A Brief Discussion of the Fuji AC Servomotor System on a CNC Lathe

A CNC lathe has at least two servo axes: the Z carriage that travels along the bed and the X cross slide that feeds the tool into the work. Many turning centers add a C axis on the spindle for milling and a Y axis for off-center features. Each axis is a closed loop made of a permanent-magnet AC servomotor, a drive amplifier, an encoder, and the mechanical train of ballscrew, coupling, and bearing block. The controller closes the position loop, the drive closes the velocity loop, and the encoder reports actual position back to both.

Fuji Electric builds these motors and drives as a matched pair. The amplifier expects a defined motor inductance, a defined encoder resolution, and a defined continuous and peak current. Swap one half without the other and the loop still turns, but not at the gain the machine was designed around. That mismatch is the source of most complaints people bring to us: a lathe that holds size on light finishing passes and loses it on a roughing cut.

The failure modes are worth naming early. Bearing noise on the motor shaft, encoder signal loss, drive overcurrent trips, and rising following error on a warm machine are the four we see most. Each points at a different part of the loop, and none of them is fixed by simply raising the gain.

Sizing starts with the load, not the catalog. We measure the moving mass of the slide, the screw pitch and diameter, the friction of the linear guides, and the cutting force the axis must push against. A turning operation with a 4 mm depth of cut in 4140 steel loads the X axis very differently from a finishing pass in 6061 aluminum. The motor has to deliver peak torque for the cut and continuous torque for the duty cycle without saturating the drive.

Inertia ratio is the number that decides how the axis will feel. It compares the reflected load inertia to the motor rotor inertia. A ratio near 1:1 gives crisp response and easy tuning. Push past 5:1 and the loop becomes slow and nervous at the same time, because the gain that keeps it stable is too low to reject cutting disturbance. Machine builders who fit a larger motor often do it for inertia reasons, not torque reasons.

Belt reduction changes both numbers. A 2:1 belt halves the reflected inertia and doubles the resolution at the encoder, at the cost of belt stiffness and a small amount of backlash. Direct coupling removes that compliance but demands a motor with enough torque to move the screw alone. Neither choice is universal; it follows the axis length and the surface finish the part needs.

Cooling matters on lathes that run all day. A motor rated for a 40 °C ambient and mounted in a closed cabinet will derate. Airflow through the cabinet, or a fan kit on the motor, keeps the winding temperature inside its class and stops the thermal protection from tripping mid-cut.

• 1
  Short axis, direct driveX slide on a compact lathe. Low inertia, easy tuning, good for Ra 0.8–1.6 μm work.
• 2
  Long axis, belt reductionZ travel over 1,000 mm. Cuts reflected inertia, adds a compliance to watch.
• 3
  Heavy roughing axisPeak torque and duty cycle drive the choice more than inertia ratio.

Position feedback sets the floor on accuracy. A 17-bit serial encoder resolves about 131,072 counts per revolution. On a 10 mm pitch screw that is roughly 0.08 μm of linear resolution before any mechanical error, which is far finer than the ±0.005 mm we hold on finished parts. Resolution is rarely the limit. Mechanical backlash and screw pitch error usually are.

Following error is the lag between commanded and actual position during motion. It grows with feed rate, with axis inertia, and with low velocity loop gain. If the lathe holds size at 0.1 mm/rev feed and drifts at 0.3 mm/rev, the loop is not rejecting the load well enough. Check the mechanical train first, then look at gain and feed-forward.

Absolute encoders keep the axis position after a power cycle, so the machine does not need a homing move at every start. That saves cycle time and removes a crash risk on lathes with a tailstock or a bar feeder in the way. Incremental encoders cost less but need the reference switch to be reliable, and a dirty switch shows up as an intermittent position shift.

Cable routing is an underrated part of feedback. Encoder cables carry low-level differential signals next to spindle and drive power cables. Run them in separate trays, use shielded twisted pairs, and ground the shield at one end only. A lathe that loses position only during spindle acceleration is usually a shielding problem, not an encoder failure.

Tuning order is fixed: mechanics, then velocity loop, then position loop. Clean and preload the ballscrew, check coupling runout, and confirm the bearing block is not loose before touching a gain parameter. A machine with 0.02 mm of backlash cannot be tuned to hold 0.005 mm, no matter what the drive allows.

For the velocity loop, raise proportional gain until the axis starts to hum, then back off to about 60–70 percent of that value. Set integral gain high enough to remove steady-state error but not so high that the axis overshoots on reversal. Feed-forward reduces following error on contouring moves and helps most on lathes cutting arcs and radii with a C axis.

Thermal drift is the slow error that appears 30 to 60 minutes into a shift. The ballscrew grows as it warms, so the tool position relative to the part shifts. On a lathe turning a 300 mm shaft to ±0.005 mm, that drift can exceed the tolerance on its own. Screw cooling, a warm-up cycle, or in-process gauging with offset correction are the usual answers.

Fuji drives usually expose a torque monitor and a following error display. Log both over a shift and you get a picture of the axis that no single measurement gives. A torque trace that climbs steadily points at friction or a failing bearing. A following error that widens with temperature points at the mechanical train, not the electronics.

Repair makes sense when the mechanics are sound and the fault is local. A noisy motor bearing, a cracked encoder connector, or a single failed drive module can be fixed without touching the rest of the machine. Keep spare drives and motors for the axes that run the hardest, because those fail first.

A full retrofit makes sense when the controller is obsolete, when spare drives are no longer available, or when the machine cannot hold tolerance even after the mechanics are rebuilt. At that point the choice is between a like-for-like Fuji replacement and a move to a current drive platform with modern tuning tools and absolute feedback.

Cost is not the only factor. Downtime on a lathe that runs two shifts is expensive, and a retrofit means re-commissioning every axis, re-checking the C axis if there is one, and re-cutting test parts. We usually advise a staged plan: fix the axis that fails, measure the rest, and retrofit when the failure rate crosses the point where spares and labor cost more than the upgrade.

For shops that outsource machining, the same judgment applies to incoming parts. A turned part that shows taper, chatter marks, or a size that walks across the batch often traces back to the servo loop, not the tool. Asking the supplier for the axis and drive model, plus a following error log, shortens the diagnosis.