Home › Field notes › Module Interface Control Drawing: Datums, Tolerances, Revisions

Modular Bag Pro · 2026-10-11 · 3,647 words · 17 min read

An interface control drawing is the one document that makes two modules genuinely interchangeable, because nothing about the outward appearance of an attachment face tells a buyer whether another source's part will seat, lock and stay locked. A usable sheet states the mating width with tolerance, the pitch between adjacent features, engagement depth, standoff from the wearer-facing plane, three ordered datum features and the conditions under which measurement happens. Each carried dimension then sits under revision control, with a sampling route written at AQL 2.5, a prototype window of 6–10 working days and bulk at 35–50 days attached to the same revision mark. The subject here is civilian load carriage — commuter organisers, worksite tool modules, first-aid inserts and hiking accessories — and deliberately excludes weapon mounts, ballistic carriers and any claim of defence certification.

Why Two Faces Can Look Identical and Still Refuse Each Other's Modules

Two attachment surfaces can sit side by side under identical lighting and behave completely differently the moment a pouch is threaded. Appearance carries no information about the three values that govern interchange: the spacing between mating features, the depth a part must travel before it locks, and the origin from which those measurements were taken. A workshop can copy a physical sample closely enough that the original pouch threads neatly, then discover twelve months later that a second module bought from elsewhere sits half a millimetre out and either will not enter or enters and rocks under load.

Three failure patterns account for almost every case. Progressive drift is the first: measurements taken between two neighbouring features rather than across the full width accumulate error towards the ends, so the outermost column becomes unusable even though every individual gap looks correct. Datum disagreement is the second: one party measures from the centreline while the other measures from a left-hand edge, both internally consistent, together incompatible. Incomplete definition is the third and the most common, where every visible dimension appears on the sheet but engagement depth, feature height or backing thickness — the values that decide whether a lock engages at all — were never written down.

An interface control drawing exists to close precisely those gaps. It is deliberately narrow: not the full product drawing, not a rendering, and not a sales sheet. It describes the mating zone only, declares datum features in order of precedence, carries every dimension that governs fit with an explicit tolerance, and prints a revision mark. The rest of the product may change freely as long as that sheet stays frozen, which is what protects a customer's installed base of modules.

Teams running several bodies on one platform should keep the mating definition identical to the shared attachment face used across the range, so one pouch development serves the whole catalogue rather than a single style.

Spec rule: An interface drawing belongs on its own controlled sheet covering only the mating zone, carrying every fit-governing dimension with tolerance, three ordered datum features and a revision mark, because a combined product drawing almost always leaves engagement depth undeclared.

The Minimum Dimension Set for a Soft-Goods Interface

Complaints that control drawings are excessive paperwork usually come from sheets padded with decorative dimensions. A useful interface sheet can be short, provided every line answers a question about fit. Six groups cover most cases.

Mating width leads. Where the interface is a row of webbing, 25 mm webbing is the standard channel dimension and the tolerance band around it decides whether a stiff strap threads cleanly or binds and twists. Where the interface is a slot cut into a laminate, both the slot height and the strap thickness band it accepts have to be stated, since one without the other permits a loose rattle or a forced entry that abrades the opening. Pitch is second: the step between features, published for PALS-derived grids as 38 mm vertical spacing and 50 mm horizontal repeat, while any other grid needs its own declared values rather than an inherited assumption.

Engagement depth comes third and is the most frequently omitted. This is the distance a strap or tongue must travel before its retaining element seats, and it is invisible once assembled. Feature height or standoff follows, deciding whether two modules mounted on adjacent rows collide or clear each other. Edge distance completes the list: how much material must sit outside the outermost feature before a seam, curve or boundary begins, which governs whether the last column can be used at all.

Measurement conditions deserve equal weight, because textiles behave unlike machined stock. State whether the sample rests relaxed on a flat surface, how long it has been conditioned after unpacking, what ambient temperature and relative humidity apply, whether any backing plate is inserted while reading, and how much tension the inspector may apply. Two inspectors following different unstated conventions can measure one part and reach different defensible answers.

Keep anything the mating partner does not touch off the sheet. Colour, print position and branding tolerances belong on the product drawing, where they can change without forcing an interface revision.

Judgement: Include engagement depth on every interface sheet even when it feels redundant, since that single undeclared value explains most cases where a module enters its counterpart, appears seated to the eye, and then releases under the first real pull.

Datum Features Both Parties Can Reach With a Gauge

A datum is simply the feature a measurement starts from, and the argument between two suppliers almost always reduces to a disagreement about datums rather than about dimensions. Three ordered features give enough structure to remove that argument: a primary that arrests the most degrees of freedom, a secondary that fixes rotation, and a tertiary that fixes the last remaining direction.

Practicality beats theoretical elegance on soft goods. A primary datum should be a feature every party can find on a finished article without destructive teardown — the base seam of the panel, the centreline of the body, or the wearer-facing plane where the module sits when mounted. A secondary might be the left-hand boundary of the mating zone; a tertiary can be the outermost row or the first row below the top edge. Whatever is chosen must be inspectable on a dressed sample, because a datum requiring disassembly will simply stop being checked after the first article.

Order of precedence matters as much as the choice itself. Writing the sequence on the sheet prevents the common situation where one inspector levels the sample on the primary and reads along it, while another squares it against the secondary first and lands a different number. Adding a note about how the sample is supported — flat on a bench, hung under its own mass, or held against a rigid plate — removes another source of legitimate disagreement.

Datum targets also help where surfaces are not flat by nature. A curved panel will never sit perfectly on a plate, and pretending otherwise produces arguments that cannot be resolved by measurement alone. Naming two or three contact points and the moderate force used to bring them into contact gives a repeatable setup anyone can reproduce, and repeatability is worth more than a theoretically ideal reference nobody can build.

Ranges spanning several bodies usually keep one shared datum convention, published alongside the product mounting maps that service staff already use, so nobody has to remember more than one.

Takeaway: Choose three datum features that can be found on the finished article without dismantling it, write their order of precedence and support conditions on the sheet, and accept a slightly imperfect reference because repeatable measurement ends disputes where theoretical elegance does not.

Stack-Up: When Two Drawings Each Spend Half the Budget

An interface always involves at least two parties. The carrier holds one half of the accumulated tolerance, the module holds the other, and several production processes add their own contributions along the way. Adding those contributions honestly at the design stage prevents the discovery, usually late and expensive, that the two halves were each plausible and jointly impossible.

Start from a functional allowance. Decide how much total variation the connection can absorb before engagement fails, become uncomfortable to operate, or allows movement that abrades components. Allocate that allowance between the two drawings according to who can actually hold tighter work: a dimension controlled by a cutting file tolerates less variation than one controlled by an operator folding fabric, and a price difference usually follows that difference in capability. Write the allocation into both sheets so neither party can later claim the whole budget.

Then add the process contributions that sit outside either drawing. Webbing stretches under sustained load, so pitch measured unloaded is not pitch at working tension. Backing layers compress, and foam behind a mounting zone compresses more. Sewn terminations take up slack differently between operators. Heat and humidity change textile dimensions measurably across a season. Each contributor deserves a line rather than a footnote, because hidden contributors are what turn a compliant pair of drawings into a non-fitting assembly.

A worksheet of this kind takes an hour and prevents weeks of argument. It also gives purchasing a factual basis when one side asks the other to tighten their drawing: either pay for the tighter process, or spend the shared residual.

Selection rule: Allocate the total functional allowance between the two drawings according to which party can actually hold tighter work, then list process contributors such as stretch and compression as separate lines, because an unallocated contributor silently consumes whoever did not claim it.

Geometric Controls That Pay for Themselves

Coordinate tolerancing — ordinary plus-and-minus dimensions — is sufficient for many soft-goods features and should not be replaced wholesale. It fails in three recognisable situations, and adding geometric control only there keeps the drawing economical.

The first situation is a pattern of repeated features. Square plus-and-minus zones around each slot in a row permit a zig-zag arrangement that still passes inspection individually but will not accept a straight module. A position tolerance referenced to the declared datums removes that possibility for one extra line. The second is angular error: a plus-and-minus band on a stud permits more tilt than the mating part tolerates, and perpendicularity control expressed against the primary datum expresses the requirement directly. The third is a surface that must lie within a band along its length: profile control states it in one callout where several coordinate dimensions would be needed and would still leave gaps.

Cost comes from inspection rather than from drawing symbols. A geometric callout that nobody can verify adds nothing, so each control needs a stated verification method — a gauge that can be made, a fixture already in use, or a routine on equipment the laboratory owns. Where the cost of verification exceeds the cost of the occasional failure it prevents, tolerancing is not worth having.

Soft-goods reality also limits what any control can achieve. Compliant materials under a measurement load are not rigid bodies, so every callout should carry the condition under which it is checked: relaxed, under declared tension, or supported against a plate. Stating the condition is not a weakness; it is the reason two inspectors can agree.

Verdict: Keep ordinary coordinate dimensions as the default and add position, perpendicularity or profile control only for repeated patterns, angular relationships and long surfaces, and never write a geometric callout whose verification method is not named on the sheet.

Revision Rules: What Counts as a Change and What Forces Re-Testing

Revision discipline is where most interface programmes quietly fail. A dimension shifts by a small amount during value engineering, nobody considers it a change, and two years later a customer's installed base of modules stops fitting. The discipline needed is simple: classify changes, then state what each class obliges.

Class A covers anything in the mating zone — width, pitch, engagement depth, standoff, edge distance, datum definition, or a material substitution that alters thickness or friction. Class A always gets a new revision mark, always requires re-verification, and always carries a statement about whether the previous revision's modules remain supported. Class B covers construction changes outside the mating zone that could reach it indirectly, such as a different backing stiffness; these need a review and usually a partial re-check. Class C covers cosmetic and neighbouring items — colourway, print, lining grade outside the zone — and needs a record but not re-verification.

Backward compatibility deserves an explicit answer rather than silence. Either new parts accept old modules and vice versa, or they do not; either answer is acceptable provided it is published with the revision. Silence forces customers to guess, and guesses become returns. Publishing the answer with the revision also lets customer service settle a compatibility question without reopening the drawing.

Freeze discipline supports the whole scheme. Interface dimensions should be frozen before artwork is commissioned, because a change discovered after a decorated sampling round costs several times more than one caught in the drawing stage.

In practice: Classify every prospective change as mating-zone, indirect or cosmetic, require a new revision mark and a written compatibility statement for the first two, and freeze interface dimensions before decoration begins, because the cost of the same change multiplies once samples are already printed.

First Article, Incoming Checks and a Cross-Lot Interchange Trial

Three verification activities sit at different points and answer different questions. Treating any one of them as sufficient is the standard route to a surprise.

The first-article report answers whether the drawing was understood. It should measure every dimension on the interface sheet across several pieces drawn from more than one production source within the batch, record the conditions applied, and note the equipment used. Where a feature involves repeated elements, measure across the full width rather than between neighbours, since accumulated error appears at the ends and nowhere else.

Incoming inspection answers whether a delivered batch matches the approved reference. It cannot repeat everything; instead, pick the three dimensions that most often drift, make or buy a simple go/no-go gauge for each, and check them at a stated frequency. Attribute sampling gives the plan a defensible basis, and many programmes write it at AQL 2.5 against defined defect classes, drawing on the sampling system documented in ISO 2859-1. Destructive or high-effort checks belong at approval stage and periodically afterwards, not at goods-in.

The interchange trial answers whether two populations actually work together, and it is the step most often skipped. Take modules from one lot and mount them on carriers from a different lot, ideally from a different production period, then run the real sequence: mount fully, apply working load, walk or simulate a walking cycle, remove, and remount several times. Record any part that requires force beyond what a user would apply, any rocking after seating, and any audible change between the first and tenth engagement. Ranges using several body styles can extend the trial across the small module configurations that customers combine most often.

Bottom line: Run all three activities — a drawing-wide first-article report, a short gauge-based incoming plan written at AQL 2.5, and a cross-lot mounting trial with loaded cycling — because each one detects a class of failure the other two cannot see.

Records, Ownership and Programme Gates

Documentation turns a one-off decision into an asset someone can act on two years later. Four records matter for an interface programme: the controlled sheet with its revision history, the first-article report with measurement conditions, the approved module list showing what has been verified against which revision, and the component approval board with lot references. Keeping them together under one order reference is more valuable than any individual record.

Ownership also has to be named. Someone specific needs the authority to approve a revision and the duty to notify everyone holding modules; without a named owner, changes drift in from tooling maintenance, material substitution and cost reviews without ever being assessed. A short change log next to the revision mark usually delivers more benefit than an elaborate document control procedure nobody follows.

The SGS-verified production base we work with measures 4,950 m² and houses 7 production lines carrying 149 machines, worked by 137 people, set against a monthly plan of 200,000 units; the founder entered bag production in 2004 and the business was set up in 2014. Programme work is coordinated so that the interface sheet carries a revision mark before component approval even opens, and coated or laminated face materials are reviewed against test methods such as ASTM D751 where slot panels form part of the interface. Documented management systems such as ISO 9001 describe how those records are maintained; they do not by themselves certify the conformance of any shipment.

Commercial gates follow the usual sequence. MOQ is 500 pieces. Sampling occupies 6–10 working days, widening to 12–15 where a new interface must be built from nothing; bulk production takes 35–50 days after approvals and inputs have closed. Final inspection runs at AQL 2.5 against defined defect classes, settlement runs on T/T 30/70, and the quotation basis is FOB Xiamen. Transit planning offers sea freight at 25–35 days, air at 5–8 days and courier at 3–5 days for samples and spare modules; containerised volume planning usually works from 20GP near 28 CBM or 40HQ near 68 CBM. Any figure quoted ahead of sample approval should be treated as indicative, resting on FOB Xiamen with a 500-unit floor.

In practice: Keep the revision history, the first-article report, the approved module list and the component board together under one order reference and name an owner for each revision, because an interface without an owner drifts silently through tooling repairs and material substitutions.

Omissions That Only Surface After Launch

Certain gaps appear repeatedly across otherwise well-documented programmes. Listing them and checking against the list costs a few minutes and prevents the majority of late-stage interoperability complaints.

The first is forgetting to state what the tolerance applies to: a dimension without a stated condition is unenforceable, and the most frequent version is a slot height written without saying whether it is measured on a free part. The second is omitting the strap thickness band on the module side, leaving whichever party makes the second decision to discover it late. The third is leaving reinforcement out of the interface sheet; backing thickness changes engagement depth, so it belongs in the mating zone even though it looks structural.

Two commercial gaps follow. Nobody states who pays for a re-tooling exercise when one party's drift forces the other to adjust, and nobody defines what happens to modules already in customers' hands when a revision changes the fit. Both are ordinary questions that become disputes only because they were unanswered at the start.

Teams still at the drawing stage can cross-check the sheet against the interface-controlled development route before any tool steel is cut, which is the cheapest moment to find an omission. A last gap concerns spares and replacements. If a programme sells modules separately, the sheet should carry a note stating which revision they suit and how that is marked on the product or its packaging. A small printed revision code on both carrier and module is cheaper than any explanation later, and gives customer service the means to answer a compatibility question without engineering involvement.

In practice: Check every interface sheet against a fixed omission list covering measurement condition, mating thickness band, reinforcement within the zone, cost ownership of forced re-tooling, treatment of installed modules and the printed revision code, because each missed item converts into a customer-visible interoperability complaint rather than a drawing error.