A load cell amplifier is the piece between a sensor and whatever is going to use its reading. It powers the bridge, takes the few thousandths of a volt that come back, and turns them into a signal the rest of your system can accept — a voltage a data acquisition card will digitise, or a current loop a controller will read.
Put like that it sounds like plumbing. It is not. Your load cell's performance is fixed the day it is made, and your data system's performance is fixed the day you buy it. What sits between them is the only part of the chain you actually get to decide, and on a board like this one those decisions are made with jumpers — which makes them easy to make badly.
Three of those settings determine what kind of measurement you end up with, and no amount of processing afterwards will recover a bad choice among them.
The first is bandwidth, and here it is unusually wide. This board is jumper-selectable anywhere from 100 hertz to 30 kilohertz. That is a three hundred to one range, and the top of it is genuinely fast — fast enough for impact, drop testing, crash work, machine transients and vibration, the things a load cell cannot show you through a conventional weighing instrument because the instrument filters them away before you ever see them.
But wider is not better, and this is the mistake worth avoiding. Broadband noise rises with the square root of the bandwidth you allow through. Move from 100 hertz to 30 kilohertz and you have opened the window by a factor of three hundred, which multiplies the noise arriving with your signal by roughly seventeen. If your measurement genuinely contains 30 kilohertz of information, that is a price worth paying. If it does not, you have degraded your own measurement for nothing. Published guidance on strain measurement puts the rule simply: select a low-pass cutoff that still passes the measurement frequency of the strain you are measuring — and no more than that.
The second is excitation, and most people treat it as a power setting when it is really a noise setting. A bridge's output is proportional to the voltage exciting it, so running a sensor at ten volts instead of five doubles the signal arriving at the amplifier. That means the amplifier needs half as much gain to reach the same output — and since an amplifier's own noise gets multiplied by its gain, halving the gain halves the noise it contributes. Micro-Measurements state the chain of reasoning directly: more excitation means more output, more output means less gain, and less gain means less noise. The limit is self-heating in the sensor, which shows up as drift; the practical advice is to increase excitation until you can see that beginning, then come back from it.
The third is matching the board to the sensor you actually have. Sensitivity is jumper-selectable from half a millivolt per volt up to ten, with gain adjustable from fifty to four thousand behind it. Set that correctly and your sensor's full capacity uses the full output range. Set it carelessly and you spend the rest of the installation working with a fraction of the resolution you paid for.
Get all three right and the conditioner disappears from your error budget entirely. Nonlinearity through this board is five thousandths of one percent of full scale — better than any load cell made, by a wide margin, which is exactly what you want from the component in the middle. It should never be the reason a measurement is wrong.
Excitation out, millivolts back, amplify, output. The parts worth knowing are how each stage is adjusted and what each adjustment costs you.
The input stage expects very little and must not be given too much. It works with signals from a couple of millivolts up to about a hundred, which at a sensible excitation voltage covers essentially every bridge sensor you are likely to connect. Its own offset is around ten microvolts — a hundredth of a millivolt — which on a typical load cell's twenty millivolt full scale output is a rounding error rather than a specification you need to plan around.
The output stage is where you choose how the answer leaves the board. Voltage in two ranges, or current in three, including a 4–20 milliamp loop. Voltage is simplest for a data acquisition card sitting next to the sensor. Current earns its place over distance and in electrically noisy plant, because a current loop is indifferent to voltage drop along the cable and because the live zero at the bottom of a 4–20 range means a dead loop can be told apart from a genuine reading of nothing. Both are selectable rather than fixed at manufacture.
Zero and span are trimmed with potentiometers, and the ranges are generous on purpose. Zero adjusts by at least ten percent of full scale and span by at least twenty-four percent, which is enough to absorb a fixture's dead weight, a preload built into a mounting, or a sensor whose actual sensitivity differs slightly from its nominal figure. It is also enough to disguise a mechanical problem, so use the adjustment to accommodate what you understand and investigate what you do not.
The excitation supply is adjustable from five to ten volts and will deliver up to thirty milliamps into a minimum of 350 ohms. Read that load figure carefully, because it decides what you can connect: 350 ohms is one standard bridge, not two. Two 350 ohm sensors wired in parallel present 175 ohms, which is below the minimum, so a multi-sensor arrangement needs a board per sensor or a different approach. This is a single channel conditioner and it is honest about being one.
There is a push-button on the board that checks your whole chain without touching the machine. Pressing it switches a precision resistor across the bridge, unbalancing it by a known amount and producing a predictable output — so you can confirm that the sensor, the cable, the amplifier, the wiring and the display or data system are all still doing what they did at commissioning, without applying a real load. That matters most in exactly the places where applying a real load is impractical: a sensor built into a machine, a vessel in service, a rig that has to stay assembled.
What it simulates depends on your bridge, which is why the value is worth knowing. The fitted resistor produces very close to one millivolt per volt on a 350 ohm bridge — so on a typical 2 mV/V load cell, pressing the button should give you almost exactly half of full scale. On a 1000 ohm bridge the same resistor produces a considerably larger simulated output, and on a 120 ohm bridge a considerably smaller one, because the simulated unbalance depends on the ratio between the shunt and the bridge arm. That is precisely why the board also has a socket for a calibration resistor of your own: fit a value chosen for your sensor and your preferred check point, and the button becomes a check against a number that means something to you.
Power is undemanding and deliberately industrial. Anything from about twelve to twenty-six volts DC, drawing well under a tenth of an amp, which means the same 24 volt supply that runs everything else in a control panel runs this too. A separate mains adapter is available where there is no DC supply to hand.
And the board is small enough to design in rather than to house. Under three inches long, under two wide, under an inch thick. It goes inside a machine, behind a panel, on a DIN rail with the available clip, or into a customer's own enclosure. The full specification holds from freezing to fifty-five degrees Celsius, with derated operation across a much wider range for equipment that has to live somewhere less comfortable.
Choosing a conditioner is mostly about being honest concerning what your measurement contains and where the board has to live. Our applications engineers would rather see the whole chain than one link of it.
Start with how fast the thing you are measuring actually is. This is the decision everything else follows from. A tank filling, a fixture holding, a part being weighed, a slowly applied proof load — all of that lives well below a hundred hertz and wants the narrowest setting available. A press stroke, an actuator cycling, a machine running, a web under changing tension — hundreds of hertz. An impact, a drop, a fastener failing, a structure ringing after it is struck — that is where the top of the range earns its keep. Write down a number in hertz before you touch the jumpers, and if you cannot, tell us what the event is and we will help you estimate it.
Then set the excitation as high as your sensor tolerates. It costs nothing and it buys signal-to-noise. Watch for zero drifting slowly after power-up, which is the sensor warming itself, and back off if you see it — static measurements are the most sensitive to this because they give the drift time to matter, while a short dynamic test may not care.
Choose the output for the receiving end, not for convenience. Voltage if a data acquisition card or an analog input is within a few feet. Current if the signal has to cross a plant, share a duct with power wiring, or reach a controller expecting a process signal. A 4–20 milliamp loop is the safest thing you can send down a long cable and it is available here without an extra module.
Work out what will hold the board and what will plug into it. This is an OEM component rather than a finished instrument. It has DB9 connectors, and the mating connectors are supplied by you and soldered by you — a normal arrangement for something designed into a product, and a surprise if you were expecting a plug-and-play box. Budget the time, or tell us what you are building and we will discuss the alternatives. A DIN rail clip, a replacement terminal block kit, a trim pot tool and a mains adapter are all available.
Decide now whether you want a calibration check you can trust. If the sensor is going somewhere you cannot easily load by hand afterwards, work out what the built-in resistor will simulate on your particular bridge before you install anything, and record the reading you get on day one. That single recorded number is what makes the button useful for the next ten years. If you would rather it simulated a value of your choosing, fit your own resistor in the socket provided and record that instead.
Consider whether one channel is really what you need. The excitation supply drives a single 350 ohm bridge, so several sensors means several boards. That is often the right answer — independent channels do not interact and each can be set up for its own sensor — but if you are summing several cells into one total, say so and we will point you at an arrangement built for it.
Send us the sensor, its sensitivity and bridge resistance, how fast your measurement changes, what the output has to reach and where the board will physically sit, and we will confirm the configuration before you order. We would far rather do that than have you jumper it and wonder.
LCA-9PC Load Cell Amplifier Applications.
The Transducer Techniques LCA-9PC provides low cost dedicated signal conditioning for a single bridge type load cell or torque sensor, with jumper-selectable sensitivity from 0.5 to 10 mV/V, jumper-selectable bandwidth from 100 Hz to 30 kHz, and a choice of voltage or current outputs.
- OEM Machine Integration: A board small enough to design into a machine, powered from the same DC supply as the rest of the control system.
- Dynamic Force Testing: Bandwidth up to 30 kHz captures impacts, drops and fast transients that a weighing instrument filters away before they reach the display.
- Press and Actuator Monitoring: Mid-range bandwidth settings follow a working stroke faithfully while still rejecting machine and electrical noise.
- Static Weighing and Load Holding: The narrowest bandwidth setting delivers the quietest possible output where the measurement changes slowly.
- PLC and Controller Interfacing: Selectable 4–20 mA output feeds a plant controller directly, with no separate transmitter required.
- Data Acquisition Systems: Voltage outputs at ±5 V or ±10 V match the input ranges of common acquisition cards without further scaling.
- Long Cable Runs: Current loop output travels across a plant without the voltage drop and interference that affect a low level signal.
- Torque Sensor Conditioning: The same board conditions bridge type torque sensors, with sensitivity and gain set to suit.
- Field Verification: Push-button shunt calibration confirms the whole measuring chain is intact without applying a physical load to the sensor.
- Panel and DIN Rail Installation: A DIN rail clip allows the board to be mounted alongside other control components in a standard enclosure.
The LCA-9PC works with our full range of load cells and torque sensors, and is supported by mains adapter, DIN rail clip, terminal block and trim pot tool accessories.
Frequently Asked Questions
What does the LCA-9PC do, and what does it connect to?
It provides dedicated signal conditioning for one bridge type load cell or torque sensor: it excites the bridge, amplifies the millivolt output and delivers a usable voltage or current signal to whatever comes next. Sensitivity is jumper-selectable from 0.5 to 10 mV/V, bandwidth from 100 Hz to 30 kHz, and the output can be ±5 V, ±10 V, 0–16 mA, 0–20 mA or 4–20 mA. Connection is through DB9 connectors. It is a single channel board designed to be built into equipment rather than a finished boxed instrument.
What does the bandwidth jumper actually change?
It sets how fast a signal the board will pass through to its output. At the narrowest setting anything changing faster than about a hundred times a second is filtered away, which is exactly what you want for weighing and static work — vibration, machinery and electrical interference all disappear before they reach your reading. At the widest setting the board follows changes up to thirty thousand times a second, which is what makes impact and dynamic testing possible. The two settings are not better and worse versions of each other; they are different instruments, and choosing between them is the most consequential thing you will do during setup.
How much does extra bandwidth cost me in noise?
More than most people expect. Broadband noise grows with the square root of the bandwidth you allow through, so widening the window by a factor of three hundred — which is the span from the narrowest setting to the widest — multiplies the noise arriving alongside your signal by roughly seventeen. That is why the correct approach is to pick the narrowest setting that still passes your measurement rather than the widest the board offers. Published strain measurement guidance says the same thing: choose a cutoff that still passes the frequency content you are trying to measure, and stop there. If you are unsure how fast your event is, tell us what it is and we will help you work it out.
What sensitivity and gain range does it cover?
Sensitivity is jumper-selectable from 0.5 to 10 mV/V and gain runs from fifty to four thousand, accepting inputs from a couple of millivolts up to about a hundred. That range covers essentially every bridge sensor we make and most that we do not, including the 2 and 3 mV/V load cells that make up the bulk of the catalogue. Matching the jumper to your sensor matters because it decides how much of the output range your sensor's full capacity actually reaches — get it right and capacity lands near full output, get it wrong and you are using a fraction of the resolution available to you.
Why is the excitation adjustable, and where should I set it?
Adjustable from 5 to 10 volts, and the setting is really about noise rather than power. A bridge produces output in proportion to its excitation, so a sensor run at ten volts delivers twice the signal it delivers at five — which means the amplifier needs half the gain to reach the same output, and since an amplifier multiplies its own noise by its gain, less gain means a cleaner result. The general guidance is therefore to run excitation as high as the sensor tolerates. The limit is self-heating: the bridge dissipates power and warms itself, which shows as slow zero drift after power-up. Increase until you can see that starting, then come back a little.
What outputs can the board produce?
Two voltage ranges, ±5 V and ±10 V, and three current ranges, 0–16 mA, 0–20 mA and 4–20 mA. Voltage is the straightforward choice when the receiving equipment is close by — the ±10 V range in particular matches the input span of most data acquisition hardware directly, so no further scaling is needed. Current is what you want over distance or through electrical noise, since a current loop is unaffected by voltage drop in the cable, and 4–20 mA additionally lets a broken loop be distinguished from a real reading of zero. The bipolar voltage ranges also mean tension and compression, or clockwise and counterclockwise, come out as positive and negative without anything extra.
How accurate is the conditioner itself?
Nonlinearity is a maximum of 0.005 percent of full scale and input offset is around ten microvolts. Both figures are deliberately far better than any load cell you could connect — a very good load cell manages perhaps 0.03 percent nonlinearity, so the conditioner is roughly six times better than the best sensor in the catalogue. That is the correct relationship for a component in the middle of a chain: it should be invisible in your error budget, so that when a measurement is wrong you can concentrate on the sensor and the mechanics rather than wondering about the electronics.
What does the push-button calibration check do?
It switches a precision resistor across the bridge, unbalancing it by a known amount and producing a predictable output, so that the sensor, the cable, the board and everything downstream can be verified without applying a physical load. The fitted resistor is a 0.025 percent part, and it produces very close to 1 mV/V on a 350 ohm bridge — so a 2 mV/V load cell should read almost exactly half of full scale when the button is pressed. The value it simulates changes with bridge resistance, which is why a socket is provided for a calibration resistor of your own if you would rather check at a different point or have a sensor of different resistance.
What power supply does the board need?
Anything from roughly twelve to twenty-six volts DC, drawing between forty and sixty milliamps depending on how it is configured. In practice that means the 24 volt supply already present in most control panels will run it without anything additional, which is part of why the board suits being designed into equipment. Where there is no DC supply available, a mains adapter is offered as an accessory. Note that the excitation the board provides to your sensor is separate from and independent of this supply voltage.
How is it mounted and connected?
The board is about 2.7 by 1.7 by 0.8 inches, or a little longer with the DB9 fitted, so it is intended to live inside something rather than to sit on a bench. A DIN rail clip is available for mounting alongside other control components. Connection is via DB9 male and female connectors, and the mating connectors are supplied and soldered by you — normal for a component designed into equipment, but worth knowing before it arrives. Full performance is specified from 0 to 55 degrees Celsius, with derated operation over a considerably wider range.
Questions From The Field
I moved the bandwidth jumper up and the output got noisy.
That is the board doing exactly what you told it to. A wider bandwidth passes more of everything, including all the noise that the narrower setting was removing, and the increase is steeper than it feels — noise grows with the square root of bandwidth, so a big jump in the setting produces a large jump in what you see. The question to ask is whether you actually needed the extra speed. If your measurement changes over tenths of a second, go back to the narrow setting and the noise goes with it. If you genuinely need the bandwidth, then the noise is real and the remedies are elsewhere: raise the excitation, improve the shielding and the cable routing, and make sure the signal is as large as it can be before it is amplified.
What should the push-button give me on my particular load cell?
Work it from your bridge resistance. On a 350 ohm bridge the fitted resistor simulates close to 1 mV/V, so a 2 mV/V sensor gives about half of full scale and a 3 mV/V sensor about a third. A 1000 ohm bridge produces a substantially larger simulated value from the same resistor and a 120 ohm bridge a substantially smaller one, because what is simulated depends on the ratio between the shunt and the bridge arm it is placed across. The most useful thing you can do is press the button on day one, with everything working and freshly calibrated, and write down what you get. From then on the number itself matters less than whether it has changed.
My output will not reach full scale with the sensor at capacity.
Check the jumpers before the potentiometers. If the sensitivity setting is higher than your sensor's actual figure, the board is expecting a bigger signal than it will ever receive and full capacity will land short. Confirm the sensor's real sensitivity from its certificate rather than from the model number, then confirm the excitation is where you think it is, since output scales with it and a supply set to five volts produces half of what ten would. The span adjustment covers at least twenty-four percent of full scale and is intended to trim out the remainder — if you are asking it to make up much more than that, something upstream is set wrong.
Can I use this to capture an impact or a drop test?
Yes, and it is one of the few things in the range that can. Set the bandwidth wide enough for the event, which usually means the upper part of the available range, and remember that the sensor and the fixture are now part of the frequency response too — a load cell has its own natural frequency and anything above it is not measured faithfully no matter how quick the electronics are. Whatever records the output has to be fast enough as well, since a conditioner passing thirty kilohertz into a system sampling ten times a second achieves nothing. Tell us the sensor, the event duration and the recording equipment together and we will sanity check the chain.
My 4–20 mA output behaves on the bench and not once it is installed.
Look at the loop rather than the board. A current output needs the total loop resistance — the cable there and back plus the receiving device's input resistance plus anything else in series — to stay within what the output can drive, and a bench setup with a short lead and a meter is a far easier load than a long run into a panel with a couple of devices in the loop. Add up what is actually in the circuit. Also check that only one device is powering the loop, since a board driving current into a supply that is also driving it is a common and confusing fault. If it still misbehaves with the loop simplified, come back to us with the arrangement.
Do I really have to solder my own connectors?
For this board, yes — the DB9s are on the board and the mating connectors are supplied and terminated by you. That is entirely normal for a component intended to be designed into equipment, where the cable, the connector and the routing are all part of somebody's own product, but it is a genuine surprise to anyone expecting a boxed instrument they can plug into. If soldering nine-way connectors is not something you want in your build process, tell us what you are making and we will talk through what else in the range fits, because there are packaged alternatives and the right answer depends on the volume you are building.
Can I run two load cells from one LCA-9PC?
Not from the excitation supply on this board. The specification calls for a minimum load of 350 ohms, which is exactly one standard bridge — two 350 ohm sensors in parallel present 175 ohms and that is below the limit. That is not an oversight; this is deliberately a single channel conditioner. If you need several sensors, the usual answer is a board each, which has the advantage that the channels are independent and each can be set up for its own sensor. If what you actually want is several cells summed into one total, that is a different requirement and we will point you at the right arrangement.
Everything is quiet on the bench and noisy once it is on the machine.
Then the machine is contributing something, and the useful step is to work out whether it is electrical or mechanical. Narrow the bandwidth right down: if the noise largely disappears, it was high frequency content, which points at electrical pickup or at genuine vibration. If a slow disturbance remains, something is loading the sensor for real — a guard resting against a fixture, a hose pulling, thermal movement in a frame. On the electrical side, keep the sensor cable away from motor leads and drive wiring rather than running alongside them, ground the shield at the instrument end only, and raise the excitation so the signal is larger before it is amplified. Those three together fix the large majority of installations.