Technical documentation

Systems, comparisons and specifications

This page is for people who want to see the numbers. Detection architecture, reference specifications, comparison tables, integration protocols, storage calculation and applicable standards. No invented figures: every data point states where it comes from.

01 · Architecture

Two layers, not one

Everything below rests on the same principle: the layer that detects and the layer that verifies must be different technologies.

Two-layer decision chain with an independent relay branchFlow diagram. The first layer detects and yields position, size and velocity. The second layer verifies with imaging and yields class and confidence. Only when both agree inside a time window is a notification raised. In parallel, a relay output drives siren and lighting without depending on the network. LAYER 1 · DETECTIONradar / LiDAR position · size · velocity COINCIDENCE (LOGICAL AND)Δt ≤ 5 s LAYER 2 · VERIFICATIONAI camera / thermal class · confidence over threshold NOTIFICATION push + image sequence RELAY OUTPUT · DRY CONTACThard-wired, in parallel OUTDOOR SIRENmax. 60 s · 85 dBA at 3 m ZONE LIGHTINGimmediate deterrence The relay branch works with no network, no internet and no cloud. Málaga city noise ordinance · system not connected to an alarm receiving centre
Why two layers. A single layer forces a choice between missing events and living with false alarms. With two, a notification is raised only when detection and verification agree inside a time window. The relay branch is deliberately independent: siren and light fire over wire even with the network down.

The model

Layer 1 — detection. Low threshold, contained cost, wide coverage: radar, LiDAR, beam, fence sensor or outdoor PIR. Its job is to raise the event and start recording. It does not sound the alarm on its own.

Layer 2 — verification. A camera with object classification, or thermal with analytics. It answers one question: is it a person?

Alarm. Only when both layers agree within a time window.

This design solves three problems at once: it removes false alarms from animals and vegetation, it produces an image attached to every event — which later serves as evidence — and it matches the verification scheme that Spanish regulation requires of systems connected to an alarm receiving centre.Orden INT/316/2011

EVENT CHAIN
  1. Detection — radar / LiDAR / beam / PIR marks position and time
  2. Orientation — PTZ or fixed camera covers the event zone
  3. Classification — AI decides: person, vehicle, animal, unknown
  4. Decision — layer agreement inside the time window
  5. Action — siren, light, alert with image to the owner
  6. Record — recording with pre-alarm and timestamp

Design rule. The security chain must work without internet and without a server: local siren, local recording and a relay output. Software is the intelligent layer, not the critical one.

02 · Comparison

Outdoor detection technologies

Reference values from professional-grade equipment. Ranges depend on the model, the lens and the mounting height.

TechnologyTypical range (person)Coverage Rejects small animalsRejects moving vegetation Tells person from large animalOutput
24 GHz radar3–60 m~5,600 m² (1 unit) YesYesNoIP + relay
24 GHz long-range radarup to 140 m~20,000–30,000 m² YesYesNoIP
60 GHz radar5–60 m95° · measures speed YesYesNoIP + I/O
2D LiDAR20×20 m plane95° · 8 zones Yes (by size)ConfigurableBy sizeIP + relay
Thermal 160×12091 m detect / 23 m recognise50°×37° With analyticsYesWith AIIP + relay
Outdoor PIR12 m85–90° Mounting-dependentNoNoRelay / IP
Outdoor curtain2×15 mNarrow plane Up to 80 cmPartialNoRadio / relay
Quad IR beamup to 200 mLine PartialYesNoRelay only
AI cameralens-dependentField of view YesYesYesIP + relay

The decisive column is the last detection one: only image classification tells a person from a large dog or a wild boar. Neither radar nor PIR does — radar classifies such an object as "unknown".

03 · Radar

Volumetric radio-frequency detection

FMCW radar: simultaneous coverage and object trackingPlan view of the area covered by a security radar. The radar does not rotate: it observes the whole sector at once using digital beamforming, and each measurement cycle yields range, azimuth and velocity for every object, which it follows as a track. 20 m40 m60 m-90°-60°-30°+0°+30°+60°+90° WALL RADAR SHADOW PERSON · ID 04 v = 1.4 m/s VEHICLE · ID 05 'SWAYING OBJECTS' FILTER vegetation discarded < 3 m: two objects merge into one 24 GHz FMCW RADAR NO MOVING PARTS · FULL SECTOR EVERY CYCLE AZIMUTH 180° · RANGE 60 m
Radar. There is no rotating antenna: the whole sector is measured every cycle. Each object yields range, azimuth and velocity, and from those a track is built. Classification separates person and vehicle; everything else — including a large dog or a wild boar — stays 'unknown'. Behind every solid obstacle there is a blind zone, and two objects less than 3 m apart merge into a single detection.
Reference · 24 GHz, 60 m class

Typical specification manufacturer datasheet

Band
24.05–24.25 GHz · transmitted power < 100 mW · licence-free band across the EU
Range
Person 3–60 m · vehicle 3–85 m
Coverage
~5,600 m² for a person · ~11,300 m² for a vehicle
Mounting
Recommended height 3.5 m · reference person 170 cm
Separation
Two objects closer than 3 m merge into a single detection
Classification
Person / vehicle / unknown — three categories
Filters
Ignore small objects (manufacturer documentation names cats and rabbits) · ignore swaying objects (trees, bushes)
Power
PoE IEEE 802.3at Type 2 Class 4 · typical draw 11 W
Protection
IP66 · NEMA 4X · IK08

Strengths

  • Immune to shadows, headlights, rain and insects on the lens
  • Large area from a single unit and a single IP address
  • Native filtering of wind-blown vegetation — decisive in a Mediterranean garden
  • Relay output on several models: it acts without a network

Limitations

  • Radar shadow behind every solid object: wall, hedge, retaining wall, plant room
  • Metal surfaces in the field cause reflections that degrade classification
  • Cannot tell a large dog or a wild boar from a person: both are "unknown"
  • There is no genuinely residential price tier
Variants

When to change band or range

VariantReference dataUse
24 GHz · 60 m180° · ~5,600 m² · relay outputVilla garden, medium plot, temporary site
24 GHz · long range140 m person / 200 m vehicle at 5 m mounting height · 180° ~20,000 m², 270° ~30,000 m² · PoE 802.3btLarge property, one unit instead of three
60 GHz61.00–61.50 GHz · 95° · person 5–60 m, vehicle up to 150 m in road profile · speed measurement ±2 km/h up to 200 km/h · 12 VDC output max 50 mA, no power relayStraight driveway, internal speed control

Radio compliance. Both the 24.05–24.25 GHz and the 61 GHz bands are short-range device bands that are licence-free across the European Union; datasheets declare conformity with the relevant ETSI standards.ETSI EN 300 440 / EN 305 550

04 · LiDAR

Laser-plane detection

2D LiDAR: range profile and detection by deviationPlan view. A rotating mirror sweeps the plane and measures range at every angular step. The unit learns the background profile; a deviation from that profile is a detection, and object size is computed from its angular width and its range. 123 DEVIATION = DETECTION width 62 cm · DETECTED width 26 cm · BELOW THRESHOLD 2D LiDAR · CLASS 1 ROTATING MIRROR WALL / FENCE the mirror turns 360°; the useful sector is 95° LEARNED BACKGROUND PROFILE MEASURED PROFILE OBJECT-SIZE THRESHOLD PER ZONE ZONE 1 · ≥ 60 cm ZONE 2 · ≥ 30 cm ZONE 3 · ≥ 15 cm size = angular width × range
LiDAR. The mirror rotates continuously and measures a range at every angular step: the result is a range profile. The unit learns the background profile, and what it detects is the deviation from it. Because it knows both range and angular width, it computes the real size of the object — which is why the threshold can be set in centimetres, zone by zone.
Reference · 2D security LiDAR, 20 × 20 m

Typical specification manufacturer datasheet

Area
20 × 20 m · 95° aperture
Zoning
Up to 8 independent zones, each with configurable sensitivity, minimum object size and output
Sensitivity
Detects objects of the order of 2 cm thrown through the plane
Laser class
Class 1 — eye-safe, no mandatory warning signage
Power
PoE IEEE 802.3at, 10 W · or 19.2–30 VDC (max 420 mA at 24 V)
Mounting
Recommended outdoor height 4 m · alarm delay ~2 s
Environment
IP66 · −40 to +60 °C
Larger model
30 × 60 m · 190° · fog/snow mode · EN 50131 Grade 3 · ONVIF Profile S

Strengths

  • The only technology with numerically configurable object size per zone — the technical route to ignoring a cat while detecting someone crawling
  • Centimetre precision: it can raise a vertical plane tight against a glazed facade
  • Clean cut at the boundary: unlike radar, it does not reach into the neighbour's plot
  • Class 1: no eye risk and no signage obligation

Limitations

  • Degraded performance in dense fog and heavy rain: scattering absorbs the pulse. Filter modes reduce this, they do not remove it
  • Expensive for general area surveillance: radar covers more for less
  • Requires installer configuration software
  • The manufacturer does not publish the numeric object-size range on the web datasheet: it is read in the software

Typical domestic application. A vertical plane in front of a glazed living room or a run of terrace doors; a cellar or valuables room; the top of a wall; a narrow strip along the boundary. Outside these cases it usually does not pay for itself.

05 · Thermal

Infrared radiation detection

Thermal: pixels on target at three distancesThe same person seen by a 160 by 120 thermal camera at 91, 23 and 11 metres. Angular resolution fixes how many sensor pixels the target occupies: at 91 metres only a few, enough to say a heat source is present; at 23 metres a human figure is recognisable; at 11 metres features can be told apart. SENSOR 160 × 120 · NETD < 40 mK · 3.1 mm LENS 91 m · DETECTION ≈ 3 px on target a heat source is present 23 m · RECOGNITION ≈ 9 px on target it is a person 11 m · IDENTIFICATION ≈ 18 px on target features distinguishable Angular resolution does not change: what changes is how many pixels the target covers.
Thermal. Distance does not 'blur' the image: it reduces the number of sensor pixels the target occupies. That is where the three catalogue figures come from. The detection figure is the one that gets advertised; the recognition figure is the one that decides whether the system is useful, and it is the one we size from.
Reference · Bi-spectral 160 × 120

Typical specification manufacturer datasheet

Thermal sensor
160 × 120 · NETD < 40 mK · 3.1 mm lens · 50° × 37.2° field
Visible channel
4 MP (2688 × 1520) · 4 mm lens · 84° field
Person DRI
Detection 91 m · recognition 23 m · identification 11 m
Vehicle DRI
Detection 280 m · recognition 70 m · identification 35 m
Analytics
Up to 8 VCA rules
Protection
IP66 · 6 kV surge protection

The commonest quoting error in this industry. The detection figure (91 m) is the one that appears in brochures; the recognition figure (23 m) is the one that determines whether the system is useful. Designing from the first produces systems that "see something" but do not let you decide. We size from the recognition figure.

ResolutionReferenceDomestic use
160 × 120NETD <40 mK · bi-spectralBest cost/performance on a 0.1–3 ha plot
384 × 288NETD 40 mK · 7 to 35 mm lenses (55°…10.5°)Long sight lines, beyond 150 m
640 × 480NETD 30 mK · 10/19/35 mm lenses (63°/31°/17°)Beyond what a home needs

06 · PIR & beams

Volumetric and line detectors

The cheapest layer — and the one that produces the most false alarms when it is mounted wrongly.

TypeReference dataDeclared immunityOutput
Dual-layer outdoor PIR 12 m / 90° · 16 zones · mounting 0.8–1.2 m · Grade 2, Environmental Class 4 · −30 to +60 °C · 9.5–18 VDC, 20 mA The manufacturer publishes no kg or cm figure: immunity comes from the dual-layer geometry Relay
PIR + microwave12 m / 85° · 94 zones · mounting 2.5–3.0 mQualitativeRelay
Bilateral curtain barrier 2 × 12 m · two PIRs per side, both must trigger · mounting 0.8–1.2 m (1.0 m recommended) GeometricRelay
Wireless outdoor curtain 2 × 15 m · dual optical system, two PIRs per system · ~4 years battery 80 cm + birdsRadio
Wireless outdoor detector 3–15 m · two independent IR sensors · anti-masking · −25 to +60 °C · up to 5 years battery 80 cmRadio
Wireless detector with imaging Mounting 0.8–1.3 m · 1–5 images per alarm at 320 × 176, or up to 3 at 640 × 352 · delivery in 9–30 s 80 cmRadio + image
Quad IR beam Up to 200 m outdoors · dual modulation · response time 50/100/250/500 ms · IP65 · −35 to +60 °C High against vegetation Dry contact only — no network version
IP outdoor PIR 30 × 20 m, 40 × 10 m or 40 × 4 m depending on model · 4 independent areas, each with its own output · PoE Configurable per area Native IP, integrates into a VMS

Two design consequences that are almost never explained.

1. On dual-layer PIRs, animal immunity is not a product parameter, it is the result of the mounting. Raising the sensor to 2 m for aesthetic reasons cancels the geometry and with it the immunity. That is why we document the mounting height in the handover dossier.

2. Classic IR beams have no network version: dry contact only. If the project has no alarm panel, an input module has to be allowed for — and it often costs more than simply choosing an IP sensor in the first place.

07 · Analytics

Levels of artificial intelligence

Classifier output: bounding box, class, confidence and trackingFrame from a camera with analytics. Every detected object produces a bounding box, a class, a confidence value compared against a configurable threshold, and a tracking identifier. A person raises an alarm; a vehicle is logged; an animal is logged without raising an alarm. PERSON · ID 04 ALARM VEHICLE · ID 05 LOG ANIMAL · ID 06 LOG · NO ALARM AI CAMERA · ON-DEVICE INFERENCE EMITTED METADATA classperson / vehicle / animal confidencevalue vs. configurable threshold boxx, y, width, height trackingpersistent identifier actionalarm / log only confidence threshold
AI analytics. The detector does not say 'movement': for every object it emits class, confidence, bounding box and a tracking id. The action depends on the class and on whether confidence clears the configured threshold. The animal category is what lets a wild boar be recorded without waking anyone.
LevelWhat it solvesWhere it runsLimitation
Person/vehicle filteringDiscards leaves, shadows, light changes, rainCamera or recorderNo attribute search
On-camera classificationPerson, vehicle and vehicle type; line crossing, intrusion, time in area, occupancy, tailgatingCamera, no serverNo dedicated animal class
Classification with animal categoryPerson, vehicle and animal as an independent classCameraModel-dependent
Search and attributesClothing type and colour, face, number plate, natural-language search over the archiveAnalysis unit or serverProcessing cost; higher data-protection sensitivity
Upgrade applianceAdds advanced recognition to existing ONVIF camerasApplianceDepends on the quality of the original stream

Where to run the analytics

LocationAdvantageDrawback
On the cameraNo extra hardware; analyses at full resolution and frame rate; scales linearlyUpdating the model means updating firmware device by device
On the recorderUpgrades cameras without AI; updated in one placeLimited number of AI channels; decoded stream quality is a constraint
On a serverStronger models, advanced search, cross-camera trackingCost, power draw and cooling

Domestic rule. Detection on the camera for the alarm — low latency and a reliable relay output; analysis on an appliance or server only if the client needs retrospective search. They do not compete with each other.

Why the animal category is a specification here, not a luxury. A study by the Malaga Veterinary Association with the University of Cordoba on urban wild boar captured in the province found: Marbella 40%, Malaga city 30%, Fuengirola 22.5%, Benahavis 7.5%. A wild boar far exceeds the 25 kg of declared pet immunity and moves at ground level. Neither PIR nor radar solves it.

08 · Selector

What your plot actually needs

Six questions, immediate result. Calculated in your browser: nothing is sent or stored.

1 · Outdoor area to cover
2 · Is the garden heavily subdivided? (hedges, retaining walls, pergolas, plant room)
3 · Large dog, wild boar or wildlife in the area?
4 · Is there a specific critical plane? (glazed facade, cellar, wall, wall of value)
5 · Outdoor lighting
6 · Distance to the sea

Indicative result. It does not replace a site visit: radar shadow, real sight lines and existing infrastructure can only be measured on site.

09 · Integration

Protocols and interfaces

What works across manufacturers, and what ties the project to one brand.

InterfaceCross-vendor?Real-world use
RTSPUniversalYou always get the stream. Authentication and URL scheme vary.
ONVIF Profile TYesThe current baseline. H.265 video, analytics metadata, imaging settings.
ONVIF Profile SBeing retiredDeprecated by ONVIF because its authentication methods no longer match current cybersecurity recommendations. Do not specify it on new projects.
ONVIF Profile GYesRecording, search and playback on the device itself.
ONVIF Profile MThinly implementedCross-vendor metadata and analytics: object classification, number plate, counting, MQTT support. Verify device by device.
MQTTYesThe automation glue. Local bus, no cloud.
Dry contact (relay)UniversalThe most underrated interface: siren, gate, panel zone. It works when nothing else does.
Vendor HTTP APINoVery capable, but it ties the project down and can change with firmware.
SIPYesVideo entry independent of the manufacturer's cloud.
KNX / ModbusYesOutput to existing home automation: lighting, blinds, irrigation.

Our stable layer. On multi-vendor projects we build on RTSP (image) + ONVIF T/M (event) + MQTT (automation) + relay (action). Anything beyond that ties the project to one manufacturer — a decision to be taken deliberately, not by default.

10 · Network & power

The layer that decides whether the system alerts

PoE classes IEEE 802.3

StandardTypePort (PSE)Guaranteed at the device (PD)Pairs
802.3afType 115.4 W12.95 W2
802.3atType 230 W25.5 W2
802.3btType 345 / 60 W40 / 51 W4
802.3btType 475 / 90 W62 / 71.3 W4

Power available at the device is always lower than at the port: the difference is loss over 100 m of cable. And bt classes use all four pairs: a "two-pair" installation rules them out, and that can only be fixed by pulling new cable.

Sizing rules engineering judgement

  • Plan the PoE budget to 80%. Heated cameras all start together after a winter outage — the worst possible moment.
  • UPS across the whole chain: ONT, router, switch, recorder and panel. The classic failure is powering only the recorder: the system records but never alerts.
  • Second communication path in the panel, not in the router: different carrier, different network.
  • Hourly heartbeat. It turns an outage into an alert instead of into silence.
  • Time synchronisation from a common source: without it the timestamp is arguable and multi-camera reconstruction falls apart.

Outdoor protection

  • Surge protection on every long outdoor line: gate, pool, outbuilding. Reference standard IEC/EN 62305, whose part 3 is in its 3rd edition, 2024 — specifications citing the 2010 edition are out of date.IEC 62305-3:2024
  • One single equipotential bonding system. The most expensive mistake is two separate earthing systems.
  • Optical isolation on long runs or towards a separate building.
  • Solid copper cable, UV-resistant sheath, gel-filled for buried runs. CCA is not acceptable.

11 · Recording

Storage calculation

Formula

Storage (GB) =
bitrate (Mbps) × 3600 × h/day × days × cameras / 8 / 1024

Indicative bitrate scene-dependent

ResolutionH.264H.265
2 MP4–6 Mbps2–3 Mbps
4 MP8–10 Mbps4–5 Mbps
8 MP16–20 Mbps8–10 Mbps

Codecs

  • H.264 → H.265: roughly 50% better compression efficiency.manufacturer technical paper
  • H.265 → AV1: theoretically 20–30% better, with variable real-world results.
  • The same paper warns: a codec change on its own does not absolutely determine the storage saving. What dominates is scene complexity, motion, night-time noise, frame rate and GOP length.

How we size it. On H.265 with a 30–40% reserve, and always to 30 days: that is the legal maximum retention in Spain, not a technical limitation. A night scene with rain and trees pushes the bitrate past any calculator.LOPDGDD art. 22.3

12 · Coastal

Material specification on the Costa del Sol

The section almost no quote includes, and the one that decides whether the system lasts five years or two.

Corrosivity ISO 12944

CategoryEnvironment
C3Moderate urban or industrial
C4Industrial, or close to the sea
C5-MCoastal, high humidity and salinity — salt spray test of 720 to 1,000 h
CXExtreme — offshore, chemical

Practical rule: front line and coastal developments, C5-M; the band a few hundred metres inland, C4.

Material consequences engineering judgement

  • 316L, not 304. The molybdenum in 316L is what resists chloride pitting. Most "stainless" brackets and screws on the market are 304 or worse.
  • The weak point is never the camera. It is the bracket, the screw, the cable gland and the connector. A 316L camera with 304 fixings accelerates galvanic corrosion.
  • Marine anodising rather than powder coating. Under damaged paint, corrosion spreads unseen.
  • Heat. The usual upper limit of professional equipment is around +55 °C, and that does not guarantee operation in full sun on a south-facing facade. Sun shields, orientation and a ventilated location for the UPS.

Mechanical protection IEC 62262 / 60529

IK08 = 5 J · IK09 = 10 J · IK10 = 20 J. On a perimeter, IP66 minimum; IP67 where flooding is possible. Any device installed in the open without shelter should meet Environmental Class IV of EN 50131 (−25 to +60 °C, 85–95% humidity).

13 · Cybersecurity

A badly exposed system is a risk, not a protection

Why we don't open ports or use P2P

After a critical unauthenticated command-injection vulnerability was published (CVSS 9.8), more than 80,000 exploitable cameras were documented as directly reachable from the internet.

And the manufacturer's P2P cloud is not the safe alternative: in a campaign documented in 2026, more than 14,500 devices were compromised; on the P2P path the relay established the route without prior authentication, and 89.4% of the serial numbers queried returned an open, unauthenticated channel.

Two authentication-bypass vulnerabilities from 2021 were still in CISA's known exploited vulnerabilities catalogue in August 2026, five years on. And in 2025, stack overflows in the ONVIF handler itself were published, exploitable without authentication.

What we apply

  • No port forwarding and no P2P: disabled by default in the handover configuration
  • Separate VLAN for cameras, with no internet egress: the single most effective measure
  • Remote access over VPN with an access list: the owner reaches the viewer, not the camera network
  • Factory reset, current firmware, an admin account separate from the client account
  • HTTPS with TLS 1.2/1.3, unused services disabled, anonymous access prohibited
  • Microphones disabled by default and documented in the handover record
  • Firmware inventory and an update window for critical vulnerabilities

Wider responsibility. A compromised system doesn't just stop protecting its owner: it becomes a tool for attacking third parties. That is why network configuration is part of the project, not an extra.

14 · Standards

Applicable technical and legal framework

StandardSubjectProject implication
EN IEC 62676-4:2025Video surveillance application guidelines Replaces DORI terminology with seven pixel-density levels: 20, 40, 80, 125, 250, 500 and 1,500 px/m. Specifications citing the 2015 edition are out of date.
EN 50131Intrusion systems Grades 1 to 4 and environmental classes I to IV. Outdoors, Class IV.
Orden INT/316/2011Operation of alarm systems Grade 1 cannot be connected to a receiving centre. Grade 2 for dwellings intending to connect to one. Sequential verification: three or more signals from different detection elements within 30 minutes. Video verification: one image from the moment of the alarm and two afterwards. Two false alarms in a month can suspend the system.
LOPDGDD art. 22Video surveillance and personal data Retention of one month maximum; where an offence is involved, handover to the authorities within 72 hours. Informative signage. Strict limits on the public road and a prohibition on capturing a neighbour's property.
IEC/EN 62305Lightning protectionPart 3 in its 3rd edition, 2024.
ISO 12944Atmospheric corrosivityC5-M on the coastal front line.
IEC 62262 / 60529Impact and ingressIK08/09/10 · IP66 minimum on a perimeter.
ONVIFInteroperabilityProfile T as the baseline; Profile S being retired.
Ley 5/2014Private security (Spain) Defines which activities require authorisation. We install systems not connected to an alarm receiving centre or control centre, outside the scope of the activities subject to authorisation. We provide no guarding, keyholder response or monitoring.

Methodology note. The specifications on this page are taken from manufacturer datasheets and from standards texts, and have been selected as product-class references to illustrate each technology: they are not an offer of a particular brand and do not imply official distribution. Ranges, power draw and performance vary with the model, firmware version, lens and mounting height, and are verified against the datasheet of the exact device before it goes into a quote.

Values marked engineering judgement are our own sizing rules, not figures published by a manufacturer. Those marked standard refer to the cited standard, whose current version must be checked on each project.