Controlling Milling Depth and Surface Roughness: How the Hardware Actually Works

• 1
  What the Device Controls and What It Cannot
• 2
  Mechanical Hard Stops for Milling Depth
• 3
  Feedback Loops and In-Process Depth Measurement
• 4
  How Depth Control Changes Surface Roughness
• 5
  Where a Depth-Control Device Stops Working
• 6
  Depth-Control Method vs Achievable Result

A device for controlling milling depth does one job: it fixes the axial position of the cutting edge relative to a reference surface. That reference may be the fixture, the machine table, or a previously machined face. Once the reference is set, the depth of a pocket, slot, or step becomes a number the machine can repeat, not something an operator dials in by ear.

The same hardware usually carries a second function. By holding the axial position rigid, it also limits how much the tool can deflect under cutting load. Tool deflection is the main source of mid-range surface roughness on floors and walls. A stiff depth stop cuts that deflection, so the floor comes out at Ra 0.8–1.6 μm instead of the Ra 1.6–3.2 μm you get from a light, chattering pass.

What the device cannot do is fix a bad setup. If the part moves in the fixture, the reference is gone and every depth on that face is wrong by the same amount. If the tool is dull, the finish degrades even with a perfect depth stop. The device controls one variable. It does not control rigidity of the part, coolant delivery, or spindle runout.

So the first question is always: what is actually moving? If the answer is the tool relative to the workpiece, a depth-control device helps. If the answer is the workpiece relative to the table, the money is better spent on fixturing.

• 1
  Reference surfaceFixture face, table, or pre-machined datum.
• 2
  Axial lockHolds Z position, limits tool push-back.
• 3
  Not coveredFixture slip, tool wear, spindle runout.

The oldest form is a mechanical stop: a shoulder, a collar, or a set of shims that physically limits how far the tool can travel. A collar on the shank bottoms out against a bushing, and the depth is set by the collar height. Shims in 0.01 mm steps let you trim the number without re-machining the collar.

Hard stops are cheap and they do not drift. Once the stack is set, the depth repeats to within the flatness of the contact faces, often ±0.01 mm. That is enough for air conditioning frames, brackets, and most stamped-then-machined parts. It is not enough when the print calls for ±0.005 mm.

The weak point is the contact face itself. Chips caught between the collar and the bushing add 0.02–0.05 mm to the depth in one hit. A chip brush and an air blast before each cycle cost nothing and prevent most of these errors. On horizontal setups, orient the stop so gravity drops chips away from the face.

Hard stops also load the spindle in one direction. On a long reach tool, the reaction can tilt the tool and cut a tapered wall. Keep the stop close to the cutting edge when the tool overhang is more than 4× diameter.

• 1
  RepeatabilityAbout ±0.01 mm with clean faces.
• 2
  Best forFrames, brackets, medium-tolerance pockets.
• 3
  Failure modeChip trapped on the contact face.

When the print tightens, the depth stop becomes a sensor. A probe or a contact gauge touches the reference face, the control stores the offset, and every subsequent Z move is corrected for that number. This removes the stack-up of fixture height, tool length, and thermal growth.

Thermal growth is the reason feedback wins on long runs. A 100 mm steel part grows about 0.012 mm over a 10 °C rise. On a 4,000 mm frame, that same rise moves the far end by nearly 0.5 mm. A probe re-zero before each part cancels most of it.

The trade-off is cycle time. Each probing move costs 5–20 seconds. On a 10,000-part run, that is real money. The usual compromise is to probe the first part of a batch, then run on the machine's own repeatability, and re-probe after any tool change.

Feedback also lets you machine to a floor that is not flat. If the raw casting is out of parallel by 0.3 mm, the probe maps it and the control follows, so the finished floor is uniform while the wall heights vary. That is often the only way to hold a thin floor on a warped casting.

• 1
  RepeatabilityDown to ±0.005 mm on a stable machine.
• 2
  Cost5–20 seconds per probing cycle.
• 3
  Best forLong runs, warped castings, tight floors.

Surface roughness on a milled floor comes from three sources: the scallop left by the cutter geometry, the vibration of the tool, and the smear of a worn edge. Depth control touches the middle one. When the axial position is rigid, the tool cannot bounce, so the scallop height is set by feed per tooth and tool radius instead of by chatter.

For a bull-nose or ball tool, the theoretical scallop height is roughly feed² divided by 8×radius. A 12 mm ball tool at 0.15 mm per tooth and 2 teeth gives about 0.9 μm of scallop, which is already near the Ra 0.8 μm floor. Beyond that, you are fighting vibration, not geometry.

That is where the device pays off. A rigid depth stop lets you push feed without the floor going rough, because the axial load path is short. In practice, shops report moving from Ra 1.6–3.2 μm to Ra 0.8–1.6 μm on the same tool and speed, just by removing the bounce.

To reach Ra 0.2–0.8 μm, depth control is necessary but not sufficient. You also need a balanced holder, through-spindle coolant, and a finishing pass that removes less than 0.2 mm. The device keeps the depth; the rest of the setup keeps the finish.

• 1
  Scallop heightFeed² / (8 × tool radius).
• 2
  ChatterSuppressed by a rigid axial path.
• 3
  Fine finishNeeds balanced holder and coolant too.

The device assumes the reference face is stable. On a thin-wall part, the face itself deflects under clamping, so the measured depth is correct at zero load and wrong at cutting load. No depth stop fixes that. The answer is a different fixture or a different process, such as milling before the wall is thinned.

It also assumes the tool is stiff enough to carry the load. A 3 mm end mill at 40 mm overhang will bend 0.05 mm under a normal cut, regardless of how well the depth is set. Here the device is holding a number that the tool cannot deliver. Reduce overhang or step down to a smaller axial depth.

On five-axis work, the reference face may not be perpendicular to the tool axis. A depth stop set in Z is then measuring along the wrong vector. The control has to compensate with the rotary table position, and any table backlash shows up directly in the depth. This is why simultaneous five-axis floors are usually finished with a probe or an on-machine touch cycle.

Finally, the device adds mass to the spindle. On a high-speed spindle running above 15,000 rpm, that mass changes the balance and the natural frequency. If the surface finish gets worse after fitting the device, the cause may be the device itself.

• 1
  Thin wallsPart deflects; depth reads wrong under load.
• 2
  Long toolsOverhang above 4× diameter bends.
• 3
  Five-axisReference is not along Z.