I have a couple questions about the OAK-D LR camera. 

What is the mapping between sensitivity (iso) control and image sensor analog gain for the AR0234 sensor?

  1. When one specifies a sensitivity/iso setting from the SDK, does the setting only affect sensor analog gain or does it also affect sensor digital gain and/or isp digital gain? In one tutorial (link) on your website, there is a note that
    "Sensitivity will only add analog gain, which will increase the image noise." I wanted to confirm this is still the case for the OAK-D LR camera with AR0234 sensor - that only sensor analog gain (and not sensor digital gain or isp digital gain) is affected when we change the iso values.

  2. What is the AR0234 sensor analog (and/or digital gain) at iso 100? At iso 100, is the ANALOG_GAIN register set to 1x or to the vendor's minimum recommended analog gain?

  3. As sensitivity is increased (e.g. iso 200,400, 800, 1600), how does the sensor gain setting change? Is it linear - e.g. sensor gain at iso 400 = 4 * sensor gain at iso 100?


    That same tutorial linked above also shares alternate binaries for image quality tuning . Are these .bins (e.g. Color tuning for low-light environments) relevant for the AR0234 as well?


    Also, at which working distance were the OAK-D LR camera modules focused for peak MTF?

  • erik replied to this.

    Hi divya ,

    1/2. It depends - initially it will increase analog gain, and then digital gain if analog gain isn't sufficient (to converge by the Auto-Expsoure)

    1. Yes, 400 == x4
    2. I believe it depends - for limited exposure times (8.3ms bin that we have), it should work for both ov9282 and imx378, but that wouldn't be the case for low-ligh performance.
    3. We would need to check with the camera manufacturer about MTF.

    I hope this helps!
    Thanks, Erik

      4 days later

      Hi Erik,

      Thanks for your reply!

      Can you clarify the mixture of sensor analog and sensor digital gain applied across the ISO setting range for the OAK-D LR? We are capturing the raw images (not ISP out images) - therefore any digital gain applied by the ISP does not matter.

      • Can you confirm that ISO 100 corresponds to 1.0x analog gain vs the vendor specified minimum gain for this sensor of 1.68x?
      • What is the crossover point at which increasing ISO shifts from purely analog gain to both analog and digital gain?
      • After that crossover point, can you provide any more detail on what the relationship between ISO value, sensor analog gain, and sensor digital gain is as ISO increases?

      Best,

      Divya

      Hi @divya ,

      Currently we don't apply the vendor specified minimum analog gain of 1.68x, that is the gain could go down to 1.0x if requested by 3A or manual exposure. Main reason for that is camera tuning not done to take this minimum into account, but we haven't noticed the potential "pixel well saturation" artifacts. Have you noticed any issues with that? Can confirm by checking ImgFrame exposure metadata, or setting manual exposure.

      For combining analog and digital gain for AR0234 when converting from ISO, it's a little more complicated vs other sensors. That's because analog gain setting it's quite coarse for this sensor, so what we do is: apply the maximum analog gain possible to achieve the given ISO (total_gain = ISO / 100), and the rest is applied by sensor digital gain (digital_gain = total_gain / analog_gain), usually this digital gain gets to a small value. Can give more details if needed.

      Alex

        8 days later

        Hi Luxonis-Alex ,

        Many thanks for the explanation!

        So if I understand correctly, since sensor analog gain can be set to 16x, only analog gain is applied up to ISO 1600 and above that, sensor digital gain starts to get added — is that correct?

        We have not yet noticed an issue with pixel well saturation, but we are still early in our testing. We observe that the autoexposure algorithm seems to reach a maximum 8ms exposure time. Is that consistent with your expectation?

        Do you plan to do any tuning updates for the AR0234? And if so, do you expect to adjust the gain bounds so that they respect the 1.68x recommended minimum and/or allow longer exposure times in autoexposure mode?

        Touching back on Erik’s reply, is there any update on the focused distance for the LR array (for peak MTF)? We are looking into the image sharpness, including at shorter range (<1m). It would be helpful to know the lens MTF specification and also what your team has measured in the lab.

        Best,

        Divya