A platen hanging its own weight on the cell holds the signal 0.2 mV below zero, and a range bounded below by 0 mV shows nothing until compression passes the bias. Configure a bipolar millivolt range and tare the bias in logic. Where the card has no bipolar span, support the platen weight or preload the cell into compression.
A press-style test stand shows zero load right up to the moment the screw is already well into the sample, then the trend line jumps. The cell, the wiring and the analog card are usually fine. The input range bottoms out at 0 mV, and the platen's own weight holds the signal on the wrong side of zero between strokes. This article works the numbers for the common version of the fault: a 2 mV/V, 500 lb cell at 10 VDC excitation carrying a 5 lb platen. It puts the three fixes in the order to try them: a bipolar range with a logic tare, a mechanical change to the load path, or two channels wired with opposite polarity.
Why does the load reading stay at zero until the force is already there?
The platen hangs on the cell in tension, which puts the bridge output 0.2 mV below zero. A range configured as 0-39.0625 mVDC has no code for a negative signal, so the display sits at zero until compression force cancels the platen weight.
A strain gauge bridge drives its output positive in one loading direction and negative in the other. Tension and compression are the two signs of one differential signal. A range bounded below by 0 mV represents only the positive side, so everything below 0 mV reads as no load. On this stand the platen hangs its weight on the cell between test strokes, so the idle signal is about -0.2 mV. The first pounds of pressing only cancel the platen weight before the signal crosses into the readable side.
That also explains the two inches of travel before counts appear: it is the sample's compliance building up that same force. The cell responds to force, so a travel interval points to force buildup or to a gap in the load train, and the raw input value tells you which.
What does a 2 mV/V load cell put out at 10 VDC excitation?
Full-scale output in millivolts equals the mV/V rating times the excitation voltage. So 2 mV/V at 10 VDC gives 20 mV across the cell's 500 lb rating, which works out to 0.04 mV per lb.
The same relation moves with the supply. If excitation sags to 9 VDC, full scale falls to 18 mV and the scale becomes 0.036 mV/lb, so meter the excitation at the cell connector before trusting any scaling. Output scales with the load fraction, which is what makes the straight-line math usable at every point in the stroke. The formula and its parts are laid out on a page on load cell mV/V ratings and excitation voltage.
Worked example: what does the platen weight do to the displayed force?
At 0.04 mV per lb the platen bias is 5 lb × 0.04 mV/lb = 0.2 mV below zero. A unipolar input hides the first pounds of force and then reports every load exactly the platen weight light.
Take the screw pressing the sample with a total force F. The cell carries F minus the platen weight, the signal is that net load times 0.04 mV/lb, and a unipolar display divides the signal by 0.04 mV/lb from a zero reference:
| Screw presses the sample with | Cell sees | Signal | Unipolar display | Bipolar display, tare at idle |
|---|---|---|---|---|
| 0 lb, platen hanging free | -5 lb | -0.2 mV | 0 lb | 0 lb after tare |
| 3 lb | -2 lb | -0.08 mV | 0 lb | 3 lb |
| 45 lb | +40 lb | +1.6 mV | 40 lb | 45 lb |
| 205 lb | +200 lb | +8.0 mV | 200 lb | 205 lb |
The unipolar column is what the operator sees on this stand now: nothing while the platen weight alone is on the cell, then a reading that trails the sample's true force by exactly that amount at every point on the stroke. The bipolar column is the same signal with the idle value subtracted once in logic, and it is correct from the first pound.
How do you prove the range is the problem before touching the wiring?
Put the stand in a safe state, stop the drive, and read the raw input with the platen hanging free. A value parked at zero or at the bottom of the range while the platen is loading the cell is the signature of a unipolar range hiding a tension bias.
Then work down this list of symptoms:
| Raw input at idle and during travel | What it points to | Check next |
|---|---|---|
| Parked at 0 mV or clipped at the range bottom, platen hanging free | Tension bias outside a unipolar range | Configure a bipolar millivolt range, or move the bias mechanically |
| Rises only after contact, then tracks net force | Sign clipping | Verify against a known weight set on the platen |
| Flat through the whole travel, moves only at contact | Gap or lost motion in the load train | Compare encoder position at contact with the position where load first appears |
| Pinned at the top limit or jumping wildly | Wiring, excitation or range mismatch | Stop and meter the 10 VDC excitation and the differential pair against the module manual |
| Jitter or slow wander near zero | Noisy millivolt signal or shared reference potential | Check the analog base wiring before the scaling |
Keep the sample clear of the platen during these checks so test force and platen bias stay out of the same reading. On this stand the input is an F0-04THM millivolt card on its unipolar millivolt range. The card-side checks for that class of input are the ones in a troubleshooting guide for millivolt analog inputs. If the raw value itself jitters or wanders, treat that as its own fault: erratic analog input readings usually trace to the analog base wiring, and the raw input at idle is the place to see it.
Which fix reads tension and compression on one input?
Configure the analog card for a bipolar millivolt range, let the idle reading show the platen as a negative number, and subtract that bias once in logic. The displayed force is then correct from the first pound.
The bipolar entry lives in the same range configuration where the unipolar millivolt range was set. Bridge-input front ends on industrial analog cards commonly offer bipolar spans as small as ±10 mV, so the range is a standard offering. The subtraction at idle is the same operation as taring a scale signal from an analog input in PLC logic, applied to the load channel once after setup or at the start of each stroke.
Two things to check before committing:
- Sign handling downstream. Logic and displays that treat the raw integer as unsigned drop the negative idle reading before the tare ever runs. The same fault shows up where an HMI drops the sign on bipolar RTD temperatures.
- Resolution. A symmetric bipolar span shares the converter's counts across twice the millivolts, so one count is worth twice as much force. Size it against the test: a 1 lb step occupies 1 lb × 0.04 mV/lb = 0.04 mV, so the bipolar span's millivolts per count must sit below that with margin. If it does not, take the mechanical route and keep the unipolar range.
When do you fix it mechanically instead?
Go mechanical when the card has no bipolar millivolt range or when the halved counts can no longer resolve the smallest force the test records. Carry the platen weight off the cell, preload it with springs, or turn the stand over so the cell stays in compression.
Three arrangements do the job:
- Carry the platen weight on pins or a bracket. A one-direction restraint lets the platen move freely in the test direction and takes its weight in the other. The cell idles near 0 mV and the existing unipolar range keeps its full counts per pound. Tare the small residual zero offset after fitting, and check that the restraint carries weight outside the measured force path.
- Preload the cell into compression with springs. Springs pull the platen up into the cell, the idle output sits at a positive tare equal to spring force minus the platen weight, and the sample force only adds compression from there. Keep the preload well above the cell's zero drift. Keep preload plus maximum test force inside the 500 lb rating, and use springs stable enough that the preload stays part of a repeatable zero.
- Turn the stand over. Put the moving platen and its cell at the bottom and press up into a fixed platen. The platen's weight now loads the cell in compression, the signal never crosses zero in normal operation, and the unipolar range reads the whole stroke. For a new build this is the clean arrangement.
Can two input channels bring the full resolution back?
Wire the cell's signal pair to two input channels at once with the polarity reversed on the second. Each channel then sees a positive signal for one loading direction, and the original counts per pound survive on both.
Logic selects whichever channel reads positive and assembles the signed force value. Do three things before trusting the combined number. Verify each channel on its own against a known weight in its own direction. Decide the crossover, meaning which channel owns readings within a few counts of zero. And damp or reject noise right at the crossover, because a sign or scale error between the channels shows up as a false step at zero load. The price is a second channel and its wiring on every cell, so this earns its place only when a bipolar range and a mechanical change are both out of reach.
What should you change first on this stand?
Set the millivolt input to its bipolar range and tare the platen bias in logic. Keep the mechanical options for a unipolar-only card or a test that needs the original counts per pound.
Recommendation: Set the millivolt input to its bipolar range and tare the unloaded platen value in logic whenever the card offers that range and the doubled force per count still resolves the smallest step the test records. When the card is unipolar-only, fit a bracket or pins that carry the platen weight so the cell idles near 0 mV, and leave the input on its configured unipolar millivolt range. Use spring preload when the idle reading must sit positive of zero by design. Reserve the two-channel reversed-polarity wiring for the case where a bipolar range and a mechanical change are both unavailable.
Check the result against a calibrated weight set on the platen. From the first pound of force to the last, the displayed value should match it.
Questions
Does swapping the load cell sense leads fix a unipolar input?
Reversing the plus and minus sense leads only chooses which loading direction reads positive. A range bounded below by 0 mV still has no code for the opposite sign, so the platen bias simply moves to the other side of zero and the hidden force problem stays. Reading both directions on electronics takes a bipolar range, a second channel wired with reversed polarity, or a mechanical arrangement that keeps the cell on one side of zero at idle.
How much force does the 0-39.0625 mV range cover with this cell?
At 0.04 mV per lb (20 mV across 500 lb at 10 VDC), a 39.0625 mV span corresponds to 39.0625 ÷ 0.04, about 977 lb, well above the cell's 500 lb rating. The span is not the limiting factor here; the missing negative sign is. Meter the excitation at the connector too, because output falls in proportion to excitation voltage and the scaling follows it.
What resolution does the bipolar range cost on a 500 lb test?
On a symmetric bipolar span, twice the millivolts share the same converter counts, so one count is worth twice as much force. If one count is worth 0.05 lb on the unipolar span, it is worth 0.1 lb on the bipolar span. A test recording 1 lb steps is unaffected. A test recording 0.1 lb steps should read the converter's count span in its manual before committing to the wider range.
Can the platen weight itself serve as the tare reference?
Yes, and it doubles as a verification weight. With a bipolar range the idle display should read the hanging platen as a negative number, about -5 lb at 0.04 mV/lb. The tare subtracts exactly that value. After taring, a known weight set on the platen should read its true value, which confirms sign, scale and tare together before the first test stroke.
Planning information, not legal or engineering advice. Rules differ by place: check codes and permits with your local authority and your licensed trades.