 dowel pins

Same rack grade. Same total length. The twelve-section option is 3.2 times worse than the three-section option.

Break down where the error comes from in the worst case: 12 × 0.026 = 0.312 mm from the racks, and 11 × 0.025 = 0.275 mm from the joints. Nearly half the total error is created at the joints - error that does not exist at all in a two-joint layout.

The dowel pin count in the last row is not a detail either. Twenty-four dowel pins means 24 holes drilled through rack and base together, in situ, on the machine. That is a day of skilled work, and every one of them is a chance to make things worse.

The assembly kit option

Using the tighter joint method (< 0.005 mm per joint), the same three layouts give:

Racks supplied in 1.25 m, 1.5 m and 2 m sections

The more joints you have, the more a better joining method helps - which makes sense, but look at the absolute numbers. Twelve short racks with the best available joint method (0.367 mm) still cannot match three long racks with the ordinary method (0.182 mm).

You cannot fix a bad layout with a better joining tool. Fewer, longer racks wins.

Joint pitch tolerance: always negative

When racks are butted, the pitch spanning the joint should be slightly tight, never loose.

Standard guidance: meshing will be poor if the pitch straddling the connection carries a positive tolerance. Specify a minus tolerance on pitch at the joint.

Their published tolerances:

The asymmetry has a physical reason. A slightly tight joint means the pinion meets a tooth marginally early - it takes up backlash momentarily and keeps rolling. A slightly loose joint means the pinion drops into a wider gap, producing a position jump, a noise, and a load spike as it catches the next flank. One is a minor disturbance; the other is a defect you will hear from across the shop.

Racks supplied with machined ends for joining typically carry a pitch tolerance of −0.05 to −0.4 mm at the end face - the negative tolerance is built into the product.

Mounting, in the order that matters

Reference from the bottom surface. Rack pitch lines are controlled using the bottom surface as the reference datum, together with over-pin measurement of tooth thickness. Machining that bottom surface affects the certified precision. Skimming it to fit a pocket destroys the datum the accuracy was measured against.

CNC hobbing and grinding capacity for long rack production

Do not reduce face width. "To avoid problems of gear precision, do not reduce the face width." Narrowing a rack to fit costs both load capacity and the certified accuracy.

Dowel the racks. "If the racks are not secured properly to the base, they could shift during operation... It is very important to insure firm mounting by the use of dowel pins or similar devices." Screws alone rely on friction; under reversing load, racks creep.

Drill dowel holes in situ. "To attach the racks to the base and drill both simultaneously." Pre-drilled holes in both parts will never line up well enough to act as location features.

Align the rack to the linear guides, not to the machine bed. Rack and pinion backlash depends on the alignment of the rack to the linear guides. The pinion travels on the carriage, so it follows the guides. If the rack is parallel to the bed but the guides are not, centre distance varies over the stroke and so does backlash.

Set backlash at the high point. "The backlash level should be set at the high point of the rack's pitchline, which will set the minimum backlash value and prevent any binding along the entire rack travel length." Find the tightest position on the whole travel, set your minimum backlash there, and everywhere else will be looser but free.

When the error is still too large: map it

Sometimes the specification demands more than the mechanics can deliver. Linear error can be minimised either by using higher quality gearing or by mapping the errors and compensating in the control. In favourable cases the mapped error can be compensated across the full length of the axis.

Most modern CNC controls support linear error compensation tables. You measure actual position against commanded position along the travel with a laser interferometer, then load the correction table into the control.

Two caveats. Compensation corrects repeatable error only - if the rack shifts or the joint moves, the map is wrong. And it works best when the underlying mechanical error is already modest; correcting 0.6 mm of accumulated error with a table is a losing battle against thermal drift and wear.

Practical recommendations

Use the longest racks the machine and logistics allow. Three 2-metre sections beat twelve 500 mm sections by more than 3× on the same grade.

Budget error properly: total = (number of racks × per-rack GTf) + (number of joints × joint error). Both terms matter.

Specify negative pitch tolerance at joints, per the module table.

Use a proper joining gauge or assembly kit - the difference between 0.025 mm and 0.005 mm per joint is significant once you have five or more.

Dowel in situ, drilling rack and base together.

Align the rack to the linear guides.

Consider error mapping as a final refinement, not as a substitute for layout.

FAQ

Why is a 2 m rack more accurate per metre than a 1 m rack?

Cumulative pitch error does not grow linearly with length on a ground rack. Published GTf figures show a 2 m DIN 6 rack at 0.044 mm total, i.e. 0.022 mm/m, versus 0.034 mm/m for a 1 m rack.

Can I mix rack lengths on one axis?

Yes, but every joint adds error. Minimise the joint count first, then work out lengths.

Should the joint pitch be tight or loose?

Tight - always a minus tolerance. A loose joint causes a position jump and a load spike; a tight one is absorbed by backlash.

Do I need dowel pins if I use high-strength bolts?

Yes. Bolts clamp; dowels locate. Under reversing drive loads, a bolted-only rack will eventually creep.

Can error mapping replace a good rack?

Only for repeatable error, and only when the base mechanical error is already reasonable. It does not compensate for a joint that moves.

Tianjin OuNaiDa Transmissions Technology supplies gear racks up to 2 metres per section with machined ends for end-to-end joining, in straight and helical form. Tell us your total travel and accuracy target and we will propose a layout with the error budget worked out.