Fan CNC Machining: What the Controller Does to Your Part
This page explains how fan CNC machining actually works at the machine level, from G-code buffering to macro variables and 5-axis kinematic compensation. It is written for design engineers and buyers who need to know which controller features change part cost, tolerance, and lead time — and which ones are just marketing.

In this article
- 1
- 2
- 3
- 4
- 5
- 6
- 7
Key takeaways
How fan CNC machining executes a toolpath
Every fan CNC machining cycle starts the same way. The CAM system posts a text file of G-code blocks. The control reads ahead, fills its buffer, and converts each block into servo commands for X, Y, Z, and any rotary axis. What matters to your part is not the language itself but how fast the control can process blocks and how far ahead it looks. A control that only sees two blocks ahead will slow down in a corner even if the programmed feed rate says otherwise.
This is why the same toolpath can run at different speeds on different machines. Look-ahead, sometimes called block processing time, is quoted in milliseconds per block. A control that processes 1,000 blocks per second can hold feed through a dense 3D surface where an older control stutters. The visible result is not just cycle time. It is also surface finish on curved walls, because a stuttering control leaves dwell marks that no amount of polishing fully removes.
For the engineer sending a model out for quote, the practical question is whether the geometry needs that continuous processing. A part made of flat faces, drilled holes, and a few pockets will run well on almost any current control. A part with organic curves, thin ribs, or a blended fillet that runs into a wall is where look-ahead and servo tuning start to matter.
One number worth asking about is the smallest programmable increment. On our 5-axis centers the rotary table is Ø400 mm, and tolerance holds at ±0.005 mm. Those two facts together mean the control must resolve very small angular moves to keep a cutter on a curved wall. If the increment is coarse, the machine will round off corners that the model shows as sharp.
- 1Look-aheadMeasured in blocks per second. Higher means smoother feed through dense toolpaths.
- 2Servo tuningMatches the drive to the load. Poor tuning shows up as chatter on corners.
- 3Program incrementThe smallest move the control can command. Coarse increments round off sharp transitions.
Work offsets and tool offsets in fan CNC machining
A machine does not know where your part is. It knows where its axes are. Work offsets bridge that gap. When a setup technician touches off a vise or a fixture, the control stores the distance from machine zero to part zero. Every programmed coordinate is then measured from that stored point. Get it wrong by 0.1 mm and every feature on the part shifts by 0.1 mm, even though the toolpath is perfect.
Tool offsets work the same way in the other direction. Each cutter has a length and a radius that differ from the nominal value. The control adds those differences to the programmed path. On a 5-axis job the radius offset becomes more involved, because the cutter is no longer always perpendicular to the surface. The control has to apply the offset in the tool axis frame, not the machine frame.
This is where most scrap on complex parts comes from. Not from a bad model, but from a stale offset. A tool change that was not re-measured, a fixture that moved after clamping, a thermal shift after four hours of cutting. The control will follow the numbers you gave it, right into the fixture.
We handle this with in-process probing and a first-article check before the run continues. Raw material is verified on receipt, dimensions are monitored during the cut, and every part gets a final inspection before shipment. Reports are available on request. That sequence catches offset drift before it becomes a batch of scrap.
- 1Work offsetMachine zero to part zero. Set once per setup, verified with a probe.
- 2Tool length offsetRe-measured after every tool change on tight-tolerance jobs.
- 3Cutter radius offsetApplied in the tool axis frame on 5-axis work, not the machine frame.
When macro variables save real money
Macro programming lets a program calculate instead of just execute. Instead of hard-coding a hole pattern at fixed coordinates, the program reads a variable for hole count and spacing, then loops. Change the variable and the same program cuts a different part. For a family of brackets that differ only in length and hole count, this removes a whole class of programming errors.
The gain is not only engineering time. A macro-driven program is easier to verify, because the logic is in one place. If the pattern is wrong, it is wrong the same way on every part, and the first article shows it. A hand-edited program for each variant hides errors in the places nobody re-reads.
Macros are also used for probing routines, tool breakage checks, and pallet scheduling. These are housekeeping tasks that do not change the part geometry, but they change how confidently a shop can run unattended. A probe that checks a datum after clamping and updates the work offset can absorb small fixture variation without an operator standing at the door.
Not every job needs this. A one-off prototype with five holes gains nothing from a macro. The break-even comes when a part repeats, or when a family of similar parts shares a fixture. Below that threshold, plain G-code is faster to write and easier to debug.
Why 5-axis fan CNC machining needs kinematic data
On a 3-axis machine, the relationship between the tool tip and the part is simple. Move X and the tool moves in X. On a 5-axis machine with a trunnion and a rotary table, the tool tip position depends on the angles of both rotary axes. To place the tip correctly, the control needs to know where the rotary centers sit relative to the spindle, and how far the pivot points are from each other.
Those numbers are the kinematic model. They are measured when the machine is installed and after any crash or major service. If they drift, the tool tip follows a path that is offset from the one in the CAM file. The error is not constant. It grows as the rotary axes tilt further from zero, which is why a part can be perfect on top and out of tolerance on a steep wall.
This is the difference between indexed 5-axis work and simultaneous work. Indexed work positions the part at an angle, locks the rotaries, and cuts with three axes. Kinematic error barely matters because the rotaries are not moving during the cut. Simultaneous work keeps all five axes in motion, and the kinematic model is used on every block.
For the buyer, the useful distinction is this. If your part has features on five faces but each face is machined flat, indexed work will usually hold tolerance with less setup risk. If your part has a curved surface that wraps around the part, or a port that must be blended into a contoured wall, simultaneous work is the only way to cut it in one pass.
- 1Indexed 5-axisRotaries lock during the cut. Simpler, lower risk, more setups per part.
- 2Simultaneous 5-axisAll axes move together. One setup, better blends, stronger kinematic dependency.
Alarms, overtravel, and the limits of the control
Most alarms that stop a run are not mysteries. Overtravel means an axis reached its soft limit, usually because a work offset is wrong or a fixture is taller than expected. A servo alarm often points to a load spike, which can be a dull tool, a chip pack, or a feed rate that is too aggressive for the material. A spindle alarm may be a temperature threshold or a drive fault.
The control reports what it sees. It does not know why. That is the job of the setup sheet and the operator. A shop that records the alarm number, the block it occurred on, and the offset values at the time will solve the next occurrence in minutes instead of hours.
There are also hard limits the control cannot work around. Maximum processing size on our machines is 4,000 mm, with travel of 4,000 × 400 × 150 mm on the largest platform. Rotary work is limited by the Ø400 mm table. A part that needs more swing than the table allows has to be repositioned, which adds a setup and a re-datum.
Tool access is the other boundary. A deep pocket with a small corner radius may be geometrically possible but practically unreachable, because the holder shank contacts the wall before the cutter reaches the floor. This is a tooling constraint, not a controller constraint, but it shows up in the same place: the part does not match the model.
Which setup fits your part
Match the geometry to the machining strategy before quoting
| Part feature | Recommended setup | Why | Watch out for |
|---|---|---|---|
| Flat faces on 5 sides | Indexed 5-axis | Rotaries lock, tolerance is stable | Extra setups add cycle time |
| Curved wall blending into a boss | Simultaneous 5-axis | One continuous pass, no blend line | Kinematic offsets must be current |
| Deep pocket, small corner | 3-axis with long-reach tool | Simpler, cheaper, easier to inspect | Holder shank may hit the wall |
| Family of similar brackets | Macro-driven program | One program covers all variants | Needs a shared fixture |
| Single prototype | Plain G-code, 3-axis | Fastest to program and debug | No gain from macros |
| Part over 4,000 mm | Split into two setups | Exceeds machine travel | Re-datum adds tolerance stack |
| High-volume run, 10,000+ | Mill-turn or dedicated fixture | Fewer setups per part | Fixture cost must be amortized |
The trade-off in one line
If the geometry is flat and the volume is low, use the simplest setup that reaches every face. If the surface is curved and wraps around the part, pay for simultaneous 5-axis and keep the kinematic offsets current. There is no middle option that gives you both.
Questions engineers ask about fan CNC machining
Does the controller brand change the tolerance I can expect?
Not directly. Tolerance comes from the machine structure, the spindle, the thermal stability of the shop, and the inspection method. The controller affects how smoothly the toolpath runs and how accurately small moves are resolved.
A well-tuned machine with a modest control can hold ±0.005 mm on a flat part. A poorly tuned machine with a high-end control will not.
What is the smallest feature you can cut?
It depends on the tool, not the control. A cutter needs enough stiffness to reach the feature without chattering. Deep slots with small corner radii are the usual limit, because the holder shank runs out of clearance before the cutter reaches depth.
Send the model and we will flag features that need a different approach before quoting.
Do I need to supply a specific post-processor output?
No. Send the 3D model and the 2D drawing with tolerances and finish callouts. We generate the toolpath and the G-code from our own CAM setup.
If you have a specific datum scheme or inspection plan, include it. That affects how we set work offsets.
How do you handle a part that exceeds machine travel?
We split it into two or more setups and re-datum between them. Each additional setup adds a small tolerance stack, so we plan the split at a feature that is not tolerance-critical.
Our largest platform handles 4,000 × 400 × 150 mm. Anything beyond that needs a different process.
What causes a good program to cut a bad part?
Offsets and thermal drift, in that order. A stale tool length offset shifts every feature in Z. A fixture that creeps after clamping shifts the part in X and Y.
We verify raw material on receipt and run in-process checks, with a full inspection before shipment.
Can you run a part family from one program?
Yes, when the variants share a fixture and differ only in a few dimensions. We use macro variables for hole counts, spacing, and depth, so one program covers the whole family.
For a single prototype, plain G-code is faster to write and easier to verify.
Send the model, get a manufacturability answer
Upload your files and we will return a quotation with a free DFM analysis, flagging any feature that needs a different setup before you commit to a run.
12-hour quote and DFM100% inspection before shipmentNDA available on request