AOV vs PIR: What a Motion-Triggered Camera Physically Cannot Record

The short answer: PIR misses slow, cold, off-axis and early

A PIR camera does not record continuously. It sleeps, and it wakes only when a passive infrared sensor registers a fast enough change in heat inside its detection cone. That mechanism has four physical blind spots: movement too slow to cross the sensor's threshold, a target whose surface temperature is close to the background, anything happening outside the cone, and the seconds before the trigger fires, because the sensor must detect before the recorder can start. An AOV camera — Always-On Video, a generic industry term rather than anyone's trademark — records continuously and not selectively, so the frame exists whether or not a sensor ever agreed that something happened.

The distinction is not about video quality. It is about whether a file exists at all. When a PIR camera misses, it does not produce a poor clip; it produces nothing, and nothing looks identical to a quiet night. This article covers what is physically lost, when it matters commercially, and how to test it on a sample before committing to a container.

What a PIR sensor is actually doing

A PIR sensor does not see an image. It is a differential heat detector: a segmented lens splits its field into zones, and the sensor fires when infrared energy shifts between zones quickly enough. Three conditions must hold simultaneously for a trigger:

1. Thermal contrast between subject and background.
2. Lateral movement across the zones — walking straight at the lens produces a far weaker signal than walking across it.
3. Presence inside the detection cone, which is narrower than the video lens's field of view.

Break any one and there is no event. This is not a firmware limitation that a better app fixes; it is how the sensing element works. Understanding that is what separates a realistic deployment from a disappointed customer six weeks after handover.

The four gaps in table form

The pre-roll problem

On a trigger-based camera, the clip begins when the sensor concedes that something is happening. Everything before that moment does not exist anywhere.

That missing minute is usually the useful one. It contains the approach — where they came over the fence, how long they watched the yard first, whether a vehicle waited outside the frame, whether they walked in through an unlocked gate that someone left open. A clip that opens with a person already standing beside the plant answers that something happened but not how, and "how" is the question the insurer asks and the only one that lets you close the gap so it does not recur.

With continuous recording nothing has to be reconstructed. You scrub back on the timeline and watch it.

Where the gap costs money

Construction sites over a shutdown

Site theft is rarely dramatic. It is cable, fittings, formwork, small plant and hand tools, removed at walking pace by people who look like they belong there. Over a long weekend or a public holiday the site is unattended from Friday afternoon to Tuesday morning, and the loss is usually discovered at the next material count, not at the moment it happens. Reconstructing four days from a handful of triggered clips is not an investigation; it is a guess.

Farms, livestock and machinery

Livestock movement is slow by nature, warm animals move among other warm animals, and the ground holds the day's heat well past dark. That is the worst combination a differential heat sensor can be handed. Discovery is delayed too — a head count is done on a schedule, not continuously.

Warehouse yards and stock reconciliation

Yard shrinkage surfaces at month end, weeks after the event. By then the only question that matters is whether the footage exists. Continuous recording makes that a storage-window question with a numeric answer, rather than a hope that a sensor happened to agree.

Second properties and idle sites

A property nobody visits for weeks is the extreme case. The delay between event and discovery is so long that trigger reliability stops being the variable at all. What matters is whether the timeline covers the date.

The power trade-off, stated honestly

PIR exists in solar cameras for a real reason: sleeping costs less energy than recording. That was always the bargain. What changed is that AI scene analysis now allows continuous capture at low standby power instead of holding the entire imaging chain awake at full draw.

It is worth naming the opposite failure mode too. A PIR camera on a flat battery does not announce that it has stopped protecting anything. It still looks installed on the pole, and the absence of clips reads as "a quiet week" rather than "no power to respond". Continuous video is more demanding of the power design, which is exactly why its health is easier to audit: either there is a timeline for last Tuesday or there is not.

Every SolaGuard model runs on a solar panel, a battery and a 4G SIM — no mains power and no WiFi. Resolution is 1080p–2K as standard, with 4K available only as a custom or OEM build. Outdoor models are IP66 sealed, storage is microSD up to 128 GB as standard with larger capacities on custom orders, and installation takes 5–10 minutes per camera because there is no cabling to run.

How to test the difference on a sample

Do not take anyone's word for it, including ours. Five tests on a single sample unit tell you more than any datasheet:

1. Slow-walk test. Approach the camera very slowly, pausing every few steps. Then check whether footage exists for the whole approach or only the last part of it.
2. Edge test. Move along the far side of the frame rather than through the middle, and compare what appears in the timeline with what appears in the event list.
3. Midday test. Repeat the slow-walk at the hottest hour. Thermal contrast is lowest then, and the two behaviours separate most clearly.
4. Pre-roll test. Trigger an obvious event, then try to scrub back one minute before it. If it is not there, it never will be.
5. Removed-object test. Leave an item in frame, take it from outside the detection cone, and see whether there is any record of who took it.

Four questions to put to any supplier

• Is continuous recording on by default, or is it a setting a customer has to find?
• What happens as battery falls — does video stop, does quality drop, or is event capture prioritised?
• Does the AI tag moments inside footage that already exists, or does it decide whether to record?
• Are the event list and the timeline two separate things in the app, or is the event list all there is?

The third question is the one that matters. If AI decides whether to record, the gap has not been closed — only the trigger criterion has changed, and a criterion can still be wrong at exactly the wrong moment. If AI tags an existing continuous file, then a missed tag costs you search time, not evidence.

Buying notes

PIR answers "did something warm cross the sensor quickly?" AOV answers "what happened here between 2am and 5am?" The second is the question a site manager, a farm owner or a loss adjuster actually asks. The first is a technical precondition that may or may not have been met during the exact minutes you care about — and you find out which, weeks later, when you go looking.

The catalogue runs to 150 SKUs with wholesale prices quoted per order, with OEM and private label available on all of them. Manufacturing is by Shenzhen Leksell, TÜV Rheinland certified, established 2012, with CE, FCC and RoHS. Dealer margin starts at 20% from around 10 units and reaches 30% from around 50; MOQ is 10 for a trial order and 100 for OEM/ODM. Samples ship 7–10 days by air, bulk 25–40 days by sea, and Vietnam orders in 7–10 business days from the Ho Chi Minh City office.

Browse the range at /en/products/, see dealer terms at /en/dealer/, or read the technical background in the AOV solar camera guide. For a sample or a per-SKU quote, message us on WhatsApp at +86 188 1869 9870 or email info@solaguard.net.