A load cell power supply provides the excitation voltage a strain gauge bridge needs in order to produce a signal at all. Without it a load cell does nothing: the bridge sits there, unpowered, producing no output no matter how hard you press on it.
But calling it a power supply undersells what it does, and the misunderstanding costs people accuracy.
A strain gauge bridge does not produce a voltage. It produces a proportion. That is what the sensitivity figure on your calibration certificate actually means — two millivolts per volt is not two millivolts, it is two thousandths of whatever you excite the bridge with. Excite it with ten volts and full capacity gives twenty millivolts. Excite it with five and the same load gives ten.
So the excitation voltage is not powering the measurement. It is one of the two numbers being multiplied together to make it.
Follow that through and the consequence is uncomfortable. Every error in the excitation appears in the reading, one for one. If the supply is a tenth of a percent high, the reading is a tenth of a percent high. If it drifts through the afternoon, the reading drifts with it. If it sags when a second sensor is connected, everything reads low from that moment onward.
And none of it can be filtered, averaged or calibrated away, because your calibration certificate was written at a particular excitation voltage. Change that voltage and the certificate is describing a system you no longer have. The error is not noise; it is a scale factor, and scale factor errors are steady, believable and invisible.
Which is why a bridge excitation supply is really a voltage reference that your measurement is multiplied by — and why the numbers on this specification sheet mean something quite different from the same numbers on a general purpose supply.
Regulation of five hundredths of one percent is not a comment about how well the supply holds up. It is a direct entry in your measurement error budget, because that is exactly how much your reading can move for reasons that have nothing to do with the load.
Fifty parts per million per degree Celsius works out at five thousandths of one percent for every degree the room moves. Across a twenty degree swing between a cold morning and a warm afternoon that is a tenth of a percent, arriving straight in the reading, with nothing to indicate it happened.
Half a millivolt of noise and ripple sounds like nothing until you notice it is being compared against ten volts — five thousandths of one percent of everything you measure, all the time.
None of those figures are impressive in isolation. What makes them the right figures is that they are all comfortably smaller than the load cell they will be feeding, which is precisely where a supply belongs: invisible, so that when a measurement disappoints you can go and look at the sensor and the mechanics without wondering about the wall socket.
Line voltage in, a transformer, a regulator, a stable low voltage DC out. Two of those choices were made deliberately for bridge work and both are worth understanding, because they are the reason this is not simply a small power supply with a different label.
The transformer is specified for low interwinding capacitance, and that phrase is doing a great deal of work.
Inside any transformer the primary and secondary windings sit next to each other, separated by insulation. That is what provides the isolation — but two conductors separated by an insulator are also, unavoidably, a capacitor. There is a small parasitic capacitance between the two windings whether anyone wanted one or not.
At the mains frequency that capacitance passes almost nothing, which is why it is usually ignored. The problem is that the mains does not only carry the mains. It carries everything the building's electrical system produces: switching transients from variable frequency drives, the arc when a contactor closes, high frequency hash from every switch-mode supply in the plant. And capacitance is a low impedance to high frequencies. The junk goes straight across the winding-to-winding capacitance and out the other side, bypassing the isolation completely, and arrives at your sensor as common mode noise.
A low interwinding capacitance transformer closes that path. The construction — usually an electrostatic shield between the windings — gives the high frequency energy somewhere else to go instead of onto the secondary.
That matters more on a bridge than on almost anything else you could power. The excitation supply connects to both ends of the bridge, so noise riding on it appears at the sensor and then travels back down the signal wires as common mode. Your amplifier has to reject it, and however good its rejection is, it is better not to have sent the noise in the first place. Every stage downstream benefits from a supply that is quiet where it matters.
The regulator is linear rather than switching, which is a deliberate and increasingly unusual choice.
A switching regulator works by chopping the input at high frequency and filtering the result. It is efficient, it runs cool and it is what almost every inexpensive supply on the market uses. It also generates high frequency noise as an inherent part of how it operates — not as a defect, but as a description of the mechanism. Filtering reduces it; nothing removes it.
A linear regulator does not chop anything. It simply drops the surplus voltage across a device and turns it into heat, which is wasteful and warm and produces essentially no switching noise at all. For most applications that is a poor trade. For a measurement made in millivolts, it is the right one.
The output is adjustable across a wide range, and the reason is compatibility rather than performance. Four volts to fifteen covers essentially every excitation voltage anything has ever been specified at — five, eight, ten and twelve are all common, and older equipment can want something less usual. That means one supply can serve a mixed bench, and more importantly it means you can set the voltage your calibration certificate was written at rather than accepting whatever a fixed supply happens to produce and living with the discrepancy.
The current available sets how many bridges you can feed, and the answer depends on the voltage you chose. This catches people out because both variables move. A standard 350 ohm bridge draws about three times as much current at fifteen volts as it does at five. At ten volts, five of them fit comfortably within the supply's capability; at fifteen volts, three; at five volts, ten. A low resistance bridge changes the arithmetic sharply — a 120 ohm bridge at ten volts uses more than half the available current on its own. Do the division before the system is built rather than after.
Most of our instruments generate their own excitation, so the first question is whether you need a separate supply at all. Our applications engineers will tell you honestly if you do not.
You need one when nothing else in the chain is providing excitation. A data acquisition card with differential voltage inputs and no bridge completion. A chart recorder. A test setup built around a voltmeter. An OEM amplifier that expects the excitation to arrive from elsewhere. In all of those the sensor has no power source and something has to supply one.
You also need one when the built-in supply has run out. Most conditioners and meters can drive one or two bridges. If you are feeding four sensors, or a low resistance bridge, or several instruments that must all see the same excitation, a dedicated supply solves it and takes the load off equipment that was not designed to carry it.
And you want one when the existing supply is the suspect. If a measurement drifts and the sensor has been eliminated, powering the bridge from a known-good reference is the fastest way to find out whether the excitation was the problem. That is worth having on a shelf in any laboratory that troubleshoots other people's installations.
Decide the voltage from the calibration certificate, not from habit. Set the supply to whatever the sensor was certified at and the certificate applies directly, with no correction and no explaining. Set it to something else and every number needs adjusting, which is a step somebody will eventually forget. If you are commissioning a system from scratch, choose a voltage, write it down, and record it alongside the calibration data.
Count the bridges against the current, at your chosen voltage. Resistance and voltage both matter and people usually remember only one of them. Work out the current one bridge draws, multiply by how many you have, and leave margin — a supply running at its limit regulates less well than one running comfortably, and margin is what lets somebody add a sensor next year without redesigning anything.
Use the sense connections if the sensor is any distance away. They exist to remove the error that cable resistance introduces between the supply and the bridge, and connecting them costs two extra conductors and a few minutes. Skipping them on a long run is one of the more common reasons a system reads low for years without anyone noticing.
Give it somewhere sensible to live. Physically it is compact — roughly four inches by two by three — with a line cord, and it expects to sit on a bench top, a shelf or a rack shelf rather than inside anything. It runs warm because a linear regulator turns surplus voltage into heat, so leave it some air rather than burying it under something. Its temperature range is wide, but a supply that is being cooked will not hold its output as well as one that is not.
Tell us the sensors you are powering, their resistance and how many there are, what excitation your certificates specify and what is going to read the signal, and we will confirm whether you need a separate supply and which one.
PSM-R Load Cell Power Supply Applications.
The Transducer Techniques PSM-R is a self-contained AC powered bridge excitation supply using a low interwinding capacitance transformer and a high stability adjustable linear regulator, designed specifically for driving transducer bridges, with remote sensing to eliminate line drop errors.
- Data Acquisition Systems: Supplies excitation where an acquisition card provides differential voltage inputs but no bridge power of its own.
- Laboratory Test Benches: One adjustable supply serves a mixed bench of transducers specified at different excitation voltages.
- Calibration Work: Setting the exact voltage a certificate was written at allows calibration data to be applied without correction.
- Multi-Sensor Installations: Powers several bridges from a single reference so every sensor sees the same excitation.
- Long Cable Runs: Remote sensing removes the error that cable resistance introduces between the supply and the transducer.
- Electrically Noisy Environments: A low interwinding capacitance transformer blocks the high frequency path that carries line noise through an ordinary transformer.
- Low Level Measurement: A linear regulator avoids the switching noise that is inherent to a switch-mode supply.
- OEM Amplifier Support: Provides excitation to signal conditioning boards and modules that expect it from an external source.
- Chart Recorders and Voltmeters: Allows a transducer to be read directly by instrumentation that has no bridge excitation capability.
- Troubleshooting and Fault Finding: A known-good reference supply isolates whether a drifting measurement originates in the excitation.
The PSM-R works with our full range of load cells, torque sensors and pressure transducers, driving full Wheatstone bridge transducers from 4 to 15 Vdc.
Frequently Asked Questions
What does a separate supply do that an instrument's built-in excitation does not?
Three things. It provides excitation where nothing else does — a data acquisition card with voltage inputs, a chart recorder, a voltmeter setup, an OEM board expecting external excitation. It provides more current than a built-in supply can, which matters when several bridges are involved or when a low resistance sensor uses most of an instrument's budget on its own. And it provides a reference you can trust when you are troubleshooting, because powering a suspect system from a known-good supply is the quickest way to find out whether the excitation was ever the problem. If your instrument already excites your sensor adequately, you do not need this.
Why does excitation accuracy translate directly into measurement error?
Because a strain gauge bridge is ratiometric — its output is a proportion of its excitation rather than a fixed voltage. That is what the sensitivity figure means: two millivolts per volt is two thousandths of whatever you supply. So the reading is the excitation multiplied by the sensitivity multiplied by the applied load, and an error in any one of the three appears in the answer at full strength. A supply half a percent high produces readings half a percent high. Nothing downstream can detect it, because the signal is exactly what a correctly working system would produce if the load really were that much larger.
What does a low interwinding capacitance transformer do?
It blocks a noise path that ordinary transformers leave wide open. Inside any transformer the primary and secondary windings sit adjacent, separated by insulation — which makes them, unavoidably, a small capacitor. At sixty hertz that capacitance passes almost nothing, so it is usually ignored. But the mains carries far more than sixty hertz: drive switching, contactor arcs and high frequency hash from every switch-mode supply in the building. Capacitance is a low impedance to high frequencies, so all of that couples straight across to the secondary and bypasses the isolation. A low interwinding capacitance design closes the path, and on a bridge that matters especially, because noise on the excitation returns down the signal wires for your amplifier to deal with.
Why a linear regulator rather than a switching supply?
Because a switching regulator makes high frequency noise as an inherent part of how it works. It chops the input at high frequency and filters the result — efficient, cool-running and what nearly every inexpensive supply uses, but the noise is a description of the mechanism rather than a defect, and filtering reduces it without removing it. A linear regulator chops nothing; it drops the surplus voltage across a device and turns it into heat. That is wasteful and it runs warm, and for most applications it would be the wrong choice. For a measurement made in millivolts it is the right one.
Why is the output adjustable from 4 to 15 volts?
So that one supply can match whatever your equipment and your paperwork expect. Five, eight, ten and twelve volts are all common excitation figures across different manufacturers and different eras, and older instrumentation sometimes wants something less usual still. A wide adjustable range means a single supply serves a mixed bench rather than needing one per sensor type. More importantly it means you can set the exact voltage your calibration certificate was written at, so the certificate applies directly with no correction factor for somebody to forget.
How many bridges can it power?
It depends on two things and people usually remember only one: the bridge resistance and the voltage you have selected. A 350 ohm bridge draws roughly three times as much current at fifteen volts as at five. Set to ten volts, five standard 350 ohm bridges fit within the supply's capability; at fifteen volts about three; at five volts around ten. A 120 ohm bridge at ten volts uses more than half the available current by itself. Work out the current one bridge draws at your chosen voltage, multiply by the number you have, and leave margin — a supply working at its limit regulates less well than one with room to spare.
Why would a power supply need sense connections?
Because what matters is the voltage at the sensor, not the voltage at the supply, and cable resistance makes those two different. Current flowing out to the bridge and back drops voltage in the wires, so the bridge receives slightly less than the supply is producing, and since output is proportional to excitation the readings come out proportionally low. The sense connections are two additional wires that measure the voltage actually arriving at the transducer and let the regulator correct for the difference. They carry almost no current, so they lose nothing themselves. On a short cable the effect is negligible; on a long one it is the difference between a system that is right and one that is quietly and consistently wrong.
How stable is the output with temperature?
Fifty parts per million per degree Celsius, which is five thousandths of one percent per degree. Put that in context: a twenty degree swing between a cold morning and a warm afternoon moves the excitation by about a tenth of a percent, and therefore moves every reading by the same amount. That is a good figure — comfortably smaller than the drift of most load cells it will be powering, which is where a supply belongs. It is also a reminder that a measurement expected to hold across a wide temperature range needs every element of the chain considered, not just the sensor.
What do the ripple and regulation figures mean for my reading?
They are error budget entries rather than abstract quality numbers. Line and load regulation of five hundredths of one percent means the excitation can move by that much as the supply voltage varies or as the load on it changes — and because output is proportional to excitation, your reading moves by the same fraction, for reasons nothing to do with force. Half a millivolt of ripple compared against ten volts is another five thousandths of one percent riding on everything you measure. Both are small, and being small is the whole point: they should be well below the sensor's own uncertainty so that the supply never becomes the thing you have to investigate.
How is it powered and where should it sit?
It plugs into line power and is a small enclosed unit of roughly four by two by three inches, intended for a bench, a shelf or an equipment rack. Line isolation is well over a thousand volts and the operating temperature range is generously wide. Two practical points: leave it some air, because a linear regulator produces heat by design and a supply that is being cooked will not hold its output as well as one that is breathing; and keep it away from the sensor cable run where you conveniently can, since there is no reason to place a transformer next to a millivolt signal.
Questions From The Field
My readings changed after the power supply was replaced.
Almost certainly the excitation voltage is not the same as it was, and that changes every reading proportionally. Measure the actual voltage at the sensor terminals rather than trusting the setting or the label, and compare it against what the system was calibrated at. A supply set to twelve volts where the previous one produced ten will make everything read twenty percent high, steadily and believably, with nothing that looks like a fault. Set the new supply to the documented voltage, re-check at the sensor, and confirm against a known load before trusting anything. This is also the argument for writing the excitation voltage on the calibration record in the first place.
Can I use a laboratory bench supply instead?
You can get a reading, and for a rough check that may be all you need. What a general purpose bench supply usually lacks is the specification that matters here: regulation and temperature stability tight enough not to appear in your measurement, ripple low enough to be irrelevant against a millivolt signal, sense connections to remove cable error, and a transformer built to keep line noise out of the secondary. Many bench supplies are also switch-mode, which puts high frequency noise onto the very thing your amplifier is trying to reject. For setting up and troubleshooting, use what you have. For a measurement anyone will rely on, use a supply intended for bridges.
Can I run it at 15 volts to get a bigger signal?
Check the sensor before you do. Higher excitation does produce a proportionally larger signal, but the bridge dissipates the extra as heat inside the sensing element, and a load cell that is warming itself drifts. Most sensors have a maximum excitation figure and it is worth finding it rather than assuming fifteen volts is available to you. Watch for the symptom too: a zero that walks slowly for several minutes after power-up and then settles is a bridge heating up. There is also a paperwork consequence — if your certificate was written at ten volts, running at fifteen means every number needs correcting, which is a step that eventually gets missed.
I set 10 volts at the supply but the sensor is receiving less.
That is cable resistance and it is exactly the problem the sense connections solve. The excitation current has to flow through the wires to the sensor and back, and every ohm of conductor drops a little voltage on the way, so the bridge sees less than the supply is producing. Wire the sense connections through to the transducer terminals and the regulator corrects for it automatically. If sense wiring is not practical, the alternatives are heavier conductors, a shorter run, or calibrating the system as installed and never changing the cable — but sense wiring is the right answer and it costs two wires.
Can one supply feed several sensors read by separate instruments?
Yes, provided the total current is within the supply's capability, and there is a real advantage to it: every sensor sees exactly the same excitation, so a drift in the supply moves all the channels together rather than pulling them apart. That makes comparisons between channels more trustworthy than if each had its own supply. The things to watch are the current budget at your chosen voltage, and grounding — several instruments sharing one excitation supply are now connected to each other through it, so think about where the system is grounded and avoid creating a loop. Send us the arrangement if you would like it checked.
There is mains hum on my measurement.
Work out whether it arrived through the excitation or was picked up on the signal wiring, because the remedies are different. Try running the sensor from a battery briefly: if the hum disappears, it is coming in through the supply and the answer is a supply built to keep it out. If it remains, the signal cable is picking it up, and the fixes are routing away from power wiring, grounding the shield at one end only, and keeping the run as short as practical. Also check whether the hum is at the line frequency or twice it — twice line frequency usually points at a rectifier and therefore at the supply rather than at pickup.
What excitation was my load cell calibrated at, and does it matter?
It is on the calibration certificate, and yes it matters, though perhaps less than people fear. Because the sensor is ratiometric, its sensitivity in millivolts per volt is largely independent of the excitation you use — that is the whole point of expressing it per volt. What must match is the number your instrument was scaled against: if your readout was set up assuming ten volts and the supply now produces twelve, the readings are wrong by that ratio. So the rule is not that you must always use the certified voltage, but that the excitation, the sensitivity and the instrument's scaling must all agree with each other. Record all three together.
The unit runs warm. Should I be concerned?
Warm is expected. A linear regulator works by dropping the surplus voltage across a device and converting it to heat, so the further you set the output below its maximum and the more current you draw, the more heat there is to get rid of — it is the direct cost of the low noise you bought it for. Give it clear air rather than enclosing it or stacking something on top, and keep it out of direct sun and away from other heat sources. What would concern me is hot rather than warm, a smell, or an output that sags as the unit heats up. Any of those is worth reporting to us rather than working around.