Blogs

Detailed Calculation of MIPI SerDes Bandwidth

NoseDisplay-Blog-MIPI-Serdes-F

2025-11-21

***Take the 4K Display and 40-megapixel Camera as examples.

In all high-speed interface designs, bandwidth calculation is not a formality but directly determines:

  • Whether the interface is affordable
  • How many lanes to use
  • Whether to reserve upgrade headroom
  • Whether it will be rendered obsolete by future “resolution upgrades”


Below we calculate using two most typical engineering scenarios:
👉 4K Display (DSI / SerDes)
👉 40MP Camera (CSI / SerDes)

NoseDisplay-Blog-MIPI-Serdes-F2

I. Standardize a “project-ready” calculation method.

Whether it’s DSI, CSI, GMSL, FPD-Link, or MIPI M-PHY,
The underlying logic remains consistent:

Effective data volume × Protocol overhead ÷ Number of lanes ÷ Encoding efficiency

Universal Formula (Engineering Edition)

Single Lane Rate (bps) =(Resolution × Frame Rate × Bit Depth per Pixel × Protocol Coefficient)/ (Number of Lanes × Encoding Efficiency)

Among these:
Protocol Coefficient: Empty lines, Blank, Header, CRC, etc.
Typically set to 1.2–1.3 in engineering practice
Encoding Efficiency:
*8b/10b → 0.812
*8b/130b → 0.985
*RAW → 1.0

📌 Note:
The minimum standard value ≠ the engineering safety value.
Design must always include a margin of safety.

II. 4K Display Bandwidth Calculation

1️⃣ Basic Parameter Assumptions (Very Typical)

Let’s select a common specification for an in-car infotainment system:

  • Project Resolution: 3840 × 2160
  • Frame Rate: 60 fps
  • Pixel Format: RGB888
  • Bit Depth per Pixel: 24 bits

2️⃣ Calculation of Raw Video Data Volume

  • Pixels per frame = 3840 × 2160 = 8,294,400 pixels
  • Pixels per second = 8,294,400 × 60 ≈ 497,664,000 pixels/s
  • Raw data volume = 497,664,000 × 24 ≈ 11.94 Gbps

3️⃣ Plus display interface overhead / synchronization overhead

Display interface (DSI/SerDes) must include:
>> HSYNC
>> VSYNC
>> Front/Back Porch

Commonly used in engineering:

  • Agreement coefficient ≈ 1.25
  • Actual required bandwidth ≈ 11.94 × 1.25 ≈ 14.9 Gbps

4️⃣ Assuming the use of 4-lane SerDes / DSI

  • Load per lane ≈ 14.9 / 4 ≈ 3.7 Gbps

If the interface uses 8b/10b encoding:

  • Actual lane rate ≈ 3.7 / 0.8 ≈ 4.6 Gbps

📌 Engineering Conclusions:

4K@60Hz RGB888, 4-lane system, Each lane must support at least 5 Gbps for safety.

This is why:

  • D-PHY is struggling
  • M-PHY / GMSL2 / FPD-Link III are just starting to feel comfortable

III. 40-Megapixel Camera Bandwidth Calculation

Now let’s calculate a real bandwidth hog.

1️⃣ Camera Parameter Settings

Project

  • Resolution: 8000 × 5000 pixels
  • Number of Pixels: 40 MP
  • Frame Rate: 30 fps
  • Data Format: RAW10

2️⃣ Calculation of Raw Data Volume

  • Pixels per frame = 8000 × 5000 = 40,000,000 pixels
  • Pixels per second = 40,000,000 × 30 = 1.2 × 10⁹ pixels/s
  • Raw data volume = 1.2 × 10⁹ × 10 = 12 Gbps

👉 Note: RAW10 is 10 bits per pixel.

3️⃣ Plus CSI/SerDes protocol overhead

CSI has:

  • Packet Header
  • ECC
  • CRC
  • Line blank

Engineering Experience:

  • Agreement coefficient ≈ 1.2
  • Actual data volume ≈ 12 × 1.2 = 14.4 Gbps

4️⃣ Assuming the use of a 2-lane high-speed SerDes

  • Load per lane ≈ 14.4 / 2 = 7.2 Gbps

If 8b/10b encoding is used:

  • Actual lane rate ≈ 7.2 / 0.8 = 9 Gbps

📌 The engineering conclusion is brutally harsh:

  • 40MP@30fps, 2-lane system, A single lane requires nearly 10 Gbps

Here’s why:

  • Traditional CSI-2 simply can’t handle it
  • Must implement: GMSL2 / GMSL3
  • FPD-Link IV
  • or M-PHY High Gear

IV Engineer’s Safety Factor Principle

  • Design Bandwidth ≥ Theoretical Demand × 1.3

4K Display:
Calculated at 4.6 Gbps, select 6 Gbps PHY
40MP Camera:
Calculated at 9 Gbps, directly select 10~12 Gbps range

V. Translate the calculation results into the language of solution selection.

4K Display

Solution Evaluation

  • D-PHY 4-lane: Marginal performance
  • D-PHY 8-lane: High cost
  • M-PHY: Cost-effective
  • GMSL2 / FPD-Link III: Engineering-friendly

40MP Camera

Conclusion of the Proposal

  • CSI-2 D-PHY: Not feasible
  • M-PHY High Gear: Challenging
  • GMSL2: Marginal
  • GMSL3 / FPD-Link IV: The correct solution

VI. Conclusion

Using the Bandwidth Result in a Real Design

A first-pass lane-rate estimate begins with pixel clock × bits per pixel ÷ lane count. The final requirement must also account for panel timing, blanking behavior, protocol overhead, display mode, host limitations and margin. A calculation is therefore a screening tool, not a substitute for reviewing the selected panel data sheet and processor documentation.

Input to confirmWhy it matters
Resolution and refresh targetDefines the active pixel throughput
Pixel formatChanges bits per pixel and lane requirement
Lane count and rate supportMust be supported by both host and panel
Timing and marginPrevents a nominal calculation from becoming an unstable link

Read the MIPI DSI interface guide, then compare SPI, RGB, LVDS and MIPI before selecting a TFT module.