How MIPI CSI-2 Lanes, Carrier Boards, and Sensors Affect Frame Rate

The Imaging Source offers and supports MIPI CSI-2 cameras on different embedded platforms, e.g.:

- 36S Series MIPI CSI-2 cameras

- NVIDIA Jetson Orin

- NXP i.MX 8M Plus

- Raspberry Pi 5

What determines the frame rate of a MIPI CSI-2 camera system?

The achievable frame rate depends on three main factors:

• Image sensor: Each sensor has a native maximum output (resolution × fps).

• Number of MIPI CSI-2 lanes: More lanes = more data bandwidth.

• Lane speed (per lane bandwidth): Depends on the carrier board/platform (e.g., 2.5 Gbit/s vs 1.5 Gbit/s per lane).

In short:
Frame rate is limited by the lowest bottleneck between sensor output and available MIPI bandwidth.

Why does the same sensor run at different frame rates on different platforms?

Different carrier boards support different MIPI configurations:

Platform	Typical MIPI Configuration
Jetson Orin (2-ch)	4 lanes @ 2.5 Gbit/s
Jetson Orin (4-ch)	2 lanes @ 2.5 Gbit/s
Raspberry Pi 5	4 lanes @ 1.5 Gbit/s
NXP i.MX 8M Plus	2 lanes @ 1.5 Gbit/s

This directly affects total bandwidth:

• 4 lanes @ 2.5 Gbit/s → highest throughput

• 2 lanes @ 1.5 Gbit/s → lowest throughput

If the available bandwidth is insufficient, the frame rate must be reduced.

How do MIPI lanes affect performance?

Each MIPI lane carries part of the image data stream.
Reducing the number of lanes reduces total bandwidth proportionally.

Example:

• A sensor running at 120 fps with 4 lanes may drop to 60 fps with 2 lanes

• This behavior is visible in sensors like:

  • DxM 36SX290-ML

  • DxM 36SX462-ML

  • DxM 36SR0234-ML

How does lane speed influence frame rate?

Lane speed varies by platform:

• 2.5 Gbit/s per lane (e.g., NVIDIA Jetson Orin)

• 1.5 Gbit/s per lane (e.g., Raspberry Pi 5, NXP i.MX 8M Plus)

Lower lane speed reduces total throughput, which can limit frame rate even if lane count is high.

Why do some sensors maintain the same frame rate across platforms?

Some sensors do not saturate the available bandwidth.
In these cases, the platform is not the limiting factor.

Examples:

• DxM 36SX297-ML → 120 fps on all platforms

• DxM 36SX296-ML → 60 fps on all platforms

These sensors operate below the bandwidth limits even on slower interfaces.

Why do some sensors scale exactly with bandwidth?

Certain sensors are bandwidth-limited, meaning:

• Halving bandwidth → roughly halves frame rate

Examples:

• DxM 36SX568-ML:

  • 80 fps (4 lanes) → 40 fps (2 lanes)

• DxM 36SR0521-ML:

  • 60 fps → 30 fps

This is typical for high-data-rate sensors.

Do these frame rates change with resolution or binning?

Yes.

The listed frame rates always refer to:
Full sensor resolution

If you use:

• Cropping

• Binning

• Subsampling

Higher frame rates may be possible.

However, these configurations are project-specific and should be evaluated individually.

How can I estimate frame rate for my setup?

A simplified rule:

Frame rate scales with available MIPI bandwidth.

Approximate bandwidth comparison:

• 4 lanes @ 2.5 Gbit/s = 100% reference

• 2 lanes @ 2.5 Gbit/s = ~50%

• 4 lanes @ 1.5 Gbit/s = ~60%

• 2 lanes @ 1.5 Gbit/s = ~30%

If your sensor is bandwidth-limited, expect frame rate to scale similarly.

Still unsure which configuration fits your application?

Frame rate optimization can depend on:

• Exact use case

• Resolution settings

• Multi-camera setup

• Driver tuning

Contact us with your requirements - we'll help you find the optimal configuration.