Chip load is the programmed advance of each effective cutting edge during one spindle revolution. It can be selected from tooling data to calculate feed rate, or recovered from an existing program to audit how much feed each edge receives.
Chip load per tooth = table feed ÷ (RPM × effective flutes)
The same unit must be used for chip load and table feed: mm/tooth with mm/min, or in/tooth with IPM.
Inch-unit reverse audit — worked checks and practical limits.
Chip load and reverse formulas
Rearranging the relationship gives table feed: vf = fz × n × zc. Feed per revolution is fn = fz × zc. These are geometric conversions; none of them decides whether the chosen chip load is appropriate for the tool and cut.
Worked example: audit an existing program
A four-flute tool runs at 6,000 RPM and 1,200 mm/min.
- Cutting-edge passes per minute:
6,000 × 4 = 24,000. - Chip load:
1,200 ÷ 24,000 = 0.05 mm/tooth. - Feed per revolution:
1,200 ÷ 6,000 = 0.20 mm/rev.
If the same chip load is required at 5,000 RPM, the matching table feed is 0.05 × 5,000 × 4 = 1,000 mm/min.
Selecting a starting chip load
Use current data for the exact tool family, diameter, geometry, grade, coating, and work material. Toolmaker ranges often assume a particular radial width, axial depth, holder condition, coolant method, and machine stability. Moving a value between a small solid-carbide end mill and a large indexable cutter can be misleading even when both have four flutes.
Use the number of effective edges. A nominal flute count is normally correct for a conventional end mill, but special pitch, wiper, or insert arrangements can require the manufacturer’s definition.
Programmed chip load is not always chip thickness
At less than 50% radial engagement, the maximum geometric chip thickness can be lower than the programmed feed per tooth. Constant-engagement toolpaths may compensate by increasing programmed feed. The correction must be removed where engagement rises, including full-width slots and tight internal corners.
Runout also prevents perfectly equal chips: one edge can remove more material than the average calculation predicts. Tool overhang, spindle and holder condition, workholding, acceleration, coolant, and interrupted cutting all affect the usable value.
Common mistakes
- Using the number of toolpaths or passes as the flute count.
- Mixing IPM with millimeter chip load.
- Multiplying a drill’s feed-per-revolution recommendation by its cutting lips.
- Treating a calculated chip load as proof that the cut is safe.
- Applying radial compensation where engagement is not constant.
Technical references
Published and formula-checked by MfgWorkbench. Last reviewed October 19, 2026. See our editorial method.
Audit a program after feed and spindle overrides
Consider a hypothetical three-flute tool with programmed speed 8,000 RPM and feed 720 mm/min. Programmed chip load is 720/(8,000 × 3) = 0.030 mm/tooth. What happens after an override depends on which quantity actually changes.
| Steady-state case | Actual RPM | Actual feed | Calculated advance per tooth |
|---|---|---|---|
| No override | 8,000 | 720 mm/min | 0.0300 mm |
| Feed reduced to 80%; RPM unchanged | 8,000 | 576 mm/min | 0.0240 mm |
| RPM reduced to 80%; feed unchanged | 6,400 | 720 mm/min | 0.0375 mm |
| Both reduced to 80% | 6,400 | 576 mm/min | 0.0300 mm |
The third case increases advance per tooth by 25%, despite a slower spindle. The last case preserves the arithmetic ratio but changes cutting speed. These examples assume feed-per-minute behavior and independently applied overrides; verify your control’s feed mode and synchronization before interpreting a real program.
What the reverse calculation can diagnose
Use actual stable speed/feed readings where possible. A commanded feed on a short move may never be reached. Then check radial engagement, effective tooth count and runout: the average programmed advance is not a measurement of the chip carried by every edge. A worn edge or one projecting flute can invalidate an equal-sharing assumption.
This page is for diagnosing an existing program. To construct a new feed command, use the forward feed-rate workflow; for low radial engagement, check the chip-thinning model only within its stated limits. Stop an unstable cut rather than treating an override calculation as a guarantee of safety.
Reverse-check an actual feed reading without mixing units
Use a stable cutting segment, not a rapid or an accelerating corner. In this hypothetical audit, a four-flute end mill has a displayed cutting feed of 24.0 in/min at 6,000 RPM. The program originally commanded 30.0 in/min at the same speed. Treat these as feed-per-minute values; verify the control mode and whether the display is actual or commanded feed.
| Audit quantity | Commanded case | Displayed-feed case |
|---|---|---|
| Feed per minute | 30.0 in/min | 24.0 in/min |
| Feed per revolution | 30 ÷ 6,000 = 0.0050 in/rev | 24 ÷ 6,000 = 0.0040 in/rev |
| Advance per effective tooth | 0.00125 in/tooth | 0.00100 in/tooth |
| Metric cross-check | 0.03175 mm/tooth | 0.02540 mm/tooth |
The displayed-feed case is 20% lower, provided the spindle really remains at 6,000 RPM. The cross-check is 24 × 25.4 ÷ (6,000 × 4) = 0.0254 mm/tooth. Multiplying by 25.4 converts inches to millimeters; it does not change the tooth count.
Record the context before changing the program
- Record tool identification, effective edge count, RPM and feed from the same steady segment.
- Record radial and axial engagement, overrides, feed mode and whether CAM applied chip-thinning compensation.
- Compare the recovered value with the supplier’s range for that exact engagement and tool. A mismatch identifies a question, not an automatic instruction to raise feed.
- If the values agree but the cut is unstable, investigate unequal edge loading, runout and workholding. The average advance calculation cannot identify which flute is overloaded.
This audit does not measure instantaneous chip thickness. Use the radial chip-thinning guide for its separate geometry assumptions; use the forward feed-rate guide when building a new command from approved cutting data.
Content and numerical examples reviewed October 19, 2026. Examples are not production test results.
Content updated October 19, 2026. Sources and numerical examples are explained above; examples are not production test results. Editorial method · Report a correction.