Guides / Sourcing / Design for Manufacture

How Tolerances Add Up Across an Audio Assembly

September 13, 2026

Most visible defects on a finished audio product are not caused by a part being out of tolerance. They are caused by several parts being comfortably in tolerance at the same time, in the same direction.

The basic idea

Every dimension has a range. When features on different parts have to line up, their ranges add.

Take a knob passing through a panel cutout. The gap you see depends on:

• The cutout position on the panel
• The cutout diameter
• The potentiometer position on the PCB
• The PCB position in the chassis
• The shaft perpendicularity
• The knob bore concentricity

Six contributors. If each is plus or minus 0.1 mm, the worst case offset is 0.6 mm, which on a 1 mm nominal gap is the difference between looking precise and looking broken.

Worst case against statistical

Worst case adds the tolerances arithmetically. Every part is assumed to be at its unluckiest limit at the same time. It guarantees assembly and it produces tight, expensive tolerances.

Statistical, or root sum square, takes the square root of the sum of the squares. It assumes the contributors are independent and normally distributed, so all being at their extreme simultaneously is very unlikely. It produces looser, cheaper tolerances and accepts a small failure rate.

For six contributors each at plus or minus 0.1 mm:

• Worst case: plus or minus 0.6 mm
• Root sum square: plus or minus 0.245 mm

The difference is large, which is why the method is worth choosing deliberately rather than by default.

For hi-fi volumes, worst case is usually right. Statistical methods earn their keep at thousands of units where a small reject rate is cheaper than tighter tolerances. At 200 units a year, a two percent problem is four units, each one an unhappy customer and a return.

Choose the datum, then dimension from it

This is the single highest leverage decision on a drawing, and it costs nothing.

Chained dimensioning measures each feature from the previous one. Errors accumulate along the chain.

Datum dimensioning measures every feature from one reference. Errors do not accumulate. Every feature is within its own tolerance of the datum, independently.

On a front panel with eight connector cutouts in a row, chaining puts the eighth cutout eight tolerances away from the first. The row will visibly drift. Dimensioning all eight from the panel centreline keeps every one within its own tolerance, and the row looks even.

Rules that follow:

• Pick the datum that matters functionally. On a platter it is the bearing bore. On a panel it is usually the centreline in both axes. On a chassis it is usually a machined face, not a folded edge.
• Use the same datum across related parts. If the panel is dimensioned from its centreline and the sub-panel behind it from a corner, the two will not agree.
• Never use a folded edge as a datum if a machined face is available, because folded dimensions carry much larger tolerances.

Reduce the number of contributors

Tightening tolerances is the expensive way to fix a stack. Removing contributors is the cheap way.

• Locate by a machined register, a step or a rebate, rather than by clearance holes and screws. Clearance holes add the hole tolerance plus the clearance itself.
• Combine parts. Two features machined in one setup on one part have one tolerance between them instead of several.
• Let one part be adjustable. Slotted holes on the hidden part move the problem from tolerance control to assembly, which at low volume is often the cheapest place for it.
• Design the gap to be forgiving. A 1.5 mm shadow gap absorbs the same absolute error as a 0.4 mm gap with a quarter of the proportional effect. The eye reads proportion.
• Break the relationship visually. A chamfer or a step at a joint creates a shadow line that hides misalignment, which is why so many well regarded enclosures have one on every visible edge.

Tolerance only what matters

The corollary is as important. Once you know which features are in the stack that a customer sees, everything else can be loose.

A workable pattern for an audio panel:

GENERAL TOLERANCE: +/-0.2 UNLESS OTHERWISE STATED
CRITICAL FEATURES, PER TABLE:
  KNOB CUTOUT POSITION       +/-0.05 FROM DATUM A AND B
  PANEL WIDTH AND HEIGHT     +/-0.10
  MOUNTING SLOT POSITION     +/-0.10 FROM DATUM A AND B
ALL OTHER FEATURES AT GENERAL TOLERANCE

Three tight features and a loose general tolerance prices dramatically better than plus or minus 0.05 mm across the whole drawing, and it delivers the same product.

Do the arithmetic before you release

It takes twenty minutes and it catches the expensive problems.

1. List every visible gap and fit in the product
2. For each, list every contributor, on every part
3. Add them worst case
4. Compare the result with what the design can tolerate visually
5. Where it fails, reduce contributors first, tighten tolerances second
6. Write the surviving critical features into the drawing as called out tolerances

Step 2 is where most of the value is. People are usually surprised by how many parts contribute to a gap they thought belonged to one part. The panel gaps guide works a full example through.

Frequently asked questions

Keep reading

• How to Choose a Machine Shop for Cosmetic Audio Parts
• Telling a Machine Shop Which Faces the Customer Will See
• Running a First Article on a Part Where Appearance Is the Spec

This guide is part of Sourcing and Quality Control for Audio Parts, one of four sections in the guide library.