niles ordinar gear grinder: how CNC changes the grinding loop

• 1
  What the niles ordinar gear grinder actually does
• 2
  How CNC closes the loop on a gear grinder
• 3
  What CNC improves and what it does not
• 4
  Size, thermal, and rigidity limits to check first
• 5
  Choosing between retrofit, rebuild, and replacement
• 6
  What to record before and after the retrofit
• 7
  Mechanical index vs CNC retrofit: which fits the job

The machine grinds the flanks of a cylindrical gear with a formed wheel. The wheel is dressed to the tooth space profile, then fed into the gap while the workpiece rotates slowly. On the original design, that rotation came from a mechanical index train: change gears, a worm drive, and a differential that tied the table rotation to the vertical stroke of the wheel head.

Two motions decide the tooth form. The first is the rotary position of the workpiece, which sets the angular location of each flank. The second is the depth and direction of wheel feed, which sets the profile. In the mechanical machine both motions are locked together by gearing, so the ratio between them is fixed by the gear set you bolted in.

That mechanical link has a real advantage: it cannot drift. If the change gears are correct, the index is correct, and it stays correct for the whole batch. The weakness is everything around it. Wear in the worm, thermal growth in the table, and backlash in the differential all show up as lead error or profile error, and none of them are visible from the control panel.

A niles ordinar gear grinder of this generation was built heavy on purpose. The bed and column carry enough mass to absorb grinding forces on gears up to roughly 600 mm diameter and modules in the mid-single digits. The frame is rarely the limiting factor after a retrofit. The control architecture usually is.

A CNC retrofit replaces the index train with two servo axes. The C axis turns the workpiece table. The X or Z axis moves the wheel head. A motion controller interpolates them so that one revolution of the table corresponds to one tooth pitch of vertical travel, exactly as the change gears used to do. The difference is that the ratio is now a number in a parameter file rather than a physical gear.

The loop closes through encoders. A rotary encoder on the table reports actual angular position; a linear scale on the wheel head reports actual depth. The controller compares commanded and actual values every control cycle, typically in the low milliseconds, and corrects the following error. This is what makes lead error visible and correctable rather than a slow drift you discover at inspection.

Servo stiffness matters more than resolution for gear grinding. A high-count encoder with a soft drive will still lag under grinding force, and that lag shows up as a thickness variation across the face width. Shops that retrofit an old machine usually keep the original heavy table drive and add the encoder there, rather than swapping in a smaller modern servo.

The controller also has to manage the dresser. On a form grinding machine, the diamond roll or dresser shapes the wheel profile, and that profile is copied into the tooth space. If the dresser position drifts by 0.01 mm, the tooth profile drifts with it. CNC lets you compensate dresser position per dress cycle instead of shimming the dresser by hand.

CNC improves three things directly. First, setup time. Changing from one gear to another becomes a parameter change rather than a change-gear swap, which on a large machine can take hours. Second, lead correction. The controller can apply a crown or a lead modification along the face width by varying the interpolation ratio, something a fixed gear train cannot do. Third, dressing consistency, because the dress cycle is programmed and repeatable.

CNC does not improve the grinding burn limit. The heat entering the flank is set by wheel speed, depth of cut, coolant delivery, and the thermal mass of the workpiece. A controller cannot remove heat it did not create. If the original machine burned the flanks at a given material removal rate, the retrofit will burn them at the same rate unless the wheel specification or coolant changes.

CNC also does not fix a worn wheel spindle or a table with damaged ways. Those are geometric errors below the control loop. A machine that cannot hold a straight line mechanically will not hold one because a servo is driving it. Retrofit projects that skip the mechanical survey tend to end with a machine that indexes perfectly and grinds poorly.

Where CNC pays back is batch flexibility and traceability. Small and medium batches with frequent changeovers benefit most, because the setup saving repeats every time. High-volume single-part runs benefit less, since the mechanical machine was already fast once it was set. The retrofit is a flexibility purchase, not a speed purchase.

Workpiece size is the first gate. A machine in this class handles gears in the range of a few hundred millimeters diameter, with face widths that fit the table stroke. If the part you want to grind is larger than the original design envelope, no control change will make it fit. Check the table load rating and the maximum swing before anything else.

Thermal stability is the second gate. Gear grinding pushes heat into the workpiece, and a large steel gear grows measurably during a long grind. A CNC loop can compensate for a known thermal model if the machine has temperature sensors and the controller supports the compensation, but that is an added engineering task, not a default feature. Most retrofits run without it and manage heat through coolant and spark-out instead.

Rigidity sets the achievable profile. Form grinding puts a line contact between wheel and flank, so the specific grinding force is high. A machine with a flexible wheel head will deflect and produce a profile that follows the deflection rather than the dress form. Look at the wheel head slide clearance and the spindle bearing condition before assuming the control is the problem.

Accuracy targets should be written down before the retrofit starts. If the drawing calls for a lead tolerance and a profile tolerance, the mechanical survey has to show that the machine can meet them with margin. A retrofit cannot create accuracy the frame does not have.

A retrofit makes sense when the frame and spindle are sound and the pain is in setup time, lead control, or documentation. The mechanical survey has to confirm the ways, the table bearing, and the wheel spindle before money is committed. If those pass, the control work is predictable and the machine keeps its original rigidity.

A rebuild makes sense when the geometry is worn but the control is adequate. Scraping the ways, replacing the table bearing, and regrinding the spindle restore the machine's ability to hold a straight line. This work is independent of the control system and often costs more than the CNC package itself.

Replacement makes sense when the part envelope has grown or when the accuracy target is beyond the frame. A modern gear grinder with a built-in measuring system and thermal compensation solves problems a retrofit cannot reach. The decision is usually driven by the tolerance on the drawing, not by the age of the machine.

For contract shops, the practical line is this: retrofit when the machine is the right size and the tolerance is reachable, rebuild when the geometry is the problem, and replace when both are. Mixing the three leads to a machine that has new electronics and the same old errors.

Record a baseline before anything is changed. Grind one representative gear on the mechanical setup, measure lead and profile over the full face width, and keep the chart. Without a baseline, there is no way to show that the retrofit helped, and no way to separate control error from mechanical error later.

Record the thermal drift as part of the baseline. Grind three parts in sequence without a cooldown and measure each one. If the third part drifts outside tolerance, the machine has a thermal problem that a control change will not remove. This single test explains most retrofit disappointments.

After the retrofit, repeat the same three-part sequence and compare. A successful retrofit shows tighter part-to-part spread and a flatter lead trace, not necessarily a lower single-part error. The gain is repeatability first, absolute accuracy second.

Keep the parameter file under version control. Index ratio, dresser offsets, and compensation tables are the machine's calibration. Losing them means re-establishing the setup from scratch, and on a large gear grinder that is a full day of work.