CNC Milling Regenerative Chatter Stability Lobe Diagram Simulator

Parameters

Spindle speed N

rpm

Current operating point — shown as red × on the lobe diagram

Teeth count z

Tool natural frequency f_n

First resonance of the tool tip, measured by Tap test

Modal stiffness k

N/m

Modal stiffness at the tool tip. Shorter stickout dramatically raises it

Damping ratio ζ

Plain tools 0.02-0.05; tools with internal damper 0.08+

Specific cutting force K_t

N/mm²

Al ≈ 800, steel ≈ 2000, Ti ≈ 3000, hardened steel ≈ 4000

Radial engagement a_e/D

Full slotting = 1.0; finishing pass 0.05-0.10

Results

—

Stable axial depth b_lim_min (mm)

—

Tooth-pass freq. f_TP (Hz)

—

Chatter frequency f_c (Hz)

—

Optimum spindle speed (rpm)

—

Material removal rate (cm³/min)

—

Modal stiffness k (N/m)

—

Machining view — end mill and chip-thickness wave

The end mill rotates and each tooth meets the wavy surface left by the previous tooth. Colour shows whether the current point is stable (blue) or unstable (red).

Stability Lobe Diagram — b_lim vs spindle speed N

Tool frequency response — |G(f)| and Re[G(f)]

Theory & Key Formulas

$$b_{lim,min} = \frac{-1}{2\,K_t\,G_{min}\,\rho},\qquad G_{min} = \frac{-1}{4\,k\,\zeta\,(1+\zeta)}$$

The stable axial depth b_lim_min is set by the most negative real part of the tool FRF (Tlusty/Altintas). K_t: specific cutting force, ρ: radial engagement ratio, k: modal stiffness, ζ: damping ratio.

$$G(j\omega) = \frac{1/k}{1 - r^{2} + 2\,i\,\zeta\,r},\qquad r = \frac{\omega}{\omega_n}$$

Single-mode tool frequency response function (FRF). r=1 is the resonance and the Re[G]<0 band (r>1 weak side) is where chatter is excited.

$$f_{TP} = \frac{N\,z}{60},\qquad N_{opt} = \frac{60\,f_n}{z\,(n+1)},\quad n=0,1,2,\dots$$

Tooth-passing frequency and sweet-spot spindle speeds. The lowest (n=0) lobe sits at f_TP=f_n and offers the largest stable depth of cut.

CNC Milling Regenerative Chatter Stability — Stability Lobe Diagram

🙋

People in the shop keep talking about "chatter" in end-mill cutting. How is it different from ordinary vibration, and why do they say raising the spindle speed can actually make it disappear?

🎓

Great question. That nasty high-pitched whine and the rough machined surface is "regenerative chatter", a self-excited vibration. The loop is: the previous tooth leaves a wave on the surface while vibrating → when the next tooth arrives, the chip thickness changes depending on the phase between its own vibration and the previous wave → the cutting force varies → the tool vibrates more → the wave gets deeper. It is positive feedback. Once the conditions are bad it explodes in seconds and the surface roughness jumps past Rz 50 µm. Tlusty cracked it in 1965.

🙋

So it is about the wave from the previous tooth meeting the current tooth's vibration. Then can we explain "raising the speed kills it" with the same phase logic?

🎓

Exactly — that is the core of the Stability Lobe Diagram. If you make the tooth-passing frequency f_TP = N·z/60 an integer ratio of the chatter frequency f_c, the previous wave and the current vibration come in phase and chip-thickness modulation cancels out. For f_n = 800 Hz and z = 4, N = 12000 rpm gives f_TP = 800 Hz = f_n — the sweet spot. There you can take 3-5× the safe minimum depth. Between sweet spots, around N = 9000 rpm, the phase is the worst and the lobe wall is highest. Move the slider on the chart on the right and you can see the valleys and walls.

🙋

So just spinning faster is not enough — you have to hit the valley. The default 5000 rpm is off the valley then?

🎓

Right, with the default N = 5000 rpm we get f_TP = 333 Hz, less than half of f_n = 800 Hz. It is on the wall side, even below the n=0 lobe — the conservative regime dominated by b_lim_min. The verdict on this tool comes out as "warn" for that reason. In practice you first run a Tap test to get f_n, then aim for N_opt = 60·f_n/z = 12000 rpm. Aluminium high-speed machining sometimes goes above 20000 rpm — but past a point spindle bearing life and active stiffness become the new bottleneck.

🙋

The "modal stiffness k" slider on the left — how does that differ from ordinary static stiffness? When I push it up, b_lim stretches a lot.

🎓

Important distinction. Static stiffness is "how many N/mm sinks under a slow push", but modal stiffness is set by the response amplitude when you shake the tool tip at its resonance f_n. As a rule modal stiffness is much smaller than static stiffness, and on a long or slim tool it can be 10× smaller. The fixes are (i) minimise stickout — stiffness goes as L³, (ii) use a heat-shrink or HSK holder for better joint stiffness, (iii) use damped tools to push ζ from 0.03 to 0.08+. Thin-wall titanium blade machining or aerospace honeycomb work always relies on damped carbide tools for that reason.

🙋

One last one: it is intuitive that a higher specific cutting force K_t lowers b_lim, but how do you measure K_t?

🎓

K_t is "cutting force per unit chip thickness" and it depends on the material and the tool rake angle. You measure it with a cutting-force dynamometer (Kistler etc.), capture F_t and F_n, and divide by the chip thickness to identify it. Typical book values are Al6061 ≈ 800, mild steel ≈ 1800-2200, SUS304 ≈ 2400, Ti6Al4V ≈ 2800-3200, hardened SKD11 ≈ 3500-4000 N/mm². Titanium is "hard to machine" because K_t is high and the thermal conductivity is low so the tool overheats. Try switching materials in the simulator and watch how the sweet spot moves.

Frequently Asked Questions

Regenerative chatter is a self-excited vibration produced when the wavy surface left by the previous tooth and the vibration of the current tooth overlap with a phase shift, periodically modulating chip thickness. The cutting force varies, which feeds back into more tool vibration. Tlusty (1965) and Tobias established the analytical model; Altintas generalised it in the frequency domain. Unlike friction-driven stick-slip, the cutting process itself acts as an amplifier, so once triggered it grows explosively in seconds, leaves chatter marks and slashes tool life. The Stability Lobe Diagram maps this on the spindle-speed vs axial-depth plane.

When the tooth-passing frequency f_TP = N·z/60 forms an integer ratio with the tool's chatter frequency f_c, the wavy surface from the previous tooth and the vibration of the current tooth align in phase, the chip-thickness variation cancels, and the stable depth of cut rises sharply. These are the valleys, the sweet spots. Between them the phase mismatch is worst, the regenerative effect is maximised, and the depth limit collapses (lobe walls). The lowest lobe (n=0) sits at f_TP ≈ f_c, i.e. N_opt = 60·f_n/z, and the stable depth there can be 2-5× the worst-case minimum. High-speed machining is the technique of hitting that sweet spot.

The standard procedure is a Tap test (impulse hammer test). With the tool mounted on the spindle, the operator taps the tool tip in the XY directions with an impulse hammer and an accelerometer records the response. FFT yields the frequency response function (FRF), from which the resonance peak gives the natural frequency f_n, the half-power width gives the damping ratio ζ, and the peak magnitude gives the modal stiffness k = 1/(2ζ·|G(f_n)|). Commercial software such as CutPro, MetalMAX and Harmonizer perform this in 5-30 minutes per machine-tool combination. Because each tool length and holder gives a different FRF, ideally every long/slim/tapered tool is measured separately.

(1) Tune the spindle speed to the sweet spot (the main purpose of this tool). (2) Reduce the axial depth of cut — most reliable but hurts MRR. (3) Use variable-pitch or variable-helix tools that scatter the frequency content and weaken regeneration. (4) Shorten the tool stickout — stiffness scales with L³. (5) Switch to a heat-shrink or HSK holder for better joint stiffness. (6) Use damped tools (internal mass dampers) — popular in aerospace titanium milling. (7) Adjust coolant flow and direction to suppress friction-driven excitation. Start with (1) and (4); these alone solve most cases.

Real-World Applications

Thin-wall aerospace structural parts: Pocketing wing spars, blades and airframe frames leaves final wall thicknesses of 1-2 mm milled with long tools. Modal stiffness is extremely low and chatter is the bottleneck. CAM packages (Mastercam, NX CAM) take Tap-test FRFs as input and automatically optimise N and b, generating NC paths that walk through the valleys of the lobe diagram. Al7050 examples report MRR jumping from 20 to 80 cm³/min.

Mould and die finishing: Finishing hardened SKD11/SKH51 (HRC 60+) with small ball end mills (Ø6-Ø3). K_t = 4000 combined with slender tools makes chatter unavoidable. With this simulator you would push radialEngagementRatio down to 0.05-0.10 and bump the spindle past 25000 rpm to ride a higher-order lobe valley. Side-step optimisation comes from Altintas-based packages such as CutPro.

Medical devices and orthopedic implants: Femoral heads and joint cups in Ti6Al4V or CoCrMo are 5-axis machined to ±10 µm with long stickouts. Titanium has K_t ≈ 3000 and a thermal conductivity of just 6.7 W/mK, so chatter goes straight to tool breakage. The standard answer is damped carbide end mills from Sandvik or Mitsubishi that push ζ above 0.08, plus minimum stickout.

Automotive mass production (engine blocks, transmission cases): Robot lines machining Al-Si (ADC12) alloy are dominated by cycle time. As this simulator shows, Al has K_t ≈ 800, so with enough stiffness you can run 15000-20000 rpm with deep cuts and reach MRR around 200 cm³/min. The latest MAG/MAZAK machines even include chatter-detection sensors and AI-based spindle-speed compensation (Active Chatter Suppression).

Common Misconceptions and Pitfalls

First, assuming the Stability Lobe Diagram solves everything is dangerous. The Tlusty/Altintas model is a single-mode, single-direction, linear analytical model. Reality includes (a) coupled tool-and-part systems where both vibrate, (b) the tool's natural frequency changing under centrifugal load at high rpm, (c) non-linear chip separation at deep cuts, (d) per-tooth force variation in multi-flute tools. Treat the prediction as "a first read on whether chatter converges in seconds" and pair it with real-time acoustic or accelerometer monitoring (Sandvik CoroPlus, Bluestreak Production Studio, etc.).

Second, doing the Tap test once and calling it done. The modal stiffness k and natural frequency f_n shift with stickout length, holder type, spindle-bearing wear and temperature. A FRF measured with a brand-new holder can be 10-30% off after six months on the same machine. Re-run Tap tests on a schedule (every three months, or with each tool swap) and update the lobe map on the CAM side. Even a 5 mm change in stickout inside the same collet can shift f_n by several tens of Hz.

Third, believing that hitting the sweet spot always raises productivity. The optimum N_opt may exceed your spindle's max rpm, cooling capacity, or the coating's heat limit. If the tool returns N_opt = 75000 rpm for f_n = 2500 Hz and z = 2 but your machine tops out at 20000 rpm, you simply cap there. And very high rpm wears tools exponentially faster — the gain in MRR may be eaten by tool cost. Always balance MRR improvement against tool and energy cost in a separate calculation. This simulator gives "physical stability", not "economic optimum".

CNC Milling Regenerative Chatter Stability Lobe Diagram Simulator

CNC Milling Regenerative Chatter Stability — Stability Lobe Diagram

Frequently Asked Questions

Real-World Applications

Common Misconceptions and Pitfalls

How to Use

Worked Example

Practical Notes