Video surveillance and CCTV system design field guide

What a video surveillance system is, and what the design actually decides

A video surveillance system, still called CCTV out of habit, is four things working together: the cameras, a recorder or video management software, the storage that holds the footage, and the network that carries it. The hardware is the easy part. The design that matters is matching the camera, the resolution, and the lens to the scene in front of it and the goal you have for that scene, because a camera that records a face it cannot resolve is just expensive video.

That goal is measured in pixels per foot at the target distance, and it is the number that drives almost every other decision. Too many installs go up as a count of cameras on a floor plan, aimed roughly at doors, recording around the clock to a box in a closet. Then an incident happens, somebody pulls the footage, and the person at the register or the door is a gray smear nobody can identify. The cameras worked. The design failed.

Designing to the pixels-per-foot goal, the field of view, the available light, the retention you have to keep, and the network the cameras live on is the real work. This guide treats the cameras as one part of a system that also includes the structured cabling and the power, which we cover in the low-voltage and Class 2 cabling guide, and that often integrates with the life-safety systems covered in the fire alarm guide.

A camera that cannot identify is just video

This is the point that the rest of the guide hangs on, so it goes first and blunt. A camera produces footage. Whether that footage is useful evidence or a fuzzy waste of disk depends entirely on whether the pixels landing on the target are enough to do the job you bought the camera for.

The failure is almost always the same. Somebody put a 4 MP camera on a wide lens, mounted it 30 ft from the entrance, and assumed megapixels meant detail. At that distance, across that field of view, the face at the door might get 15 pixels per foot, which is enough to see that a person walked in and nothing more. You can see a coat. You cannot see a face well enough to identify anyone, and identify is what the owner thinks they bought.

The fix is not a bigger number on the spec sheet. It is designing backward from the goal: decide what each camera has to accomplish at a stated distance, calculate the pixels per foot that goal requires, then pick the resolution and lens that deliver it across the scene. Get that right and the rest of the system has something worth recording. Get it wrong and no amount of storage or analytics saves you.

Detect, recognize, or identify: which is the goal?

Every camera in a design has one of three jobs, and the job sets the resolution you need at that distance. The industry sorts them as detect, recognize, and identify, and the European framing adds observe and monitor for a five-level scale often called DORI. The distinction is the most useful design tool you have, so set it for every camera before you pick the hardware.

Detect means you can tell a person or vehicle is present in the scene. Recognize means you can tell whether it is someone you know, or match a person seen on one camera to the same person on another. Identify means you can pick a stranger out from the footage well enough to stand behind it later. A parking lot perimeter might only need detect. A cash register, a pharmacy counter, or a controlled door needs identify, and that is a far more demanding target.

Set the goal per camera, in writing, on the design. A camera asked to identify at 40 ft is a different camera, a different lens, and a different bill than one asked to detect across the same 40 ft. The manufacturer's pixel-density tables and the project's operational requirements should confirm the goal-to-resolution mapping for the specific model, because the thresholds shift with sensor, lens, and scene.

What is pixels per foot in CCTV?

Pixels per foot, abbreviated ppf, is the number of horizontal pixels the camera lands on each foot of the scene at the target distance. You get it by dividing the camera's horizontal pixel count by the width of the field of view in feet at that distance. A 1920-pixel-wide image covering a 20 ft wide scene gives 96 ppf at that plane. The same camera covering 96 ft gives 20 ppf. Same camera, very different evidence.

The working rule of thumb most designers carry: roughly 20 to 25 ppf to detect a person, around 40 to 50 ppf to recognize someone, and 80 ppf or more to identify a stranger, with license-plate capture often wanting 120 ppf or higher. The European DORI scale states the same idea in pixels per meter, where identification is commonly set near 250 ppm, which works out to about 76 ppf. Treat these as design targets to confirm against the manufacturer's figures for the specific camera, not as hard law.

The number that catches people is that pixels per foot falls off as the scene gets wider or deeper. A camera is not a fixed quality. It is a quality at a distance and a field of view. Design to the ppf at the spot where the action happens, the door threshold or the counter, not to the average across the whole frame.

IP or analog cameras?

IP cameras are the standard for new commercial work, and have been for years. They are network devices: megapixel sensors that encode video and send it over twisted-pair Ethernet, usually powered by the same cable through PoE. They reach high resolutions, run analytics on the camera, integrate with software, and ride the structured cabling you are already pulling for everything else.

Analog still exists in two forms. True legacy analog runs composite video over coax to a DVR at standard-definition quality, and it belongs to systems built a decade or more ago. HD-over-coax, sold as HD-TVI, HD-CVI, and AHD, pushes megapixel images down existing coax and earns its place on one job: a retrofit where good coax is already in the walls and re-pulling Cat6 to every camera is not worth the cost or the disruption. Power over coax keeps those cameras on a single cable the way PoE does for IP.

For a new building, IP is the default and the question is which IP cameras, not whether. For a retrofit with sound coax already run, HD-over-coax can be the honest call. The manufacturer's compatibility and the existing cabling decide it, so confirm the coax type and condition before you promise a coax upgrade will hold the resolution you are selling.

Resolution, and why more is not always better

Resolution is the sensor's pixel count, given in megapixels or by a label like 1080p (about 2 MP), 4 MP, 5 MP, or 4K (about 8 MP). It sets the ceiling on how much detail a camera can deliver, which feeds straight into pixels per foot. So far the spec sheet and reality agree: more pixels, more potential detail.

They part ways fast. Cramming more pixels onto the same sensor size shrinks each pixel, and smaller pixels gather less light, so a high-megapixel camera can perform worse at night than a lower-resolution one with a larger sensor. More pixels also cost more storage and more network bandwidth, every hour of every day. And a high-resolution camera on a wide lens spreads those pixels across a huge scene, so the ppf at the target can be lower than a modest camera aimed tighter.

Match resolution to the ppf goal at the distance, not to the biggest number the budget allows. The right answer is sometimes a 4 MP camera on the correct lens rather than a 4K camera on a wide one, because the 4 MP camera puts more pixels on the face that matters and costs less to store. Let the calculation pick the sensor, with the manufacturer's low-light specs as the tiebreaker when two options hit the same ppf.

Camera types: dome, bullet, PTZ, and multisensor

Form factor follows function. The four common types each answer a different problem, and mixing them on one site is normal rather than a compromise. Pick the type from what the location has to do, then size the lens and resolution to the ppf goal.

Dome cameras sit low-profile against a ceiling or soffit, are harder to tell where they are aimed, and resist casual tampering, which suits interiors and entries. Bullet cameras are visible and directional, easier to aim at a long throw, and read as a deterrent, which suits perimeters and parking. PTZ cameras pan, tilt, and zoom on command and earn their cost only where a live operator drives them or a guard tour is programmed, because an unmanned PTZ is usually pointed the wrong way when the incident happens. Multisensor cameras carry several imagers in one housing to cover a wide area or a full 360 from a single mounting point and one cable drop, which cuts the device count on a large open space.

The lens and the field of view

The lens sets the field of view, and the field of view sets the pixels per foot at distance. Focal length is the lever: a short focal length gives a wide angle that covers more of the scene with less detail per foot, and a long focal length gives a narrow angle that covers less but packs more pixels onto the target. You cannot have both wide coverage and high detail from one fixed camera. That tradeoff is the heart of lens selection.

A varifocal lens lets you set the angle on site, which matters because the as-built mounting position rarely lands exactly where the drawing put it. Motorized varifocal and autofocus models let you tune the view from the software without a ladder, which is worth the premium on high or awkward mounts.

Run the geometry before you order. A field-of-view or lens calculator from the camera manufacturer, fed the mounting distance and the target width, tells you the ppf you will actually get from a given focal length. Pick the lens that hits the ppf goal at the target plane, then confirm the horizontal coverage still spans the area the camera is responsible for. Wide enough to cover, long enough to identify, is the balance, and on a demanding scene that balance forces a second camera rather than one camera trying to do both jobs.

Camera layout and coverage

Coverage is the plan that places the cameras so the scene the owner cares about is captured at the ppf each spot needs, with no gap where it counts. It starts from the goals, not from a grid. Walk the choke points first: entries and exits, the path to the safe or the server room, the register, the loading dock, the points where anyone has to pass. Those are where you spend your identify-grade cameras.

Blind spots are the quiet killer. A camera covers a cone, and people, racking, signage, and opening doors create shadows inside that cone. The classic miss is a camera that watches the sales floor beautifully and sees nothing of the doorway directly beneath it, so everyone entering is captured only from behind. Stand where the subject will stand and look back at the camera, on the plan and again on site, to find the angles the floor plan hides.

Overlap at the boundaries on purpose. Where two cameras meet, a little shared view means a person stays on at least one camera while crossing between them, and it covers the moment one camera is blocked. The goal is continuous capture along the paths that matter, not a camera in every corner. Cameras are not free to buy, power, cable, or store, so place them where the goal lives and leave the empty corners empty.

Low light, IR, and backlight

Most cameras look fine in the showroom and in daylight. They earn their keep, or fail, at night and against glare, and that is where cheap cameras fall apart. Low-light performance is set by sensor size, lens aperture, and the camera's processing, and a high-megapixel sensor with tiny pixels often does worse in the dark than a lower-resolution camera with larger ones.

Infrared illuminators built into the camera light the scene with near-IR the sensor sees and the eye does not, giving a usable black-and-white image in full darkness. IR has limits: a stated effective range that drops with weather, hot spots and washout up close, flare off a dome bubble or a nearby wall, and a real power draw that pushes the camera into a higher PoE class. Confirm the IR range covers the distance to the target, not just the foreground.

Wide dynamic range, WDR, handles the other hard case: a bright background behind a dark subject, the silhouette at a glass entrance with sunlight behind it. Without WDR the face goes black against the window and you have a shadow nobody can identify. The catch is that low-light and WDR ratings on a data sheet are measured under ideal conditions. Walk the site at the worst hour, dusk and full dark, with sun on the glass, and confirm the camera holds the ppf goal then, because the worst light is when the incident happens.

The recorder: NVR versus VMS

Where the video lives and how it is managed comes in two forms. An NVR, a network video recorder, is a dedicated appliance that records the IP cameras to its own drives, sized by channel count, and it suits small and mid-size sites where the camera count is bounded and simplicity wins. A DVR is the older equivalent for analog and HD-over-coax systems.

A VMS, video management software, runs on standard servers or in the cloud, scales to far more cameras across multiple sites, and integrates with access control, alarms, and analytics under one interface. Enterprise and multi-building work points to a VMS for the scale and the integration. The line between the two has blurred as cloud-managed recorders pick up VMS features, so the real question is the count, the number of sites, and what the system has to talk to.

Interoperability runs on ONVIF, the standard that lets cameras and recorders from different makers work together. ONVIF Profile S covers basic streaming, Profile T adds H.265 and analytics events, and Profile M carries analytics metadata. A manufacturer's NVR often only fully supports that manufacturer's cameras, while a VMS is usually camera-agnostic. Confirm the specific ONVIF profile support before you mix brands, because partial support is where features quietly go missing.

How much storage does a camera system need?

Storage is the number one sizing miss in the trade, and it is pure arithmetic, so there is no excuse for guessing. Total storage is the sum across cameras of bitrate times recording hours times retention days. The clean shortcut: one megabit per second of continuous video is about 10.8 gigabytes per day, so multiply each camera's bitrate in Mbps by 10.8 for daily gigabytes, multiply by retention days, sum the cameras, then add headroom.

The drivers are resolution, frame rate, compression, hours recorded, and how many days you keep. A 1080p H.265 stream might run 2 to 3 Mbps, a 4 MP stream more, and 4K more again. Frame rate scales it nearly linearly: 15 fps is the surveillance default, dropping to 10 fps saves roughly a quarter to a third, and 30 fps roughly doubles the load. Recording on motion or schedule instead of around the clock cuts it further but makes the daily number variable, so size for the busy day.

Retention is usually the input you do not get to choose. A project spec, an insurer, a franchise agreement, or a local rule sets the days, commonly 30, sometimes 60 or 90. Pin that number down before anything else, because it multiplies the entire system. Add 10 to 15 percent overhead for filesystem and future cameras, use surveillance-rated drives built for continuous write, and plan the RAID or redundancy so one failed drive is not lost footage. The manufacturer's calculator and the project's retention requirement control the final size, so confirm both against the design.

Compression: H.264, H.265, and smart codecs

Compression decides how many bits each camera spends to send a given image, and it lands directly on both storage and network. H.264 is the long-standing codec. H.265, also called HEVC, delivers similar image quality at roughly 30 to 50 percent less bitrate, which is real money across dozens of cameras recording every hour.

On top of the codec, manufacturers layer smart or adaptive schemes under names like Smart Codec, H.264+, H.265+, and Zipstream. They cut the bitrate further in static scenes by spending bits only where something changes, then ramping back up when the scene gets busy. The savings are largest on a quiet hallway and smallest on a windy parking lot full of motion, so the marketing number is a best case, not a guarantee.

Use H.265 with a smart codec where the whole chain supports it, the camera, the recorder or VMS, and the client that plays it back. Mismatched support means a stream that records but will not display, or one that silently falls back to H.264 and blows your storage budget. Confirm end-to-end codec support with the manufacturer before you size storage on the smaller number.

The cameras live on a network

Every IP camera is a networked device, and the network is part of the design from the first day, not an afterthought handed to IT at the end. Each camera generates a steady stream, and dozens of them add up to real bandwidth that has to reach the recorder without choking the rest of the building. Size the switch uplinks and any inter-building links to the total camera bitrate plus the live-view and playback traffic, with margin.

Keep the cameras on their own segment. A separate VLAN for surveillance isolates the camera traffic from the business network, contains a compromised camera so it cannot reach the file server, and keeps the heavy video stream off the staff segment. This is both a performance move and a security one, and it is the single most common thing missed on installs where the electrician ran the cable and nobody owned the network. Coordinate the VLAN, the IP scheme, and the switch capacity with IT early.

The physical layer, the Cat6 and the switches and the pathways, is structured cabling like any other low-voltage system, and the bend radius, separation from power, and termination standards in the low-voltage and Class 2 cabling guide apply here too. The cameras only perform if the cabling underneath them was pulled and tested right.

Power over Ethernet and the power budget

PoE delivers power and data to the camera on one cable, which is most of why IP cameras are quick to install. The standard comes in tiers: 802.3af (Type 1) supplies about 15.4 W at the switch port, 802.3at (PoE+, Type 2) about 30 W, and 802.3bt (PoE++, Types 3 and 4) up to 60 or 90 W. A fixed dome may sip under af. A camera running IR illuminators, a heater for an outdoor housing, or a PTZ motor needs at, and the heavy outdoor and multisensor units reach into bt.

Two numbers bite in the field. First, the run limit: Ethernet, and therefore PoE, is rated to 100 m of cable, and a long run loses several percent of the power to cable resistance, so a camera near the end of a 100 m run can drop into low-power mode and shut off its IR exactly when you need it. Second, the switch power budget: a switch rated at, say, 370 W total cannot actually drive 24 ports all pulling 30 W. Add the per-camera draw, leave margin, and confirm the switch's total budget covers it.

Where the distance or the power exceeds what the switch gives, a midspan PoE injector sits between the switch and the camera and supplies the power, or a higher-class switch solves it. Pull the per-camera power class and the cable distance from the manufacturer and check both against the switch budget, the same way the cabling guide treats PoE on the structured-cabling side.

Structured cabling to every camera

The cabling to the cameras is structured cabling, and it follows the same rules as any data run. Cat6 is the common choice for PoE cameras, with Cat6a where the run is long, the power class is high, or the bundle is large enough that heat in the cable bundle matters. The 100 m channel limit is the hard distance for copper, and that includes the patch cords at both ends, not just the horizontal run.

Test every run before the camera goes up. A camera that boots but resets randomly is often a marginal cable, not a bad camera, and you do not want to be diagnosing that from a lift after the ceiling is closed. Certify or at least verify each link for wiremap, length, and the performance the PoE class needs.

Beyond 100 m you change the medium rather than stretch the copper. Fiber with a media converter or a fiber-fed switch in a remote enclosure carries the camera out to the far corner of a site. The pathways, supports, separation from power, and firestopping for all of this are covered in the low-voltage and Class 2 cabling guide, and they apply to surveillance cable the same as to any other system.

Video analytics and AI

Analytics turn a camera from a passive recorder into a sensor that flags events. The early generation was pixel-change motion detection, which alarmed on wind, rain, headlights, and shadows, and trained operators to ignore it. Modern deep-learning analytics classify what moved, person, vehicle, or animal, so a blowing bag no longer triggers a guard call. That false-alarm reduction is the practical reason to care about analytics.

The features range from line-crossing and intrusion zones to loitering, object-left-behind, people counting for retail, and license plate recognition. LPR is its own discipline: it wants a dedicated camera, a tight lens, the right shutter and IR for moving plates, and a far higher ppf at the lane than general surveillance. Do not expect a general-purpose camera to read plates as a side job.

Analytics run either on the camera, at the edge, or on a server or in the cloud. Edge analytics cut bandwidth because the camera sends an alert and metadata instead of full video for review, and they keep working when the link is down. Server and cloud analytics bring heavier models and cross-camera reasoning the camera cannot do alone. The accuracy, the supported objects, and the licensing vary widely by manufacturer, so confirm what a given camera or VMS actually does against your scene rather than the brochure.

Site lighting for the cameras

Lighting is part of the camera design, not a separate trade you ignore. A camera is only as good as the light it has, and many night failures are a lighting problem the camera gets blamed for. Decide early whether a scene is lit by the camera's own IR, by existing site lighting, or by added white light, and design the coverage around the answer.

White light gives color at night, which IR cannot, and color is often what makes a description usable, the red jacket, the blue car. White light also doubles as a deterrent. IR is covert and avoids light-trespass complaints from neighbors but yields black-and-white only. Watch for glare and hot spots: a fixture or a bright sign inside the frame can wash the sensor and bury the subject in bloom, so aim the camera and place the light so the source is not staring back into the lens.

Camera cybersecurity

Cameras are computers on your network, and an insecure camera is a way into the building's data, not just a privacy problem. Hijacked cameras have powered some of the largest botnet attacks on record, and the entry point is almost always the same: a default password left in place and a service like Telnet left open. Change every default credential on every device before it goes into service, and do not reuse one password across the fleet.

Keep firmware current, because camera vulnerabilities are found and patched on a steady cadence, and an unpatched camera is a known door left standing open. Put the cameras on their own VLAN so a compromised one cannot reach the business network, and be aware that everything on the camera VLAN can still reach everything else on that VLAN, so one breached camera can pivot to its neighbors. Restrict and log who can reach the management interface.

Procurement is part of security now. Under Section 889 of the 2019 NDAA, federal agencies and recipients of federal funds are barred from buying gear from several named manufacturers, so any job touching federal money has to specify NDAA-compliant cameras. Whether or not federal money is involved, treat the manufacturer's security advisories and the building's IT security policy as governing, and coordinate the camera network with IT rather than bolting it on.

Privacy and recording law

This is where a good install becomes a lawsuit, so treat it as a design constraint, not an afterthought. Video and audio surveillance are governed by privacy law that varies by state and by country, and it is stricter than most installers assume. The hard rule that survives everywhere: do not place cameras where a person has a reasonable expectation of privacy. Restrooms, locker rooms, changing areas, and private medical spaces are off limits, and getting that wrong is not a fine, it is a serious claim.

Audio is the trap. Recording audio is regulated far more tightly than video under the federal Wiretap Act and state law, and many states require all-party consent to record a conversation. A camera with a live microphone can break wiretap law in a place where the silent video is perfectly legal, which is why many commercial systems disable audio entirely. Do not enable a microphone without confirming the law for that location.

Beyond placement and audio, signage is often required to put people on notice that recording is in use, some jurisdictions require written notice to employees about monitoring, and labor law restricts using cameras to watch union or protected activity. Pointing a camera at a neighbor's property invites its own complaints. None of this is electrical knowledge, and it is exactly the part installers get wrong. Hedge hard: confirm the placement, the signage, the audio, and the notice requirements with local privacy law, counsel, and the AHJ before the cameras record a single frame.

Integrating the VMS with access control and alarms

Surveillance rarely lives alone on a commercial site. The value climbs when the video ties into the other systems: an access-control badge swipe or a door-forced alarm pulls up the camera on that door, an intrusion alarm calls a camera to a preset, and an operator sees the event and the picture together instead of hunting for the clip after the fact. A VMS is usually where that integration lives.

Tie-ins to the fire alarm system are a special case and a regulated one. A fire event can be set to call cameras and release the egress doors, but the fire alarm itself is life-safety equipment that the surveillance system must not interfere with, and the interface has to respect the listing and the rules covered in the fire alarm installation and testing guide. Coordinate any cross-system interface so the security side never compromises the life-safety side, and confirm the integration with both manufacturers and the AHJ.

Commissioning the system

Commissioning is where you prove the design on the actual building, and it is the step that separates a working system from a closet full of recording cameras. Camera by camera, set the focus and the aim, confirm the field of view matches the plan, and verify the pixels per foot at the target plane, not in the foreground. Walk a person to the door, the counter, the gate, and confirm you can actually do the job that camera was assigned, detect, recognize, or identify.

Check the recording end too. Confirm every camera is recording to the NVR or VMS at the resolution, frame rate, and codec the design called for, and that the calculated retention is really achievable on the installed storage by letting it run and watching the oldest footage age out at the right number of days. Confirm motion or scheduled recording fires when it should. Test the failover and any alarm integration.

Then write it down. The as-built that records each camera's location, model, lens, aim, ppf at the target, and recording settings is the document the owner and the next technician live on. A system that was never commissioned to its goals is a system nobody can trust when the footage is finally needed.

Keeping the system working

A surveillance system degrades quietly, and the day you need the footage is the worst day to discover a camera has been down for a month. Set a maintenance cadence rather than waiting for an incident to find the failures.

Clean the lenses and domes: dust, spider webs, condensation, and grime soften the image and ruin night performance, and a dome bubble that has hazed over kills the IR. Check that every camera is online and recording, on a schedule, because cameras drop off and nobody notices until the clip is missing. Keep firmware patched for the security reasons above. Watch the storage so a full or failing drive does not silently stop retention or shorten it below the required days, and replace failed drives promptly in a RAID before a second one goes. Verify retention periodically against the spec, since added cameras or raised bitrates quietly eat into the days you are keeping.

What to document

The record is what makes the system defensible and maintainable, and a field tool such as FieldOS keeps the camera schedule, the field-of-view notes, the storage math, and the as-built in one place instead of scattered across a drawing, an email, and somebody's memory. Capture enough that the next technician, the owner, and a reviewer can reproduce and trust the design.

Record each camera's location and model, the lens and field of view, the target distance and the ppf goal with the goal type, the recording resolution, frame rate, and codec, the storage calculation and the required retention, the network details including the VLAN and the switch and PoE class, and the privacy and signage decisions for the site. Tie the storage and retention numbers to the spec or authority that set them, so the next person knows why the disk is sized the way it is.

Common mistakes

• Resolution too low to identify anyone at the actual target distance, because the camera was picked by megapixels instead of by ppf at the scene.
• Storage undersized for the required retention, so footage ages out before the incident is even reported.
• Blind spots from a layout drawn on a grid instead of walked at the choke points, leaving the doorway under the camera uncovered.
• Poor low-light or backlit cameras that look fine in daylight and produce a black silhouette at the hour the incident happens.
• Default passwords and open services left in place, turning the cameras into a network breach vector.
• Audio enabled or cameras placed where privacy and recording law forbid it, without signage or the notice the jurisdiction requires.

Field checklist

Standards and references

Surveillance design pulls from several authorities, and the honest move is to cite the one that controls the point and hedge the rest to the people who own it. The pixels-per-foot targets and the camera and lens selection are the manufacturer's territory: use the camera maker's pixel-density and field-of-view tables to confirm the ppf for the specific model and scene, because the detect, recognize, and identify thresholds shift with sensor and optics. The European DORI framing gives the recognized vocabulary for those levels.

Interoperability runs on ONVIF profiles, S, T, and M, which decide whether mixed-brand cameras and recorders actually cooperate. The network and power side follows the IT and cabling standards: the IEEE 802.3 PoE classes (af, at, and bt), the 100 m structured-cabling channel limit, and the Cat6 and Cat6a performance covered in the low-voltage and Class 2 cabling guide. Coordinate the bandwidth, the VLAN, and the switch budget with IT.

Storage and retention are driven by the project specification, an insurer, or a local requirement, so confirm the days against the document that sets them rather than a rule of thumb. Camera cybersecurity follows the manufacturer's security advisories, the building's IT policy, and, where federal funds are involved, NDAA Section 889 sourcing limits. Privacy and audio-recording law, signage, and employee-notice requirements are set by local and state law and the federal Wiretap Act, and they vary enough that you confirm them with counsel and the AHJ for the jurisdiction, not from a guide. Design to the ppf goal, size the storage and the network and the PoE, secure the cameras, and follow the privacy law that applies where the cameras actually point.

Units and terms

Surveillance carries its own vocabulary, and the same idea shows up under several names across a camera sheet, a drawing, and a spec. These are the terms a design hangs on.

Pixels per foot is the working detail metric; the metric world uses pixels per meter, where roughly 3.3 ppm equals 1 ppf. Resolution comes as megapixels or as labels like 1080p and 4K. Bitrate in Mbps sets the storage and bandwidth load, and one Mbps is about 10.8 GB per day of continuous recording. Keep the goal, the ppf, and the retention attached to every camera, and the rest of the design follows from them.

Video surveillance / CCTV

Cameras, a recorder or VMS, storage, and the network that carries the video; designed to capture a scene at a chosen detail level

Pixels per foot (ppf)

Horizontal pixels landed on each foot of the scene at the target distance; the metric that decides whether footage is usable

Detect / recognize / identify

The three goal levels: detect a presence (about 20 to 25 ppf), recognize a known person (about 40 to 50), identify a stranger (80 or more)

IP vs analog camera

IP cameras are networked, megapixel, PoE-powered, and the new-build standard; analog and HD-over-coax run over coax and fit coax retrofits

NVR / VMS

NVR is a dedicated recording appliance for small to mid sites; VMS is server or cloud software that scales and integrates with other systems

PoE

Power over Ethernet; power and data on one cable, in classes af (about 15.4 W), at (30 W), and bt (60 to 90 W), to 100 m

Retention

How many days of footage the system keeps, commonly 30, 60, or 90; usually set by spec, insurer, or local rule, and it multiplies the storage

Field of view (FOV)

The angle and width of the scene the lens captures; set by focal length, and it determines the ppf at a given distance

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• What is a low-voltage system?
• What is the difference between conventional and addressable fire alarm systems?
• What is an intrusion alarm system?
• What pipe is used for compressed air?
• When do you need a fire pump?