Security Guides — Checklists, Decision Trees & Buyer Tools

Guide 01 · decision tree

Choosing a camera for a use case.

Start from the job, not the product. Answer two or three questions and you get a design target, the geometry that has to be true, and the traps specific to that job.

Read the whole tree as text

Every branch of the chooser, written out. If JavaScript is off, this is the guide.

What is this camera actually for?

Identify a stranger — Produce an image good enough to identify someone you have never met.
Does everyone pass through a narrow point you control?
- Yes — a door, turnstile, till or window — There is a place every subject must cross, one or two at a time.
  A dedicated identification camera at the choke point
  Give the identification job its own camera, aimed along the direction of travel at face height, and let a second wider camera cover the rest of the room. One camera asked to do both does neither well.
  - Design to the identify class — roughly 250 pixels per metre, which is about 76 pixels per foot across the scene at the face plane.
  - Work backwards from a scene width of about 4 to 6 ft at that plane, then choose the lens. Never pick the camera first.
  - Mount low enough to see a face rather than the top of a head. Excessive down-tilt is the single most common cause of unusable identification footage.
  - Fix the backlight. A glass entrance behind the subject defeats a correctly specified camera; wide dynamic range helps, changing the lighting or the aiming direction helps more.
  - Raise shutter speed enough to freeze a walking pace, and accept that you will need more light to do it.
  Class figures come from the published operational requirement classes in IEC 62676-4. They are design targets, not guarantees — lens quality, lighting, compression and subject motion all move real performance.
- No — an open lobby or forecourt — People cross a wide area on unpredictable paths.
  Identification across an open area needs more than one camera
  No single fixed camera identifies strangers across an open lobby or forecourt. Accept that and split the problem: wide cameras for context and movement, dedicated identification cameras at every point people must pass.
  - Find the real choke points — the doors, the lift lobby, the reception desk, the turnstile line — and put an identification camera on each.
  - Cover the open floor at recognise or observe level with a multi-sensor or panoramic camera so you can follow a track between identification points.
  - If there is genuinely no choke point, say so in the design rather than promising identification the layout cannot deliver.
  - Consider whether a modest change to the layout, a rope line or a reception desk position creates the choke point for a fraction of the camera cost.
  This is the most common gap between what an owner expects from a system and what the geometry can deliver. Settle it in the design phase, in writing.
Recognise someone known — Confirm that a person in the frame is a specific employee, resident or regular visitor.
How wide is the area one camera has to cover?
- Under about 20 ft — A corridor, an aisle, a doorway, a single bay.
  One fixed camera, sized to the width
  A corridor or single doorway is the easy case. Target the recognise class — roughly 125 pixels per metre, about 38 pixels per foot — and choose the lens so the far end of the useful range still meets it.
  - Verify the number at the far end of the coverage, not at the camera. Pixels per foot falls off with distance.
  - A common mainstream sensor covers this comfortably; you are buying lens and light, not megapixels.
  - Watch for doors that swing into the field of view and shelving that changes the geometry after the tenant moves in.
  - If the corridor is also an entrance, add an identification camera at the door rather than upgrading this one.
- About 20 to 50 ft — A room, a sales floor section, a dock bay group.
  Check the arithmetic before you commit to one camera
  Recognition across a wide room is where sensor resolution starts to bind. Multiply your target pixels per foot by the scene width and compare that with the camera's horizontal pixel count.
  - 38 pixels per foot across 50 ft needs about 1,900 pixels of horizontal resolution — which puts a 1080p camera exactly at its limit, with nothing left for the corners.
  - Two cameras from opposite corners usually beat one high-resolution camera in the middle: you get two angles, and faces are not all pointed away.
  - A higher-resolution sensor raises bitrate and storage. Run the storage math before the decision, not after.
  - Multi-sensor cameras cover the width on one cable and one licence, but check the per-sensor resolution rather than the headline total.
- Over 50 ft — A yard, a car park, a warehouse span.
  Split the job: wide coverage plus targeted detail
  Recognition across a yard from a single fixed camera is rarely achievable at a sensible price. Design a layered answer instead.
  - Fixed cameras provide continuous wide coverage and the recorded track.
  - A pan-tilt-zoom camera provides detail on demand — but only detail that someone drove it to, so it is not a substitute for recording.
  - Put fixed recognition or identification cameras on the gates, doors and pinch points where subjects have to slow down.
  - Outdoors, lighting and weather move performance more than the sensor does. Budget for both.
See what happened — General awareness — an event occurred, roughly here, involving roughly this.
Indoors or outdoors?
- Indoors — Controlled light, controlled temperature, controlled access to the camera.
  Wide indoor coverage at observe level
  General awareness indoors is the cheapest coverage there is. Target the observe class — roughly 62.5 pixels per metre, about 19 pixels per foot — and spend the savings on the entrances.
  - Fewer cameras with wider lenses, placed so the fields of view overlap enough to follow a person between them.
  - Mind the ceiling: mounting height and light fixtures decide what you can actually achieve.
  - Keep every entrance on its own recognition or identification camera. The wide cameras answer "what happened"; the door cameras answer "who".
  - Lower pixel density means lower bitrate, which is real money at scale.
- Outdoors — Weather, sun angle, wind, vandalism and no useful light at 03:00.
  Outdoor awareness — design for the worst hour, not the site visit
  An outdoor camera is a lighting, weather and mounting problem that happens to contain a sensor. Specify for 03:00 in the rain, not for the afternoon you walked the site.
  - Check the sun path. A camera aimed east or west spends part of every clear day looking into the sun.
  - Mounting matters: pole sway, mast vibration and thermal movement all blur a long-lens shot. Rigid mounts and shorter lenses are more reliable than a heroic zoom.
  - Integrated infrared gives you a monochrome image, a hotspot near the lens and less range than the datasheet implies. Site lighting usually beats it.
  - Specify the enclosure for the environment — ingress protection, impact resistance where reachable, heaters where it freezes, sunshields where it bakes.
  - Verify the cable run against the 100 m channel limit before you place the camera, not after.
Read a licence plate — Capture plate characters reliably on moving vehicles.
Licence plate capture is its own discipline
Plate reading is not a general camera with a zoom on it. It is a constrained-geometry problem, and the constraints are unforgiving.
- Constrain the lane and the speed. A plate camera aimed at a wide, multi-lane, unrestricted approach is a plate camera that fails.
- Keep the angle shallow, horizontally and vertically. Steep angles distort characters; manufacturers publish limits — stay inside them.
- Use a fast shutter to freeze the plate, and dedicated infrared tuned to the plate's retroreflective surface rather than general scene lighting.
- Expect two cameras per lane: one for the plate, one for the vehicle and occupants. One camera rarely does both.
- Reading a plate is not identifying a driver. If you need both, say both in the specification.
Detect presence over a wide area — Know that something is in a yard, aisle or perimeter, and trigger a look.
Detection and counting over a wide area
The detect class — roughly 25 pixels per metre, about 8 pixels per foot — is enough to know something is there and to trigger a look. It is not enough to describe what it was.
- Pair wide detection coverage with a small number of higher-density cameras where subjects must pass, so an alert has somewhere useful to go.
- Analytics need pixels on target too. Check the manufacturer's stated minimum object size in pixels and design to it rather than to the marketing claim.
- Outdoors in poor visibility, thermal imaging detects reliably at ranges where a visible camera sees nothing — and identifies nothing at any range.
- Decide in advance who or what receives the alert at 03:00. Detection with no response path is an expensive log file.

A note on the class figures. The pixel densities quoted above come from the operational requirement classes in IEC 62676-4 and are stated in pixels per metre; the pixels-per-foot equivalents are rounded conversions. They are design targets for specifying a system, not a promise about any particular scene. See Pixels per foot for the arithmetic and IP video 101 for the background.

Guide 03 · how to write it

Writing a scope of work you can bid.

A scope of work exists to make three bids comparable. If two bidders can read it and reach different quantities, it is not finished — and the difference will surface as a change order after you have picked the cheapest one.

The eight sections

Purpose and objective — what the system must accomplish, in plain language. “Identify persons entering the north lobby” beats “install cameras.”
Site and existing conditions — what is there now, what stays, what is removed, and who verified it.
Scope of supply — the bill of material, by drawing reference. State who furnishes and who installs each line.
Scope of work — the labour: pathways, cabling, mounting, terminations, programming, commissioning.
Exclusions and work by others — conduit, line-voltage circuits, firestop, patching, painting, network ports, licences.
Standards and codes — the editions actually adopted, plus any owner standard the work must match.
Submittals and closeout — what you require before work starts and before final payment.
Acceptance — the test that decides whether it is done, written before anyone bids.

Five sentences that save projects

“Quantities are per the drawings; the drawings govern.” Stops a bidder from pricing their own smaller count.
“Substitutions require written approval and a side-by-side comparison of salient characteristics.” Stops the quiet downgrade between award and delivery.
“Pathway, backboxes and line-voltage power are by [named party].” The single most expensive ambiguity in low-voltage work.
“Acceptance requires a witnessed test per the attached procedure.” Turns “it is working” into a signed document.
“Final payment requires as-builts, test results, O&M documentation, credentials and training.” The only reliable way to get closeout documents.

And one to avoid“Contractor shall provide a complete and operational system” on its own. It sounds strong and means nothing; every bidder will interpret “complete” at their own price point.

Who writes it matters. LA CCTV Supply provides security consulting and system design, sells the equipment and trains your people. We are not an installing contractor: installation is performed by your licensed contractor, except for small non-permitted work under $1,000 all-inclusive, which we can handle directly.

Because we do not bid the labour, a scope of work we write is not quietly shaped to suit one installer’s crew or one manufacturer’s programme. You own the document, and you can send it to as many contractors as you like.

Guide 05 · planning

Planning an IDF.

The intermediate distribution frame is the room every security device eventually depends on. It is decided early, by people thinking about something else, and it is the hardest thing on the project to move later.

Work through it in this order

Draw the 100 m circles first. Balanced twisted-pair cabling is a 100 m channel — up to 90 m of permanent link plus patch cords. Measure as the cable routes, with slack and vertical runs, not in a straight line.
Count what lands there — cameras, readers, controllers, intercoms, wireless access points, and whatever the next tenant adds.
Size the rack, then add a third. Patch panels, switches, controllers, power supplies, a UPS and cable management take more rack units than the first sketch shows.
Budget the power. Total PoE draw plus head-room, plus the line-voltage circuits to feed it, plus the UPS. Confirm the circuit exists and is dedicated.
Solve the heat. Everything you just budgeted becomes heat inside a small room. Cooling and ventilation are a design item, not an afterthought.
Get the pathways in. Sleeves, conduit, tray and firestop, sized for the cable you will add later — conduit fill limits come from the NEC, and future pulls need room.
Bond and ground it to the telecommunications bonding infrastructure per ANSI/TIA-607.
Secure the room itself. A locked, access-controlled, alarmed door with its own camera. An IDF with a shared key is the weakest point in the system you just designed.
Label everything per ANSI/TIA-606 and hand over the records.

Questions that decide the room

Who owns it? IT, facilities or security — and who holds the key. Answer before construction, not after.
Is it shared? Shared rooms are normal and fine, provided the security equipment is in its own lockable cabinet with its own access record.
What is above and beside it? No plumbing above a rack. Ever.
Where does the backbone come from? Fiber count and path diversity are cheap now and expensive later.
How long must it run without utility power? That question sizes the UPS and decides whether it is on the generator.
Can a person work in it? Clearance in front and behind the rack, a light that works, and somewhere to put a laptop.

The recurring mistakeAn IDF sized exactly for day one, in a closet with no cooling, on a shared circuit, behind a door with a building master key. Every one of those is cheap to fix on paper and expensive to fix in concrete.

Reference set

SCIF & SAPF downloads.

If your project involves an accredited space, these go deeper than anything on this page — all public-level, all source-cited, no sign-in. The full set and its context live on the SCIF & SAPF Security page.

Have a question instead of a download?

SCIF Answers is a searchable knowledge base of source-cited answers on ICD 705, the IC Tech Spec, UL 2050 Extent 3, doors and locks, acoustics, RED/BLACK and accreditation. Ask it the way you would ask a colleague on the jobsite.

Popular starting pointsStart here · IDS & UL 2050 · Doors & locks · Reference library

Not sure whether you need a SCIF at all?

A great many projects described as “we need a SCIF” turn out to need a well-built secure room, an open storage area, or a policy change. The distinction is set by what information you handle and who accredits the space — not by construction.

Read firstSCIF vs. a locked room: what accreditation actually requires

Tools for the person writing the cheque.