Measure live-video delay before operators must act

Source facts: live-video latency is an end-to-end property

Axis Communications’ Latency in live network video surveillance white paper defines sensor-to-screen latency as the time from capture of a frame until that frame is visible on the video display. It divides the path into camera processing and encoding, network transport, and receiver-side buffering, decoding, and display. Each stage adds delay.

The source says the network portion is often the most unpredictable, while a client receive buffer can add up to seconds. Resolution, frame rate, image processing, compression, bitrate, number of streams, available throughput, congestion, protocol choice, client hardware, graphics processing, and display behavior can all affect the result. Network latency alone therefore does not describe what an operator experiences.

Axis also describes a practical measurement using a camera timestamp overlay and a view of the displayed stream, with accuracy bounded by the frame interval. That is one manufacturer-specific technique; the correct method depends on the installed products. The important boundary is the same: measure from a real scene change to its presentation on the actual operator display.

DSE recommendation: make delay part of acceptance

Start with the operational consequence. A receptionist checking a visitor, a remote guard assessing an alarm, and an investigator reviewing yesterday’s footage do not have the same latency need. Document a target for each time-sensitive workflow and identify who approves it. Do not invent one universal number or confuse a smooth picture with a timely picture.

1. Map the full path. Record camera and firmware, stream profile, encoder, switch path, VLAN, router or firewall, WAN or VPN, VMS services, transcoding or proxy stages, client version, workstation, graphics adapter, monitor, and any cloud relay. Note whether live and recorded video follow different paths.
2. Measure what the operator sees. Use a manufacturer-supported timestamp or another controlled visual event and compare the scene time with the displayed time. Measure several samples, report typical and worst observed delay, and retain the method so results can be repeated.
3. Test every operating location. Repeat at the local control room, remote office, guard station, mobile or browser client where approved, and during a PTZ or intercom workflow if those functions are in scope. A good result on the server-room console does not validate a remote user.
4. Exercise load and degradation. Test during normal busy periods, simultaneous multi-camera viewing, incident playback, analytics events, export activity, WAN congestion, and approved network failover. Capture packet loss, throughput, bitrate, workstation utilization, and buffering behavior alongside latency.
5. Verify control alignment. For PTZ, doors, speakers, or intercoms, determine whether video, audio, and control commands arrive with mismatched delay. An operator should not act on a picture that materially trails the command or conversation.
6. Protect forensic quality. Do not lower bitrate, resolution, or compression quality merely to improve the live display unless the change is evaluated against the recording requirement. Where supported, use separate, governed live and recording profiles instead of degrading the evidence stream.

When delay exceeds the target, isolate stages instead of changing settings at random. Compare direct camera viewing with the VMS client; local with remote; one stream with a video wall; and one workstation with another. Review encoder load, unexpected transcoding, saturated links, quality-of-service behavior, client buffers, decoder and graphics capacity, monitor processing, and unsupported stream combinations. Make one controlled change at a time and repeat the measurement.

Store a latency baseline with the system acceptance record and monitor material changes after camera or VMS upgrades, firewall inspection changes, WAN migrations, new proxy or cloud paths, graphics-driver updates, or expansion of a viewing wall. An alert that appears eventually may still be an operational failure. Live video earns a response role only when its delay is measured through the same path and conditions the operator will actually use.

Official reference

• Axis Communications, Latency in live network video surveillance, June 2024.