Natural frequency, chatter and why a stiff frame can still ring

A frame can pass every static deflection target and still produce a machine that chatters. Static stiffness answers the question “how far does it move under a steady load”. Chatter is a dynamic problem, and the relevant question is “what does it do when the load arrives at a frequency close to one of its modes”.

Machine builders discover this in a specific and expensive way: the prototype cuts beautifully in light passes and howls at the depth of cut the machine was sold on. Dynamics is one decision in our guide to machine frame design.

The mechanism, briefly

Regenerative chatter is a feedback loop. The tool leaves a wavy surface. On the next revolution or the next tooth pass, the tool encounters that waviness, so the chip thickness varies, so the cutting force varies. If the structure responds to that varying force with a vibration that is out of phase with the existing waviness, it deepens it. The loop grows until something saturates — usually the tool leaving the cut.

The structure’s contribution to this loop is its frequency response: how much it moves per unit of force, at each frequency. Two properties govern that.

Natural frequency determines where the structure is vulnerable. A mode at 90 Hz means the loop can close when the tooth passing frequency lands near 90 Hz or a sub-multiple of it.

Damping determines how badly. At resonance, the response amplitude is roughly the static compliance multiplied by the amplification factor, which is inversely related to damping. A lightly damped structure can move ten or thirty times more at resonance than the same force would produce statically.

Why stiffness alone does not settle it

Natural frequency goes as the square root of stiffness divided by mass. So adding stiffness raises the frequency — good — but adding material to achieve that stiffness also adds mass, which lowers it. The two partly cancel.

What actually raises natural frequency is stiffness added without proportionate mass: depth instead of thickness, closed sections instead of open, a diagonal instead of a thicker wall. This is the same geometric argument as in rib and box-section design for stiff machine frames, and it pays twice — once statically and once dynamically.

Adding plate thickness everywhere raises stiffness and mass roughly together and can leave the natural frequency almost where it was. This is why “we made it heavier and it still chatters” is such a common report.

Moving the mode versus damping the mode

Two strategies, and they suit different situations.

Move the mode out of the excitation range. If your spindle runs 2,000 to 8,000 rpm with a four-tooth cutter, tooth passing frequency runs roughly 133 to 533 Hz. A structural mode at 400 Hz sits squarely in that band. Raising it to 700 Hz, or lowering it to 80 Hz, takes it out of the loop.

Raising is usually preferred — a stiff, light structure with high modes. Lowering can work for very heavy machines where the modes are far below the excitation and the structure behaves as a mass.

Damp the mode. If the mode cannot be moved, reduce the amplification at resonance. Options, roughly in order of cost-effectiveness on a welded frame:

• Fill the box sections. Sand, concrete, or epoxy-aggregate inside the hollow members. Cheap, effective, widely used, and it adds damping through internal friction in the fill and at the fill-to-wall interface. Frequently the best return available on a steel weldment.
• Bolted joints with controlled preload. Friction at a bolted interface dissipates energy. A structure with well-designed bolted joints damps better than the same structure fully welded — one of several reasons the bolted versus welded question is not obvious.
• Constrained-layer damping. A viscoelastic layer sandwiched between the structure and a constraining plate. Very effective on panels that ring, adds little mass, costs more to engineer.
• Tuned mass dampers. A small mass on a spring tuned to the problem mode. Powerful when one specific mode dominates, useless against a broad problem, and it needs the mode to be identified accurately first.
• Change the material. Cast iron damps roughly an order of magnitude better than steel in the bulk, which is the historical argument for iron beds — see cast iron or welded steel. Mineral casting does substantially better again, at a cost discussed in mineral casting versus welded steel.

Where the modes usually are on a machine frame

Predictable troublemakers, worth checking on any new design:

• Column bending — the first mode on any machine with a tall column, usually the lowest and most consequential.
• Torsion of the base — twisting between the ends of a long base, common on open-fronted structures without a diagonal or closed section.
• Crossrail bending on a gantry — the beam between the columns, which carries the spindle mass at the worst possible place.
• Panel drumming — large unribbed plates acting as diaphragms. Rarely affects cutting directly, often makes the machine sound worse than it is, and occasionally couples into a structural mode.
• The machine on its foundation — a rigid-body mode on the mounts, typically low frequency. Often the actual culprit when a machine that tested well at the builder behaves differently at the customer.

That last one is worth emphasising because it is a frame supplier’s problem only indirectly. The foundation, the levelling mounts and the grout are part of the dynamic system. A machine that was quiet in the assembly hall and chatters on a customer’s slab has usually not developed a structural fault.

Modelling and measuring

Modal FEA gives you the mode shapes and approximate frequencies of a design before it exists. Treat it the way you treat static FEA — reliable for comparing options, approximate in absolute terms, because joint stiffness and boundary conditions are where the error lives. Model welded joints as joints, not as bonded continua, or you will predict frequencies significantly higher than reality.

Experimental modal analysis — tap testing with an instrumented hammer and an accelerometer — is the ground truth, and it is cheap. A morning of tap testing on a prototype frame tells you where the real modes are, which is the input every other decision needs. The stability lobe relationship that turns those measured modes into usable spindle speeds is well established, and the practical payoff is often immediate: a machine that chatters at 3,000 rpm may cut cleanly at 3,600 with no hardware change at all. How to run the tap test is in impact hammer test on a machine frame.

What a frame designer can actually do

1. Close every section that carries torque.
2. Buy stiffness with depth and diagonals, not with wall thickness.
3. Keep mass low and put what mass exists where it does not hurt — away from the moving structure’s compliance.
4. Plan for fill in the main box sections, even if you do not use it on unit one. The access ports are free at design stage and expensive later.
5. Run a modal FEA early and check nothing sits in the tooth passing band.
6. Tap test the first frame, before the machine is built around it.
7. Design the bolted interfaces deliberately — preload and contact area are damping parameters, not just fastening details.

None of this costs much at design stage. All of it is expensive to retrofit into a machine that is already shipping.

If you have a frame design and want it reviewed for section closure, fill access and machining practicality before it goes out to quote, send the drawing through — the review comes back within 24 business hours.