Author Topic: Chronos 2.1 Mono clipping at less than full 12 bit at faster framerates  (Read 256 times)

LimaKilo

  • Newbie
  • *
  • Posts: 11
    • View Profile
Hi, we're imaging electric arcs with our Chronos 2.1 monochromatic. These are very bright and frequently overexpose the high-speed frames. However, when filtering the overexposed parts of the images, I noticed that the camera starts clipping the pixels at less than the maximum value the 12 bit depth allows and at differing levels for each pixel. Intentionally overexposing the whole image with a lamp shows the pattern in the attached image (colorcoded from blue to yellow for better visibilty).

This issue seems to begin at around 7 kfps at 640x240. In multiple overexposed frames recorded in the same recording, the pattern is exactly the same for each frame.

Is there a way around this or a way to calibrate for this?
« Last Edit: May 29, 2020, 12:39:39 PM by LimaKilo »

Nikon1

  • Sr. Member
  • ****
  • Posts: 337
    • View Profile
No specialist about the Whole details, how a sensor works, but i think crazy bright Flashes like those that get generated when filming arcs can not only seriously Mess with the Way the sensor works (If you throw enough Light at electronic Sensors, like those that sense the Brightness in the Imaging Sensor of the Camera, at some point they start to do things they are not supposed to do, like Loose data or generate funky images), but also actually damage the Sensor itself Permanently (if strong enough, keep in mind, you have a crazy Bright Flash there, and the Lens will focus that onto a very small area. You basically built yourself a pulse laser. Depending on Energy or Pulse time there can be damage Caused to the Sensor). If it is so bright that it overexposes so badly, maybe try using an ND-Filter (Neutral-Density, like Sunglasses for your Lens. They come in different Densitys, the higher the number is, the darker it usualy makes your image), those filters will help make your image Darker, and maybe this allready solves the Problem? I think this could be worth a Try.
« Last Edit: May 29, 2020, 07:07:34 AM by Nikon1 »

tesla500

  • Krontech
  • Full Member
  • *****
  • Posts: 209
    • View Profile
    • krontech.ca
The simplest way to remove this is to simply gain up the images enough that the darkest pixels clip to white. You could also use the digital gain option on the record settings screen, but this doesn't have particularly good granularity. Doing it in postprocessing would be the most accurate.

LimaKilo

  • Newbie
  • *
  • Posts: 11
    • View Profile
The simplest way to remove this is to simply gain up the images enough that the darkest pixels clip to white. You could also use the digital gain option on the record settings screen, but this doesn't have particularly good granularity. Doing it in postprocessing would be the most accurate.

But this means losing at least a third of dynamic range, if taking into account the first 32 columns, more than half.

Do you know what is causing this "premature clipping" and is there a way to fix it?

Also, do you know if this is just uneven clipping or if the individual pixel gains are also affected? Meaning, if I would homogenously expose the sensor with light below white threshold, would I get a homogeneous image or the same pattern as above?

tesla500

  • Krontech
  • Full Member
  • *****
  • Posts: 209
    • View Profile
    • krontech.ca
When the camera is operated at low horizontal resolutions, the horizontal line readout period is reduced, and there is less time for the sensor's analog-domain correlated double sampling operation to complete. This causes the pixel black levels to be offset upwards, so black comes out as dark grey. This is all invisible to the user, the black calibration corrects for this increased offset automatically, but it does decrease the dynamic range as you mentioned. The offsets are variable across the 32 analog to digital converter (ADC) channels in the image sensor, so different amounts are subtracted from each column.

When the image is saturated to white, the ADCs are all clipped, and so output the same value. When the column offsets are subtracted from this clipped output, it results in this vertical banding you see, since some columns have more subtracted than others. I don't believe there is any significant improvement to be made here unfortunately, it may be possible to have some improvement in a future software update by optimizing the sensor register settings, but it's largely a limitation of the image sensor itself.

To answer your question on homogeneous exposure, if you stay below the point where the vertical banding starts to occur, a homogeneous light pattern illuminating the sensor should result in a homogeneous image output. This is an offset issue only, there should not be any significant variance of gains across columns.
« Last Edit: June 15, 2020, 08:08:45 PM by tesla500 »