What Is an IP Camera? Imaging, Network Protocols, Storage, and Camera Types

An IP camera combines an imaging device with a small networked computer. It captures a scene, processes the image, compresses the video, and sends data using Internet Protocol. The system includes a lens, an image sensor, processing and encoding functions, memory, and a network interface. Several of those functions may share one chip rather than exist as separate processors.

The word “IP” describes communication, not image quality, a particular housing, or a requirement to use the public internet. A network camera can operate entirely on a local network. Remote access is an additional design choice that requires connectivity, compatible software, authentication, and appropriate security controls.

How an IP camera produces video

The lens focuses light onto the sensor. Photosensitive elements convert that light into electrical signals; processing then adjusts exposure, color, brightness, noise, and other image characteristics. An encoder compresses the resulting video so the network can carry it efficiently. The network interface packages and transmits the data to a recorder, server, or viewing application.

Each active IP interface needs suitable addressing. A recorder can discover a camera or connect to its configured address, subject to credentials and supported protocols. Addressing makes communication possible; it does not itself synchronize timestamps. Configure time synchronization separately so recorded events from different cameras can be compared reliably.

Software expands a camera's functions, but not without limits. Processor capacity, memory, firmware support, licenses, optics, and sensor performance constrain what a model can do. An internet connection also does not guarantee that a camera is reachable from anywhere: firewalls, provider restrictions, service availability, and account permissions still apply.

What cameras connect to and how they receive power

A camera may send video to a network video recorder, a VMS server, a computer running compatible software, or a supported cloud service. A cloud subscription is often delivered as software as a service, but an arbitrary IP camera cannot necessarily connect to any cloud platform. Some systems require a gateway, an approved camera model, or a specific firmware version.

Multiple cameras normally connect through an Ethernet switch. A recorder with dedicated camera ports may contain its own switch, sometimes with a separate camera subnet. A router connects networks. DHCP assigns addresses when a DHCP server is present; an ordinary Ethernet switch does not automatically provide that service. Use unique static addresses or DHCP reservations according to the system design.

One camera can connect directly to a computer's Ethernet interface for configuration or testing, provided their addresses and subnet settings allow communication. Several cameras usually connect through a switch shared with the computer. A direct cable does not supply power unless the equipment explicitly supports an appropriate power-delivery arrangement.

Power may come from a specified adapter, a PoE switch, a compatible injector, or a PoE recorder port. Power over Ethernet carries data and power over the same twisted-pair cable. Confirm the camera's PoE type and maximum demand, the switch's per-port rating, and its total budget. A network connector alone does not imply PoE support, and passive power systems should not be treated as interchangeable with standards-based PoE.

The network protocols behind the picture

A useful TCP/IP model has four layers. At the application layer, protocols such as HTTP or HTTPS, RTSP, DNS, and file-transfer protocols perform specific services. At the transport layer, TCP and UDP carry application traffic. IP provides addressing and routing at the internet layer; ICMP supports control and diagnostics, while IGMP supports IPv4 multicast group management. Ethernet or Wi-Fi provides local network access.

ARP maps IPv4 addresses to local link-layer addresses. Other transport protocols, including SCTP and DCCP, exist but are not typical camera setup choices. Routing protocols should not be grouped solely by the service they perform: for example, RIP uses UDP, while OSPF runs directly over IP. Older technologies such as Token Ring, SLIP, ATM, and T1/E1 are historical context rather than a shopping list for a new camera LAN.

TCP, UDP, and RTP

TCP establishes a connection and provides an ordered byte stream with loss recovery, duplicate handling, flow control, and congestion control. Retransmission can recover missing data, but TCP cannot keep a failed network alive or guarantee a maximum delivery delay. The source's story about a packet traveling 150,000 kilometers is not a useful basis for specifying surveillance reliability.

UDP sends datagrams without TCP's connection and retransmission mechanisms. A receiver may need to tolerate or conceal missing packets. Lower recovery overhead can suit live video, but UDP is not automatically faster in every network, and packets are not guaranteed to arrive in order.

RTP supplies media timestamps, sequence numbers, and other information useful for real-time playback. It commonly runs over UDP, so “RTP versus UDP” is not a simple choice between equivalent layers. RTP alone does not retransmit every missing packet. Select the transport supported by the camera and client, then test latency, loss, and recovery under actual conditions.

RTSP controls a streaming session

RTSP controls media sessions; it is not a compression codec. Its setup exchanges can negotiate how the media will be carried. The media may use separate RTP connections or, in supported implementations, be interleaved with control traffic over TCP. Therefore, control and media do not always require completely separate connections.

The following methods preserve the terminology of the original RTSP 1.0 discussion. Support varies by device and protocol version. The RTSP 1.0 specification is a historical reference, superseded by RTSP 2.0; many installed devices still implement 1.0 behavior.

ONVIF and interoperability

Products from one manufacturer are not automatically compatible in every combination. Check the camera model, recorder or client version, firmware, codec, stream limits, and required features. A basic RTSP picture does not prove that audio, events, analytics, PTZ control, or archive retrieval will work.

ONVIF profiles group interoperability requirements for defined functions. The original four-profile list is outdated. For video projects, Profile T covers advanced streaming, Profile G addresses recording and retrieval, and Profile M addresses analytics metadata and events. Profile S remains relevant to installed basic-streaming systems. Profiles C, A, and D address different access-control functions.

ONVIF states that Profile Q was deprecated on April 1, 2022. Its published Profile S deprecation guidance gives March 31, 2027 as the end of support for new conformance submissions. Existing Profile S devices do not simply stop operating on that date. Review the official profiles overview and the product's conformance record when selecting equipment.

Discovery, network configuration, stream settings, events, PTZ, and security features involve mandatory and conditional requirements. A camera with a memory-card slot does not automatically conform to Profile G, nor does an ONVIF label guarantee every advanced feature. Confirm conformance and test the specific functions needed.

Wired, wireless, and combined connections

Wired Ethernet provides a predictable local connection when cable quality, distance, switching, and power are correctly designed. Wi-Fi avoids a data cable but still needs power and sufficient radio coverage. Walls, interference, shared airtime, and access-point placement affect performance. Cellular cameras require supported carrier service, suitable antennas, and a data plan; verify current network compatibility rather than buying around an obsolete generation.

Some cameras support both wired and wireless interfaces. That flexibility does not automatically provide seamless failover or redundant recording. Confirm how the device selects interfaces and what happens when the active link fails. A local recording option can help during communication loss, provided the camera remains powered.

Ethernet over twisted pair

Ethernet is standardized in the IEEE 802.3 family. IP cameras can share infrastructure with office computers, but that does not eliminate the need to check capacity, segmentation, security, and switch uplinks. Several streams can share a cable only while the aggregate demand stays within the useful capacity of that link.

The usual 100 m channel includes patch cords, not just the permanent cable. PoE voltage drop, conductor material, temperature, and connector condition also matter. A camera may need only a 100 Mbps port while a switch aggregating many cameras needs a much faster uplink.

Coaxial and fiber links

The original text reversed two historic coax Ethernet names. 10BASE2 used thin coax with a maximum segment of 185 m; 10BASE5 used thick coax with a 500 m segment. These obsolete bus networks are different from modern Ethernet-over-coax adapters. For an existing CCTV cable, use a compatible converter pair and its own distance, bitrate, and power specifications.

Fiber reach depends on the optical standard, transceiver, fiber type, wavelength, connector losses, and optical power budget. “Fiber works for 100 km” is not a general camera-network specification. Multimode and single-mode fiber have different propagation characteristics and require appropriate optics.

Legacy 10 Mbps fiber technologies included FOIRL and 10BASE-FL; 100 Mbps installations commonly used 100BASE-FX. Their distance limits vary with the defined implementation and duplex mode. The source's mixed 100BASE-SX and WDM figures should not be used for procurement. At Gigabit rates, 1000BASE-SX normally uses multimode fiber, while 1000BASE-LX supports longer single-mode links and certain multimode arrangements. Consult the exact module's supported fiber table.

For 10 Gigabit links, SR, LR, and ER serve different reach requirements. Cisco's optics datasheet, for example, specifies 10 km for its 10GBASE-LR modules on suitable single-mode fiber. Higher-rate uplinks can also serve large surveillance networks; the old claim that surveillance never uses anything above 10 Gigabit is not a design limit.

Full duplex permits simultaneous sending and receiving; half duplex alternates access. Modern switched camera Ethernet normally operates full duplex. Fiber is commonly introduced at switches through SFP or SFP+ modules, although some cameras have an optical interface. Match modules, ports, and cabling rather than assuming every pluggable module fits every link.

Image processing, analytics, and compression

Network cameras can offer wide dynamic range, backlight compensation, highlight handling, electronic stabilization, and digital noise reduction. These functions solve different problems and can introduce tradeoffs. A high frame rate can show motion more smoothly, but usable detail still depends on exposure, focus, lighting, and compression.

Digital transmission avoids the gradual picture degradation associated with some analog paths when packets arrive correctly. It is not immune to network faults: loss or congestion can cause missing frames, artifacts, or delays. Encoding is also not encryption. Protect credentials, management access, and video transport using the mechanisms supported by the actual system.

Analytics range from pixel-change motion detection to line crossing, object classification, license-plate recognition, and other specialized functions. Event-triggered recording can reduce stored footage, but the camera may still continuously analyze or transmit video. Savings depend on configuration and real activity. Use pre-event recording and test missed events and nuisance alarms before relying on event-only archives.

IP cameras generally encode video onboard. Analog cameras usually send a signal to a DVR that digitizes and encodes it, often using dedicated hardware. It is inaccurate to assume every analog system therefore requires a heavily loaded general-purpose server.

JPEG stores still images; Motion JPEG represents video as successive JPEG images. H.264 and H.265 are standardized coding systems with licensing considerations, not merely proprietary camera-brand formats. H.265 can reduce bitrate at a comparable quality, including at 2 MP, but savings vary by scene and implementation. Check decoder support and inspect moving subjects rather than assuming one codec always wins.

Storage, multiple streams, corridor views, and sound

Edge storage records inside or near the camera, commonly on a supported memory card. It can provide an independent archive or temporary coverage during network loss. Card capacity, endurance, health reporting, and recorder retrieval support matter. Not every camera has a card slot or supports USB recording, and a stolen or failed camera can take its local archive with it.

Multiple streams let the same camera provide a detailed recording stream and a smaller stream for live grids or remote viewing. Some models offer additional streams, but their number and combinations are model-dependent. Changing one stream can affect another because encoding resources are shared. Analytics do not universally require a separate transmitted stream for every detector.

Corridor mode uses a tall image orientation, often 9:16 instead of 16:9, to fit hallways or tunnels. Some cameras rotate the image internally; compatible clients can also rotate or crop a picture. Software rotation cannot add missing sensor coverage or guarantee the same efficiency as a native mode. Confirm orientation support, resolution, and installation instructions.

Audio hardware varies: a camera may have no audio, a built-in microphone, an external input, an output, or two-way communication. Match microphone signal level and power requirements to the input; “audio input” does not mean every professional microphone can connect directly. An output may need an amplifier and speaker.

Sound analytics can detect level changes or classify selected sounds, but available functions and accuracy differ. A sound alert is not proof of the underlying event and is not a substitute for required life-safety equipment. Before enabling audio recording, determine the consent and privacy requirements that apply to the location and use.

Sensor size, progressive scanning, and resolution

CMOS sensors convert light to electrical signals and can integrate readout and processing functions efficiently. Their speed, power consumption, cost, and image quality vary by design. CMOS does not automatically mean lower network bitrate, superior color, or better low-light performance than every other sensor. Encoding and scene settings determine the transmitted stream.

Progressive scanning represents complete frames instead of alternating interlaced fields, avoiding interlace combing on motion. Its bandwidth is not universally “twice as much”: compare actual resolution, frame rate, field rate, and compression. Enlarging a progressive image can improve presentation, but it does not create captured detail that was absent.

Optical-format labels such as 1/2, 1/3, and 1/4 inch describe sensor classes rather than literal diagonal measurements. A larger active area can collect more light for comparable designs and changes field of view for a given lens. Pixel size, sensor technology, aperture, exposure, and processing also matter. No one format guarantees adequate sensitivity at every site. Multi-sensor cameras combine several imaging channels for panoramic coverage or separate views.

More pixels can improve detail when the optics and lighting support them. They do not guarantee identification. A face occupying only a few pixels remains unsuitable for identification even in a large frame. Design around target distance, scene width, motion, and the evidence required. High resolution can help broad coverage and later cropping, not only specialized tasks such as reading plates or banknotes.

Lenses and camera housings

A lens projects the scene onto the sensor. Focal length is measured in millimeters; for the same sensor, shorter focal lengths generally provide wider views and longer focal lengths narrower views. It is not measured from the lens axis to an arbitrary focus point. A 2.8 mm lens does not always provide exactly 100 degrees because sensor size and optical design also matter.

• Fixed focal length: the angle of view is selected with the camera or lens model. Focus may still be adjustable.
• Varifocal: a range such as 2.7–13.5 mm allows framing during installation. Refocusing may be necessary.
• Motorized: remote lens adjustment simplifies commissioning. Continuous optical zoom, autofocus speed, iris control, and stabilization are separate specifications; there is no universal five-second limit.
• Fisheye: a very wide view can be dewarped into useful layouts. One camera may cover several directions, but it cannot remove occlusions or guarantee the same detail as several purpose-positioned cameras.

Los Axis lens guide provides a useful reference for focal length, field of view, iris control, and sensor matching.

Dome cameras enclose the lens behind a dome and support approved ceiling or wall mounts. Bullet cameras use an adjustable bracket and often a sunshield. Box cameras may accept interchangeable lenses and protective housings. Cube cameras commonly suit indoor shelves or desks. Turret or eyeball designs allow manual aiming without a transparent outer dome. These shapes do not by themselves establish weather or impact resistance.

PTZ cameras add powered pan, tilt, and zoom. Presets and automatic tracking can follow selected activity, but the camera cannot simultaneously record every direction at full detail. Specialized modular cameras with separate sensor units serve installations where a conventional housing is unsuitable.

Indoor, outdoor, and transportation requirements

For outdoor use, check the exact ingress-protection rating, operating and startup temperatures, mounting orientation, cable glands, and connector sealing. IP ratings describe tested resistance to solids and water, not image quality or immunity to corrosion. A higher second digit does not always imply every lower-numbered water test. Use the manufacturer's stated exposure limits rather than treating IP54 through IP68 as one interchangeable outdoor category.

Temperature suitability depends on the actual site. Ratings such as −10, −20, −40, or −60°C cannot be assigned automatically to entire countries or regions. Consider solar heating, humidity, icing, wind, salt exposure, and power needed by heaters. Impact ratings such as IK08 or IK10 are separate from weather ratings.

Vehicle cameras may require vibration resistance, secure connectors such as M12, and application-specific electrical and environmental approvals. Certification requirements depend on the vehicle and project; there is no single universal transport-camera certificate. Compare documented specifications and test the finished installation. Start with Safsale's security camera catalog, then select a complete recording and network arrangement that preserves the required image under real operating conditions.