TECH NOTE 07 / Robotics and acquisition
Connecting load cells in parallel: total force versus independent channels
Compatible parallel bridges produce one combined signal; individual forces cannot be recovered from that signal alone.
Compatible parallel bridges produce one combined signal; individual forces cannot be recovered from that signal alone.
Matched load-sharing sensors can be calibrated for total force even though their raw voltage is a weighted average.
Mechanical series sensors see the same force; multiplying their combined reading by sensor count can overcount it.
01 / ENGINEERING DECISION
Separate the mechanical load path from the electrical connection.
Connecting load cells in parallel is an electrical choice. It does not tell you whether the sensors share a supported load or sit sequentially in one force path. Define the physical quantity first: total force, distribution among supports, or independent tendon forces.
For load-sharing supports, an approved matched summing arrangement can be useful. For individual tendon tensions, use independently recoverable measurements. Separate channels do not automatically mean simultaneous conversion or electrical isolation; specify those requirements independently.
| Mechanical task | Desired result | Measurement arrangement |
|---|---|---|
| Load distributed among supports | Total applied force | Approved matched summing system or separate channels plus validated sum. |
| Separate robotic tendons | Each local tendon force | Independent measurement values and suitable timing. |
| Sensors in one axial series path | The same transmitted force | Appropriate local readings; do not count the same force repeatedly. |
Supporting scope: Beckhoff ED336x: commissioning a measuring bridge ↗
02 / ENGINEERING DECISION
A compatible bridge network has a weighted output.
As a simplified differential Thevenin model, let each compatible sensor have an unloaded output voltage V_i and output resistance r_i. With a sufficiently high-impedance input and the approved common excitation/reference arrangement, the combined voltage is approximately sum(V_i/r_i) divided by sum(1/r_i). Equal output resistances give an arithmetic average.
This is a model of compatible passive bridge sources, not permission to join arbitrary outputs. Input bridge resistance determines excitation loading; output resistance determines the weighting in this approximation. Neither should be inferred from an unrelated sensor specification.
V_combined ≈ sum(V_i / r_i) / sum(1 / r_i)
Supporting scope: Beckhoff ED336x: commissioning a measuring bridge ↗
03 / ENGINEERING DECISION
Worked example A: average voltage can represent total force.
Assume four matched sensors, each 100 N capacity with a 10 mV full-range change and equal output resistance. One sensor at 80 N and three at zero produce 8, 0, 0 and 0 mV. Their average is 2 mV. Calibrating the matched load-sharing system for 400 N at 10 mV maps that combined signal to 80 N total.
Four sensors at 20 N each also produce a 2 mV combined output and the same 80 N total. The two force distributions cannot be distinguished from this one output. This is the information lost by summing, even when the total-force scale is correct.
| Individual forces / N | Individual outputs / mV | Combined output | Scaled total |
|---|---|---|---|
| 80, 0, 0, 0 | 8, 0, 0, 0 | 2 mV | 80 N |
| 20, 20, 20, 20 | 2, 2, 2, 2 | 2 mV | 80 N |
| 80, 80, 80, 80 | 8, 8, 8, 8 | 8 mV | 320 N |
04 / ENGINEERING DECISION
Worked example B: the same wire connection can overcount a series path.
Now place those four ideal sensors sequentially in a single axial path carrying 80 N. Ignoring inertia and other intervening loads, each reads 80 N. The parallel electrical output is 8 mV. Applying the previous 400 N load-sharing calibration would display 320 N even though the transmitted axial force is 80 N.
Nothing in the connector tells the readout which mechanical arrangement exists. The error is the wrong physical calibration model, not a failure of the averaging equation. Preserve the distinction in system drawings, software names and acceptance tests.
| Arrangement | Meaning of four 80 N readings | Correct quantity |
|---|---|---|
| Four load-sharing supports | Four separate contributions | 320 N total when the model applies. |
| Four sensors in one series path | The same force observed four times | 80 N transmitted force, under the stated static assumptions. |
| Four independent tendons | Four independently useful values | Retain the four channels; a sum loses distribution. |
05 / ENGINEERING DECISION
Check supply loading separately from output weighting.
For four hypothetical 350 ohm bridge inputs in parallel, equivalent input resistance is 87.5 ohm. At 5 V, ideal excitation current is about 57.1 mA, before other loads and instrument details. That is a supply calculation using input resistance, not the output-resistance weighting model.
Beckhoff’s ED336x instructions explicitly require attention to supply power and manufacturer-approved, matched parallel sensors. Its published current limit belongs to that terminal. Confirm the actual instrument’s voltage, current, sense arrangement, thermal behavior and fault handling rather than copying a permissible sensor count.
1 / R_input,total = sum(1 / R_input,i); excitation current ≈ V_exc / R_input,total
| Item | Calculation / review |
|---|---|
| Four equal 350 ohm inputs | 350 / 4 = 87.5 ohm |
| Assumed 5 V supply | 5 / 87.5 A ≈ 57.1 mA |
| Actual instrument | Check available current, cable drop, sense and temperature. |
| Matched sensors | Check sensitivity per force, zero and output impedance. |
Supporting scope: Beckhoff ED336x: commissioning a measuring bridge ↗
06 / ENGINEERING DECISION
Nominal sensitivity alone does not create a matched sum.
Equal mV/V at different force capacities does not mean equal volts per newton. Unequal output impedances, zero offsets, excitation drops and sensitivity differences change the aggregate result. Use manufacturer-approved matching/trim methods and calibrate the installed load-sharing system, including its relevant force distributions.
Do not parallel independent active voltage outputs, current-loop transmitters or digital lines as if they were passive bridge sources. Follow the exact device and bus documentation; separate addressing or synchronization is a different architecture from passive summing.
| Shortcut | Why it fails | Better request |
|---|---|---|
| Same mV/V label only | Capacity and output weighting can differ | Approved matched sensitivity/impedance and system calibration. |
| More sensors than supply supports | Excitation voltage/current limits may be exceeded | Input-resistance/current budget and approved wiring. |
| Short active outputs together | They are not compatible passive sources | Independent inputs or a documented aggregation interface. |
| Infer four forces from one sum | Multiple distributions give the same output | Separate measurement values. |
Supporting scope: Beckhoff ED336x: commissioning a measuring bridge ↗
07 / ENGINEERING DECISION
Independent readings still need a timing specification.
For a hypothetical sequential scan of 16 channels at 16,000 conversions/s, the per-channel rate is 1,000 samples/s and the first-to-last nominal sample-time separation is 15/16,000 = 0.9375 ms. A simultaneous converter has a different timing structure. Neither case is fully described by saying “16 kS/s”.
For a moving multi-tendon mechanism, retain channel identity, calibration association, timestamps, filtering and saturation/dropout state. A software total can be computed from independent values where the physical model allows it; the independent values cannot in general be reconstructed after a one-output passive sum.
- Specify whether the goal is total force, force distribution or individual tendon feedback.
- Choose approved bridge summing or independent acquisition accordingly.
- Define per-channel rate, skew, isolation and total control latency separately.
- Calibrate the actual sensor associations and validate loading distributions and fault cases.
Supporting scope: NI-DAQmx: Bridge Sensor Scaling ↗
08 / ENGINEERING DECISION
Send a mechanical sketch and an acquisition goal with the RFQ.
Include the number and capacities of sensors, their load path, desired individual or total readings, bridge input/output resistance where known, excitation and instrument input. Add rate, synchronization, output protocol and calibration requirements.
The paired-kit and tendon application pages are the next step for independent force channels. Use the downloadable worksheet to record the architecture; it is an owned planning aid rather than a generic approved wiring diagram.
Resolve the fit, signal and calibration questions.
Does connecting two load cells in parallel add their output voltages?
Not in the simplified passive bridge model described here. Compatible equal-output-resistance sources produce an average voltage; an appropriately matched load-sharing system can then be calibrated for total force.
Can a parallel connection ever measure total force?
Yes, with a suitable mechanical load-sharing arrangement, approved matched sensors, excitation and input, and system calibration. Averaging raw voltage does not prevent a correct total-force scale.
Can I recover each tendon force from one combined signal?
Not in general. Different force distributions can produce the same combined voltage. Retain independently recoverable measurement values.
Are sensors in mechanical series measuring separate force contributions?
In the stated ideal static axial path, they observe the same transmitted force. Intervening inertia or external loads can change local forces, so model the actual mechanism rather than simply summing sensor count.
Can I use bridge input resistance as output weighting resistance?
Not automatically. Input resistance describes excitation loading; the output-source model uses output resistance. Obtain the relevant values and approved summing arrangement.
Do independent channels guarantee simultaneous sampling?
No. A converter can multiplex channels. Specify per-channel rate, scan skew and synchronization, as well as electrical isolation if needed.
ENGINEERING RFQ / SAMPLE REQUEST
Send the load case.
Define the interfaces.
Include both thread ends, working and peak force, available space and the signal your controller expects.
Company and university email domains are preferred. Prototype and research inquiries are welcome.
inquiry@inlineloadcell.com ↗