Against the Specification Race

Camera modules are sold on megapixels, on maximum frame rate and on minimum illumination — three figures that describe the sensor rather than whether your feature will be legible, and that are comparable between suppliers only under conditions nobody states.

A separate team-process reference can be found learn more.

This site is written against that, and the reason is practical rather than moral.

Why catalogues are built this way

Because a single number ranks.

A buyer comparing eight modules wants an ordering, and a specification sheet supplies one — whereas the honest answer requires knowing the buyer's working distance, their feature size and their processing.

Further background is available from Arm.

Which the supplier does not know and frequently cannot ask.

So the catalogue offers the figures that can be stated without knowing anything about the application, and those are the ones that rank badly against real requirements.

Nobody is being deceived. The information that would decide is not available to the person writing the page.

What it costs a buyer

Modules chosen on a number and rejected at integration, after the mechanics were designed around them.

Specifications written as answers — "we need a five-megapixel module" — rather than as requirements, which prevents anybody proposing something better.

And a systematic bias toward more: more pixels, more frame rate, more sensitivity, each costing data rate, light per pixel, power and money for capability the task does not use.

The commonest expensive error on this site is a sensor bought to solve a lighting problem.

What this site does instead

Explains the mechanisms, so that a reader can work out what they need and compare candidates against it — including candidates this site has never seen.

The technical material is generic deliberately. It applies to any module from any supplier, and a reader who can specify properly can compare properly, which is the only basis on which a comparison is worth having.

No competitor is named, and no product is ranked.

The specification worth writing

Four lines.

The feature and its size. The working distance and required coverage. The pixels the processing needs across the feature. The environment.

Which a supplier can answer specifically, which two candidates can be measured against, and which makes a test have a pass condition.

"We need a high-resolution module" specifies nothing, and it is what most enquiries contain.

Where a number genuinely ranks

Some do, and dismissing all of them would be its own error.

Physical dimensions. A module either fits or it does not.

Interface and lane count. A gating compatibility fact rather than a matter of degree.

Operating temperature range, as a limit.

And published part availability, which is commercial and definite.

Those four rank because they are constraints rather than performance claims — and a shortlist built from them is a shortlist worth measuring.

The pressure on suppliers

Worth acknowledging, because a page criticising catalogues should say why they persist.

A supplier publishing a modest figure alongside competitors publishing optimistic ones under looser conditions loses comparisons it should win.

Which makes honest specification a commercial disadvantage unless conditions are stated and buyers read them — and most buyers do not.

The way out is on the buyer's side: asking under what conditions, which is one of the four questions and which changes the conversation immediately.

A supplier asked a specific question answers specifically. The catalogue is what exists in the absence of the question.

Where this site is exposed

It is published by a module supplier.

Which means it has an interest in modules being chosen over board-level and boxed cameras, and in readers believing that specifying carefully is worth the effort.

The correction is that the technical material is generic and the page on integration levels concludes against modules in several named cases — and the limits page states the rest.

A reader should notice the interest and check the reasoning, which is the same thing this page asks of a catalogue.

What a good enquiry looks like

Four lines and a question, and it changes what comes back.

"We need to distinguish a 0.3 mm feature on a 60 mm part, at 150 mm working distance, under fixed LED illumination, in an enclosure reaching 50 degrees. Which variants would you suggest, and what would you expect the millimetres per pixel to be?"

A supplier given that answers specifically and frequently suggests something the buyer had not considered — including, occasionally, that no module in the range fits.

Which is a useful answer and one a catalogue cannot give.

Where the honest figures live

Not on the front of a catalogue and frequently available on request.

Signal-to-noise against illumination, rather than a minimum illumination figure.

Readout time, rather than a maximum frame rate.

Pixel size in micrometres alongside the resolution.

And a distortion profile rather than a percentage.

Each is a specific question and each has a definite answer — which is the practical response to everything on this page.

The number this site would put on a catalogue

Millimetres per pixel at a stated working distance, for each variant.

Which is derivable from the sensor and lens figures already published and which nobody states, because it requires assuming a distance and a supplier does not know yours.

Publishing it at two or three representative distances would let a buyer read the useful quantity directly instead of computing it.

Whether that is commercially sensible is a separate question — it makes a module look narrower and more specific, which is accurate and does not rank well.

In one line

Write the requirement rather than the answer, and the catalogue stops being the thing you compare.

The short version

• Modules are sold on megapixels, frame rate and minimum illumination, which describe the sensor and not your task
• Catalogues are built that way because a single number ranks, and the information that would actually decide is not available to the supplier
• It costs buyers modules rejected at integration, specifications written as answers, and a systematic bias toward capability nobody uses
• The commonest expensive error is a sensor bought to solve a lighting problem
• This site explains mechanisms generically so that a reader can compare candidates it has never seen
• Four lines specify a system: the feature and size, the distance and coverage, the pixels required, and the environment