In short

A potentiometer measures a potential difference by balancing it against the drop along a uniform wire — no current flows from the cell under test at balance, so its internal resistance does not contaminate the reading. Its sensitivity is the smallest EMF detectable; sharper sensitivity means a smaller potential gradient k = V_\text{wire}/L on a longer wire.

\text{sensitivity } \propto \frac{1}{k} \propto \frac{L}{V_\text{wire}}.

A Kelvin double bridge extends the Wheatstone idea to resistances below 1\ \Omega (milli-ohms, micro-ohms) by adding a second ratio arm that cancels the resistance of the lead wires — the lead resistance is what would otherwise swamp the reading. The balance condition is

X \;=\; S\cdot\frac{P}{Q} \;+\; \frac{p\, r}{p + q + r}\left(\frac{P}{Q} - \frac{p}{q}\right),

and the clever step is arranging P/Q = p/q so that the second (lead-resistance) term vanishes, leaving the pure ratio X = S\, P/Q.

The Post Office box is a fixed-ratio Wheatstone bridge: plug-type resistance arms P, Q (ratio) and R (balancing), a built-in galvanometer key and cell key, and a set of resistance plugs from 1\ \Omega to 5000\ \Omega. For unknown X in the fourth arm, balance gives X = R\cdot P/Q — and the whole apparatus fits in a wooden box the size of a textbook.

Error rules, in one line. Use the middle third of a potentiometer or bridge wire (25\text{ cm} \le \ell \le 75\text{ cm}) for best sensitivity, check zero before every reading, and use the shortest thickest leads available — the instrument is only as good as its wires.

A high-precision strain gauge in the wind-tunnel at IIT Kanpur changes its resistance by about 0.2\ \text{m}\Omega for every newton of force it feels. The total gauge resistance is 350\ \Omega. How do you measure a 0.0002\ \Omega change on top of a 350\ \Omega baseline? You cannot. No voltmeter has that resolution. No ammeter has that resolution. Your ₹150 digital multimeter reads to 0.1\ \Omega — five hundred times too coarse.

You can measure it — using a bridge. A Wheatstone bridge built with three 350\ \Omega precision resistors and the gauge as the fourth arm will show a tiny off-balance voltage of about 0.14\ \mu\text{V} per newton across a galvanometer that detects 0.01\ \mu\text{V}. That is the bridge doing the impossible: measuring a change that is a million times smaller than the background, by subtracting the background away before you measure anything at all.

This is what null methods are for. The meter bridge and potentiometer article built the basics: why a null reading is independent of the galvanometer's calibration, why the balance length on a uniform wire gives a pure ratio, why no current is drawn from a cell at potentiometer balance. This article goes a layer deeper. You will see what sensitivity really means and how to tune it, meet the Kelvin double bridge that measures milli-ohms (the lead-resistance problem and the elegant trick that cancels it), the Post Office box that sat on every engineering student's desk from 1910 to 1990, and the error sources that decide whether your reading is trustworthy to three significant figures or to five.

Potentiometer sensitivity — the core design quantity

You already know how a potentiometer works: a long uniform wire carries a steady current from a driver cell, producing a uniform potential drop along its length. A jockey slides until the galvanometer between the unknown EMF and the wire shows zero. At that moment, the unknown EMF equals the drop along the balanced length.

The sensitivity is not a single fact about the instrument — it is a design choice you tune. Two potentiometers can both "work" in the sense of producing a balance, but one might resolve 1\ \text{mV} and the other only 10\ \text{mV} because of how they are set up.

Defining sensitivity

Sensitivity is the smallest change in measured EMF that produces a detectable galvanometer deflection. If moving the jockey by 1\ \text{mm} changes the wire-side voltage by an amount smaller than the galvanometer's threshold, you cannot distinguish two EMFs that differ by less than that. So the sensitivity is determined by two things:

  1. The potential gradient k along the wire — how many volts per metre.
  2. The galvanometer's minimum detectable current i_\text{min} — how tiny a nudge it can register.

Define the potential gradient explicitly:

k \;=\; \frac{V_\text{wire}}{L},

where V_\text{wire} is the voltage dropped across the full wire and L is its length.

Why: on a uniform wire the drop is linear in distance, so the gradient k (volts per metre) is a single number that characterises the whole wire. A 4\ \text{m} wire with 2\ \text{V} across it has k = 0.5\ \text{V/m} — every centimetre along the wire contributes 5\ \text{mV}.

The sensitivity formula

You want the smallest EMF \delta\varepsilon that produces a visible deflection. A displacement \delta\ell of the jockey changes the balance voltage by k\cdot\delta\ell. So:

\delta\varepsilon_\text{min} \;=\; k\cdot \delta\ell_\text{min}.

Why: near balance, the galvanometer current is proportional to the imbalance voltage. A jockey displacement \delta\ell creates an imbalance k\,\delta\ell, and the galvanometer threshold converts that to an observable kick.

This is the whole story. Smaller k means smaller \delta\varepsilon_\text{min} means finer sensitivity. To make a potentiometer more sensitive you want a smaller potential gradient.

How to lower the gradient

Three knobs are available to you:

1. Increase the wire length L. For a fixed driver voltage, a longer wire spreads the same drop over more metres. Doubling L halves k. A school-lab "four-wire potentiometer" is exactly four one-metre segments joined end-to-end — effectively a four-metre wire — precisely to cut k by a factor of four versus a simple one-metre instrument.

2. Decrease the current through the driver wire. The driver cell forces a current I through the wire; the drop across the wire is V_\text{wire} = I R_\text{wire}. Put a rheostat in series with the driver cell to reduce I. Halving I halves V_\text{wire} and halves k.

3. Use a lower-EMF driver cell, or a fraction of its EMF. Some precision potentiometers use a helper "standard" circuit that drops the driver's EMF across a precision divider before applying it to the wire. If you only need to measure EMFs up to 0.5\ \text{V} (thermocouples, say), there is no reason to put 2\ \text{V} across the wire.

The trade-off. You cannot drop k without limit. If k is too small, the wire cannot reach the EMF you are trying to measure — the balance point would lie beyond the wire's end. A practical rule: choose k so that the EMF you want to measure balances somewhere between 25\% and 75\% of the wire's length. Too close to either end and small non-uniformities in the wire dominate the reading.

Sensitivity, quantified

Consider two potentiometers:

If both galvanometers can detect a jockey displacement as small as \delta\ell_\text{min} = 1\ \text{mm}, then:

Instrument B resolves four times finer — just because its wire is four times as long. That is why physics labs have long potentiometers and why pocket voltmeters cannot touch them.

Interactive: tune the gradient yourself

Drag the wire length L or the driver current I and watch the potential gradient k and the smallest resolvable EMF respond. The wire resistance per metre is held fixed at R_\text{pm} = 1\ \Omega/\text{m}; the galvanometer detects jockey displacements of 1\ \text{mm} or larger.

Interactive: potentiometer sensitivity vs wire length A curve showing potential gradient k versus wire length L for a fixed driver current, with a draggable length parameter and live readouts of k and smallest detectable EMF. wire length L (metres) potential gradient k (V/m) 1 2 3 4 5 1 2 3 k = 2/(L+1) (driver 2 V, rheostat 1 Ω, wire 1 Ω/m) drag the red point along the x-axis
Drag the red point to change the wire length $L$. A longer wire means a lower gradient $k$ and a smaller minimum detectable EMF — the resolution improves linearly with length until the rheostat resistance dominates. This is why precision potentiometers use long wires wound back and forth on a board.

Calibrating the gradient — the standard cell trick

Before you measure an unknown EMF you need to know k to high accuracy. You could compute it from V_\text{wire} = IR_\text{wire} and L, but both I and R_\text{wire} drift with temperature and contact resistance. The standard procedure is to balance a known EMF against the wire.

A standard cell — traditionally a Weston cadmium cell at 1.01864\ \text{V} — has an EMF known to five significant figures and essentially constant over time. Balance the standard cell on the wire; call the balance length \ell_\text{std}. Then

k \;=\; \frac{\varepsilon_\text{std}}{\ell_\text{std}}.

Every subsequent measurement uses this calibrated k. If the driver current drifts between readings, recalibrate. A good protocol recalibrates before every unknown measurement. This is the reason a Class 12 potentiometer experiment at a CBSE lab specifies "balance the standard cell first, then the test cell, then the standard cell again — if the two standard-cell readings agree within 0.5\ \text{cm}, accept the test reading". That consistency check is a drift check on k.

Kelvin double bridge — measuring milli-ohms

The Wheatstone bridge is clean for resistances roughly from 1\ \Omega to 1\ \text{M}\Omega. Below 1\ \Omega it fails, not because the physics breaks but because the wires connecting the unknown to the bridge start to matter. A copper lead wire of diameter 1\ \text{mm} and length 30\ \text{cm} has a resistance of about 6\ \text{m}\Omega. If you are trying to measure an unknown of 2\ \text{m}\Omega, the lead contributes three times more than the thing you want to measure. The bridge reads the sum.

The physical problem

Imagine the unknown X is a short rod of copper busbar carrying a large current in a power station. Its intrinsic resistance might be 0.5\ \text{m}\Omega. You connect it into a Wheatstone bridge in the standard way, with two leads running from the bridge to the rod's two ends. Each lead has resistance of a few milli-ohms. The bridge balance condition

\frac{X + r_\text{lead,1} + r_\text{lead,2}}{S} \;=\; \frac{P}{Q}

contains X and the lead resistances summed together. You cannot untangle them.

The Kelvin insight

Kelvin — William Thomson, building electrical measurement apparatus for the early Atlantic telegraph cables in the 1860s — noticed that you can separate current-carrying connections from voltage-sensing connections. Use four terminals on the unknown: two heavy current leads that carry the bridge current, and two fine voltage-sensing leads that pick off the potential difference across exactly the right part of the unknown. The current leads can have any resistance; they only have to carry current. The voltage leads carry no current at balance (the galvanometer is a null detector), so their resistance does not matter either.

This reduces the lead-resistance problem but does not eliminate it entirely. There is still the short bit of wire between the voltage-sense contact on the unknown and the voltage-sense contact on the standard S — the "yoke" wire, of some small resistance r. Kelvin's bridge deals with this last r by adding a second ratio pair p, q that spans it.

Kelvin double bridge circuitA battery drives current through the unknown X (four-terminal, shown as a box), through a yoke wire r, and through the standard S (four-terminal). Above, two ratio arms P and Q join the top of X and the top of S to a midpoint M; a second ratio pair p and q span the yoke and join to a point N. A galvanometer connects M to N. + − E X (unknown) r (yoke) S (standard) a b c d P Q M p q N G
Kelvin double bridge. The unknown $X$ and the standard $S$ are connected as four-terminal elements: bulk current flows through the end terminals, while fine voltage-sensing wires (red dots $a, b, c, d$) pick off the potentials at the boundaries of each bulk resistor. The yoke $r$ is the inevitable stretch of wire between $b$ and $c$. A *second* ratio pair $p, q$ bridges $b$ and $c$, meeting at $N$, while the primary ratio $P, Q$ meets at $M$. The galvanometer reads between $M$ and $N$.

Deriving the balance condition

Let I be the current flowing along the main circuit through X, through the yoke r, through S. Let i be the current through the P–Q ratio arms. At balance, no current flows through the galvanometer. Kirchhoff's laws give:

Step 1. The potential difference from a to M (via P) equals the potential difference from a to M via the direct path a \to b \to N \to M (but N \to M carries no current at balance, so V_N = V_M). Therefore:

V_a - V_M \;=\; iP.

And also:

V_a - V_N \;=\; (V_a - V_b) + (V_b - V_N) \;=\; IX + i_2 p,

where i_2 is the current through p (and q).

Step 2. Setting V_M = V_N at balance:

iP \;=\; IX + i_2 p. \tag{*}

Step 3. The same argument on the right side (d to M via Q, and d to N via c \to N):

iQ \;=\; IS + i_2 q. \tag{**}

Step 4. You need one more relation. The current i_2 flows through p and q in series (they meet only at N and nowhere else at balance). The potential difference driving i_2 is the drop from b to c across the yoke:

i_2(p + q) \;=\; V_b - V_c \;=\; Ir.

Therefore:

i_2 \;=\; \frac{Ir}{p + q}. \tag{***}

Step 5. Divide equation (*) by equation (**):

\frac{P}{Q} \;=\; \frac{IX + i_2 p}{IS + i_2 q}.

Why: the unknown i cancels when you take the ratio, leaving X and S on top and bottom, with corrections from the yoke current i_2. Now substitute (***) for i_2.

Step 6. Substitute i_2 = Ir/(p+q) and simplify:

\frac{P}{Q} \;=\; \frac{X + \dfrac{pr}{p+q}}{S + \dfrac{qr}{p+q}}.

Cross-multiply:

P\left(S + \frac{qr}{p+q}\right) \;=\; Q\left(X + \frac{pr}{p+q}\right).

Expand and solve for X:

X \;=\; \frac{P}{Q}\,S \;+\; \frac{pr}{p+q}\left(\frac{P}{Q} - \frac{p}{q}\right).

Why: this is the Kelvin double-bridge balance condition. The first term is the "ideal" answer — exactly the Wheatstone ratio X = (P/Q)\,S. The second term is the yoke-resistance correction, and it has a remarkable property.

Step 7. The correction term vanishes if you build the instrument so that P/Q = p/q:

\boxed{\;\text{if } \frac{P}{Q} = \frac{p}{q}, \quad X = S\cdot\frac{P}{Q}\;}

Why: the factor (P/Q - p/q) in the correction becomes zero when the two ratios match. The yoke resistance drops out of the result entirely. You still have the yoke wire there, carrying its current, causing its drop — but the symmetry between the two ratio pairs cancels its effect on the balance.

How the cancellation is achieved in practice

A Kelvin bridge is built as a mechanical unit: P and p are mounted on the same shaft of a precision resistance box, as are Q and q. When you dial in a new value of P, the value of p tracks it in lock-step, and the ratio P/Q = p/q is maintained automatically. The user adjusts the standard S until balance; the final answer X = S\cdot P/Q does not even know that the yoke exists.

The practical range is 10\ \mu\Omega to 1\ \Omega. Below 10\ \mu\Omega, thermoelectric EMFs at the bridge contacts start to compete with the real signal, and you need to reverse the current and average two readings. Above 1\ \Omega, a plain Wheatstone works fine and the extra complexity is not worth it.

The Post Office box — a portable Wheatstone

Before pocket multimeters, an engineer inspecting a railway signal line or a telegraph cable carried a Post Office box (named because the Post Office telegraph department commissioned the first mass-produced model). It is a Wheatstone bridge compressed into a wooden box the size of a hardback book, with brass plugs and dials that let you set the ratio and balance arms.

The layout

Three arms are built in: two "ratio arms" P and Q, each a plug-type resistance box with values 10, 100, 1000\ \Omega (you select one value per arm), and a "balancing arm" R with plugs from 1\ \Omega up to 5000\ \Omega in a binary or decimal sequence. The fourth arm — the unknown X — is connected to a pair of external terminals. A galvanometer terminal and a battery terminal, each with its own plug-type key, complete the setup.

Post Office box layoutSchematic: a Wheatstone bridge square with P in the top-left arm (labelled ratio), Q top-right (ratio), R bottom-right (balancing), X bottom-left (unknown). Galvanometer across the horizontal diagonal; battery across the vertical diagonal. Each arm is drawn as a resistor symbol. A B C D P (ratio) Q (ratio) X (unknown) R (balance) G E
Post Office box as a Wheatstone square: $P$ and $Q$ are the ratio arms, $R$ is the balancing arm, and the unknown $X$ sits in the fourth corner. A galvanometer spans one diagonal, the battery the other. You pick $P$ and $Q$ to set a ratio (e.g., $P = 100, Q = 10$ for a $10\!:\!1$ multiplier), then dial $R$ until the galvanometer reads zero.

The balance condition and range

The Wheatstone law applies unchanged:

X \;=\; R\cdot\frac{P}{Q}.

Why: at balance, V_A = V_C, so the IR drop across P equals that across Q on the top path, and the drop across X equals that across R on the bottom. Taking ratios of the adjacent arms gives X/R = P/Q, hence X = R\,P/Q.

The ratio multiplier P/Q can be 0.01, 0.1, 1, 10, 100 depending on which plugs you remove, and R ranges from 1\ \Omega to 5000\ \Omega. The useful measurement range is roughly:

X_\text{min} \approx 0.01\ \Omega, \qquad X_\text{max} \approx 500\ \text{k}\Omega.

The procedure (and why it is always done in this order)

  1. Start with P = Q = 10\ \Omega (ratio 1:1). Close the battery key first, briefly close the galvanometer key. If the deflection is too large to read, you have a gross-range problem; go to step 2.
  2. If the galvanometer pegs full-scale, you know X is outside the range of R. If the deflection was positive, X > 5000\ \Omega and you switch to ratio 1:100 (multiplier 1/100, so the effective R range becomes 100\ \Omega to 500\ \text{k}\Omega). If it was negative, X < 1\ \Omega and you switch to 100:1 (multiplier 100, effective range 0.01\ \Omega to 50\ \Omega).
  3. Now dial R by bisection. Try R = 2500\ \Omega. If the galvanometer kicks one way, halve; if the other way, halve the other half. Converge to a single-ohm resolution in five or six steps.
  4. Battery-key-first rule. Always close the battery key before the galvanometer key and open the galvanometer key before opening the battery key. This prevents induced transients in long leads (especially to an unknown inductive load like a motor winding) from slamming the galvanometer needle against its stop.

This procedure sounds antiquated and it is — since the 1990s every working engineer reaches for a digital multimeter instead. But the Post Office box still sits in every CBSE lab, and its logic is the same logic every precision resistance measurement uses today: a ratio-based null method, selected decade by decade, with a bisection search that converges in a handful of readings.

Errors and precautions — where the reading gets contaminated

Every measurement has error. A thoughtful experimenter knows where the errors come from, which are random and which are systematic, and which of the precautions in the lab manual actually matter.

Lead and contact resistance

Any wire between the instrument and the unknown has resistance. A half-metre copper lead of 0.5\ \text{mm} diameter has about 40\ \text{m}\Omega. A poorly crimped terminal adds another 10\!-\!100\ \text{m}\Omega. A greasy thumbprint on a brass contact is worse than a resistor — it adds a small voltage from electrochemistry on top of the resistance.

Precaution. Use the shortest thickest leads you can. Clean every brass plug with fine emery paper before a precision measurement. In the Kelvin bridge, insist on true four-terminal connections to the unknown — bolted where possible, not just touched.

Thermoelectric EMFs — the silent error

Whenever two dissimilar metals meet at a junction, the junction generates a small thermoelectric EMF proportional to the temperature difference between that junction and a distant reference junction. A copper-to-brass contact at 25\ ^\circ\text{C} can produce a few microvolts. For a Wheatstone measuring 100\ \Omega driven by 1\ \text{V}, a few microvolts is nothing. For a Kelvin bridge measuring 10\ \mu\Omega driven by a few millivolts, a few microvolts is everything.

Precaution. For any measurement below 1\ \Omega, take two readings — one with the current in one direction and one reversed. Average them. The thermoelectric EMF is independent of the current direction, so it cancels out. This is called the current-reversal method.

Non-uniformity of the bridge wire

A metre-bridge or potentiometer wire is supposed to have uniform resistance per unit length. Real wires vary by a few tenths of a percent along their length because of slight thickness variations, and more than that near the end terminals where the wire is crimped into a brass end-piece. Reading near 0\ \text{cm} or near 100\ \text{cm} puts you in the non-uniform region.

Precaution. Arrange the experiment so the balance length falls between 25\ \text{cm} and 75\ \text{cm}. For a meter bridge, swap the unknown and standard resistor between trials — this interchanges the roles of the two wire halves, and the average of the two measurements cancels any asymmetry in the wire.

Heating of the driver wire

A potentiometer wire carrying a steady current heats up. Heating raises its resistance, which changes the current and therefore the potential gradient k. In a five-minute experiment on a hot day, k can drift by 0.5\%. The standard-cell calibration and drift-check procedure catches this, but you should not press the jockey for a full second when a brief touch gives the same reading — every touch dumps energy into the wire.

Precaution. Press the jockey gently and briefly. Recalibrate with the standard cell every couple of minutes during a long run.

Contact-pressure and end-effect errors

A meter bridge's jockey rides on a bare wire; too much pressure dents the wire, changes its local cross-section, and over time creates a groove that makes the resistance-per-unit-length non-uniform. Too little pressure and the contact resistance is erratic. End-effects — the tiny bits of wire between the end-piece and the scale's zero mark — are systematic and about 1\!-\!3\ \text{mm} on a typical instrument.

Precaution. Press just hard enough to hear a soft tap. Calibrate the end-effects by measuring a known resistor in the left gap at two positions on the scale and solving for the corrections.

The summary table

Error source Affects Fix
Lead resistance Low-R measurements Four-terminal wiring; Kelvin bridge
Thermoelectric EMF Sub-ohm, sub-mV readings Current reversal; average two readings
Wire non-uniformity Meter bridge, potentiometer Balance in the middle third; swap arms
Driver heating Potentiometer, long measurements Gentle brief contact; recalibrate k often
Contact pressure Jockey-based instruments Light steady touch; clean contacts

Worked examples

Example 1: Sensitivity of a 4-metre potentiometer

A four-wire potentiometer has a total wire length L = 4\ \text{m}. The driver cell has EMF 2.0\ \text{V} and negligible internal resistance; a series rheostat is adjusted so the current through the wire is I = 0.5\ \text{A}. The wire resistance is 1.0\ \Omega/\text{m}. The galvanometer detects a jockey displacement as small as 1\ \text{mm}. Find the potential gradient k and the smallest EMF the instrument can resolve.

Schematic of a four-wire potentiometerA 4-metre uniform wire laid out in four parallel horizontal metre-segments connected end-to-end. A driver cell with a rheostat and ammeter sits on the left. The potential gradient k = V_wire / L is annotated as 0.5 V per metre or 5 mV per cm. 2 V rheostat segment 1 (0–100 cm) segment 2 (100–200 cm) segment 3 (200–300 cm) segment 4 (300–400 cm) Total wire length L = 4.0 m, wire current I = 0.5 A, wire resistance 1.0 Ω/m
Four metres of uniform wire, driven by a $2\ \text{V}$ cell through a series rheostat. The voltage dropped across the wire is $V_\text{wire} = I\cdot R_\text{wire} = 0.5\times 4 = 2\ \text{V}$, so the gradient is $k = V_\text{wire}/L = 0.5\ \text{V/m}$.

Step 1. Compute the wire's total resistance and voltage drop.

R_\text{wire} \;=\; (1.0\ \Omega/\text{m})\times(4.0\ \text{m}) \;=\; 4.0\ \Omega.
V_\text{wire} \;=\; I\cdot R_\text{wire} \;=\; 0.5\ \text{A}\times 4.0\ \Omega \;=\; 2.0\ \text{V}.

Why: the rheostat is set so all 2\ \text{V} drops across the wire, with the driver cell's negligible internal resistance absorbing nothing. If the rheostat were set to drop, say, 1\ \text{V}, only 1\ \text{V} would go across the wire and k would halve.

Step 2. Compute the potential gradient.

k \;=\; \frac{V_\text{wire}}{L} \;=\; \frac{2.0\ \text{V}}{4.0\ \text{m}} \;=\; 0.50\ \text{V/m} \;=\; 5.0\ \text{mV/cm}.

Why: on a uniform wire the drop is linear, so dividing the total drop by the total length gives the drop per metre.

Step 3. Convert the galvanometer's jockey-displacement threshold into an EMF resolution.

\delta\ell_\text{min} \;=\; 1\ \text{mm} \;=\; 0.001\ \text{m}.
\delta\varepsilon_\text{min} \;=\; k\cdot \delta\ell_\text{min} \;=\; 0.50\ \text{V/m}\times 0.001\ \text{m} \;=\; 0.50\ \text{mV}.

Why: the smallest detectable shift in the balance length maps to the smallest detectable change in EMF through the gradient.

Step 4. Comment on what this buys you.

A standard thermocouple junction between copper and constantan produces about 40\ \mu\text{V} per degree Celsius. With \delta\varepsilon_\text{min} = 500\ \mu\text{V}, this potentiometer cannot resolve 1\ ^\circ\text{C} on a single junction — it resolves about 12\ ^\circ\text{C}. To measure finer temperature differences you would need either a longer wire, a smaller current (higher-ratio rheostat), or a more sensitive galvanometer.

Result: The potential gradient is k = 0.50\ \text{V/m} and the smallest detectable EMF is 0.50\ \text{mV}.

What this shows: Potentiometer sensitivity is set by the interplay of wire length and driver current. A four-metre wire at 0.5\ \text{A} resolves half a millivolt — good enough for cell-EMF comparisons but not for sub-degree thermocouple thermometry. Designing a better potentiometer means dropping k, and that means either a longer wire or a smaller current.

Example 2: A Kelvin bridge reading on a busbar

A power-station busbar carries a steady current. An engineer wants to measure its contact-resistance to the terminal bolt. A Kelvin double bridge is set up with ratio arms P = Q = 1000\ \Omega and secondary arms p = q = 1000\ \Omega. The standard resistor is S = 0.01\ \Omega (a precision manganin shunt). The bridge balances when P = 100\ \Omega, Q = 10000\ \Omega, leaving p and q to track so that p/q = P/Q. Find the unknown resistance X.

Simplified Kelvin balance schematicUnknown X on the left, standard S on the right, connected by a short yoke. Above them, primary ratio arms P (1:100) and Q (1:100) meet at a galvanometer node. The balance condition X = S × P/Q is annotated. X (unknown) S = 0.01 Ω yoke r P = 100 Ω Q = 10 000 Ω G X = S·(P/Q) = 0.01·(100/10 000) = 1.0×10⁻⁴ Ω = 100 µΩ
The Kelvin bridge balance for a busbar contact-resistance measurement. The ratio $P/Q = 1/100$ multiplies down the $0.01\ \Omega$ standard to give an answer at the hundred-microhm scale.

Step 1. Check the yoke-cancellation condition.

\frac{P}{Q} \;=\; \frac{100}{10000} \;=\; \frac{1}{100}, \qquad \frac{p}{q} \;=\; \frac{1}{100}.

The two ratios match: the yoke correction vanishes.

Why: Kelvin's trick only works when P/Q = p/q. If the two ratios disagree, the second bracket in the full formula contributes a spurious term of order r\cdot|P/Q - p/q|, which for a yoke of 5\ \text{m}\Omega and a ratio mismatch of 1\% would be 50\ \mu\Omega — the same order as X itself. Always verify the ratios are tracked.

Step 2. Apply the clean balance condition.

X \;=\; S\cdot\frac{P}{Q} \;=\; 0.01\ \Omega\times\frac{1}{100} \;=\; 1.0\times 10^{-4}\ \Omega \;=\; 100\ \mu\Omega.

Why: with the yoke term cancelled, the bridge is effectively a plain Wheatstone with ratio P/Q multiplying S. The massive ratio 1\!:\!100 is what lets a 0.01\ \Omega precision standard measure a 100\ \mu\Omega unknown — you could not stock a precision 100\ \mu\Omega resistor, but you can stock a 0.01\ \Omega one and scale it.

Step 3. Sanity-check by a current-reversal diagnostic.

In a real measurement you would now reverse the battery current, rebalance, and take the average of the two X values. If the two readings differ by 5\ \mu\Omega, the thermoelectric EMF is contributing about 2.5\ \mu\Omega — reportable, probably tolerable, but logged.

Result: The contact resistance is X = 100\ \mu\Omega = 0.0001\ \Omega.

What this shows: The Kelvin double bridge converts a hundred-microhm measurement — impossible with any handheld meter — into a problem of dialling a precision resistance box and balancing a galvanometer. The yoke resistance, which would dominate the answer if left unchecked, is cancelled by a symmetry built into the bridge. The ratio multiplier scales the precision of a good standard resistor down into the sub-ohm range where standard resistors do not exist.

Common confusions

If your goal was to understand why null methods give cleaner answers than direct methods and how to tune a potentiometer, you have what you need. What follows is the JEE-Advanced-level treatment of sensitivity at finite galvanometer resistance, error propagation for a Wheatstone bridge, and the connection between the Kelvin bridge and the modern Kelvin–Varley precision voltage divider used today in IIT-Madras and ISRO calibration laboratories.

Sensitivity at finite galvanometer resistance

So far the sensitivity analysis assumed the galvanometer detects any nonzero current — a true null detector. Real galvanometers have a threshold current i_\text{g,min} and a resistance R_g. Near balance in a Wheatstone bridge with arm resistances P, Q, R, X, an off-balance perturbation \Delta R on one arm drives a small galvanometer current. Applying Kirchhoff to the off-balance network (and expanding to first order in \Delta R) gives:

i_g \;\approx\; \frac{E}{(P+Q)(R+X) + R_g(P+Q+R+X)}\cdot\frac{\Delta R}{R}\cdot\frac{PQ}{P+Q},

where E is the battery EMF.

Why: this is a first-order perturbation of the bridge network. The \Delta R/R factor encodes the fractional imbalance; the rest is the Thévenin-equivalent current delivered to the galvanometer by the unbalanced network.

The sensitivity is highest when the arms are equal (P = Q = R = X) and the galvanometer resistance matches the output impedance of the bridge, R_g \sim (P+Q)(R+X)/(P+Q+R+X) = R/2 (for equal arms). This is the impedance-matching condition for a bridge. For a potentiometer, the equivalent condition is that the test galvanometer's resistance match the parallel combination of the wire-segment resistance to the jockey and the rest-of-circuit resistance.

The practical consequence: a bridge designed to measure a specific unknown (strain gauge, platinum resistance thermometer) is built with all four arms nominally equal, so that at balance the sensitivity is maximised.

Error propagation in a Wheatstone bridge

If X = R\cdot P/Q and you know R, P, Q with relative uncertainties \delta R/R, \delta P/P, \delta Q/Q (all uncorrelated), the standard error-propagation formula gives:

\frac{\delta X}{X} \;=\; \sqrt{\left(\frac{\delta R}{R}\right)^2 + \left(\frac{\delta P}{P}\right)^2 + \left(\frac{\delta Q}{Q}\right)^2}.

Why: for a product X = R P/Q, the relative uncertainties add in quadrature when the sources are independent. Take log and differentiate: d(\ln X) = d(\ln R) + d(\ln P) - d(\ln Q); square and take expectation.

For a Post Office box with 1\%-tolerance resistors, all three uncertainties are 1\%, and \delta X/X = \sqrt{3}\%\approx 1.7\%. For a precision box with 0.01\% tolerances, \delta X/X \approx 0.017\% — about 2\ \Omega uncertainty on a 10\ \text{k}\Omega measurement.

The strain-gauge bridge and IIT labs

A strain gauge is a grid of fine foil bonded to a specimen. When the specimen stretches, the foil stretches, and its resistance increases by a fraction proportional to the strain:

\frac{\Delta R}{R} \;=\; G\cdot\varepsilon,

where G \approx 2 is the gauge factor and \varepsilon is the strain (dimensionless). A typical 350\ \Omega gauge under 1000\ \mu\text{strain} changes by 0.7\ \Omega — small, but now a Wheatstone bridge turns it into a voltage.

Put the gauge in one arm of a bridge with three 350\ \Omega precision resistors as the other arms, drive the bridge with 5\ \text{V}, and the off-balance output is:

V_\text{out} \;\approx\; \frac{V_\text{in}}{4}\cdot\frac{\Delta R}{R} \;=\; \frac{5}{4}\cdot\frac{G\varepsilon}{1} \;=\; 2.5\ \text{mV per } 1000\,\mu\text{strain}.

An instrumentation amplifier with a gain of 1000 produces 2.5\ \text{V} full-scale — directly readable by a microcontroller's ADC. This is the backbone of every load cell, every wind-tunnel force balance at the IIT Kanpur aerospace department, every weighbridge at a factory gate.

The Kelvin–Varley divider

The modern precision voltage reference device is a Kelvin–Varley divider — a ladder of resistor decades that divide an input voltage by a factor dialled from 0.00001 to 1.00000 in million-count resolution. It is the direct descendant of the Post Office box and the Kelvin bridge. National metrology institutes (NPL India at New Delhi, NIST in the US, PTB in Germany) use Kelvin–Varley dividers to transfer voltage standards between laboratories, calibrated against the Josephson-junction voltage standard. A thousand-rupee bench multimeter you buy today calibrates its voltage scale against a circuit whose logic was laid down in a Glasgow workshop in 1861.

Why "bridge methods" still dominate precision measurement

Null methods look old. They are. And yet every precision electronic measurement today — impedance analyzers, automatic-balance capacitance bridges, LCR meters, modern digital multimeters' auto-zero cycles — uses the same logic. The reason is deep: a measurement that finds a zero only requires the null detector to be linear near zero. It does not require the detector to be calibrated over a wide range. A meter with a 5\% gain error can still find a zero to 0.01\%.

This is why a wooden-board potentiometer in a dusty CBSE lab at Kanpur can outperform a ₹50,000 digital multimeter on an EMF measurement: the potentiometer asks the galvanometer only to say "am I at zero?", while the multimeter asks its electronics to be linear over six decades. The first question is easy. The second is a marvel of engineering and still only accurate to 0.1\%.

Where this leads next