A ratiometric bridge measurement cancels a shared excitation scale factor; it does not cancel every error in the sensor, cable, amplifier or ADC. The useful design question is where the reference voltage is sensed, how it reaches the converter, and which disturbances remain different between the signal and reference paths. A board can look ratiometric on a block diagram while losing that advantage at the connector or input filter.
This review addresses cancellation and the residual error budget. For broader placement, excitation and low-level routing decisions, use the strain-gauge PCB front-end review. For a different sensor type or an offset output range, start with mapping the sensor range to the ADC input. The calculations here are illustrative design checks, not measurements of a 吃瓜tv assembly.
Trace the ratio from the bridge terminals to the ADC
Write the transfer equation before choosing a reference
Let the bridge output be Vsig = S 脳 x 脳 Vexc, where S is its full-scale sensitivity in volts per volt, x is normalized applied input, and Vexc is the actual differential voltage across the bridge excitation terminals. Ignoring offsets and nonlinearities, an ADC result after fixed scaling is proportional to G 脳 Vsig / Vref. If Vref equals Vexc, the common excitation factor cancels and the result is proportional to G 脳 S 脳 x.
This is an algebraic relationship with conditions. The bridge must remain within its specified operating range, the ADC must accept the input and reference voltages, and the two paths must represent the same excitation at the relevant time. explains the basic ratio and the distinction from an independent-reference measurement. It is not a reason to ignore supply limits, interference or sensor heating.
For an illustrative sensor rated 2 mV/V at full scale, 4 V excitation produces 8 mV. A 1% excitation increase produces 8.08 mV. With the reference rising from 4 V to 4.04 V, both ratios equal 0.002 before gain. With a fixed 4 V reference, the ratio becomes 0.00202, which is a 1% change. This example describes cancellation of excitation gain variation, not total system accuracy or ADC resolution.
Distinguish the power source from the voltage at the sensor
A four-wire bridge uses two excitation conductors and two signal conductors. If the ADC reference is measured at the board connector, the reference includes the local excitation voltage but excludes the drop along the excitation cable. That cable drop changes the voltage actually applied to the bridge. It can therefore introduce a scale error even when the board uses one supply for excitation and reference.
Suppose the excitation source is 4 V, the bridge presents an equivalent 350 ohm load across its excitation terminals, and the two excitation wires contribute 2 ohms in total. Under these simplified assumptions, the bridge sees 4 脳 350 / 352 = 3.9773 V. Referencing the ADC to the source rather than the bridge gives a scale factor of 350 / 352, about 0.568% below the ideal value. A calibration may remove that fixed initial factor, but a changing cable resistance can reintroduce error.
A six-wire connection adds a separate sense pair at the bridge excitation terminals. Routing that pair to a suitable reference input allows the ADC to observe the actual excitation voltage rather than the supply setting. Verify reference input loading, voltage range, common-mode range, protection and settling. Sense wires are not magic: leakage, open connections or reference input current through resistance can create another error mechanism. Define how the system detects a disconnected or implausible reference.
Document the connector pins and cable option in the PCB design review. A board assembled for a four-wire sensor may need explicit sense jumpers; a six-wire variant should not accidentally short the remote sense pair to local excitation through a default option. List fitted and not-fitted links for each variant and make them visible in the test procedure. A correct schematic with the wrong assembly option can defeat the intended measurement topology.
Check input range and common mode separately
A small differential signal does not guarantee that both ADC input pins are within their permitted range. For a balanced bridge, the signal terminals may sit near the midpoint of excitation, but offset, fault conditions and the actual bridge arrangement must be considered. Verify each pin relative to the converter supplies as well as the differential span. Include the PGA restrictions at the selected gain, especially if the sensor and ADC use different supply voltages.
Likewise, reference full scale is device-specific. Do not assume that every converter accepts a differential input of plus or minus Vref divided by gain. The , for example, defines its own full-scale relationship, reference limits and operating conditions. Use those definitions when translating a bridge’s millivolts-per-volt specification into headroom. Leave room for offset and overload recovery instead of choosing gain solely to fill the nominal span.
| Design boundary | Question to answer | Failure hidden by a simple ratio diagram |
|---|---|---|
| Excitation | Which terminals define Vexc? | Cable drop excluded from the reference |
| Reference | Can the input accept and settle to that voltage? | Loading, clipping or disconnected sense line |
| Signal | Are differential and common-mode limits both satisfied? | PGA overload despite a small bridge output |
| Assembly | Which wiring and jumper option is fitted? | Remote sense defeated by a local connection |
Budget the errors that remain and test their mechanisms
Match the disturbance paths, not just the nominal voltages
At DC, two nodes may have the same voltage. During a transient, different filtering can make them disagree. If the bridge signal path responds through one transfer function and the reference through another, a common excitation disturbance can appear with different amplitude or phase at conversion time. The ratio then contains a residual transient rather than complete cancellation.
For a small fractional excitation disturbance e, a simplified linearized model gives a fractional ratio disturbance approximately equal to e 脳 (Hs 鈭� Hr), where Hs and Hr are the normalized signal and reference responses at that disturbance frequency. This is a diagnostic model, not a complete switched-capacitor ADC simulation. It shows why equal DC gains alone are insufficient. Buffer response, source impedance, sampling behavior and digital filter timing may also matter.
The ADS1232 application circuit specifically calls for matched signal and reference RC time constants when those filters are used. Follow the device-specific network guidance and calculate the effective differential response, rather than copying identical printed component values without examining the topology. Then verify settling after excitation switching, channel changes and wakeup. A large reference capacitor can reduce noise while making startup data invalid for longer than expected.
Keep offset, gain and sensor behavior in separate columns
An additive input offset is not proportional to excitation, so ordinary ratiometric division does not remove it. With an 8 mV full-scale bridge signal, a 4 microvolt input-referred offset is 0.05% of full scale. Its percentage of a near-zero reading is much larger and can be a misleading metric. Specify low-end behavior in engineering units or as a fraction of full scale, and distinguish zero calibration from the stability of zero over time and temperature.
Amplifier or ADC gain error multiplies the ratio and therefore remains after excitation cancellation. For an illustrative gain drift of 2 ppm per degree Celsius over a 40 degree change, the first-order gain shift is 80 ppm, or 0.008%. This arithmetic is not a claim about the selected converter. Use its specified drift conditions and calibration behavior, then include bridge sensitivity drift and mechanical effects separately.
Excitation also changes sensor power. For the simplified 350 ohm bridge above, 4 V across the bridge dissipates about 45.7 mW; 5 V would dissipate about 71.4 mW. The ratio can cancel the immediate electrical scaling while self-heating changes the physical sensor behavior over a longer interval. Check the sensor’s rated excitation and mounting conditions. More excitation may improve signal amplitude without improving the complete measurement once heating and power consumption are included.
Creep, hysteresis, mounting stress, humidity, connector thermoelectric effects and cable movement are other possible contributors. Do not assign them arbitrary numerical values merely to complete a spreadsheet. Identify which are relevant, obtain applicable component data or measurements, and retain an explicit unknown where evidence is missing. An honest uncertainty budget is more useful than a precise-looking sum of unsupported assumptions.
Use controlled tests to identify the residual error
Begin with a stable electrical bridge simulator or resistor network to isolate the electronics, then repeat with the actual sensor and installed cable. Record fixture stability and temperature. Apply a small permitted excitation change while holding the simulated input ratio constant. Observe both the transient and settled output. If only the transient moves, inspect path settling; if the settled ratio moves, inspect reference location, loading and other supply-dependent terms.
Next vary the cable arrangement within the intended product options. Compare local reference and remote sense only with a controlled, documented configuration. Do not insert an undocumented jumper during a demonstration and leave production to infer its purpose. Use thermal tests to distinguish electronic zero drift from sensor heating, and repeat zero and span checks after returning to the starting condition. Preserve raw ADC codes alongside filtered engineering values so software smoothing does not hide recovery time.
AC excitation can help reject some offset mechanisms, but it introduces switching, synchronization and settling requirements. provides a concrete AC-excited weigh-scale example. Treat it as a reference architecture to evaluate, not as a drop-in upgrade for every bridge. Confirm that the sensor wiring, converter and application bandwidth support the proposed scheme before changing the board.
Make the prototype and production checks reproducible
For prototype PCB assembly, provide the exact reference network, optional components, connector pinout and sensor simulator requirements. Test pads should support the intended measurement without adding large probe loops or disturbing sensitive nodes. Record which checks are performed with a sensor attached and which use a fixture. A bare-board continuity test cannot establish bridge accuracy or validate firmware scaling.
When preparing the assembly and test scope, separate workmanship inspection from electrical calibration and product-level sensor validation. Define serial-number association for calibration data, allowable adjustment ranges and the response to an out-of-range result. Do not use calibration to conceal a wrong component or assembly option. Preserve a known-good configuration and its limits so rework can be followed by the appropriate retest.
Include stackup, copper and fabrication notes in the PCB manufacturing package, while keeping the reference topology and acceptance procedure under engineering revision control. The file-format checklist helps make that handoff unambiguous. For a quotation or build review, share the bridge-interface requirements with the sensor sensitivity, wiring, quantity, assembly options and proposed test scope. State unresolved accuracy questions explicitly rather than treating the word ratiometric as a performance guarantee.