Ask anyone who grew up carrying things in the Chinese countryside and you will get the same sentence, delivered with total confidence and zero elaboration: a bamboo carrying pole is easier on you than a rigid one. 省力, in the one word they always use — not “more comfortable”, not “traditional”, but easier, as a claim about your body. My grandmother says it the way you would state that water is wet.
And the strange part is that the thing everyone points to as the reason is the thing that looks most like a defect. A carrying pole under load visibly bends, and while you walk it bounces — the buckets bob up and down, the whole assembly looks one pothole away from structural failure. If you handed the problem to a first-year engineering student they would call that compliance a bug and reach for a stiffer material.
So: is the folk claim actually true? Not “does it feel nicer”, but is there a mechanism, and does the mechanism have conditions? That is the kind of question you can hand to a differential equation, and this post is my attempt to do exactly that. (Full disclosure: I met this problem while hunting for an EE topic and could not climb back out. What follows is the part I think is genuinely beautiful, minus the forty pages of uncertainty propagation that only an IB examiner would ever want to read.)
The short version of the answer, so you know where we are going: the claim is true, but only past a sharp threshold, and most of the pole’s properties have nothing to do with where that threshold is. The threshold is a single pure number — $\sqrt2$ — independent of the bamboo, the length, the load, and even the internal friction of the material. It is a property of the equation, not of the pole. Below it, a flexible pole is not merely useless; it is measurably worse than an iron bar.
Let me build it.
1. What does “easier” even mean?
Here is the first problem, and it is not a physics problem. “Easier” is doing a lot of quiet work in that folk claim. It could mean at least three different things: the pole reduces the energy you burn over a kilometre; the pole reduces the peak force slamming into your shoulder once per step; or the pole simply feels better, which is a statement about nerve endings and not obviously about either of the first two.
These are not the same quantity, and — this is the punchline I will come back to at the very end — optimising one of them can actively hurt another. Picking the wrong one is how you end up with a beautiful model that answers a question nobody was asking.
So I am going to commit to exactly one, out loud, before any equations:
The quantity we care about is the peak dynamic force transmitted to the shoulder while walking.
Not energy. Not metabolic cost. Not comfort. Just the biggest spike of force your trapezius absorbs, once per step, forever. It is a defensible reading of “easier” — peak force is what leaves a mark on your shoulder — and it has the enormous practical advantage of being something a model can actually compute. But it is a choice, and I am flagging it now so that when it collides with somebody else’s choice in the last section, you will see the collision coming.
2. The model
The configuration. One person, shoulder at the midpoint, mass $m$ hanging at each end.
Strictly there are two low-order modes here. In the symmetric mode both ends move up and down together, driven by the shoulder translating vertically. In the antisymmetric mode the ends move in opposite directions, driven by the shoulder rotating. I am only going to study the symmetric one, because that is what sets the vertical force, and because in real life people damp the antisymmetric mode by steadying the front end with a hand. (Real carriers do exactly this. It is not a coincidence.)
In the symmetric mode the two halves are mirror images, so I only need to analyse one end and double at the finish — and since everything we want is a ratio, that factor of 2 will cancel anyway.
The stiffness. Treat the half-pole from shoulder to load as a cantilever beam: the fixed end is at the shoulder (in the symmetric mode the slope at the midpoint is zero, which is precisely the cantilever boundary condition), and the free end carries a point load $P=mg$. Standard Euler–Bernoulli result:
$$y_{\text{tip}}=\frac{P\,x_L^{3}}{3EI}\quad\Longrightarrow\quad k=\frac{3EI}{x_L^{3}}$$
with $x_L$ the shoulder-to-load distance, $E$ Young’s modulus, $I$ the second moment of area.
Why one degree of freedom is allowed. A bamboo strip is a continuous elastic body with infinitely many modes, and I am about to replace it with a single spring. Two reasons this is legal rather than lazy:
- Frequency separation. The highest meaningful harmonic of walking is around $8$ Hz, far below the pole’s second bending mode. Only the first mode gets excited.
- Mass concentration. By Rayleigh’s method the effective mass of a cantilever with a tip load is $m_{\text{eff}} = m + \frac{33}{140}m_{\text{beam,half}}$. A bamboo pole weighs $1$–$2$ kg total; the load is $20$ kg per end. The correction is under about $1\%$.
So: massless spring $k$, dashpot $c$, mass $m$, base excited by the shoulder. A textbook base-isolation problem, which is a nice thing to discover, because it means the answer will generalise to car suspensions and vibration-isolated optical tables and anything else that hangs a mass off a spring attached to something that shakes.
3. Gravity cancels — and leaves a present behind
This step gets skipped in most treatments and I think that is a shame, because the most useful formula in the whole post falls out of it.
Let $X_{\text{abs}}$ be the load’s absolute height and $Y_{\text{abs}}$ the shoulder’s. Newton’s second law on the load:
$$m\ddot{X}_{\text{abs}}=-mg-k\big[(X_{\text{abs}}-Y_{\text{abs}})-\ell_0\big]-c(\dot X_{\text{abs}}-\dot Y_{\text{abs}})$$
where $\ell_0$ is the spring’s natural length. At static equilibrium everything with a dot vanishes, and if we call the resulting compression $\delta$:
$$0=-mg+k\delta\quad\Longrightarrow\quad \boxed{\;\delta=\frac{mg}{k}\;}$$
That $\delta$ is the static sag: how much lower the load hangs once you have picked it up, compared to the unloaded pole. Hold onto it; it is going to become the punchline.
Now move the origin to static equilibrium, $x = X_{\text{abs}} – X_{\text{abs}}^{\text{eq}}$ and $y = Y_{\text{abs}} – Y_{\text{abs}}^{\text{eq}}$. The constant $-mg$ and the constant $+k\delta$ annihilate each other exactly, and what is left is
$$m\ddot{x}+c(\dot{x}-\dot{y})+k(x-y)=0 \tag{1}$$
Gravity has vanished from the dynamics entirely. But it did not vanish for nothing — it left behind $\delta = mg/k$, and later that one equation will let us throw out $E$, $I$, $x_L$, $k$ and $m$ all at once and replace them with something you can measure with a ruler.
A warning about notation: equation $(1)$ ends in “$=0$”, which makes it look homogeneous. It is not. $y(t)$ is a known input, not an unknown. Move it across and its true face appears:
$$m\ddot{x}+c\dot{x}+kx=c\dot{y}+ky$$
A perfectly ordinary forced linear ODE. Two different things happened in this section and it is worth keeping them apart: shifting to equilibrium coordinates eliminated something (the constant $mg$, in exchange for $\delta$); moving $y$ to the right merely tidied up.
4. The quantity to minimise, in one line
The force your shoulder feels is whatever the spring and dashpot push up with:
$$F_{\text{transmitted}}=k(x-y)+c(\dot{x}-\dot{y})$$
But look at $(1)$ again — that entire right-hand side is $-m\ddot{x}$. So
$$\boxed{\;F_{\text{transmitted}}=-m\ddot{x}\;} \tag{2}$$
I find this identity delightful and it is worth a paragraph. First, it says the force on your shoulder is nothing but the acceleration of the load. If you want to measure the force on a real carrier’s shoulder, you do not need a load cell — tape a phone to the bucket and read the accelerometer. Second, it means the thing we are minimising depends on $x$ alone, and not separately on how the stiffness and damping split the job between them.
In steady state the load moves harmonically, $x(t)=X\cos(\Omega t+\varphi)$, so $|\ddot x|_{\max}=\Omega^2 X$ and
$$F^{\text{flexible}}_{\text{peak}}=m\Omega^{2}X$$
Now the comparison. Swap the bamboo for a rigid iron bar: the load is forced to follow the shoulder exactly, $x\equiv y$, so
$$F^{\text{rigid}}_{\text{peak}}=m\Omega^{2}Y$$
and the ratio of the two is
$$T\;\equiv\;\frac{F^{\text{flexible}}_{\text{peak}}}{F^{\text{rigid}}_{\text{peak}}}\;=\;\frac{m\Omega^2X}{m\Omega^2Y}\;=\;\frac{X}{Y} \tag{3}$$
This ratio $T$ is called the transmissibility, and I want to stress that it is not a third concept sneaking in. It is the force on your shoulder, divided by a reference, written as a pure number. $T=0.5$ means the force is halved. $T=2$ means it is doubled. $T=1$ means your bamboo pole is doing precisely as much for you as an iron bar, which is to say nothing.
(One genuinely nice accident: in base-excitation problems the force transmissibility and the displacement transmissibility are numerically identical, because the $m\Omega^2$ cancelled. In force-excitation problems they are different functions. So this is not automatic and it is worth noticing when it happens.)
5. Solving it, complex-style
I will do this with complex exponentials. Not because it is fancier but because it converts a differential equation into one division.
Why $e^{i\Omega t}$ and not $\cos\Omega t$: differentiation eats $\cos$ and spits out $\sin$, so you have to carry two unknowns and solve a $2\times2$ system. But
$$\frac{d}{dt}e^{i\Omega t}=i\Omega\,e^{i\Omega t}$$
Differentiation becomes multiplication by $i\Omega$. In linear-algebra language: $\cos$ and $\sin$ span a two-dimensional invariant subspace of $d/dt$ but neither is an eigenfunction, whereas over $\mathbb{C}$ that subspace splits into two one-dimensional eigenspaces with bases $e^{\pm i\Omega t}$. The complex method is just diagonalising.
Why taking the real part at the end is legal: this deserves a proof, because it is the one step that could actually be wrong. Let $L[x]=m\ddot x+c\dot x+kx$ with $m,c,k\in\mathbb{R}$, and write a complex function as $\tilde x=p+iq$ with $p,q$ real. Differentiation is a real-linear operation, so
$$L[\tilde x]=(m\ddot p+c\dot p+kp)+i(m\ddot q+c\dot q+kq)=L[p]+i\,L[q]$$
Taking real parts, $\operatorname{Re}L[\tilde x]=L[\operatorname{Re}\tilde x]$. Hence if $L[\tilde x]=\tilde F$ then $L[\operatorname{Re}\tilde x]=\operatorname{Re}\tilde F$, and solving the complex problem and discarding the imaginary part really does solve the real one. Notice that this used $m,c,k$ being real. With complex coefficients the real and imaginary parts get stirred together and the whole trick collapses.
Also, one more thing we are entitled to assume and one we are not. We are not assuming the load moves sinusoidally — we try a sinusoid, find it fits, and invoke uniqueness. And uniqueness is easy: if $x_1$ and $x_2$ both solve the forced equation, their difference solves the homogeneous one, which for $c>0$ decays like $e^{-\zeta\omega_n t}$. So there is exactly one solution that neither decays nor blows up, and that is the steady state. This matters more than it looks: it is what makes “the $T$ of this pole” a property of the pole rather than of how you happened to heave it onto your shoulder.
(The transient itself is not negligible in real life, by the way. Bamboo’s damping is tiny, so the decay time is something like $8$ seconds — roughly 50 steps. Start-stop carrying spends a serious fraction of its time out of steady state. I am ignoring that here and I want to be honest that I am ignoring it.)
The algebra. Put $\tilde y=Ye^{i\Omega t}$, $\tilde x=\hat X e^{i\Omega t}$ into $(1)$:
$$\big(-m\Omega^{2}+ic\Omega+k\big)\hat X=\big(ic\Omega+k\big)Y$$
$$\frac{\hat X}{Y}=\frac{k+ic\Omega}{k-m\Omega^{2}+ic\Omega}$$
One complex division and we are done. A real-variable treatment would be solving a linear system here and then a separate arctangent for the phase; the complex answer carries amplitude in its modulus and phase in its argument, both for free.
6. Five parameters become two
The expression above contains $m$, $c$, $k$, $\Omega$, $Y$. Divide numerator and denominator by $k$ and define
$$\omega_n=\sqrt{\frac{k}{m}},\qquad r=\frac{\Omega}{\omega_n},\qquad \zeta=\frac{c}{2\sqrt{km}}$$
($r$ = how fast you walk relative to the pole’s natural rhythm; $\zeta$ = damping ratio.) Then $m\Omega^2/k = r^2$ and $c\Omega/k = 2\zeta r$, and everything collapses to
$$\frac{\hat X}{Y}=\frac{1+2i\zeta r}{\left(1-r^{2}\right)+2i\zeta r} \tag{4}$$
Taking the modulus:
$$\boxed{\;T(r,\zeta)=\sqrt{\frac{1+(2\zeta r)^{2}}{\left(1-r^{2}\right)^{2}+(2\zeta r)^{2}}}\;} \tag{5}$$
Two sanity checks, both of which it passes: as $r\to0$ (infinitely stiff pole, or infinitely slow walking) $T\to1$, which is right, because a stiff pole is an iron bar. And at $\zeta\to0,\ r\to1$, $T\to\infty$: undamped resonance.
Five physical quantities went in; two dimensionless numbers came out. That is not an accident of algebra — by Buckingham’s $\pi$ theorem, five parameters with three independent dimensions must reduce to $5-3=2$. And the payoff is specific: it is why the threshold below is a pure number rather than “the critical stiffness for this particular bamboo at this particular length.”
7. The structure hiding in $(5)$, and Theorem 1
Before differentiating anything, look at the shape of $(5)$. Set
$$u=(2\zeta r)^{2}\ge 0,\qquad a=\left(1-r^{2}\right)^{2}\ge 0$$
and the whole thing is
$$T^{2}=\frac{1+u}{a+u} \tag{6}$$
$T^2$ is a Möbius transformation of $u$ — a linear fractional function. All the damping enters through $u$, and $u$ only appears here. Which means, without any calculus at all:
Theorem 1 (the √2 threshold). For every damping ratio $\zeta\ge0$, $$T<1\iff r>\sqrt{2}$$ and $T=1$ exactly when $r=0$ or $r=\sqrt2$. The threshold does not depend on damping.
Proof. Since $a+u>0$,
$$T^{2}<1\iff 1+u<a+u\iff 1<a$$
The damping term cancels off both sides. So the condition is $1<(1-r^{2})^{2}$, i.e. $|1-r^{2}|>1$, which splits into $1-r^2>1$ (giving $r^2<0$, impossible) and $1-r^2<-1$, i.e. $r^2>2$. Equality gives $|1-r^2| = 1$, so $r^2\in\{0,2\}$. $\blacksquare$
That is the whole proof. Three lines, no derivatives, and the damping — the messiest, least measurable, most material-dependent thing in the model — deletes itself.
The geometric reading is the one I like best. Think of $\{T(\cdot,\zeta)\}_{\zeta\ge 0}$ as a whole family of curves, one for each damping ratio. Theorem 1 says every curve in the family passes through the same two points, $(0,1)$ and $(\sqrt2,1)$. No matter how lossy your material is, the curve is pinned at $r=\sqrt2$. It is a fixed point of the family. (If you have met Den Hartog’s fixed-point argument for tuned mass dampers, this is the same structure, just degenerate enough to solve in one line.)
And physically: your pole’s natural frequency must be below $1/\sqrt2\approx0.707$ times your step frequency. The pole has to wobble at least 41% slower than you walk. Anything stiffer and you are not just failing to help — you are amplifying.
8. What √2 looks like in centimetres
Here is where $\delta=mg/k$ comes back. Since $\omega_n^2=k/m$ and $k/m=g/\delta$,
$$\omega_n=\sqrt{\frac{g}{\delta}}$$
Every material parameter is gone. So is the configuration: do the same calculation for two people carrying one load at the midpoint of a simply supported beam and you get $k=48EI/L^3$ with the whole mass attached, totally different numbers — and $\omega_n^2 = g/\delta$ again. $E$, $I$, $L$, $m$, the number of carriers: all of it cancels.
Feed that into $r>\sqrt2$, with $\Omega=2\pi f$ and $f$ the step frequency in Hz:
$$\boxed{\;\delta>\frac{g}{2\pi^{2}f^{2}}\;}$$
Numerically, $\delta_{\min}\,[\text{cm}]=12.4/f^{2}$. At a normal walking cadence of $f = 2$ Hz:
The pole must sag at least 12.4 cm when you load it, or it is making things worse.
Twelve centimetres. That is a lot of droop — it is exactly the visibly-alarming sag that the engineering student wanted to design out. The folk claim’s mechanism and the engineer’s bug report are the same observation, read with opposite signs. Three consequences fall straight out:
- Walking faster is better. $\delta_{\min}\propto f^{-2}$, so shuffling along slowly pushes you toward the amplification region. Nobody’s intuition predicts this.
- A carrying pole is a heavy-load tool. Rewriting via $\delta = mg/k$ gives $m > k/(2\pi^2f^2)$: the same pole reduces your shoulder force at 40 kg and amplifies it at 10 kg. Nobody carries two cabbages on a carrying pole, and it turns out that is not just inconvenience — it is physically worse than a rigid stick.
- “Alarmingly floppy” is a necessary condition. A pole that stays straight under load is, provably, amplifying the force on your shoulder.
9. So why not a rubber band?
If softer is better, the mathematically optimal carrying pole is infinitely soft. This is obviously wrong, and the reason it is wrong is instructive.
$T$ is monotonic in stiffness on the useful side of the threshold — it just keeps decreasing as $k\to0$, with no interior minimum. And here is a general principle worth stating outright:
When the objective is monotonic in the design variable, the optimum is not determined by the objective at all. It is determined entirely by the constraints. Until you write down a constraint, the optimisation problem has no solution.
The binding constraint is embarrassingly mundane: the buckets must not hit the ground. Call the largest tolerable sag $\delta_{\max}$, fixed by your height, your shoulder height, and how tall the buckets are.
Theorem 5 (optimal stiffness). Subject to $0<\delta\le\delta_{\max}$, the minimum of $T$ occurs at the boundary $\delta=\delta_{\max}$, with $$k_{\text{opt}}=\frac{mg}{\delta_{\max}},\qquad \omega_n^{\text{opt}}=\sqrt{\frac{g}{\delta_{\max}}}$$
Proof. From $\omega_n^2 = g/\delta$ we get $r^2=\Omega^2\delta/g$, so $r$ is strictly increasing in $\delta$. And $T$ is strictly decreasing in $r$ throughout the isolation region $r>\sqrt2$ (differentiate $(5)$ with respect to $r^2$; the peak sits just below $r=1$, so everything past $\sqrt 2$ is on the falling branch). Composing, $T$ is a strictly decreasing function of $\delta$, and a strictly decreasing function on a closed interval attains its minimum at the right endpoint. Then $k_{\text{opt}}=mg/\delta_{\max}$ follows from $\delta = mg/k$. $\blacksquare$
This is an actual design formula. Give me your height and the load you usually carry, and the stiffness of your pole is determined — and from $k=3EI/x_L^3$, so is $EI$:
$$EI=\frac{mg\,x_L^{3}}{3\delta_{\max}}$$
Which finally answers “why not steel.” For the same required $EI$, steel’s modulus is about $200$ GPa against bamboo’s $10$–$22$ GPa, so a steel pole would need its second moment of area smaller by an order of magnitude — meaning a rod so thin it could not survive the static load. Bamboo’s low stiffness is not a limitation you tolerate. It is the specification.
And then the prettiest corollary. Substitute $k_{\text{opt}}$ back, take the lightly-damped limit (for bamboo, $\zeta\approx0.015$, and the error here is well under $1\%$), and the best achievable performance is
$$T_{\min}=\frac{1}{r^{2}-1}=\frac{g}{4\pi^{2}f^{2}\delta_{\max}-g}$$
$m$, $k$, $E$, $I$, the length, the cross-section — all gone. How good a carrying pole can possibly be depends on exactly two things: how much sag you can live with, and how fast you walk.
| $\delta_{\max}$ | 0.15 m | 0.20 m | 0.30 m | 0.40 m |
|---|---|---|---|---|
| $T_{\min}$ at $f=2$ Hz | 0.71 | 0.45 | 0.26 | 0.18 |
And the condition for this optimum to be worth having at all, $T_{\min}<1$, requires $4\pi^2f^2\delta_{\max}-g>g$, that is
$$\delta_{\max}>\frac{g}{2\pi^{2}f^{2}}$$
which is precisely the threshold from §8. The theory closes on itself: the critical condition reappears inside the optimum, wearing a different hat, as the question of whether the optimum is worth taking.
Closing, with a caveat I refuse to bury
I have skipped a lot — what damping does on either side of the threshold (it is harmful in the isolation region, which took me a while to believe), what happens when you admit that real gait is not a pure sinusoid, and where the linear model detonates. Maybe next time.
But one caveat belongs here rather than in a footnote, because it is the most interesting thing in the whole project. When you plug in measured stiffnesses from real carrying poles, most of them fail the criterion. They sit at or below the threshold — some almost exactly at resonance. Centuries of use did not converge on $r>\sqrt2$.
I do not think this breaks the model. I think it means my grandmother and I are not talking about the same word. The literature puts minimum metabolic cost at around $r\approx1.1$–$1.3$ — on the wrong side of $\sqrt2$, in the region where the peak force on your shoulder is being amplified. So the two readings of “easier” are not merely different; in the range where real poles actually live, they point in opposite directions. Centuries of use selected for the energy one. I selected for the force one. Both are defensible readings of the same folk sentence; only one of them has a $\sqrt2$ in it.
Which means the honest verdict on the folk claim is: true, and for a reason nobody who says it has in mind. The bamboo really is doing something real and mechanical — but the thing it is optimising is not the thing that makes your shoulder hurt, and a pole tuned to spare your shoulder would be a worse pole to walk a kilometre with.
That is why §1 came before the physics, and why I would rather have spent a page on a definition than on another theorem. The answer to “is a bamboo pole easier?” was settled the moment I decided what easier meant — everything after that was just algebra finding out.
Leave a Reply
You must be logged in to post a comment.