Coaxiality after the second setup: how not to lose the reference

Why the datum is lost at the second setup

In the first setup a part usually rests on clear geometry: a face, center holes, an external diameter or a fit. That support becomes the reference for the rest of the machining. At the second setup the reference is often changed not because it’s correct but because it’s easier to clamp the part that way.

That’s when the datum is lost. If a shaft was first turned from centers and later clamped by a rough diameter, the axis no longer has to match the axis of the first operation. The same happens with housings: you may locate the part on a face and bore, then flip it and clamp by the outer surface with a slight tilt.

Even a small error quickly creates a noticeable shift. A bit of swarf in the jaws, a burr on the face, or a slight misalignment when tightening the clamp is enough. On a short part this gives a few hundredths of a millimeter; on a long shaft the axis visibly wanders.

The problem seldom lives in a single place. It grows along the chain of setups. The first setup creates an initial axis, the second shifts it a little, the third inherits that shift and adds its own. Each step looks acceptable on its own, but the final coaxiality doesn’t hold.

The worst part is that dimensions can still be within tolerance. Journal diameters, step lengths and bore depths will be correct relative to the new datum. But checked against the working axis, you may find the bearing seat, a thread or mating bore no longer line up.

On CNC lathes this is especially deceptive. The machine repeats the program exactly, so it’s easy to assume the problem isn’t fixturing. But the machine cuts where you defined the axis with the clamp. If the second setup changed that axis, the program merely locks the error into the part.

Losing the datum after the second setup is not a rare fault but a typical result of a small change of support. The error usually starts unnoticed: the part sits conveniently, dimensions look right, and the overall axis has already shifted.

Where the error accumulates

The error rarely appears from one big cause. More often it accumulates bit by bit at each transition and by the second setup gives a noticeable offset. So the problem is usually created before cutting: when the part is removed, cleaned, flipped and clamped again.

The most common cause is simple: dirt on the datum. Fine swarf, oil plus dust, or a small burr under the face easily shift the seating by hundredths. It’s barely visible to the eye. The machine then repeats the incorrect seating and the error propagates through the operation.

Clamping force behaves the same way. One time the part was tightened lightly, the next time more firmly, and it sits differently. A thin shaft will bend slightly; on a housing the position relative to the face, prism or mandrel changes. Individual dimensions may still meet tolerance, but the overall axis moves.

When an operator flips a part they almost always change contact points. Even if the datum seems the same, contact occurs on different spots. This is clear on housings after roughing, where the surface is uneven. The part can look correct but actually support itself differently.

Switching to different tooling also easily breaks the common datum. The part might be in the chuck for the first operation and placed in a mandrel or between centers for the second. Each fixture imposes its own position. If the process plan hasn’t linked these transitions, errors simply add up.

The tooling itself can be at fault. Worn jaws, a mandrel with runout, a damaged center or tired chuck fingers introduce offsets even when the operator acts carefully. In series production such small faults quickly become repeated defects.

A simple example shows it immediately. On a shaft you machine one journal, then flip the part for the second. If the face wasn’t cleaned and the second clamping was tighter, the two journals will no longer align on the same axis. For a housing the same happens when boring two holes from different setups. Each small detail seems tolerable alone, but together they cause the drift that later people hunt down with an indicator.

How to choose the datum in advance

You decide the datum not when the part is already in the chuck, but earlier—at the routing stage. If you don’t, the first operation will pass quietly and then the axis will start wandering from small things: different clamp, different support, different zero.

For a shaft a simple rule usually works: keep one axis until the end of machining. Most often this is the centers or a single reliable journal that can be repeated. If the first journal was made “as it comes,” you can’t confidently use it afterwards. The error looks small, but in the next transition it combines with chuck runout and eats into the size allowance.

For housings the logic is similar, but the datum is often a plane and a bore. Choose surfaces that the operator can easily find and repeat at each setup. A good datum for a housing isn’t the nicest surface after the first operation, but the one that consistently produces the same position without long shimming.

Another common mistake is removing all stock at once and then realizing that surface is needed as a support later. If the datum will be required in a subsequent operation, leave a reasonable stock on it or keep an area that the finishing cut won’t touch prematurely.

A practical order is simple: first pick the surface or axis you can repeat across setups, then check it won’t disappear after the first operation, next reduce the number of re‑setups, and only then distribute stock removal across operations.

On CNC lathes this is very noticeable with long shafts. If you started between centers, keep that method as long as possible. For housings, it’s often better to establish one stable plane and a primary bore, then build the rest of the operations from those.

A good datum saves more than inspection time. It removes hesitation from the operator about where to register, what to control and which surface to trust after the second setup.

How to keep the datum without extra measurements

The datum usually shifts not from one big mistake but from several small moves between setups. The operator moves a stop, checks a different surface, clamps the part a little differently and the axis creeps.

The simplest way to hold the datum is to tie both clamping and control to the same surface. If in the first operation you used a shaft journal or a housing bore as the datum, don’t switch to the neighboring diameter on the next operation just because it’s easier to reach the indicator. It’s easier now, but takes longer overall.

Checking the whole part around isn’t always necessary. Often one control point is enough if it’s in the right place and repeated every time. For a shaft it’s sensible to check runout on one finished journal that defines the fit. For a housing use one bore rather than running the indicator over multiple surfaces and adding noise.

Change the clamping scheme only when unavoidable. If the part originally sat on a face and registered the same basic surface, keep that principle. When the second setup moves from a face stop to clamping on the outer diameter, variation grows even if the operator is careful.

A stop that mimics the real assembly seating works well. Not a temporary pin, but a support that has the same meaning as the finished assembly. The part then finds a predictable position and the operator doesn’t have to hunt for zero with the indicator every time.

It helps to fix the sequence of actions for the operation: clean the datum, seat the part in the stop, clamp in the same order, then check runout at one point. If the sequence isn’t recorded, each operator will build their own version of the process and errors begin to accumulate even on identical machines.

When shaft and housing fixturing is set up this way, extra measurements almost disappear. What’s needed isn’t frequent checking but repeatable checking.

Order of transitions between operations

Sequence matters more than extra measuring. If you change setups without a clear reference to the same datum, the error quietly grows: hundredths at each step until the axis noticeably wanders.

First machine the surface that will become the working datum. For a shaft this is often a journal, center or face from which lengths are taken. For a housing it may be the locating bore or bearing face. Don’t try to take all dimensions in the first setup. First create a support you can trust.

Then take the dimensions that are tied to that datum in the same setup. This order maintains coaxiality better than a scheme where some features reference a rough surface and others reference a finished one. When the datum is consistent, the operator can repeat the setup and the inspector can more easily find the source of any offset.

A usual working sequence is:

1. Prepare and machine the working datum.
2. In the same setup, machine dimensions referenced to that datum.
3. Before re‑clamping, clean the datum, stops and clamping tools.
4. Repeat the same clamping scheme and clamping force.
5. Check runout and only then start cutting.

Cleaning before the second setup seems trivial, but it’s often where accuracy is lost. One chip on a stop, a thin oil film or a burr changes the seating more than many expect. So before a new setup, wipe the datum, blow out the tooling and quickly check for marks from the previous clamp.

Clamping force must also be repeatable. If you held a shaft lightly the first time and then overtighten on the second, the part may bend slightly. For housings a weak clamp can allow micro‑movement during cutting. It’s better to keep the same clamp regime and contact points than to adjust by feel each time.

Do a short runout check before cutting. It takes less than a minute and immediately shows if the part seated correctly. If the indicator shows a shift, don’t start machining hoping the cutter will fix it. It won’t; it will only carry the error further down the process.

Examples for a shaft and a housing

Errors usually don’t appear immediately as large displacements. They build gradually: one flip, a different datum, an extra clamp, and the axis wanders. This is clearly seen on shafts and housings.

For a shaft a straightforward order usually works. First center and rough the initial journal to establish a trusted axis. Leave a small but sufficient stock for the finish pass.

If you then re‑clamp the shaft on the same axis, coaxiality holds much better. Practically this means: don’t look for a new datum where one already exists. If you started from centers, finish the journal from those same centers rather than clamping the shaft on a random external diameter.

A common innocent error is flipping the shaft for easier tool access and thinking one short pass will fix it. In reality that flip often introduces more offset than another finish pass without changing the setup.

For housings the logic is similar but the datum differs. First establish a locating plane. It gives a clear stop and removes much uncertainty for the next operation.

Then bore the mating hole from the finished plane. The hole’s axis will be tied to the part, not to a random position in the jaws. If there is a second bore or a face, it’s better to produce them from the already finished datum rather than start a new chain of measurements.

A good example is a housing with two coaxial bearing bores. If the first bore is made after establishing the locating plane, the second bore should be done in the same setup or with support on the finished first bore and the same plane. Adding an extra flip between them quickly creates axis divergence even if individual dimensions remain in tolerance.

So on shafts maintain one axis, and on housings first create a plane from which to build bores and faces. This order seems simple but often removes half the extra checks on the shop floor. When the datum isn’t changed without reason, fewer measurements are needed and results are steadier.

Mistakes that shift the axis

After the second setup the axis usually shifts not from one big fault but from several small decisions. Separately they seem harmless; together they produce a displacement visible at assembly or during finishing.

The first common mistake is choosing a non‑repeatable datum simply because it’s convenient. The part quickly sits in the chuck or prism, but next time it sits slightly differently. If the datum doesn’t hold the same position from operation to operation, the axis wanders even when drawing dimensions stay in tolerance.

Clamping after a flip causes no less trouble. On the first setup the part was held gently, on the second it was tightened more. A thin sleeve, a long shaft or a housing with uneven stiffness will deform under clamping. The operator measures a diameter and sees a correct value, so thinks everything is fine, while runout has already increased.

Dirt on the tooling also moves the axis faster than expected. One small burr on a jaw, swarf in a prism or film on a face can tilt the part by a few hundredths. That may be acceptable for roughing, but not for a bearing seat or a pair of coaxial bores.

Another typical mistake is mixing rough and finish datums without a plan. First the part is referenced from a cast face or an unmachined diameter, then suddenly you switch to a finished journal or bore without checking how the datums relate. Errors then stack: each operation holds its own datum, but the part’s overall axis drifts.

It also doesn’t help to look only at size. Diameters and lengths can be correct while no one checks runout or the axis position after transitions; scrap hides until the end.

Before the second setup it’s useful to check five things: cleanliness of the jaws, cleanliness of the stop, clamping force, whether the chosen datum matches the process plan, and runout on the control surface. That takes a couple of minutes and saves the trouble of finding the operation where the axis shifted by 0.03 mm.

Short checklist

Before the second setup spend two minutes on a simple check rather than chasing an offset later with an indicator and rework. Most often the axis shifts due to small things nobody checked before clamping.

It’s easier to follow the same short routine every time. The operator won’t have to remember what to check first and will simply follow the sequence.

First inspect the datum. There should be no chips, dents or small burrs on the support. Even a thin speck can cause a noticeable offset, especially on a long shaft or a housing with a bore.

Then check the tooling state. Jaws, prism, mandrel or chuck should not have play. If the clamp moves by hand, further measurement is nearly pointless.

Next verify the clamping scheme. If the previous step used one datum and now you clamp on a different surface without reason, the axis can shift by tens of microns or more.

Choose the indicator control point before clamping, not after. That way the measurement shows the feature you need to hold between setups, not a convenient place you can reach with the dial stand.

Finally, the operator must understand which single size defines the datum. Not the entire drawing at once, but one concrete dimension or fit that you cannot lose the axis for.

On the shop floor this works simply. For a shaft check the cleanliness of the center hole or the locating journal, then verify clamp repeatability. For a housing the common error is the bore datum: a surface looks clean, but a thin burr from the previous step already lifted the part.

If you use CNC lathes or machining centers, the machine won’t save you from this mistake. It will execute the program precisely, but only from the datum you gave it. So the short checklist before start is often more useful than one more measurement after machining.

What to do next on the shop floor

Focus first on the parts where the axis drifts most often: long shafts after flipping, housings with bores from both sides, and parts where the initial datum cannot be used directly later. If coaxiality changes from batch to batch, the cause is almost always the route, the datum or the clamping.

Pick 3–5 problematic items and walk the part’s path end to end. Don’t rely on memory—watch how it’s clamped, where it seats, which surface defines dimensions and where the operator has to find the position by hand. This step shows where actual fixturing differs from the process plan.

For each operation check four things: what datum the routing sheet specifies and what is actually used in the machine, whether the clamping scheme changes between operations without reason, whether there’s a fixture element that repeats the part position consistently, and whether the machine holds repeatability across a series, not just on a trial part.

If the paper says one datum but the shop clamps a different surface, the error accumulates quietly. On a shaft this often shows as increased runout after the second turning. On a housing the problem emerges later when bores are finished but the overall axis doesn’t line up.

Check tooling separately. Worn jaws, a floating stop, dirt under the support, weak clamping and play in the fixture cause shifts that the operator may not notice. Be realistic about the machine too: if it doesn’t hold repeatability, extra measurements won’t save you—they’ll only slow production.

How to formalize the next step

After the check make a short plan for the shop. For each problematic part record one fixed datum, one acceptable clamping scheme and one quick check after re‑clamping. Often that’s enough to remove redundant measurements and make transitions calmer.

If the cause is equipment capability rather than operator discipline, address it immediately. For these cases the experience of EAST CNC can help: the company supplies CNC lathes and machining centers and assists with selection, commissioning and service. east-cnc.kz also has a blog with equipment reviews and practical machining materials, which is useful when you need to align your approach to fixturing, datuming and routing.