Planning a measurement system study for polyphosphate cartridge fill weight begins with one decision that is easy to get wrong: define exactly what “fill weight” means for your product. If your study weighs a sealed cartridge and subtracts a nominal tare, but the empty housing weight varies from piece to piece, the study will report housing variation as measurement error. Fix the measurand — net crystal weight, gross cartridge weight, or in-process dispensed weight — before you select samples, operators, or a gauge.
This article is written for quality engineers, process engineers, and cartridge-filling supervisors who already have a filling process and need a defensible measurement system study (often a gage R&R or GRR) on fill weight. It covers study definition, sample selection, design choice, and how to interpret the results for polyphosphate crystal cartridges and similar media-filled water-treatment cartridges.
Why fill weight is a harder measurand than it looks
Fill weight is a mass, but the measurement system that produces it is not just a scale. It includes the operator, the tare value or empty-container handling, the filling or transfer step, the environment, and the data-recording method. Any of those can dominate the result.
Three complications are specific to crystal-filled cartridges:
- Tare variation. Cartridge housings, end caps, screens, and labels have their own weight distribution. If housing weight varies by a meaningful fraction of the fill tolerance, gross weighing cannot resolve fill weight.
- Material state. Polyphosphate crystals are generally irregular, free-flowing granules or beads, but bulk density and flow behaviour differ between products. Siliphos is a specific branded material; other polyphosphate crystal media are not necessarily identical in composition, granulation, or bulk density. Confirm what your actual media does before assuming a fill target behaves the same as another product.
- Process dynamics. Vibratory or auger fillers settle material after filling. A cartridge weighed immediately after filling and again after handling may read differently on a sufficiently precise scale. Your procedure must state when the weight is taken.
Decide what the study must answer
A single GRR does not cover everything. A complete measurement system evaluation typically has four separate components, and they answer different questions.
| Component | Question it answers | Typical cause of a bad result |
|---|---|---|
| Repeatability | Does one operator get the same result on the same cartridge? | Scale resolution, drift, unstable placement, vibration, airflow |
| Reproducibility | Do different operators or shifts get the same result? | Handling technique, tare entry, zeroing habit, reading position |
| Bias | Does the system read consistently high or low versus a reference? | Tare assumption, calibration offset, fixture weight included by mistake |
| Linearity and stability | Is the error the same across the weight range, and does it hold over time? | Scale non-linearity near range limits; drift over a shift or week |
If the process weighs every cartridge on an inline checkweigher, the “operator” factor may be weak or absent. In that case, replace it with the factors that actually vary: fixture position, line, shift, product orientation, or crystal lot. A GRR that reproduces a factor which never changes in production produces a falsely reassuring number.
Step 1 — Freeze the procedure before collecting data
Write down the measurement procedure and follow it exactly for the duration of the study. At minimum, document:
- Whether the reported value is net fill weight or gross weight, and how tare is determined (fixed tare, measured empty housing, or automatic tare).
- The exact scale or checkweigher used, its capacity and readability, and its warm-up and zeroing routine.
- How the cartridge is placed on the pan — centred, orientated, and at what point in the fill cycle.
- Any settling, cooling, or holding time between filling and weighing.
- How the value is recorded and rounded.
Changing any of these mid-study invalidates the comparison between trials. If the procedure genuinely cannot be fixed, that instability is itself the finding.
Step 2 — Verify resolution and calibration first
Calibration and a measurement system study are not the same activity. Calibration confirms that the instrument reads correctly against a traceable standard. A GRR study quantifies the total variation of the whole system, including operator and procedure, in units that matter to the product tolerance.
Before running a study, check two things:
- Resolution. A common industry convention is the “rule of ten”: readability should be at most one tenth of the fill weight tolerance. If your tolerance is tight and your scale reads in coarse increments, the study will be dominated by rounding before any operator effect appears.
- Calibration status. Use current, traceable check weights and confirm that the range used in the study is actually covered. A scale calibrated at the top of its range may behave differently near the fill target.
Step 3 — Select samples that span the real process
The most common planning error is choosing ten cartridges that all weigh nearly the same. If part-to-part variation is artificially small, %GRR as a percentage of total variation looks terrible even when the gauge is fine, and the study tells you nothing about whether the system can sort good from bad.
Select samples that represent the range the process actually produces:
- Include cartridges near the lower and upper specification limits. If you cannot deliberately produce them, use cartridges from the tails of recent production data rather than from the centre.
- Avoid samples that are physically damaged, partially filled, or from a trial run that is not representative.
- Confirm that the crystal lot used is representative. If your supplier’s bulk density has shifted, the fill behaviour shifts with it.
Step 4 — Choose crossed or nested design
If weighing is non-destructive and the cartridge can be re-weighed, use a crossed design: every operator weighs every sample, several times each. This is the standard approach and allows repeatability and reproducibility to be separated.
If determining fill weight requires opening, emptying, or destroying the cartridge, a crossed design is impossible because each measurement destroys the sample. Use a nested design instead, where each operator measures a different set of cartridges, and accept that it estimates reproducibility differently.
A conventional starting point for a manual filling operation is ten samples, three operators, and three trials per operator, run in randomized order. This is a planning convention, not a rule; increase the number of samples if you need tighter confidence in the variance components, or if your customer’s quality agreement specifies a different design.
Step 5 — Control the environment and the order
- Run trials in randomized order so that operator fatigue or scale drift does not align with sample sequence.
- Keep the scale on a stable surface away from filler vibration, fans, and draughts.
- If the area has static problems, address them before the study — static affects both crystal handling and electronic balances.
- Record ambient conditions if your procedure or customer requires it.
- Train operators on the written procedure and let them practise before data collection. A first attempt during the study measures learning, not the system.
Interpreting the results
Widely used MSA guidance (AIAG) applies the following conventions when %GRR is expressed against the tolerance:
| Result | Conventional interpretation | Practical response |
|---|---|---|
| %GRR under 10% | Generally acceptable | Keep the system under periodic verification |
| %GRR between 10% and 30% | Conditionally acceptable | Accept only with justification based on cost, risk, and customer requirements |
| %GRR above 30% | Generally not acceptable | Find and fix the dominant source before relying on the data |
| Number of distinct categories below 5 | Insufficient discrimination | The system cannot reliably separate good parts from bad across the tolerance |
Always check your customer’s or sector’s own acceptance criteria, which may be stricter or expressed differently. Also report the variance components, not just the percentage, so that readers can see whether repeatability or reproducibility dominates.
Planning checklist
- Measurand defined: net fill weight, gross weight, or dispensed weight.
- Tare method defined and its variation measured separately.
- Scale or checkweigher resolution verified against the tolerance.
- Calibration current and traceable, covering the working range.
- Samples selected across the real production range, not clustered at nominal.
- Study design chosen: crossed for non-destructive weighing, nested if destructive.
- Operators trained, trials randomized, environmental conditions stable.
- Analysis method agreed in advance (ANOVA is generally preferred for variance components).
- Acceptance criteria agreed with the customer before results are reviewed.
Frequently asked questions
How many cartridges do I need for a fill weight study?
Ten samples with three operators and three trials each is a common starting point for manual weighing. It is a convention rather than a requirement. Increase the sample count when the tolerance is tight, when you need confidence intervals on the variance components, or when your customer specifies a particular design.
Can the same cartridge be weighed three times?
Yes, if weighing does not alter the cartridge or its contents. Re-weighing the same sealed cartridge is exactly what a crossed GRR requires. If the crystal settles or shifts inside the housing between weighings, document the handling and keep it consistent, because that variation becomes part of your repeatability estimate.
Is a calibration certificate enough to qualify the measurement system?
No. Calibration addresses instrument accuracy against a standard. It does not address operator technique, tare handling, sample presentation, or the effect of the production environment. A calibrated scale can still belong to an unacceptable measurement system.
Do I need to repeat the study after changing crystal supplier or grade?
Not automatically, but it is worth reviewing. A change in granulation or bulk density can change settling behaviour and therefore the weight reading at the moment of measurement. If the change is significant, re-run at least the repeatability and stability portions rather than assuming the previous result still applies.
When the study points at the material instead of the measurement
Occasionally a carefully planned study comes back clean, yet fill weight still drifts. That usually means the variation is real and lives in the process or the material — filler settings, feed consistency, or lot-to-lot differences in the crystal itself. In that situation the study has done its job by ruling out the measurement system, and the next step is material and process investigation, not another GRR.
If your study is being planned alongside a crystal specification review, ENVO CHEMICAL can discuss the polyphosphate crystal parameters relevant to your filling method and target weight. To respond usefully, we need your cartridge type and dimensions, target fill weight and tolerance, filling equipment type, the measurement method you intend to study, required crystal grade or specification, estimated annual quantity, and destination market. Share those details and we can address the material side of the fill weight question directly.
