Load Cell Amplifier Gain Selection

A load cell converts a force into a small differential voltage, and the amplifier that follows decides how much of that voltage the system can actually use. Gain selection is not a single number: it depends on the full scale output of the bridge, the resolution the system needs and the noise and drift of the amplifier itself.

The Signal the Bridge Produces

A strain gauge load cell is a bridge of four resistors on a metal body. The rated output is quoted in millivolts per volt of excitation, usually between one and three, so a two millivolt per volt cell fed with five volts produces ten millivolts at full load.

That is the entire signal range, and the resolution required might be one part in ten thousand of it, which is a microvolt. Both figures come from the same bridge, so the front end has to work with a signal that is tiny in absolute terms and demanding in relative terms.

The bridge impedance also matters. A three hundred and fifty ohm bridge draws about fourteen milliamps from a five volt supply, and the excitation source has to hold that current steady while the load changes. Any variation in the excitation appears directly in the reading.

Choosing the Gain

gain selection starts from the full scale output and the input range of the converter. A ten millivolt signal and a two point five volt reference need a gain of two hundred and fifty to use the whole range, which is a high but practical figure for an instrumentation amplifier.

Using the whole range is not always right. Leaving headroom for overload, for zero drift and for the tolerance of the cell itself keeps the system from clipping during a shock load, and a gain that uses sixty to eighty percent of the range is a reasonable compromise.

The gain also amplifies the noise and the offset of the amplifier along with the signal. Gain bandwidth product and input noise density both matter, and a part selected only for its gain accuracy may have a noise that limits the effective resolution to far less than the converter could deliver.

The instrumentation amplifier and Its Limits

An instrumentation amplifier is three operational amplifiers in a fixed topology with laser trimmed resistors, giving a high common mode rejection and a gain set by one external resistor. That resistor sets the gain and also its temperature stability, so a low drift type is worth the price.

Common mode rejection is the reason the topology exists. The bridge sits at half the excitation voltage, which is two and a half volts in the example, and the amplifier must ignore that while measuring ten millivolts. A rejection of a hundred decibels leaves a residual of tens of microvolts, which is already a visible fraction of the signal.

Rejection falls with frequency, so a switching excitation or a nearby motor can inject a common mode signal at a frequency where the rejection is much lower. The excitation source should therefore be quiet, and the bridge wiring should be symmetric.

Reference, Offset and Drift

The reference input of the amplifier sets the output for a zero differential input. In a single supply system it is usually set to half the converter range so that both tension and compression can be measured, and its stability appears directly in the reading.

Offset drift is the error that dominates at zero load. An amplifier with a drift of one microvolt per degree produces an apparent weight change of several grams on a scale with a two millivolt per volt cell, which is why a chopper stabilised front end is common in precision weighing.

temperature drift in the load cell itself adds to the problem. The cell has a specified temperature effect on zero and on span, and the firmware can correct for it with a temperature sensor on the cell body. Measuring the effect, rather than assuming a figure, is what makes the correction useful.

Excitation and Ratiometric Measurement

bridge excitation sets the sensitivity, so a variation in it becomes a scale error. A ratiometric arrangement, in which the converter reference is derived from the same excitation supply, cancels that error because both the signal and the reference change together.

A constant current excitation has the advantage that it removes the effect of the lead resistance on the bridge, at the cost of a more complex supply. A constant voltage supply is simpler but the lead resistance drops some of the voltage before it reaches the bridge, and that drop changes with temperature.

The excitation level trades sensitivity against self heating. A higher voltage gives more signal, and it also dissipates more power in the strain gauges, which raises their temperature and shifts the zero. The datasheet gives a recommended value, and exceeding it is a common cause of a scale that drifts while it is loaded.

Filtering and the Sample Rate

A low pass filter ahead of the converter removes the noise above the band of interest and prevents aliasing. A weighing system needs a bandwidth of a few hertz at most, so a two pole filter with a corner near ten hertz is generous and still removes a great deal of noise.

Mains hum at fifty or sixty hertz is the largest interfering signal, and an integrating converter that rejects the mains period is a good partner to the analog filter. The rejection is only complete if the integration time is an exact multiple of the mains period, which is a setting rather than an accident.

Filtering in the digital domain is cheaper and can be sharper, but it cannot remove what has already been aliased. Analog filtering first, digital filtering second, is the order that works. The measurement discipline behind that split is described in our guide to mixed signal board design.

Layout and Guarding

Keep the amplifier close to the connector and the bridge wiring short and symmetric. The differential input is a high impedance node at a high common mode voltage, so any asymmetry in the two paths converts common mode into differential error.

Guard the input traces with a track at the same potential as the common mode, or at least keep the two input traces equal in length and parallel. A ground plane underneath helps, but it should be a quiet analog plane rather than the digital ground of the rest of the board.

The excitation return carries the full bridge current, so it must not share a trace with the signal return. A separate pair for excitation, brought to the connector and returned on its own conductor, keeps that current out of the measurement path. The assembly and release points that catch a mistake here are collected in our PCB design release checklist.

Verification and Calibration

Verify with a known weight or a dead weight tester and record the reading at several points across the range. A plot of error against load shows the non linearity of the cell and the gain error of the amplifier separately, because they have different shapes.

Check the zero stability over an hour with the load removed. A slow rise points to self heating in the bridge or to a thermal gradient across the amplifier, and moving air across the board is a quick way to confirm the second.

Check the gain stability with a calibrated shunt resistor placed across one arm of the bridge, which simulates a known load electrically. That standard is more repeatable than a physical weight and is the one to use for a production test. The wider reliability questions are covered in our guide to component tolerance and reliability design and the board quality points in judging PCB quality.

FAQ

What gain should I use for a load cell? Set it so that full load uses sixty to eighty percent of the converter range. That leaves headroom for overload and for the tolerance of the cell.

Why does my scale drift with temperature? The amplifier offset drift and the cell temperature effect on zero both contribute. Use a chopper amplifier and correct the cell error with a temperature measurement.

Is constant current excitation better than constant voltage? It removes the lead resistance error, but needs a more complex supply. Constant voltage is simpler and reads well when the wiring is short and symmetric.