STM32N6ISP: Need help with Eliminating Auto Exposure convergence delay using external OPT3001 (Feed-forward LUT)

[vedantgg]

Hello everyone,

I am currently developing with the STM32N6570-DK and utilizing the hardware ISP. I am facing an issue with the embedded Auto Exposure (AE) algorithm's convergence time. It currently takes about 20-30 frames (~1 second) for the AE to settle on the correct exposure and gain. For my application, it is critical that the very first frame upon wake-up is perfectly exposed, meaning a 1-second delay is not acceptable.

To bypass this feedback loop, I have integrated a Texas Instruments OPT3001 Ambient Light Sensor (I2C) to implement a feed-forward solution. My intended sequence is:

1. Read the absolute ambient lux from the OPT3001 on boot.

2. Map this lux value to a pre-calibrated Lookup Table (LUT) to determine the ideal Exposure Time and Sensor Gain.

3. Force the STM32 ISP into manual exposure mode and inject these values directly.

4. Start the camera stream to achieve a zero-delay, perfectly exposed first frame.

I have a few questions regarding the best way to implement this within the ST ecosystem:

• Calibration via IQTune: I am currently using the STM32 ISP IQTune tool to manually find the ideal exposure/gain at various lux levels (by turning the AE Algorithm OFF). Is this manual sweep the recommended way to build this LUT, or does ST provide a more automated procedure/tool for mapping external lux to sensor parameters?

• API Injection Sequence: What is the recommended API sequence (via the STM32 BSP or V4L2 framework) to inject manual Exposure Time and Gain before the image stream starts? I want to ensure the AE algorithm doesn't override these values on the first frame.

• Seeding the AE: Once I inject the manual values for the first frame, is there a seamless way to hand control back over to the closed-loop AE algorithm (using my injected values as the new starting seed), or is it safer to leave the ISP in manual mode and continually update it via the OPT3001?