Time To Balance — The Technical Details

The complete physics behind the BS-8 Balancing Stand and the stopwatch method. This is the page for people who want to know why it works, not just that it does. Nothing here is required to balance a wheel — but everything here is free, and we think you'll enjoy it.

1. The Idea in One Paragraph

An unbalanced grinding wheel resting on level, precision ways is a pendulum. Displace it and it oscillates, and the period of that oscillation depends directly on the amount of imbalance: the better the balance, the weaker the gravitational restoring torque, and the longer the period. A stopwatch therefore becomes a balance-measuring instrument. Better still, the two error sources that cripple the traditional static ("mark the heavy side") method — rolling friction and imperfect leveling — corrupt the period only at second order. Timing one full swing cycle is a more honest measurement than watching where the wheel comes to rest, and it keeps working long after the static method has gone blind.

2. Pendulum Fundamentals

2.1 The simple pendulum

A point mass $m$ on a massless arm of length $L$, swinging through small angles:

$$T = 2\pi\sqrt{\frac{L}{g}}$$

The mass is absent from the formula. Gravitational torque and rotational inertia both scale linearly with $m$, so it cancels — the same reason all masses free-fall at the same rate.

2.2 The compound (physical) pendulum

A rigid body of mass $m$ pivoting about a fixed axis, with its center of mass a distance $d$ from the pivot:

$$T = 2\pi\sqrt{\frac{I_p}{m\,g\,d}}$$

where $I_p$ is the moment of inertia about the pivot. Mass no longer cancels in general, because the mass distribution ($I_p$) and the CM offset ($d$) are now independent parameters.

2.3 The unbalanced wheel about a fixed axis

Model the wheel as a perfect disk of mass $M_{disk}$ and radius $R$, plus a small error mass $m_e$ at radius $r$ from the geometric axis, everything symmetric about the wheel's midplane (single-plane imbalance). If the wheel could pivot frictionlessly about its own fixed geometric axis, the perfect disk would contribute no restoring torque (its CM is on the axis); only the error mass would:

$$T = 2\pi\sqrt{\frac{\tfrac{1}{2}M_{disk}R^2 + m_e r^2}{m_e\, r\, g}}$$

Note the character of this expression: the disk's inertia dominates the numerator while the restoring torque comes entirely from the imbalance. As $m_e r \to 0$, the period diverges. A perfectly balanced wheel is a pendulum with an infinite period. That divergence is the entire working principle of the method — the stopwatch reading blows up exactly as the thing you're trying to eliminate goes to zero.

3. The Real Geometry: Rolling on Ways

The BS-8 does not pivot the wheel on a fixed axis. The wheel rides on an arbor of radius $a$, and the arbor rolls on two level ways. This changes the mechanics in an important way: the rotation axis is no longer fixed in space. As the assembly rocks through angle $\theta$, the rolling constraint translates the arbor center horizontally by $a\theta$, and the instantaneous center of rotation is the moving contact line between arbor and ways. The right tool is the Lagrangian, not torque-about-a-pivot.

3.1 Setup

Let $M$ be the total rolling mass (wheel + hub + arbor + collet), and let $\varepsilon$ be the offset of the assembly's center of mass from the geometric arbor axis. For a single error mass,

$$\varepsilon = \frac{m_e\, r}{M}$$

Define $\theta = 0$ as the equilibrium with the heavy spot at bottom dead center. With rolling without slip, the CM coordinates are:

$$x = a\theta - \varepsilon\sin\theta, \qquad y = a - \varepsilon\cos\theta$$

3.2 Energies and equation of motion

$$KE = \tfrac{1}{2}\Big[I_{cm} + M\big(a^2 + \varepsilon^2 - 2a\varepsilon\cos\theta\big)\Big]\dot\theta^2$$

$$U = -Mg\,\varepsilon\cos\theta$$

Applying Lagrange's equation gives the full equation of motion:

$$\Big[I_{cm} + M\big(a^2 + \varepsilon^2 - 2a\varepsilon\cos\theta\big)\Big]\ddot\theta \;+\; Ma\varepsilon\sin\theta\,\dot\theta^2 \;+\; Mg\,\varepsilon\sin\theta = 0$$

Two features distinguish this from the fixed-pivot pendulum: the effective inertia is $\theta$-dependent (the bracketed term "breathes" as the CM swings toward and away from the contact line), and the $\dot\theta^2$ term appears because of that changing inertia. Both are nonlinear effects that vanish at small angle — and both are numerically negligible when $\varepsilon \ll a$, which is always true near good balance.

3.3 Small-angle period

Linearizing about $\theta = 0$:

$$\boxed{\;T = 2\pi\sqrt{\frac{I_{cm} + M\,(a-\varepsilon)^2}{M\,g\,\varepsilon}}\;}$$

The parallel-axis term uses $(a-\varepsilon)$, the distance from the contact line to the CM at equilibrium. In practice it is cleaner to work with $I_{axis}$, the moment of inertia about the geometric arbor axis (the quantity you would naturally compute from the wheel's dimensions). Using $I_{cm} = I_{axis} - M\varepsilon^2$:

$$T = 2\pi\sqrt{\frac{I_{axis} + M\,a\,(a - 2\varepsilon)}{M\,g\,\varepsilon}} \;\approx\; 2\pi\sqrt{\frac{I_{axis} + M a^2}{M\,g\,\varepsilon}} \qquad (\varepsilon \ll a)$$

Physical reading: the restoring torque about the contact line, $Mg\varepsilon\sin\theta$, is identical to the fixed-axis case — gravity's lever arm is still just the horizontal CM offset. The arbor radius enters only as added inertia, $+Ma^2$, which for a ½-inch arbor under a 6–8 inch wheel is about 1% of the total. Rolling on ways, in other words, costs almost nothing dynamically — while buying an enormous reduction in friction compared to any bearing (we tried bearings; the friction was hopeless).

3.4 Inverting: from stopwatch to imbalance

Solving the boxed formula for the imbalance (with $I_{eff} = I_{axis} + Ma^2$):

$$\varepsilon = \frac{4\pi^2\, I_{eff}}{M\,g\,T^2} \qquad\Longleftrightarrow\qquad U \equiv m_e r = M\varepsilon = \frac{4\pi^2\, I_{eff}}{g\,T^2}$$

where $U$ is the imbalance in the units balancing machines use (g·mm, oz·in). Note the $1/T^2$: the period scale is quadratic. Doubling the measured time means the imbalance fell by a factor of four. Going from a typical fresh-mount reading of 3 seconds to the 25-second criterion means the imbalance was reduced by a factor of $(25/3)^2 \approx 70$ — essentially all of it.

4. Table of Variables

Variable	Description
$T$	Period of one full small-angle oscillation cycle (the stopwatch reading)
$g$	Gravitational acceleration, 9.81 m/s²
$\theta$	Angular displacement from equilibrium (heavy spot at bottom dead center)
$M$	Total rolling mass: wheel + hub + arbor + collet
$M_{disk}$	Mass of the idealized perfect (balanced) wheel
$m_e$	Error (imbalance) mass
$r$	Radius of the error mass from the geometric axis
$U = m_e r$	Imbalance, mass–radius product (g·mm)
$R$	Wheel radius
$a$	Arbor radius (¼ inch = 6.35 mm on the BS-8)
$\varepsilon$	CM offset from the geometric axis; $\varepsilon = U/M$
$I_{cm}$	Moment of inertia about the assembly CM
$I_{axis}$	Moment of inertia about the geometric arbor axis
$I_{eff}$	$I_{axis} + Ma^2$; effective inertia for rolling oscillation
$\beta$	Tilt of the ways from level (radians)
$\kappa$	Curvature (slope variation per unit length) of the way surfaces
$b$	Rolling-resistance parameter, units of length (hardened ground steel on steel: a fraction of a micron)
$\theta_0$	Release amplitude (the procedure uses 90° = $\pi/2$)
$\omega$	Wheel operating speed, rad/s
$G$	ISO 21940 balance quality grade, $G = \varepsilon\,\omega$ (mm/s)

5. Worked Example: Where 25 Seconds Comes From

Take the worst case the BS-8 is rated for: an 8 × 1 inch vitrified aluminum-oxide wheel, 1¼-inch bore (density ≈ 2.4 g/cm³).

Wheel mass: ≈ 1.9 kg. With hub, arbor, and collet: $M \approx 2.3$ kg.
$I_{axis} \approx \tfrac{1}{2}m_{wheel}(R^2 + r_{bore}^2) \approx 0.0102$ kg·m². The $Ma^2$ term adds ≈ 9×10⁻⁵ kg·m² — about 1%.

At $T = 25$ s:

$$\varepsilon = \frac{4\pi^2 (0.0103)}{(2.3)(9.81)(625)} \approx 29\ \mu\text{m}$$

$$U = M\varepsilon \approx 66\ \text{g}\!\cdot\!\text{mm} \quad(\approx 0.65\ \text{g at the rim})$$

What does that mean on the grinder? At a typical 3450–3600 RPM spindle ($\omega \approx 360$–$377$ rad/s):

Balance quality grade: $G = \varepsilon\omega \approx 10$–$11$ mm/s — i.e., roughly G10, well inside the range commercial standards assign to machine-tool grinding wheels in this class.
Residual centrifugal force: $F = U\omega^2 \approx 9$–$10$ N — about two pounds of once-per-rev force. Against the stiffness of a surface-grinder spindle system (order $10^8$ N/m), that drives synchronous motion at the wheel of well under 100 nanometers — below the finish floor of ordinary work.

One number worth appreciating: the remaining once-per-rev geometric runout gets removed by truing on the machine. Balance only has to handle the force. A wheel at 25 seconds contributes force-driven motion your grinder cannot show you.

5.1 Why one time threshold covers all wheel sizes

Since the disk inertia dominates, $T^2 \propto (R^2)/\varepsilon$, so a fixed 25-second criterion corresponds to a threshold offset $\varepsilon \propto R^2$. Meanwhile grinding wheels run at roughly constant surface speed, so $\omega \propto 1/R$, and the achieved quality grade is

$$G = \varepsilon\,\omega \propto R$$

The 8-inch wheel is therefore the worst case: every smaller wheel that passes 25 seconds is balanced to a proportionally better grade. Calibrating the criterion at the capacity limit makes it automatically conservative everywhere below it. "More is better; 25 seconds is enough" is one rule, and it is the right rule for the whole 6–8 inch range.

6. Why the Period Beats the Static Method

The traditional technique — let the wheel settle, mark the heavy side, add weight, repeat until it "stops indicating" — fails at exactly the moment it matters, and it fails at first order. The period method demotes the same error sources to second order. This section is the heart of the argument.

6.1 Way tilt: first order statically, second order in the period

Tilt the ways by a small angle $\beta$ and the rolling assembly feels a bias torque about the contact line of approximately $Mga\beta$. Statically, this shifts the rest position and masks any imbalance smaller than

$$U_{min}^{static} \sim M\,a\,\beta$$

For a 2.3 kg assembly on a ¼-inch-radius arbor, each milliradian of tilt hides ≈ 15 g·mm — a quarter of the entire 25-second budget. That is a first-order catastrophe, and it is why static balancing demands obsessive leveling.

Now consider the period. The tilt shifts the equilibrium angle to $\sin\theta_{eq} = a\beta/\varepsilon$, and the effective stiffness at the new equilibrium becomes $Mg\varepsilon\cos\theta_{eq}$. The period error is

$$\frac{\Delta T}{T} \approx \frac{1}{4}\left(\frac{a\beta}{\varepsilon}\right)^2$$

Quadratic. A tilt that shifts the static rest position by ten degrees — ruining a static measurement — changes the period by less than one percent. The no-roll leveling procedure (adjust the screw until the bare arbor stays put) gets $\beta$ far below even that.

6.2 Rolling friction: kills the static method, leaves the period alone

Real rolling contact has a resistance characterized by a length parameter $b$ (for hardened, ground steel on steel, a fraction of a micron). Statically, friction creates a dead band: the wheel can come to rest anywhere within

$$|\theta_{rest}| \lesssim \frac{M b}{U} = \frac{b}{\varepsilon}$$

When $\varepsilon$ approaches $b$, the dead band spans all angles — the wheel stops wherever it happens to stop, and the static method returns no information. This is the "it stopped telling me anything" wall every user of a static stand has hit.

The period responds differently. Coulomb-type friction decays the oscillation amplitude linearly — a loss of roughly $4b/\varepsilon$ radians per cycle — but to first order it does not shift the period. (This is a classical result: dry friction terminates the pendulum's swing but does not detune its clock.) Friction's only real cost to the method is that the oscillation must survive long enough to time. The number of cycles available before stall is roughly

$$N \approx \frac{\theta_0\,\varepsilon}{4b}$$

At the 8-inch threshold ($\theta_0 = \pi/2$, $\varepsilon \approx 29\ \mu$m, $b \approx 0.25\ \mu$m): $N \approx 45$ cycles available, of which the procedure needs one. Timing from the first reversal to the third reversal is deliberately frugal with amplitude — and it also discards the release transient, so any wobble or translation imparted by the operator's hand dies before the clock starts.

6.3 Way straightness

A uniform tilt is second order, but curvature of the ways is not: if the local slope varies as $\beta(x) = \kappa x$, then since the contact point translates ($x = a\theta$), the bias torque becomes angle-dependent — a spurious stiffness $Mga^2\kappa$ that masquerades as imbalance:

$$\varepsilon_{apparent} \sim a^2\kappa$$

This is why the BS-8's ways are precision-ground shafting rather than machined or printed surfaces. Over the ±90° swing, the arbor travels only $\pm a\theta_0 \approx \pm 10$ mm along ways whose straightness is measured in microns over their full 6-inch length; the resulting spurious $\varepsilon$ is far below the 25-second threshold.

6.4 Arbor roundness

Lobing of amplitude $\delta$ on the arbor makes the axis height vary with roll angle even for a perfectly balanced wheel — a parasitic potential of order $Mg\delta$, equivalent to a spurious imbalance $\varepsilon \sim \delta$. Precision-ground shafting holds $\delta$ to a fraction of a micron, i.e., 1–2% of the 25-second threshold. This — not the plastic structure — is the true systematic floor of the instrument, and it is why all three rods (two ways and the arbor) are the same hardened, ground, 8-microinch-finish shafting. The printed structure never touches the measurement; its job is to hold two rods parallel and still.

6.5 Amplitude correction

The pendulum period grows with release amplitude:

$$T(\theta_0) \approx T_{0}\left(1 + \frac{\theta_0^2}{16} + \frac{11\,\theta_0^4}{3072} + \cdots\right)$$

At the prescribed 90° release this correction is about +18% on the earliest cycles, decaying with the amplitude. Two things neutralize it. First, the 25-second criterion was established empirically using the same procedure — the correction is baked into the number. Second, because the pass threshold is quadratically generous, a ±15% period variation corresponds to only ±30% in $\varepsilon$, i.e., a fraction of one "correction round" with the balancing ring. The practical rule is simply consistency: release from 90° every time, time the same cycle every time. (If you ever use the stand quantitatively — comparing 26 s against 28 s across sessions — control the amplitude or extrapolate to zero; see §7.)

The reversal is also the correct place to click the stopwatch for a human: it is the moment of zero velocity, the easiest event to judge by eye. A ±0.2 s judgment error against a 25 s reading is under 1% — noise.

6.6 The honest floor

Collecting the pieces, the period method's detectability floor is set by:

Stall — friction ends the oscillation before one cycle completes when $\varepsilon \lesssim 4b/\theta_0$ (well below a micron of CM offset here);
Spurious stiffness — way curvature and arbor lobing, $\varepsilon_{floor} \sim a^2\kappa + \delta$ (a fraction of a micron with ground shafting);
Environment — as $\varepsilon$ shrinks, $T \propto 1/\sqrt{\varepsilon}$ grows, and a multi-minute oscillation is driven by micro-newton-millimeter torques that drafts and vibration can rival.

The 25-second criterion sits more than an order of magnitude above floors 1 and 2, in $\varepsilon$ terms. The instrument is not the limit; the criterion is deliberately conservative against the instrument.

7. Quantitative Use: The Trial-Mass Method

The pass/fail procedure never needs $I_{eff}$ or $M$. But if you want an absolute number for $U$ in g·mm without computing the wheel's inertia, two timings and a known trial mass suffice.

Measure the period $T_1$. Add a known trial mass $m_t$ at radius $r_t$, aligned with the marked heavy spot (so imbalances add), and measure $T_2$ ($T_2 < T_1$). Since $U T^2 = 4\pi^2 I_{eff}/g$ is the same constant in both measurements:

$$U_1\,T_1^{\,2} = \big(U_1 + m_t r_t\big)\,T_2^{\,2} \qquad\Longrightarrow\qquad \boxed{\;U_1 = m_t r_t\,\frac{T_2^{\,2}}{T_1^{\,2} - T_2^{\,2}}\;}$$

The unknown inertia cancels entirely. A single ¼-20 setscrew of known mass in the B-200 ring at its known radius is a calibrated trial mass — the correction hardware doubles as the calibration standard.

(The period alone is blind to the angular location of the residual imbalance; magnitude only. The static settle supplies the location — and at 25 seconds and beyond, friction randomizes the rest position anyway, which is just another way of saying the residual no longer matters.)

8. Stated Limitations

We claim the BS-8 plus Time To Balance is an 80–90% solution, and we mean that literally:

Single-plane (static) imbalance only. The method measures the CM offset in the wheel's midplane. A couple imbalance — equal and opposite heavy spots on the two faces of a wide wheel — produces no CM offset and is invisible to any gravity-based method. For wheels an inch wide or less running on small grinders, couple imbalance is a second-order concern; for wide wheels at high speed it is why dynamic balancing machines exist.
No phase from the period. Magnitude comes from the stopwatch; the heavy-spot location comes from the static settle early in the procedure, when imbalance is still large enough to indicate reliably.
Environment matters at the extremes. Chasing 60-second periods in a drafty shop is a losing game. 25 seconds is not at the extremes.
The criterion is calibrated at 8 inches. Larger wheels would need a longer threshold (and a bigger stand — sorry, 12-inch guys).

The comparison that matters is not against a $30,000 dynamic balancer. It is against the prevailing practice on small grinders, which is no balancing at all. Between those two conditions lies nearly all of the finish improvement, the chatter reduction, and the spindle-bearing life — and a stopwatch will take you there.

9. Summary

	Static method	Time To Balance
Measures	Rest position (heavy side)	Period of oscillation
Output	A direction	A number
Way tilt error	First order — masks $U \sim Ma\beta$	Second order — $\Delta T/T \sim (a\beta/\varepsilon)^2/4$
Rolling friction	First order — dead band $\sim b/\varepsilon$	No first-order period shift; only limits cycle count
Endpoint	"It stopped telling me anything"	"25 seconds. Done."
Scale	—	Quadratic: 2× the time = 4× the balance

$$T = 2\pi\sqrt{\frac{I_{eff}}{M\,g\,\varepsilon}} \qquad\qquad U = \frac{4\pi^2\,I_{eff}}{g\,T^{2}}$$

More is better. 25 seconds is enough.

Updated 3 JULY 2026 - Revision 1.0

TTB