Selecting a filling scale for a conveyor cell is a load-cell-and-weighbridge decision, not an integrator decision: a stainless parallelogram-style load cell sampling at 32 kHz, mounted in a single-idler carriage on a straight, flat section of belt, is the combination most OEM white papers recommend for ±0.5% to ±0.125% system accuracy [S1].
The "filling scale" name covers three different physical products — a belt scale for continuous flow on a moving conveyor, a static checkweigher for discrete packages, and a bench scale for hand-loaded fills — and the wrong pick is the single most expensive procurement mistake on a packaging line.
What "filling scale on a conveyor cell" actually means
A belt scale is "generally not well suited to discrete weighing, check weighing, or flow monitoring of batches or samples less than 10 minutes in duration," so any conveyor-cell filling duty shorter than 10 min per batch belongs on a static checkweigher or a net weigher, not a belt scale [S3]. Continuous in-line filling — bulk bag, tote, or open-mouth bag — where material flows for tens of minutes to hours is the natural fit for an integratordriven belt scale [S3].
For mining and aggregate reference points, a properly installed belt scale can hit 0.5% accuracy, with the best designs reaching 0.125% on a stable conveyor, and the loss curve is steep: a 5% accuracy drift on a 200 t/h ore conveyor running 16 h/day, 300 days/year, costs roughly $960,000/year at a $20/t extraction cost [S1]. The same arithmetic applies to a food or chemical filling cell — small percentage errors on high-tonnage conveyors compound fast.
Load cell: parallelogram, pivot, or pancake
The first selection gate is the load cell topology. A solid stainless-steel parallelogram-style load cell converts the downward belt force into a millivolt signal through a strain gauge deforming less than 1 mm, sampled at 32 kHz — that is the baseline Siemens white-paper figure for mining-class accuracy [S1]. A pivot-mounted weigh idler through a lever system "approximately doubles the response time" of the load cell because the pivot introduces a mechanical low-pass stage ahead of the sensor, so designers who need the 32 kHz dynamic behaviour have to accept a parallelogram (no-pivot) carriage [S1].
For OEM skid builders, a universal pancake-style load cell such as the FUTEK LCF400 is a common fit when the scale frame is pre-engineered and the cell capacity is matched to the idler load; capacities in the 45 kg, 100 kg, and 200 kg range are the standard build slots documented across integrator manuals [S5][S6].
The single most common field defect is misalignment: weighbridge installation guides call for the mounting pipes "MUST be centered on the holes on the load cell assemblies" and require 90° geometry on both A and B idler sides, with the warning "ALL MEASUREMENTS must be EXACT" [S7]. A 1 mm offset on a parallelogram is enough to introduce a horizontal force component that the load cell reads as extra weight.
Weighbridge: single-idler, dual-idler, or multi-idler
Single-idler weighbridges are the standard for conveyor-cell filling because one idler is the easiest geometry to keep level and aligned; dual-idler and multi-idler (idler assemblies with two or more weighed idlers) average out idler-spacing error and conveyor-troughing noise, but they cost more and need tighter installation [S3][S5]. The integrator's setup wizard explicitly asks for "Number of weigh idlers (1 for single idler scale and 2 for dual idler scale)" and "Load cell capacity found on Load cell assembly label" as the first two input parameters — every downstream accuracy figure is a function of those two numbers [S5].
Carriage rigidity is not negotiable. The Thermo Scientific belt-conveyor-scale handbook lists the carriage's three mechanical duties as: (1) rigidity with minimal deflection, (2) torsional stability, and (3) transmitting only the vertical force V to the load cell while rejecting the horizontal force H — if the carriage flexes, the load cell reads H as a false V, and the integrator sees drift that no software calibration can fix [S4]. For a filling-cell duty with frequent start-stop cycles, a single-idler welded-steel carriage with two pivot points (the parallelogram layout) outperforms a stamped plate carriage in long-term drift.
Conveyor location: flat run, no curves, no tripper
Scales must sit on a straight, flat section of conveyor with idlers equally spaced on both A and B sides of the belt; curved conveyors, convex or concave sections, ploughs, and trippers all create belt-tension and load-distribution errors that a downstream integrator cannot fully compensate [S3][S7]. Material feed should land upstream of the weigh idler at a distance that lets the belt settle into a stable cross-section — typically one to two idler spacings — and any control gate at the feed point changes the dynamic mass profile and must be flagged in the integrator setup [S3].
Belt tension is a hidden variable. A take-up device located too close to the scale, or a stacker conveyor with a moving head, varies the belt tension under the weigh idler and adds drift proportional to belt elasticity; OEM guidance is to keep the scale at least 5–6 idler spacings away from the take-up, or to move the take-up to the return run [S3]. Material turbulence on a curved conveyor is the second common cause of integrator noise; the Siemens application guidelines flag it as a layout decision to be made before the scale is ordered, not after.
Integrator outputs: 4-20 mA, HART, and what the cell PLC actually reads
The integrator outputs three values: flow rate (t/h or kg/min), belt load (kg/m), belt speed (m/s), and a totalized weight; filling-cell applications read flow rate and totalizer, not instantaneous belt load [S3][S5]. HART on a 4-20 mA analog loop is the default because the cell PLC can read flow and total on one pair of wires and accept a HART multidrop for remote calibration; Foundation Fieldbus or PROFIBUS PA are the digital alternatives on greenfield cells where the PLC already speaks PA.
Calibration cadence matters as much as sensor choice. The Siemens mining white paper quantifies the savings at 1% versus 5% accuracy as $768,000/year on a 200 t/h, 16 h/day, 300 day/year conveyor — and that assumes calibration is being done at all [S1]. A filling cell that runs a once-per-shift zero and span check plus a material-test calibration every quarter holds the published accuracy band; a cell that skips material-test calibration drifts toward the 5% error column within a few production campaigns [S1].
Who should NOT pick a belt scale for a filling cell
Any of these three duty profiles belongs on a different scale: (1) batch time under 10 minutes, (2) discrete package weighing at the end of the line, and (3) any conveyor with a curve, tripper, or stacker head within 6 idler spacings of the candidate scale location [S3]. For those duties, a static checkweigher, a net weigher, or a multi-head combiweigher on a separate feed conveyor is the right pick. Trying to force a belt scale onto a short batch or curved conveyor is the most common reason a filling line ships with 2-3% accuracy instead of 0.5%.
For a brownfield conveyor cell, the shortlist logic is: confirm straight-and-flat geometry with at least 6 idler spacings to the take-up; pick single-idler parallelogram if response time is the constraint, dual-idler if idler-spacing noise is the constraint; pick load cell capacity so live load sits at 30-70% of rated; pick 4-20 mA + HART if the existing PLC has analog cards, PA or FF if it is already on a digital segment. For a greenfield filling cell, the transformer and signal isolator upstream of the scale should be specified with the same duty cycle as the scale itself, since power-quality noise and ground loops are the two non-obvious accuracy killers on a packaged PLC skid.
Trackable signals for a buyer writing a spec this quarter: (1) the integrator's support for HART 7 device descriptors and FDI packages, which is now the default on Siemens, Thermo, and FUTEK LCF400-class cells; (2) the availability of stainless parallelogram load cells in the 45-200 kg range with a 32 kHz sample rate and a published ±0.02% combined error; (3) the OEM's documented material-test calibration interval for food-grade or chemical-grade filling, since 90 days is the industry rule of thumb and a longer interval is a procurement red flag [S1][S5][S6].
This topic is covered further in Fuse vs Motor Control Center: Spec-Driven Selection for LV Motor Branches.