Load Cell Amplifier Gain: Getting Useful Voltage Output

Why Load Cell Signals Are So Annoying

Load cells are everywhere — industrial scales, force measurement rigs, torque sensors, even that bathroom scale you keep meaning to hack. They're reliable, accurate, and based on well-understood physics. But they produce absolutely tiny signals, and that's where most projects go sideways.

A typical load cell is a Wheatstone bridge of strain gauges bonded to a metal element. When you apply force, the element deforms, the resistances change, and you get a differential voltage out. Sounds simple enough. The problem is the sensitivity: most load cells output somewhere between 1 mV/V and 3 mV/V at full scale. That "mV/V" notation trips people up constantly, so let's be clear — it means millivolts of output per volt of excitation, at full rated load.

So if you're exciting a 2 mV/V load cell with 5 V and you put the full rated load on it, you get 10 mV. That's your entire signal range. Try feeding 10 mV into a 12-bit ADC with a 3.3 V reference and you'll use about 12 counts of your 4096. Not great.

The Signal Chain Math

Let's break down what's actually happening. The full-scale output voltage from your load cell is:

where $V_{EX}$ is your excitation voltage and $S$ is the sensitivity in mV/V (which you need to convert to V/V for the math to work out — so 2 mV/V becomes 0.002 V/V).

After amplification:

where $G$ is your amplifier gain. You want $V_{AMP}$ to be close to your ADC's full-scale input without clipping.

The sensitivity of your overall system — the volts per unit of load — becomes:

where $F_{FS}$ is your full-scale load in whatever units you're working with (kg, lb, N, doesn't matter as long as you're consistent).

And if you're targeting a specific ADC input range, you can work backwards to find the required gain:

Worked Example: A 100 kg Load Cell

Let's say you've got a 100 kg load cell with 2 mV/V sensitivity. You're running 10 V excitation (pretty common for industrial cells), and you want to feed the signal into a microcontroller ADC with a 5 V reference. You've got an INA125 instrumentation amp sitting in your parts bin.

First, find the full-scale output:

That's your entire signal swing at 100 kg. At 50 kg, you'd see 10 mV. At 1 kg, you'd see 200 µV. This is why you need amplification.

To fill a 5 V ADC range, you need:

So a gain of 250 would put your full-scale load right at 5 V. In practice, you'd back off a bit — maybe aim for 4.5 V at full scale to give yourself headroom. That's a gain of 225.

Now, the INA125 sets gain with a single resistor according to $G = 4 + \frac{60\text{k}\\Omega}{R_G}$. For G = 250:

You'd probably use a 240 Ω or 249 Ω resistor depending on what's in your kit, giving you a gain of 254 or 245 respectively. Close enough.

With a gain of 250, your system sensitivity becomes:

Or equivalently, 20 counts per gram on a 12-bit ADC. That's workable resolution for most weighing applications.

Excitation Voltage Tradeoffs

Higher excitation means more signal, which is great. But there are limits.

Most load cells specify a maximum excitation voltage — often 10 V or 15 V. Go higher and you'll cook the strain gauges. The power dissipation in the bridge is P = V_EX² / R_bridge, and with a 350 Ω bridge at 10 V, that's nearly 300 mW. The heat has to go somewhere, and if it causes thermal gradients in the load cell body, you'll see drift.

Lower excitation reduces self-heating but gives you less signal to work with. For battery-powered applications, 3.3 V or 5 V excitation is common, and you just accept that you need more gain. A 2 mV/V cell at 3.3 V excitation gives only 6.6 mV full-scale — you'd need a gain of about 750 to fill a 5 V ADC range.

There's also the question of excitation stability. Your load cell output is ratiometric to the excitation voltage. If your excitation drifts by 0.1%, your reading drifts by 0.1%. Using a precision reference for excitation, or better yet, using a ratiometric ADC that references against the same supply, eliminates this error source.

Common Mistakes and Gotchas

Forgetting the mV/V Units

I see this constantly. Someone reads "2 mV/V" on a datasheet and plugs 2 into their gain calculation as if it were volts. The result is a gain that's off by a factor of 1000. Your amplifier output will be microvolts when you expected volts, and you'll spend an hour convinced the load cell is broken.

Ignoring Offset Voltage

Load cells have a zero-load offset, typically specified as a percentage of full scale. A cell with ±0.05% offset at 10 V excitation could have ±5 mV of offset before you even apply load. After a gain of 250, that's ±1.25 V of offset at your ADC input. If your ADC can't handle negative voltages (most single-supply MCU ADCs can't), you've got a problem.

The fix is either AC coupling (rarely appropriate for weighing), level shifting, or — most commonly — just accepting the offset and calibrating it out in software. But you need to make sure your amplifier output stays within the ADC's input range across the entire offset range plus signal range.

Choosing Gain Without Considering Noise

Higher gain amplifies everything, including noise. If your load cell and wiring pick up 50 µV of interference, a gain of 500 turns that into 25 mV of noise at your ADC. That might be several LSBs on a 12-bit converter.

The instrumentation amp's input-referred noise matters too. A cheap amp with 50 nV/√Hz input noise will contribute about 5 µV RMS over a 10 kHz bandwidth — not terrible, but it adds up. Good load cell amps like the INA125 or AD620 are designed for this application and have lower noise.

Not Filtering Before the ADC

Load cell signals are slow. Weight doesn't change at kilohertz rates (unless something has gone very wrong). A simple RC lowpass filter before the ADC — even just 1 kΩ and 100 nF for a 1.6 kHz cutoff — knocks down high-frequency noise and prevents aliasing. Skipping this step because "I'll filter in software" is lazy and costs you dynamic range.

Using the Wrong Amplifier Topology

A load cell outputs a differential signal. Using a single-ended op-amp configuration throws away the common-mode rejection that makes load cells work well in noisy environments. Always use an instrumentation amplifier or a proper differential-to-single-ended front end.

Overdriving the Amplifier

If someone stands on your 100 kg scale, you might see 150 kg briefly. With a gain set to put 100 kg at 5 V, that overload drives the amp to 7.5 V — except it can't go there, so it clips. Some amps recover from clipping gracefully; others take milliseconds to come back, during which your readings are garbage. Either design for the overload case or add protection circuitry.

Practical Gain Selection

Don't target exactly 100% of your ADC range. Aim for 80-90% to leave headroom for offset, overload, and component tolerances. If your calculations say you need a gain of 250, consider using 200 instead.

Also consider whether you'll be doing any analog filtering or level shifting that eats into your voltage range. If you're adding a 0.5 V offset to keep the signal positive, your usable range is now 0.5 V to 5 V, not 0 to 5 V.

And check what gain values your chosen amplifier can actually produce. The INA125's gain equation means certain gains are easier to hit than others. You might find that 200 requires a standard resistor value while 225 requires something odd.

Try It

The math isn't hard, but it's easy to make unit errors or forget a factor of 1000 somewhere. Open the Load Cell Amplifier Gain calculator to quickly check your numbers. Plug in your excitation voltage, load cell sensitivity, full-scale load, and proposed gain — you'll immediately see the full-scale output, amplified voltage, system sensitivity, and what gain you'd need to hit a 5 V ADC range. It's a sanity check that takes ten seconds and might save you an afternoon of debugging.