CNC Gear Grinding Machine: How High Precision Is Actually Made
A CNC gear grinding machine removes a few hundredths of a millimeter from a hardened tooth flank to fix profile, lead and pitch error. This page is for engineers and buyers who need to know which gears go on a grinder, which do not, and what the numbers really mean.

In this article
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
Key takeaways
What a CNC gear grinding machine actually removes
A gear grinder does not make a gear. It corrects one. By the time a part reaches the machine, the teeth have been hobbed, shaped or milled to within 0.05–0.10 mm of nominal. Grinding then takes 0.02–0.15 mm off each flank, depending on how much distortion came out of the heat treat furnace.
The cutting action is thousands of small abrasive scratches. A vitrified bonded wheel turns at 30–80 m/s and each abrasive grain acts as a tiny cutting edge. Because the grains are bonded and not a single point, the load per edge stays low. That is why a grinder can cut 60 HRC steel without the chatter a milling cutter would produce.
What the machine really controls is the position of that wheel relative to the tooth flank. Profile error, lead error, pitch error and runout are all corrected by moving the wheel along a controlled path while the work head indexes the gear. The CNC is doing kinematics, not brute force.
Heat is the limit. Roughly 80–90% of the grinding energy turns into heat at the contact zone. Without flood coolant, a 0.05 mm depth of cut can raise the surface layer above its tempering temperature and burn the flank. Coolant pressure and nozzle aim matter more than most operators expect.
- 1Stock removal per pass0.02–0.05 mm rough, 0.005–0.02 mm finish
- 2Wheel surface speed30–80 m/s, higher for small modules
- 3Heat splitMost grinding energy leaves as heat, not chips
Form grinding versus generating grinding
Form grinding uses a wheel dressed to the exact tooth space profile. The wheel plunges into the gap and grinds both flanks at once. Setup is fast, the machine is simpler, and it handles internal gears and shoulders that a generating wheel cannot reach. The trade-off is the dresser. A profile dresser has to reproduce the flank every few parts, and a worn diamond shows up directly on the tooth.
Generating grinding uses a disc or threaded wheel that meshes with the gear as it feeds. The tooth form comes from the rolling motion, not from the wheel shape. That makes it fast for spur and helical gears in volume, and it self-corrects small wheel wear. It also gives more freedom for profile crowning and lead correction, because the CNC can add motion on top of the pure generating roll.
Pick by geometry first. Module below 1 mm, internal teeth, or a flank that runs into a shoulder points to form grinding. External spur and helical gears from module 1 to 10 mm in batches points to generating.
Both methods leave a visible pattern on the flank. Form grinding tends to leave axial lines, generating grinding leaves a diagonal crosshatch. The pattern is a fingerprint of the setup, so keep a first-off part for comparison.
- 1Form grindingWheel shape defines the tooth; good for internal gears and tight shoulders
- 2Generating grindingRolling motion defines the tooth; faster for external gears in volume
- 3First-off partKeep it. The flank pattern is the fastest setup check you have
Wheel choice and dressing control the flank
The abrasive does the cutting, but the bond holds it in place and the grade decides when grains release. A wheel that is too hard glazes: the grains blunt and rub instead of cut, and the flank burns. A wheel that is too soft sheds grains early, loses its form and the tooth goes out of profile. On hardened alloy steel, most shops land between an 80 and 120 grit alumina wheel with a medium grade.
Dressing is where the process is set. A single-point diamond traverses across the wheel to restore concentricity and open the pores. If the dress is too fine, the wheel cuts cool but loads up. If it is too coarse, the wheel cuts fast but leaves a rougher flank. Depth of dress of 0.01–0.03 mm per pass is a common starting range.
Superabrasive wheels change the trade. Cubic boron nitride holds its form far longer, so a generating grinder can run hundreds of parts between dresses. The cost per wheel is high, which is why it suits production runs, not one-off repair work.
Coolant is part of the wheel system. Straight oil gives better lubrication and lower burn risk but needs fire protection. Water-based coolant is cheaper and easier to handle but evaporates at the contact zone and can leave a softer finish on some steels.
- 1GlazingWheel too hard or dress too fine; flank burns and finish drops
- 2SheddingWheel too soft; profile drifts and size wanders
- 3CBNLong wheel life for volume, high entry cost for one-offs
Where the process stops working
Grinding cannot fix a bad blank. If the gear was cut out of position, the grinder will chase the error and remove uneven stock. One flank ends up with 0.02 mm and the other with 0.15 mm, and the case depth on the thin side drops. Past a certain point the tooth is soft on one side and the part is scrap.
Case depth sets the ceiling. A carburized gear with 0.6–0.9 mm case can lose 0.10 mm of stock and still hold a usable hardened layer. A nitrided gear with a 0.2 mm case cannot. That is the single most common reason a ground gear fails a hardness check after finishing.
Grinding burn is the other hard stop. It shows as a light or dark patch on the flank, and it means the surface layer has been tempered or re-hardened. Acid etch or nital etch inspection reveals it. A burned flank loses its residual compressive stress, and a gear that should have run 10,000 hours starts pitting early.
Very small modules are difficult too. Below module 0.5 mm the wheel tip gets thin, the coolant cannot reach the root, and the risk of burning the whole tooth outweighs the accuracy gained.
- 1Uneven stockIf the pre-grind cut is off, one flank goes soft
- 2Case depthRoughly 0.6 mm case minimum if you plan to remove 0.10 mm
- 3Burn checkNital etch on a sample, not on the whole batch
How the result is proven
A ground gear is judged by numbers, not by feel. Profile total deviation, lead total deviation, pitch deviation and runout are measured on a gear measuring center against the drawing. The values are reported in micrometers, and they are what the customer's assembly actually cares about.
Tooth thickness is the other number. It is usually measured over pins or balls, because the caliper reading over teeth depends on where it lands and repeats poorly. For a 2 mm module gear, a change of 0.01 mm in tooth thickness moves the over-pin dimension by roughly 0.014 mm.
Surface finish on the flank is measured as Ra, usually 0.2–0.8 μm after grinding. A ground flank that reads Ra 0.1 μm is not automatically better; too smooth a surface can hold less oil and run hotter in a gearbox.
Documentation matters as much as the measurement. A shop that reports profile and lead values per tooth, plus a hardness check on a sample, gives you something to compare against the next batch. A single size reading does not.
- 1Profile and leadReported per flank, in micrometers
- 2Tooth thicknessMeasured over pins or balls, not over teeth
- 3Flank finishRa 0.2–0.8 μm is the normal target band
Where grinding sits in our shop
GreatLight runs 127 high-precision CNC machines across three plants, including 16 simultaneous 5-axis machining centers and a maximum processing size of 4,000 mm. Gear blanks, shafts and housings that feed a grinder are turned and milled here to ±0.005 mm, so the stock left for grinding is even and predictable.
Even stock is not a small thing. If the pre-grind blank varies by 0.05 mm around the tooth, the grinder has to take a heavier cut on one side, and that is where burn and soft flanks come from. Holding the blank tight removes that variable before the part ever reaches the wheel.
We work to ISO 9001:2015, IATF 16949:2016, ISO 13485:2016 and ISO 27001:2022, and we inspect 100% of parts before shipment. For gear work that means material certificates, in-process checks on tooth thickness, and a final report on request.
Materials we see most in gear work include 4140, 4340, 17-4PH and case-hardening grades, plus aluminium 7075 and 6082 for lightly loaded timing components. Uploads are secure and confidential, and a non-disclosure agreement is available on request.
- 1Pre-grind tolerance±0.005 mm keeps stock even on both flanks
- 2Batch rangeNo minimum order quantity, from one prototype to 10,000+ parts
- 3Lead timeQuotation and free DFM analysis within 12 hours
Which finishing route fits which gear
Judged on hardness, geometry and batch size
| Gear condition | Best route | Why |
|---|---|---|
| Soft steel, module 2, 50 parts | Hob and deburr only | Grinding adds cost with no accuracy gain |
| Hardened 58 HRC, external, 500 parts | Generating grinding | Rolling motion holds profile at volume |
| Hardened, internal teeth | Form grinding | Generating wheel cannot reach inside the ring |
| Case depth under 0.3 mm | Do not grind the flank | Risk of grinding through the hardened layer |
| Module below 0.5 mm | Cut, then lap or hone | Wheel tip too thin, burn risk too high |
| Damaged gear, single unit | Form grinding, light stock | Fast setup, removes minimal material |
The verdict
If your gear is hardened past 50 HRC and the drawing calls for profile or lead control, grind it. If the gear is soft, low volume, or has a case under 0.3 mm, cut it and finish it another way; grinding will cost more and can leave one flank soft.
Questions we get from engineers
How much stock should I leave for grinding?
For a carburized gear, plan on 0.05–0.10 mm per flank after heat treatment. That covers distortion without eating the case.
If the blank is turned and milled to ±0.005 mm before heat treat, 0.05 mm per flank is usually enough. Anything above 0.15 mm per flank usually means the pre-grind cut drifted and the part should be checked before grinding.
Can you grind a gear that is already carburized and hardened?
Yes, that is the normal sequence. We cut the blank soft, send it for heat treatment, then grind the hardened teeth.
What we cannot do is grind a gear with a nitrided case under 0.2 mm deep. Removing 0.05 mm would leave too little hardened layer on the flank.
Does grinding change the tooth thickness?
Yes, and it has to. Grinding removes material from both flanks, so the tooth gets thinner. The over-pin or over-ball dimension shifts by roughly 1.4 times the stock removed per flank.
That is why the pre-grind tooth thickness is cut oversize on purpose. The grinder brings it down to the final dimension, and the operator checks the over-pin value as the size control.
What surface finish should I specify?
Ra 0.2–0.8 μm is a normal ground flank. Ra 0.8–1.6 μm is fine for most power transmission gears.
Do not push for the smoothest possible flank. A very fine finish can hold less oil film and run hotter in a loaded gearbox. Ask for the finish that matches the lubrication and load, not the lowest number.
Will grinding leave marks I can see?
Yes. Form grinding usually leaves axial lines along the tooth, generating grinding leaves a diagonal crosshatch.
These marks are normal and not a defect. A light or dark patch with no pattern is the one to worry about; that points to grinding burn and should be checked by nital etch.
How do I know the gear was ground correctly?
Ask for profile, lead, pitch and runout values measured on a gear measuring center, plus the over-pin dimension and a hardness check on a sample.
A report with only one size number does not tell you whether the flank is sound. Per-tooth profile and lead data is what lets you compare batch to batch.
Send the gear drawing, get a grinding plan
Upload the part and we will come back with a quotation and a free DFM analysis within 12 hours, including the stock to leave for grinding and the inspection plan.
12-hour quote±0.005 mm pre-grind100% inspection