2. Verify DAQ Channel Gain and Offset with a Calibration Shunt

Why a Shunt Resistor Belongs in a Gain/Offset Check

A strain-gage bridge channel in a crash DAQ system is only trustworthy if its transfer function — volts out per microstrain in — is known and documented. The shunt-calibration method exploits a simple physical fact: placing a precision resistor across one arm of the bridge produces a predictable, calculable bridge imbalance. That imbalance is equivalent to a known mechanical strain, so you can drive the entire signal chain (bridge completion, excitation, amplifier, anti-alias filter, ADC) with a defined input and compare the recorded output against the expected value. A precision voltage source then handles the offset half of the check, because offset errors are best isolated with a defined input at the amplifier rather than through the bridge.

The deliverable is not "the channel works." It is a pair of calibration factors — a gain correction and an offset correction — plus the raw evidence (shunt value, excitation, expected versus measured counts) that lets a reviewer reproduce the numbers during data reduction.

Worked Example: 350 Ω Bridge, 120 Ω Shunt, 5 V Excitation

Take a quarter-bridge strain-gage channel with a nominal gage resistance Rg = 350 Ω, gage factor GF = 2.10, bridge excitation Vex = 5.000 V, and amplifier gain set so that ±2 mV/V maps to ±5 V at the ADC input (gain = 2500). The DAQ digitizes at 16 bits over ±5 V, giving 305.18 µV/count.

Shunt the active arm with Rsh = 120 kΩ. The equivalent strain is:

ε_shunt = -1 / (GF · (1 + Rg/Rsh))
        = -1 / (2.10 · (1 + 350/120000))
        = -1 / (2.10 · 1.002917)
        = -474.8 µε

The bridge output in mV/V follows from the standard quarter-bridge relation:

Vout/Vex = (ε · GF) / 4 = (-474.8e-6 · 2.10) / 4 = -0.2493 mV/V

At the amplifier output that becomes -0.2493 mV/V × 5 V × 2500 = -3.116 V, or in ADC counts -3.116 / 305.18e-6 ≈ -10,211 counts. Record the actual reading with the shunt applied and with it removed; the difference is the shunt response, and the ratio of measured to expected response is the gain correction factor.

Isolating Offset with the Precision Source

Offset has two contributors: amplifier input offset and ADC zero error. With the shunt removed and the bridge balanced (no load), apply a known differential voltage at the amplifier input using the precision source — for example +1.000 mV — and record the output. Then apply -1.000 mV and record again. The midpoint of the two readings is the offset in counts; the slope between them is a second, independent estimate of gain that should agree with the shunt result to within the channel's specified linearity.

If the two gain estimates disagree by more than about 0.5 %, suspect a bridge-completion resistor tolerance problem or excitation droop rather than the amplifier — the shunt path exercises the bridge, the source path does not.

Reading the Verification Diagram

The figure below traces the two stimulus paths through the same signal chain and shows where each measurement is taken. The upper path is the shunt stimulus; the lower path is the precision-source stimulus. Both converge on the same ADC, which is the point: one channel, two independent ways to characterize it.

Documenting the Calibration Factors

Two numbers leave this procedure and enter the data-reduction chain. The gain correction factor is K_gain = expected_counts / measured_counts from the shunt step. The offset correction is K_off = -offset_counts from the source step, applied before gain scaling. A minimal record that a reviewer can reproduce contains the following fields; fill them with your own measured values rather than the illustrative ones shown.

Failure Symptoms and What They Point To

• Shunt response near zero. The shunt is not actually across the active arm, or the bridge completion is open. Verify continuity from the shunt clip to the gage terminals before touching gain settings.
• Shunt response correct but source response wrong. The amplifier input path is suspect — check for a differential offset or a common-mode issue at the source connection, not the bridge.
• Both responses scale together but are off by a constant ratio. Excitation is not the assumed 5.000 V. Measure it at the bridge, not at the supply, and recompute the expected counts.
• Noise on the shunt reading exceeds a few counts. The anti-alias filter or sample rate is marginal for the bridge impedance; average a short capture window rather than trusting a single sample.

Once K_gain and K_off are recorded, the channel is characterized end to end: the shunt proves the bridge-to-ADC path scales correctly, and the precision source pins the zero. Those two factors are what the data-reduction software applies before any CFC filtering, so getting them right here keeps every downstream strain and load result traceable to a documented stimulus.