- Ffront, accessible
- Wwindow inside
- Ssensor, worst case
Cleanliness and What a Fingerprint Costs
Generic engineering material, applicable to any module from any supplier.
For more on organizing work around technical projects, read this resource.
Contamination on an optical surface scatters light, which reduces contrast across the whole image long before anything is visibly obscured — and reduced contrast is exactly what the processing cannot tolerate.
What it actually does
Scatter. Light that should form the image arrives spread across the frame instead, raising the level in dark regions and lowering it in bright ones.
Which is a contrast loss rather than a blur, and it does not look like dirt. It looks like a slightly flat image, which is easy to attribute to the sensor.
Further background is available from Mouser Electronics.
A veiling glare where a bright source is in or near the frame, since contamination gives it something to scatter from.
And in severe cases, visible marks — but by then the contrast has been degraded for a long time.
Where it collects
The outermost surface, whatever that is: the lens front element, a protective window, or a filter.
Which is where it is accessible and where it is least damaging optically, since it is furthest from the sensor and therefore most defocused.
Inside the enclosure, on the back of a window — which needs the enclosure opened and is where condensation deposits leave residue.
And on the sensor itself, which is the worst case: contamination there is sharply focused and appears as distinct dark spots that do not move with the scene.
Telling them apart
A useful diagnostic and it takes one image.
Photograph a uniform bright surface, out of focus, with a small aperture.
Distinct dark spots that stay in the same place are on or very near the sensor.
Soft, large, faint patches are on the front surface.
And a general loss of contrast with no visible features is contamination distributed evenly, which is the commonest state and the one nobody notices.
Cleaning without damage
Blow first. Loose particles dragged across a coated surface scratch it, and a scratch is permanent where a fingerprint is not.
Use appropriate materials. Lens tissue or a clean microfibre cloth, and a solvent intended for optics.
Not a shirt, not a paper towel, not a household glass cleaner — the first two scratch and the third can attack coatings.
Work from the centre outward, lightly.
And do not clean what does not need it. Every cleaning carries a small risk of damage, which makes routine unnecessary cleaning a slow way to ruin a surface.
Designing so it can be cleaned
A window that can be reached without disassembly.
A surface that is flat rather than recessed, since a recessed one collects and cannot be wiped.
And a documented procedure with the materials named, because otherwise the first person to clean it will use whatever is nearby — which is the same reasoning as any assembly instruction.
Building the check into service
A reference image taken at commissioning, and the same view compared periodically.
Contrast measured against a fixed target rather than judged makes gradual degradation visible, which by eye it is not — because the decline is slow and the observer adapts.
Which converts cleaning from a response to a complaint into a scheduled task, and prevents the period during which the system is quietly performing worse than it should.
The failure that looks like something else
A slowly rising rate of processing failures, with no change to the parts, the software or the machine.
Contamination is the first thing to check and is checked last, because it is not a component and nobody owns it.
Thirty seconds with a clean cloth resolves it where it is the cause — and where it is not, the elimination cost nothing and the investigation continues with one fewer variable.
Protective films
Some modules ship with a film over the lens or window.
Remove it — which sounds unnecessary and is the cause of a real number of first-image complaints, since a translucent film produces a soft low-contrast image that looks like a focus problem.
And where a sacrificial film is used deliberately in a dirty environment, treat it as an optical surface: it scatters, it degrades, and it needs replacing on a schedule.
The environment decides the interval
Not a calendar.
A clean assembly area and a sealed enclosure may need the window touched once a year.
A machining shop with coolant mist, a bakery with flour, a site with diesel exhaust — weekly, and the schedule should say so.
Establish it by measuring rather than by guessing: check the contrast against a reference at intervals during the first months, and set the schedule from what the decline actually looks like.
In one line
Blow before wiping, and clean only when a measurement says to.
What contamination does to a measurement
Worth separating from what it does to an image.
A scattered-light contribution adds a roughly uniform level to the whole frame, which shifts every absolute brightness measurement without changing the shape of anything.
Which means a system calibrated when clean reads differently when dirty, in a direction that looks like an illumination change.
Where absolute values matter, a reference patch in the frame corrects for both together — and is the standard remedy for a class of slow drift that would otherwise be untraceable.
The short version
- Contamination scatters light, reducing contrast across the whole image long before anything is visibly obscured
- It looks like a flat image rather than like dirt, which is why it gets attributed to the sensor
- It collects on the outermost surface, inside on the back of a window, and worst of all on the sensor itself
- Diagnose by photographing a uniform surface out of focus: sharp fixed spots are near the sensor, soft patches are at the front
- Blow before wiping, use optical materials, and never a shirt, a paper towel or household glass cleaner
- Do not clean what does not need it, since every cleaning carries a small risk of permanent damage