A sourcing engineer sent us the same aluminium housing model twice last spring, once as-is and once with a single note added to the drawing: "sculpted blend on flange must be cut with simultaneous 5-axis machining, no indexing". The first quote came back at 118 index points. The second came back at 163. Same part, same material, same tolerance block — a 38% premium paid for one 46 mm long blend radius. Three weeks later the buyer discovered the winning supplier had simply ignored the note, indexed the part into nine positions and handed the blend to two operators with abrasive paper. The parts looked fine. They measured fine. They were also 0.06 mm out of profile at the blend centre, which nobody would notice until the mating cover rocked on assembly.
3+2 vs simultaneous 5-axis machining is one of the few process questions where buyers routinely pay for capability they never receive, and occasionally refuse to pay for capability they genuinely need. Both mistakes come from the same root cause: nobody has published a pass/fail test. This article gives you one, built around three numbers — achievable facet error, tool reach requirements, and rotary-axis tolerance contribution — plus a routing method that typically removes 15-40% of the cost of a "five-axis" part without touching its function. Everything here assumes you are buying parts, not machines, which is where most published comparisons quietly switch tracks and stop being useful.
What 3+2 machining actually does
In 3+2 machining (also written 3+2 positional machining, indexed 5-axis machining or 5-axis positional machining), the two rotary axes tilt the part or spindle to a fixed angle, lock, and stay still while the three linear axes cut. Nothing rotates during the cut, so every feature produced is a three-axis feature approached from a tilted direction.
That locking behaviour is what makes indexed work attractive. Because the rotary axes are stationary during the cut, the toolpath can be generated with standard 3-axis strategies, verified quickly, and run with confidence by a shop that has not invested years into multi-axis CAM expertise. It also means that the cut is being taken against a rigid, braked rotary joint rather than against a servo that is simultaneously trying to hold position while cutting loads push it around. For flat faces, pockets, holes and chamfers — and for the threaded families covered in our precision fasteners and custom nuts and studs guides — positional rigidity usually wins outright.
What simultaneous 5-axis machining does differently
In simultaneous 5-axis machining all five axes move at once, so the tool orientation changes continuously along the path. That continuous re-orientation is what makes impeller blades, turbine airfoils, medical implants and ported engine components cuttable in one operation: surfaces no fixed angle reaches without visible steps or material that no straight tool can remove.
The cost of that flexibility is not the machine, at least not directly to you. It is the software and the human time behind it: multi-axis toolpath strategies, gouge checking against the whole holder assembly, axis-limit management, post-processor correctness and machine-specific simulation. Published estimates from shops put programming effort for a true simultaneous toolpath at two to five times the effort for the same part programmed as a set of indexed setups. Verification follows the same ratio. That is the real premium, and it is the part you are paying for when you insist on continuous motion.
The vocabulary that causes most of the confusion
Buyers and vendors use the same phrase for three different things. "Five-axis machining" can mean (1) any 5-axis machine doing positional work, (2) five-sided machining in one clamping, or (3) genuine simultaneous contouring. Only (3) carries the programming surcharge. A shop can honestly advertise all three and deliver only the first, which is why the words on the capability list matter far less than the words on your drawing. If your RFQ says nothing about which features need continuous tool motion, you have delegated a commercial decision to someone whose incentive runs the other way.
The decision test: must the tool change angle during the cut
Describe each feature using nouns, then test it. If holes, pockets, slots, flats, bosses, chamfers and threads describe it completely — a feature with a fixed axis of definition — then 3+2 machining is enough, whichever face the feature sits on. If you need the words surface, blend, transition, twisted, or an undercut the tool must reach under progressively, test that feature for simultaneous 5-axis machining, then use the math in the next section to decide.
In production terms, indexed vs simultaneous machining is settled by the feature, not the part: a hydraulic manifold with twenty-seven ports machined at nine different compound angles is a 3+2 part. The same test applies to turned-and-milled work, including CNC turning services and transmission shaft machining, where every feature still has a fixed axis. A compressor impeller with seven twisted backswept blades is a simultaneous part. Most real components, including aerospace camera mounts, drone gimbal arms and high-end electroacoustic and audio housings, fall somewhere in between — which is exactly why a feature-level decision beats a part-level one.
A ten-question scorecard
Score each machined feature separately. Two or more "yes" answers and that feature should be routed for simultaneous cutting; zero and it should be quoted as positional work.
| Question | Answer drives |
|---|---|
| Does the feature demand the tool axis to change while cutting? | Simultaneous |
| Is the surface defined as freeform, sculpted or blended in the model? | Simultaneous |
| Is the specified profile/form tolerance tighter than the achievable facet error (see next section)? | Simultaneous |
| Does reaching it require an undercut that a straight tool cannot enter? | Simultaneous |
| Is required surface finish at or below Ra 0.8 µm on a curved area? | Usually simultaneous |
| Is the feature a prismatic element on any face? | 3+2 |
| Can it be described with a simple tool axis plus depth? | 3+2 |
| Are all its reference features reachable within the same clamping? | Either |
| Is the production volume high enough that cycle time dominates setup savings? | Lean 3+2 |
| Is this the third or later revision of a design still in flux? | Lean 3+2 (cheaper to re-program) |
That split is also why our five-axis CNC machining guide treats capability as a set of feature capabilities rather than a machine specification. Notice how rarely "the part is complex" is the right answer. Complexity itself is not the trigger; continuous change of tool orientation is the trigger.
Finish math: can indexed passes fake a curved surface
It comes down to one number, and almost nobody publishes it. Indexed passes can fake a curve, but only at a cost: facet error is roughly Rr x (1 - cos(delta/2)), so on a 60 mm blend radius a 5 degree index leaves 0.057 mm of error, while holding ±0.01 mm needs about 1 degree steps, or roughly ninety separate index operations.
When you replace a continuous sweep with discrete indexes, the surface becomes a series of facets. The deviation from the true geometry at the middle of each facet is the classic sagitta error:
facet error ≈ Rr × (1 − cos(Δθ / 2))
where Rr is the local radius of revolution about the rotary centre (mm) and Δθ is the angular step between indexes (degrees). A second, independent limit comes from the ball-nose step-over inside each facet:
scallop height h ≈ s² / (8 Rt)
where s is step-over (mm) and Rt is the tool nose radius (mm). Rough rule for translating that to roughness: for a clean scallop profile, Ra lands around h/4 to h/5, before machine vibration and material effects are counted.
Take a realistic case: a blend radius with Rr = 60 mm, indexed every 5 degrees.
| Angular step Δθ | Number of indexes over 90° | Facet error (mm) | Typical outcome |
|---|---|---|---|
| 15° | 6 | 0.51 | Visible faceting, heavy manual work |
| 10° | 9 | 0.23 | Visible, cannot hold ±0.1 mm profile |
| 5° | 18 | 0.057 | Bench polishing needed, 8-15 min/part |
| 2° | 45 | 0.009 | Acceptable for most ±0.05 mm callouts |
| 1° | 90 | 0.0023 | Acceptable for most ±0.01 mm callouts |
Facet error scales with the square of the index step, so read the first and last rows. First, 3+2 machining can indeed approximate a gentle curve — but it needs dozens of orientations to hold anything tighter than about ±0.05 mm. Second, each additional orientation costs air-cutting time, a verification pass and, crucially, more chances for the indexing error discussed below. These tools and their setup procedure belong in an early DFM analysis, not in an ECN issued after tooling is cut.
The rule of eight
Eight orientations. That is the crossover. Once a single sculpted surface needs more than roughly eight index orientations to hold its specified form, continuous motion becomes cheaper *and* more accurate than the indexed workaround. Below eight orientations — meaning your part has a handful of angled flat faces, compound-angle holes and maybe one generous blend — indexed work is the better buy. That single threshold resolves the majority of arguments, and it gives procurement a sentence they can actually use in a negotiation: "this surface needs twelve indexes to hold 0.05 mm, quote it as simultaneous or quote the polishing labour separately."
Tool reach, rigidity and the length that ruins tolerance
The second reason to tilt anything is shorter tooling, and the physics here are unforgiving. A cantilevered end mill deflects in proportion to the cube of its stick-out length (δ ∝ L³), which is why a modest reduction in tool length buys a big improvement in form accuracy.
| L/D ratio | Relative deflection (normalised to 3:1) | What it means on the floor |
|---|---|---|
| 3:1 | 1× | Full slotting, reliable size control |
| 4:1 | 2.4× | Normal finishing, good wall accuracy |
| 5:1 | 4.6× | Light finishing only; taper begins to show |
| 8:1 | 18.9× | Chatter risk, poor Ra, obvious tool marks |
| 12:1 | 64× | Rarely holds tolerance; usually rejected at quoting |
Tilting the part or the spindle so that the cutter meets a wall near-perpendicular typically cuts required stick-out by 30-50% compared with reaching down to the same feature vertically. Run that through the cubic law and a move from 8:1 to 4:1 gives roughly eight times less deflection — which is why the tilted setup holds the wall flat while the long vertical tool leaves a taper you can catch with a fingernail. It is also why deep features are routinely quoted with either deep-hole drilling strategies or tilted multi-axis milling rather than long tooling.
Titanium and hardened stainless steel multiply every row in that table; aluminium forgives most of it. Cutting titanium or hardened stainless with an 8:1 tool will not produce a part that survives contact measurement, and neither will PEEK or heat-treated alloy steel in the same geometry. When you combine material stiffness with reach, review your material selection guide before you argue about whether a feature needs five simultaneous axes.
Tolerance: where simultaneous helps and where it quietly hurts
More simultaneous axes do not automatically buy accuracy. Sometimes it runs backwards. Simultaneous 5-axis machining improves form on curved surfaces, but it also adds a rotary error term that grows with distance from the pivot — about 0.010 mm per 100 mm at ±20 arcsec. For flat features already machined in one clamping, braked 3+2 positioning is usually the more accurate choice. Setup consolidation, for its part, is a genuine win: every time a part is unclamped and re-indicated you typically add 0.01-0.05 mm of positional error with good fixtures, and more with soft jaws. Machining five faces in one clamping removes most of that, and it is the single largest accuracy argument for putting a prismatic part on a 5-axis machine at all — our tolerance guide shows how those numbers stack up, and 4-axis machining covers the cheaper case where one rotary axis is enough.
Simultaneous motion adds one error source that indexed work never has: rotary accuracy amplified by distance from the rotary centre. The conversion is simple and rarely quoted:
rotary contribution (mm) ≈ distance from pivot (mm) × angular error (arcsec) ÷ 206 265
A well-maintained rotary axis indexes within roughly ±15 to ±30 arcsec, measured the way ISO 230-7 defines for rotary axes and ISO 10791-6 defines for interpolation accuracy on five-axis machines. At a feature 100 mm from the pivot, ±20 arcsec contributes about 0.010 mm. Move that same feature 250 mm away from the pivot and the identical rotary performance contributes 0.024 mm — more than many GD&T callouts allow, and it arrives before the tool has cut anything.
Five-axis machining vs 3+2 tolerances come down to where the error enters. Practical consequences worth putting on your drawing:
- Place GD&T datums and the tightest features as close to the rotary centre as the design allows, or specify that they are machined in the same fixture orientation.
- For compound-angle features, tolerance stack-up must include the rotary term, not just the cutting term; our tolerance guide shows the addition method.
- Distinguish *form* tolerance Ra/Rz/profile from *positional* tolerance. Simultaneous motion improves form on curved surfaces; it does nothing for positional relationships between two flat faces that were already in one clamping.
- Remember RTCP (rotating tool centre point). Without a working, calibrated RTCP function the machine cannot compensate the pivot offset, and every tilted move shifts the tool tip by an amount proportional to tool length. Always confirm RTCP is active on the specific machine that will run your part.
This is why a feature-level tolerance budget beats a blanket "machine on 5-axis" instruction, and why IATF 16949 documentation asks for process-level reasoning rather than machine-level claims, as does ISO 13485 on the medical side.
| Error source | 3+2 indexed | Simultaneous |
|---|---|---|
| Re-fixturing between faces | Removed (one clamping) | Removed (one clamping) |
| Rotary contribution during cut | Near zero (axes braked) | Present, scales with pivot distance |
| Form error on sculpted surface | Facet error per index step | Scallop from step-over only |
| RTCP sensitivity | Low | High |
| Best used for | Prismatic features, compound holes | Freeform surfaces, blends, undercuts |
Cost: what you are actually buying
Full 5-axis CNC machining cost is nearly always quoted as an hourly rate, and that is the wrong denominator. Avoid comparing absolute machine prices or shop rates; both are volatile and both distract from the decision. Compare the three elements that actually move your unit price. We go through the rate comparison and the set-up arithmetic in our 5-axis machining cost guide.
| Cost element | 3-axis baseline | 3+2 indexed | Simultaneous |
|---|---|---|---|
| Programming effort | 1.0× | 1.5-2.0× | 3-5× |
| CAM verification / simulation | 1.0× | 1.5× | 3-4× |
| Machine hour index | 1.0× | 1.3-1.6× | 1.6-2.2× |
| Fixtures required | One per face group | One | One |
| First-article rework risk | Moderate | Low | Moderate to high (bugs) |
| Best at | Simple prismatic parts | Multi-face parts | Genuine freeform geometry |
Read the table together with the programming ratio: for the *same* geometry, asking for a true simultaneous 5-axis toolpath where indexing would do multiplies professional labour by two to three times before a chip is made. That is the premium you must be able to justify, not the hourly machine rate. For a wider view of where money goes in a machined part, see the true cost of CNC machining.
Hybrid routing: the quote nobody offers you
Most components that get quoted as "5-axis" need continuous motion on one or two surfaces only and indexed access everywhere else. Quoting the part as a single method forces you to buy a simultaneous 5-axis process for the whole part, including the features that only needed indexing. Splitting it by feature is legitimate, easy to specify, and almost never offered unprompted.
Worked example — an aluminium bracket, 90% prismatic features, one 46 mm sculpted blend with ±0.05 mm profile:
| Routing | Programming index | Cycle index | Relative part cost | Result |
|---|---|---|---|---|
| All-simultaneous (as specified) | 100 | 100 | 100 | Correct, overpaid ~25% |
| All 3+2, blend indexed at 5° | 55 | 92 | 70 | Facet 0.057 mm, needs 12 min hand polish: +14 recurring |
| Hybrid: 3+2 base + simultaneous finishing pass on blend only | 70 | 88 | 78 | Correct profile, no manual work, no functional change |
At 30 pieces the difference between the first and third rows is a rounding error. At 300 pieces it is a budget line, and the second row quietly costs more than the hybrid every single time while adding operator-dependent variation that no CMM report will ever see. Hybrid routing is how we quote the majority of low-volume five-axis work, and it is exactly what a good prototyping review should surface before you freeze the design.
Do I need 5-axis machining? When paying for simultaneous is simply wasted
- Multi-face prismatic housings with flat faces at several angles
- Hydraulic or pneumatic manifolds with compound-angle ports
- Brackets where every critical dimension is a hole pattern position
- Parts whose only "curve" is a large-radius cosmetic fillet with no profile callout
- Anything where the draughtsman wrote "5-axis" as a synonym for "accurate"
Machine configurations: what 3+2 vs simultaneous 5-axis machining can reach
Two shops advertising identical specifications may not be able to produce identical parts. The rotary hardware defines the achievable envelope, rigidity and — importantly for the tolerance section — how far your features sit from the pivot.
| Configuration | Typical part size | Rigidity | Rotary accuracy loss when loaded | Cost index | Best for |
|---|---|---|---|---|---|
| Trunnion table | Small to medium | High | Low | Medium | General precision brackets, audio chassis, optical mounts |
| Swivel head | Large (even very large) | Medium | Medium | High | Big structural parts, large moulds |
| Head + table (hybrid) | Medium to large | Medium | Medium | Very high | Complex aerospace components |
| Horizontal + rotary | Medium to large | High | Low | High | High-volume cube work |
If your part needs genuine simultaneous 5-axis cutting on a freeform surface, the machine matters less than the process; if your part has tight features at the extremes of the envelope, the machine matters enormously. This is one of the first things worth asking during supplier selection.
How to write the drawing so you do not overpay
Add seven lines to your RFQ. Suppliers can quote sharply against them, and they remove the incentive to either over-engineer the process or quietly substitute indexed work.
- Method per feature group: state "3+2 indexed" or "simultaneous" explicitly.
- For every feature marked simultaneous: give the reason (blend / undercut / profile).
- Profile tolerance on each sculpted surface, with the datum structure named, per ASME Y14.5.
- Acceptable facet or blend witness limit, if any, and where it is not permitted.
- Surface texture requirement per surface (Ra value and lay direction if cosmetic), per ISO 21920-1.
- Machining datums and clamping faces the supplier may use.
- First-article evidence required (CMM report on the sculpted feature, or a sample coupon).
Line 7 is the one most buyers omit and the one that prevents the opening story of this article: a supplier who prices the job as 3+2 machining, indexes the sculpted surface and never mentions it. Also note where the part will eventually be finished — anodising, plating or powder coat changes what counts as an acceptable witness mark on a non-critical face. It matters most on brushed aluminium panels and amplifier front panels, where the lay direction is part of the brand, and on consumer-electronics enclosures such as an iPad hub shell.
Auditing real simultaneous 5-axis machining capability
If you are going to pay simultaneous rates, verify simultaneous capability. None of this needs a factory visit. All six checks fit into one twenty-minute call.
| Evidence requested | What real capability looks like | Red flag |
|---|---|---|
| Continuous toolpath test coupon across a dome (ISO 10791-7 publishes exactly this test piece) | No visible index steps under raking light | Facet lines or dyed highlight bands |
| Rotary axis calibration date | Within 12 months, certificate available | "We calibrate annually" with no record |
| CMM report on a compound-angle feature | Reports the feature to the drawing datum | Only reports easily reachable prismatic dims |
| CAM verification / machine simulation | Shows full machine model incl. holder | Only a wireframe toolpath video |
| RTCP demonstration | Control holds tool tip through a tilt | Operator re-posts instead of locking RTCP |
| Sample part in the same material family | Available, dated, traceable | Only aluminium samples offered for a titanium job |
Two minutes on the phone using this list is worth more than any capability brochure. If a supplier cannot produce a continuous-path coupon, they will index your part and polish it, and you will discover it during assembly.
FAQ
What is the difference between 3+2 and simultaneous 5-axis machining?
In 3+2 machining the rotary axes position the part at an angle and lock before the cut, so only the three linear axes move during material removal. In simultaneous 5-axis machining all five axes move together and the tool orientation changes continuously along the path. Indexed 3+2 suits multi-face prismatic parts; simultaneous cutting is required for freeform surfaces, blends and progressive undercuts.
Is 3+2 cheaper than simultaneous 5-axis machining?
Yes, usually 20-40% of machining cost for eligible geometry, and the saving comes mostly from programming labour rather than machine time: a positional strategy needs 1.5-2x the programming effort of 3-axis work, true simultaneous toolpaths 3-5x. The qualifier is "eligible" — forcing indexed work onto a genuine freeform surface adds hand polishing that eats the saving immediately.
Can indexed 3+2 passes replace a true simultaneous 5-axis cut?
Only while the geometry is forgiving. Facet error is roughly Rr x (1 - cos(delta/2)), so holding ±0.01 mm on a 60 mm blend radius needs about 1 degree steps, or around ninety index operations — slower and more error-prone than one continuous cut. Within about eight index positions, indexing is the better buy.
Does simultaneous 5-axis always deliver better tolerances?
No. Continuous motion improves form on curved surfaces, but it adds a rotary error term that grows with distance from the pivot: about 0.010 mm per 100 mm at ±20 arcsec. For flat features already machined in one clamping, braked 3+2 positioning is usually the more accurate choice.
Why do two shops quote very different prices for the same "5-axis" part?
Usually because one quoted indexed work and the other quoted true simultaneous cutting, and your drawing never said which features require continuous tool motion. Occasionally it is the reverse: a cheap quote assumes the sculpted surface will be indexed and hand-polished, and that labour never appears on the quotation.
Which parts genuinely require simultaneous 5-axis machining?
Impellers and blisks, turbine blades and vanes, propellers, orthopaedic implants with organic geometry, ported surfaces, freeform mould cores with continuous draft, and any surface where the tool must change its tilt progressively to avoid gouging. Multi-face prismatic parts — even complex ones such as manifolds, mould plates, robot joint housings and structural brackets — almost never do.
How do I specify this without writing a process specification?
Route features, not parts. Mark every feature group as either positional or simultaneous, attach a profile tolerance to each sculpted surface, and ask for a CMM report on that feature at first article. Seven lines are enough; the template above is designed to copy straight into an RFQ.
Is there a hybrid approach?
Yes, and it is usually the best answer. Machine all prismatic features with indexed 3+2 positioning, then finish only the sculpted surfaces with a simultaneous operation. Material, accuracy and function are unchanged; typical savings run 15-40% against quoting the whole part as simultaneous.
Getting this decision right on real parts
The whole discussion reduces to three questions you can answer before a request for quotation leaves your desk: does any feature require the tool angle to change during the cut, can indexed passes hold the specified profile within about eight orientations, and how far do the tight features sit from the rotary centre? Answer those three and the correct routing — indexed, simultaneous, or hybrid — becomes obvious, along with what you should reasonably pay.
Ruijin Fenghui Precision Technology runs 200+ CNC machines including 3-, 4- and 5-axis machining centres under IATF 16949 and ISO 9001, holding ±0.01 mm on production parts across milling, turning and the finishing operations that follow it, for audio, drone, automotive, medical and robotics customers. Send us the STEP file and the tolerance block, tell us which surfaces are cosmetic, and you will get a 24-hour quote that states whether each feature group runs as 3+2 machining or as simultaneous 5-axis machining — no unspecified "five-axis" line item, and no silent substitution later. Start your review with our DFM checklist or ask for a routing comparison before the design freezes.