(This article is a supporting part of our ongoing testing of low-light camera metering reliability)
What is the dimmest light level that a given camera can meter? For a nightscape photographer, and particularly for someone who films timelapse video with day/night transitions, this question is critical. Normally when I film a timelapse video that starts during the day and ends at night, I start with the camera in aperture priority mode, and the camera compensates as the sun sets.
However, as twilight fades into night, at some point the camera’s metering fails, and I have to switch the camera to manual mode and adjust the settings until night has fully fallen. Innumerable exposure errors have occurred because I didn’t switch to manual at the right moment, or irregularly changed settings in manual mode.
In recent years, some cameras have become able to meter at night, so switching to manual mode is unnecessary. Manufacturer metering specifications are completely wrong and useless. Therefore, while stuck in quarantine in 2020, Sean Goebel and I set out to build a standardized comparison of the low light metering abilities of different cameras.
To answer this question, we built an integrating sphere. An integrating sphere is a sphere that is painted white on the inside and has LEDs for internal illumination. The camera lens pokes into the sphere. Through the use of baffles, light from the LEDs has to bounce at minimum twice before entering the camera lens, and this causes the inside of the sphere to have almost perfectly even illumination.
In other words, it doesn’t matter where the camera is pointed inside the sphere–it’s all the same brightness. The sphere is carefully sealed so that light cannot leak inside from the outside.
Gray image showing what the camera sees
Here are the basics of the metering test: we use an Arduino (an open-source microcontroller) to control the LEDs inside the sphere and trigger the camera being tested. We calibrated the maximum light level of the sphere to be exactly 0 Exposure Value (EV0).
For reference, EV0 is about the light level of a very dim restaurant or a landscape 20 minutes after sunset, and “correct” exposure settings for this are 1/8 s, ISO 6400, f2.8. The camera is set to aperture priority mode with an f/2.8 lens and takes a photo of the inside of the sphere.
The light level is dropped by half (to EV-1), and another photo is taken. Ideally, the camera detects the dimming of the light and compensates with a longer exposure time and/or higher ISO. The light level is dropped by one stop again (to EV-2), and this repeats until the ambient light level reaches EV-10. The correct settings for EV-10 are 30 sec, ISO 25,600, f/2.8, which are appropriate for a light level achieved under a cloudy moonless night without light pollution. Most manufacturers report that their bodies can meter to light levels of EV0 or EV1, but as our tests show, most bodies can meter well darker than this.
After collecting the 11 photos in Aperture Priority mode, each with half the light level of the previous photo, we use Python’s RawPy software library to calculate the average brightness of each raw file. We plot this against the EV that the photo was taken at. If a camera was perfect, the average brightness of the raw file would be constant, since the camera would double its exposure time or ISO with each new photo to compensate for the decrease in light. In practice, plots look like this:
For the first few dimmings, the camera effectively compensates, and the average brightness of the resulting image is constant. However, beyond some point (about EV-6 here), the camera only partially compensates for the decrease in light level–when the light drops by 1 stop, the camera only adjusts by 2/3 stop. Further along, the camera can no longer detect any change in brightness at all, and it does not adjust its settings at all.
Because cameras don’t abruptly change from perfect metering to completely nonfunctional metering, we define two criteria for metering failure.
First, we report at what light level the camera adjusts its settings only ⅔ of a stop for every 1 stop decrease in light. In other words, the first time it fails to compensate by at least ⅓ of a stop.
Second, we report at what light level the camera produces images that are half the brightness of (one whole stop dimmer than) its EV0 image.
Once you know this information about your specific camera, you can translate it to a specific ambient lighting condition (using this article here) that you can safely capture in Aperture Priority with your camera.
Say, for example, if your camera can meter correctly at EV -5, but fails at EV -6, that means your camera (with an f/2.8 lens) can correctly meter and expose a moonlit scene of ~50% illumination, but at ~25% illumination your camera will fail to meter correctly, and you should probably switch to manual exposure on such dark nights.
Integrating Sphere & Camera Metering Test Project