How-to
How to Choose Servo Gearing for a Ballscrew Router Axis
Short answer
Run 3:1 to 4:1 belt reduction on the X and Y axes and 2:1 on Z. The ratio comes from required rapid speed and the continuous torque rating, not the 190 oz-in stall figure; above about 4:1 a single timing-belt stage needs pulleys that do not fit a router.
On this page
Gear the X and Y axes of a router like this at 3:1 to 4:1 with timing belts, and the Z at 2:1. Those ratios come from two numbers — the rapid speed the axis needs and the continuous torque the motor can hold — not from the 190 oz-in static torque on the data sheet, which is a near-stall peak you cannot hold. With a one-inch-lead screw and a 4504 RPM no-load motor, 4:1 still leaves a 1126 in/min ideal rapid and multiplies the continuous rating to 96 oz-in at the screw. Above about 4:1 a single belt stage needs pulleys that will not fit the machine; 10:1 means two belt stages or a worm gearbox, and nothing on this axis calls for that.
What the motor numbers actually buy you #
Four figures do the work: 4504 RPM no-load speed, 190 oz-in static (stall) torque, the maximum rated torque, and 3.4 A continuous against 6.7 A peak stall current. Each one means something narrower than it looks.
No-load speed is the speed at which the motor makes essentially no torque. Stall torque is the torque at zero speed, drawn at peak current. Between those points a brushed DC servo's torque falls close to linearly, so a usable first-order model is T ≈ 190 × (1 − N ÷ 4504) oz-in at N RPM — a calculated estimate, not a measured curve.
The number to size against is the maximum rated torque. Continuous ratings follow the S1 duty idea in IEC 60034-1: the torque the machine can carry for unlimited time at its stated cooling. The stall figure has no such rating behind it, and 6.7 A is about double the continuous current, so sitting near stall overheats the windings. One caveat: the rating was published for the 30 V winding while this motor runs on a 42 V bus, so treat it as the rating clue and confirm it with a temperature check after the first long cuts.
This is the same loaded-speed margin you apply when checking servo speed margin for a rack and pinion axis; a ballscrew only changes the lead term.
The two numbers that pick the ratio #
Speed sets the floor and torque sets the ceiling:
- Required rapid gives screw speed: screw RPM = rapid ÷ lead, then motor RPM = screw RPM × ratio. The motor must not be asked to run near its no-load speed, where little torque remains; the linear model should still show a few times the continuous rating at that speed.
- Required thrust comes from the screw converting torque to force: F = 2πTη ÷ L, with T in lb·in, L the lead in inches per revolution, and η the screw's efficiency. Use 0.90 as a ballscrew rule of thumb and the maker's figure when you have it. Whatever is left after friction and cutting force accelerates mass, F = ma — the same pass as sizing torque and reduction for a robot arm joint, with the screw lead standing in for the arm length.
Choosing the reduction, step by step #
- Write down the rapid you will actually use. On a router with a 2.5 hp spindle and 40 in of X travel, 300 in/min is a sensible target; more buys nothing in cutting.
- Screw speed = rapid ÷ lead. For the 300 in/min target on these screws, that is 300 RPM.
- Pick a trial ratio and compute motor speed = screw RPM × ratio. Check available torque from the linear model at that speed; it should beat the continuous rating by a factor of two or more.
- Screw torque = ratio × the continuous torque. Convert to thrust with F = 2πTη ÷ L.
- Reserve about 10 lbf of thrust for friction and cutting force, then divide the remainder by 0.2 g to get the moving weight the ratio covers at that acceleration. Weigh the gantry; if it is heavier, step the ratio up.
- Check belt geometry: one stage up to 4:1 with a sane driver pulley; above that, plan two stages.
- After the build, verify: commanded versus measured rapid, drive current during acceleration against 3.4 A, and motor temperature after a cutting cycle.
Worked example: the X axis at a 300 in/min rapid #
Inputs, with sources: screw lead 1 in/rev (the machine's X and Y ballscrews); rapid target 300 in/min (chosen in step 1); motor 4504 RPM no-load, 190 oz-in stall, 24 oz-in continuous, 3.4 A continuous current (motor data); screw efficiency η = 0.90 (rule of thumb); acceleration target 0.2 g (design choice).
Rule in words: the belt divides screw speed by the ratio and multiplies screw torque by it, and the screw turns torque into axial force through its lead.
Arithmetic at 4:1:
- Screw speed: 300 in/min ÷ lead = 300 RPM; motor speed 300 × 4 = 1200 RPM.
- Torque available at 1200 RPM: 190 × (1 − 1200 ÷ 4504) = 190 × 0.734 = 139 oz-in, about 5.8 times the continuous rating, so the margin is fine.
- Screw torque: 4 × the continuous rating = 96 oz-in = 6.0 lb·in.
- Thrust: F = 2π × 6.0 × 0.90 ÷ L = 33.9 ≈ 34 lbf.
- Moving weight covered at 0.2 g: since a = F ÷ m and m = W ÷ g, W = F ÷ 0.2 = 34 ÷ 0.2 = 170 lb.
At 3:1 the same pass gives 900 RPM at the motor, 152 oz-in available, 72 oz-in at the screw, 25 lbf of thrust, and 127 lb of moving weight at 0.2 g.
| Belt ratio | Ideal rapid (in/min) | Motor speed at 300 in/min | Screw torque (oz-in) | Continuous thrust (lbf) | Weight covered at 0.2 g (lb) |
|---|---|---|---|---|---|
| 1:1 | 4504 | 300 | 24 | 8.5 | 42 |
| 2:1 | 2252 | 600 | 48 | 17 | 85 |
| 3:1 | 1501 | 900 | 72 | 25 | 127 |
| 4:1 | 1126 | 1200 | 96 | 34 | 170 |
Every figure is a calculated example for these inputs — ideal belt, 0.90 screw efficiency, continuous torque only — not a measured result.
Decision: weigh the gantry stack. Under about 125 lb, run 3:1 and keep the 1501 in/min ideal rapid; heavier than that, or when you want cutting-force margin, run 4:1, which covers 170 lb. Both hold motor speed in the fat part of the torque curve at a 300 in/min rapid, and neither gives up a speed you will miss.
Why 4:1 is the ceiling for one belt stage #
The ratio is driven teeth ÷ driver teeth. A 10:1 single stage means something like 10 teeth driving 100: about five teeth in mesh on the driver at 180 degrees of wrap, and a driven pulley ten times the driver's diameter (diameter = teeth × pitch ÷ π). That is a wheel, not a pulley, sitting beside this screw. If you truly needed 10:1, two stages around 3:1 each, or a worm gearbox, would do it — but the worked example shows this axis does not, because 4:1 already covers a 170 lb gantry.
Two belt facts bite during layout. Keeping the total tooth count constant does not keep belt length constant when the pulley diameters change — length depends on both pitch diameters and the center distance, so slot the motor mount or compute the center distance for the belt you choose. And set tension to the belt maker's figure for the pitch and width; common design practice holds the slack-side tension near 10 to 30 percent of the effective tension, which is the regime the tooth engagement on the driver assumes.
Once the ratio and belt pitch are fixed, picking the timing pulleys — the tooth counts and widths that realize that ratio — is a straight catalog match to the belt you standardize on.
The Z axis wants 2:1, not 6:1 #
Matching the ratio to the 6 TPI lead feels tidy, but nothing couples them: the lead already divides speed and multiplies force by six per screw revolution, and the belt ratio just stacks on top. Size the Z from its own two numbers. At 2:1 the screw turns 4504 ÷ 2 = 2252 RPM, which on a 1/6 in/rev lead is a 375 in/min ideal rapid, far past anything a Z needs. Gravity, not acceleration, sizes this axis: holding a 15 lb router takes T = F × L ÷ (2πη) = 15 × (1/6) ÷ (2π × 0.90) = 0.44 lb·in, about 7 oz-in at the screw and 3.5 oz-in at the motor — a seventh of the continuous rating.
Before you pull a belt, pulley, or coupling on this axis, isolate the drive and block the carriage: a ballscrew back-drives, and the spindle's weight will drop the Z the moment the motor lets go.
Decision: 2:1 on Z. A 3:1 only slows the rapid to 250 in/min to buy thrust the gravity load does not need.
Verifying the choice on the machine #
Command a 300 in/min rapid and time the traverse over the full 40 in of travel: measured speed close to commanded, with current settling under 3.4 A after the acceleration spike, confirms the ratio. Current may touch the 6.7 A peak during acceleration; it must not live there. After twenty minutes of cutting cycles, put a hand on the motor case — uncomfortably hot means drop the acceleration or step the ratio up one.
If the drive closes its position loop on the motor encoder, belt backlash between motor and screw reads as following error at the tool; ballscrew encoder feedback working with backlash-free belts covers keeping that loop honest. Re-check pulley alignment and tension after the first few hours, while the belt seats.
Questions asked
Why not bolt the motor straight to the screw at 1:1?
Speed is the one thing you have in surplus: 1:1 gives a 4504 in/min ideal rapid on a one-inch-lead screw. What you lose is thrust, because the continuous torque through that screw is about 8.5 lbf, so acceleration and cutting load lean on peak torque and heat the motor. A 3:1 or 4:1 belt trades speed you will never use for roughly 25 to 34 lbf of continuous thrust.
Should the Z axis ratio match its 6 TPI screw?
No. Nothing couples ratio to lead. A 2:1 belt on a 6 TPI screw still gives a 375 in/min ideal rapid, and holding a 15 lb router against gravity takes about 7 oz-in at the screw, under 4 oz-in at the motor. Choose the Z ratio from its speed and gravity load, the same way as X and Y.
How much good is the 190 oz-in stall torque?
Only at and near zero speed, and only briefly. On a brushed DC servo the torque falls roughly linearly to zero at the 4504 RPM no-load point, and the 6.7 A stall current is about double the 3.4 A continuous rating, so sitting near stall overheats the windings. Size on the continuous rating and treat stall as a momentary allowance for acceleration.
Is the 1024 PPR encoder enough resolution at these ratios?
Yes, for routing. If the drive counts all four edges, a 1024 pulse-per-revolution disc gives 4096 counts per motor revolution; at 3:1 that is 12,288 counts per screw revolution, or about 0.00008 in per count on this screw. Confirm the drive's decoding mode before relying on the number.