IP camera bandwidth calculator

The bandwidth math behind every IP camera

Per-camera bandwidth is the product of five factors: raw resolution-frame-rate baseline, codec efficiency, GOP structure, scene complexity, and rate-control mode. Manufacturer datasheets quote a single CBR figure that almost never matches the bitrate you'll see on the wire — the spec sheet number is a controlled-lab maximum at medium scene complexity, 2 s I-frame interval, and 30 fps. Real installations run at 0.4× to 1.6× that number depending on configuration.

The resolution-frame-rate baseline scales roughly linearly with both pixel count and frame rate. A 4 MP camera at 30 fps using H.264 typically lands around 8 Mbps. Doubling frame rate to 60 fps doubles the bitrate. Quadrupling pixels to 16 MP quadruples the bitrate. This is the floor — codec and content-aware compression eat into it.

Codec efficiency relative to H.264 is the largest single lever. H.265 (HEVC) reaches roughly 50% of H.264 bandwidth at the same perceptual quality through better intra-prediction, larger coding units, and asymmetric motion partitioning. H.265+ — Hikvision and Dahua's dynamic ROI extension — drops a further 20-40% on static-camera CCTV by suppressing motion-vector encoding in unchanging background regions. AV1 reaches roughly 35% of H.264 baseline but is rare in IP cameras as of 2026; expect it to appear in chipset refreshes from 2027.

GOP structure — the cadence of full I-frames among predicted P/B frames — matters because I-frames are 5-10× larger than P-frames. A 1 s I-frame interval at 25 fps places one I-frame per 25 frames; a 5 s interval places one per 125. Halving the I-frame interval roughly increases average bitrate by 40-60%. The trade-off is seek responsiveness in the VMS playback timeline: shorter GOPs allow frame-accurate seeking but cost network and storage.

Scene complexity is the variable nobody plans for. With variable bitrate (VBR) — the default on every modern IP camera — a static parking lot at 03:00 may run at 0.4× the rated bitrate while a busy transit hub at rush hour runs at 1.6-2× rated. Lighting transitions (dawn, dusk, sunny-to-cloudy) trigger short bitrate spikes as the encoder rebuilds reference frames. Two identical cameras in different environments can produce 3× different storage and bandwidth footprints over a month.

Rate-control mode — VBR vs CBR vs constrained-VBR — is the last lever. CBR is preferable when bandwidth is constrained because peaks are bounded, but it wastes compression efficiency on quiet scenes. VBR is the default for storage-sensitive deployments. Constrained-VBR sets both a target average and a maximum cap, giving the best of both worlds at the cost of a more complex encoder configuration.

How to use this bandwidth calculator

1. Enter camera count and resolution. Pick the number of cameras streaming simultaneously to the NVR or VMS. Resolution is the camera's native sensor megapixel count — this is the dominant bandwidth driver.
2. Set frame rate and codec. Most CCTV runs at 15-25 fps; bump to 30 fps for ANPR and access-control adjacency. Choose the codec your VMS actually decodes — H.265 is now near-universal, H.265+ requires Hikvision/Dahua-aware decoding paths.
3. Pick scene complexity honestly. Don't default to medium for everything. A pole-mounted highway camera at night is static. A retail floor at 14:00 is medium. A train platform at 08:30 is complex. The 0.6× to 1.6× swing changes your switch sizing decision.
4. Read the three result cards. Per-camera Mbps drives PoE switch port selection. Total Mbps drives switch uplink and NVR write throughput. The uplink recommendation card calls out the 1 Gbps saturation threshold so you can plan distributed NVR architecture before an installation goes wrong.

Worked example: 16-camera retail store

A 600 m² fashion retail store needs a 16-camera deployment: 4× 4 MP turrets covering the sales floor, 6× 4 MP bullets for aisles and changing-room corridors, 4× 8 MP fisheyes at the entrances and tills, and 2× 4 MP IR bullets for the back-of-house and loading dock. Recording is 25 fps, H.265, 2 s I-frame interval, no audio.

The 4 MP cameras at H.265 25 fps medium-complexity run roughly 8 × 0.83 (fps scale) × 0.5 (codec) × 1.0 (scene) ≈ 3.3 Mbps each. The 8 MP fisheyes run roughly 16 × 0.83 × 0.5 × 1.0 ≈ 6.6 Mbps each. Aggregate: 12 × 3.3 + 4 × 6.6 ≈ 39.6 + 26.4 ≈ 66 Mbps. The two back-of-house cameras at static night-time complexity drop to 0.6× = 2 Mbps each. Total continuous network load ≈ 70 Mbps.

70 Mbps is comfortable on a 1 Gbps switch uplink — under 10% saturation. But peak instantaneous load when all 16 cameras emit an I-frame in the same 40 ms window can exceed 200 Mbps. A 1 Gbps trunk handles that easily; a 100 Mbps trunk would drop frames. PoE switch sizing also matters: 16× 4 MP turrets at typical 6-9 W PoE budget needs ~120 W aggregate plus dome heaters in winter — a 24-port 1 Gbps PoE+ switch with 250 W budget is the minimum.

Switching the same install to H.265+ Smart Codec drops the aggregate to about 42 Mbps — significant for remote-viewing scenarios where store HQ pulls live streams over a single 100 Mbps fibre uplink. Switching back to H.264 raises it to about 130 Mbps and starts to put pressure on the 1 Gbps switch uplink during multi-stream review windows.

Common bandwidth-planning mistakes

• Sizing for average, not peak. Aggregate average is fine for storage. For switch uplink and NVR write throughput, plan 2-3× the average to absorb I-frame bursts when many cameras synchronise.
• Trusting datasheet bitrate as gospel. Datasheet figures assume CBR at medium complexity, 2 s I-frame, 30 fps. Your actual VBR install can run at 0.4× or 1.6× depending on scene and configuration.
• Ignoring dual-stream architectures. VMS clients pull a low-resolution second stream for live wall display. Each camera emits two simultaneous streams, not one. Add the secondary stream (typically 0.5-1 Mbps) to your aggregate.
• Forgetting remote-viewing upload. Live-view from off-site uses your ISP upload, not your LAN. A 70 Mbps internal load may need only 5-10 Mbps upload if remote viewing is rare and rate-limited, but 70 Mbps if it's continuous to a NOC.
• Mixing H.265+ with non-aware VMS. If your VMS doesn't natively decode H.265+, the camera or NVR re-encodes to H.265 on egress and you lose the bandwidth saving. Verify decoder support before relying on H.265+ in your sizing math.

Continue reading