Nine out of ten emails we get about precision ball screws come down to one line on the drawing, and it's usually the line the buyer is least sure about. Not the diameter. Not the lead. The accuracy grade. Plenty of people asking for C3 don't need it; someone else every month is chasing a tolerance on a C7 screw it was never going to hold. And when two quotes for the same size come back 40% apart, the grade is usually where the difference sits.
What the Grade Number Actually Measures
The pitch of a screw is the distance between adjacent thread crests. The ball screw accuracy grade says how close the actual pitch stays to nominal as the nut travels - how precisely the screw moves a load along its axis. Rolling, grinding, hardening and straightening all leave small errors; the grade says how much is permitted.
The grade is defined over a travel length, normally 300 mm, at 20 °C, with the screw supported at its bearing journals; some catalogues quote 100 mm instead. The number also runs backwards: grade 1 is tighter than grade 5, C3 tighter than C7. ISO 3408, JIS B1192 and DIN 69051 all work that way - lower means more grinding passes, more money, more lead time.
The grade is an input, not an output. It isn't the tolerance on your part, and it doesn't cover backlash. For ±0.01 mm over 300 mm, subtract the coupling, the nut's axial play, the locknut and thermal growth first - then see what's left for the screw.
Precision Ball Screw Grades: C3, C5 and C7 Side by Side
Figures are permitted travel error over 300 mm.
| Grade | Travel error per 300 mm | How it's made | Where it belongs | Price vs C7 |
|---|---|---|---|---|
| C3 | ±0.008 mm | Ground thread, ground preloaded nut | Surface grinders, wire EDM, CMMs, medical and semiconductor machining, precision lathe Z axes | 2.0–2.5× |
| C5 | ±0.018 mm | Precision rolled, or rolled then finish-ground | CNC machining centres and lathes, servo routers cutting aluminium, laser gantries, robot seventh axes | 1.3–1.6× |
| C7 | ±0.050 mm | Rolled, standard production | Woodworking and plastic routers, plasma tables, packaging axes, 3D printers, light pick-and-place | Baseline |
The multiplier assumes the same diameter and lead, so a large C7 still costs more than a small C3. And 300 mm doesn't scale linearly - on a 3 m screw check the full-length figure.
Matching a Grade to a Specific Axis
Give the screw a third of the budget
±0.03 mm of position tolerance over 300 mm? Give the screw ±0.01 mm and you land on C5; at ±0.1 mm, C7 covers it. If the part needs ±0.01 mm overall, you're in C3 territory - and the frame, spindle and shop temperature have to hold that too.
Don't buy accuracy twice on a closed loop
On a machine with a glass scale, the controller reads the table, not the screw. Pitch error stops being a position error; the screw only adds smoothness, stiffness and repeatability. We've had customers order C3 for an axis already carrying a 1 μm scale; the C7 beside it would have run the same.
Long axes and short axes differ
A 300 mm figure is comfortable. A 3 m X axis is not. Steel grows about 11.7 μm per metre per °C, so 3 m of screw in a shop swinging 10 °C morning to afternoon moves roughly 0.35 mm - forty times the C3 allowance. Without temperature control, C3 there is precision the environment erases; C7 plus a scale is the engineering answer.
Short axes go the other way. A 250 mm Z axis never uses the full 300 mm, so the full-length figure and the nut's axial rigidity matter more than the grade. That's where C3 or C5 on a C7 machine makes sense.
Match the job, not the brochure
In precision machining - aerospace parts, medical devices, turbine blades - a few microns of lead error shows up in the finish and the dimensions, and a batch that misses spec is scrap. Robotics is the middle case: a robot arm on a smartphone assembly line placing chips and connectors has to stop in the same place every cycle, so you're buying repeatability, not accuracy. On a conveyor moving boxes from A to B, none of it matters.
The Series Is a Separate Question
Grade and series get mixed up. An SFU series rolled ball screw is the general-purpose workhorse - rolled thread, C7 standard, C5 on request. A DFU series ground ball screw is the one we grind, and it sits behind the C3 and tight C5 grades. The SFE family gets specified where high-speed motion matters, such as high-speed machining centres.
What the Extra Money Actually Buys
Higher accuracy costs more than the screw line suggests: a C3 ships with a ground preloaded nut, tighter housings and better journals, and adds 7–15 days to the lead time because every thread is ground and measured. What you buy is a smaller error budget, steadier friction and less lost motion on reversal.
Three cases where we tell customers to save the money:
- The loop is closed by a scale. Position accuracy comes from the encoder; grade adds little.
- The machine isn't in the same class. A C3 screw in a frame that deflects under cutting load is a C7 with a nicer invoice.
- You can't measure it. If your inspection room works to ±0.02 mm, then ±0.008 mm is unverifiable - and unused.
One caution on preload, which tends to arrive with the higher grades: more stiffness and less backlash, but also more drag and heat. On a long screw run fast by an open-loop stepper, over-preloading is a common cause of stalling.
How to Check the Grade Once the Screws Land
Buying C3 and assuming it arrived isn't a plan. Five checks, in order of usefulness:
- Read the paperwork before the metal. Ask for the lead deviation chart for your lot number, not just a grade stamp - a printed "C3" says nothing about the part in the crate.
- Check the marking. Grade and lot are usually laser-marked on the screw end or barrel. Compare against the certificate; mixed shipments happen.
- Set up like the machine. Support the screw at its bearing journals, at 20 °C, and let it soak. V-blocks in a cold shop invent error.
- Measure over 300 mm in steps. Zero at one end, index the nut forward in 25 mm increments, record the deviation. A laser interferometer is good to about 1 μm; a dial indicator on granite carries 3–5 μm of uncertainty, so it confirms C7 and roughly C5, not C3.
- Check backlash and repeatability separately. Lock the screw, load the nut axially, measure the movement - that's the nut and preload, not the grade, which is why a "bad grade" complaint often turns out to be a loose locknut. Then drive the axis to the same position 30 times and note the spread. Tight repeatability with a steady offset needs controller compensation, not a better screw.
Buy the grade your axis can measure, not the grade that sounds safest. On most routers and light mills, C5 does the job and the money is better spent on preload and a straighter mounting surface.
Tell us the stroke, the load and the positioning error you have to hold through the inquiry page, and we will say which grade we would ship - including when that means selling you the cheaper one.
What Customers Ask Before They Order
Do I really need C3 for a woodworking router?
Almost never. Cutting MDF and hardwood, the error budget is dominated by the gantry, the spindle and the material. A preloaded C7 is what we ship for that job; for aluminium moulds to a tight tolerance we'd talk C5.
Can I mix grades on one machine?
Yes, and often you should. Long X axis on C7 with a scale, Y on C5, short Z on C3 - each axis has its own error budget.
Is C5 a rolled screw or a ground screw?
Depends on the maker, so ask. Rolled threads reach C7 comfortably. C5 can be reached by precision rolling, but on most lines it means finish-grinding after rolling. C3 is always ground.
Will a higher grade fix my position error or my backlash?
No. Backlash lives in the nut, the preload and the locknut; position error often comes from the coupling or thermal drift. Going from C7 to C3 to chase 0.05 mm is an expensive way to learn the coupling was loose.
How much longer does C3 take to deliver?
Usually 7–15 days on top of our standard window for the same size, plus one drawing confirmation.
If you're sizing a new machine or replacing a screw, send the axis length, the load, the speed and the position tolerance - a drawing or a photo of the old screw is fine. We usually answer the same working day, and you can send that through here.
