Every new camera is tested by various trade magazines, blogs, comparison portals, online stores and many amateur and professional photographers. Mostly, however, there are no meaningful test results and test shots on the subject of long time exposure and light painting.
This article does not claim to be complete. The test series with the various cameras were not carried out under laboratory conditions. Measurement errors can therefore not be ruled out. The results are only intended to give a rough overview and answer some questions.
Is a new camera per se better suited for long exposure times than a three-year-old one? Does a higher resolution basically produce more image noise? Are cameras from Nikon and Sony basically better than models from Canon or Olympus?
WHICH TECHNICAL FEATURES OF THE CAMERA HAVE AN INFLUENCE ON THE IMAGE QUALITY?
I am truly not a pixel counter. I don’t look at my pictures in 300% magnification and count the “missing pixels”. A good picture is a good picture, no matter if it was taken with an analog camera, the 10 year old digital camera with 12 megapixels or the brand new, hip, “mirrorless” 50 megapixel bolide. The influence of the camera on the overall result is quite small in the vast majority of cases. The work in front of the camera, the idea, the quality of the lens and possibly the post processing have a much bigger effect on the finished image than the number of pixels and the technical quality of the sensor. As long as the image does not create a mood, does not convey a feeling, the special experiences and thoughts of the light painter remain invisible during the work and the viewer is not taken into the action, the technology does not matter. The three hundred and seventy-fourth unmotivated “test shot” of the great new Light Blade no one will seriously and profoundly look at anyway or hang as a print over the couch.
Most modern digital cameras are basically suitable for exposure times of several seconds to several minutes. With hardly any current SLR or so-called system camera, the image noise will be so great that the 5 minute light painting would be spoiled by it.
There are an infinite number of measurements on the topics of resolution, sharpness, etc. for every camera. So that will not be a topic of this article. We will only try to find differences in the noise behavior at long exposure times, at high ISO values in combination with long exposure times, as well as in the dynamic range of the sensor, and above all to find out whether the differences are so great that one should seriously think about changing the camera. In addition, the article is perhaps a good help for the light painting beginner when he or she is faced with the question of the first “right” camera.
The sensor records image information where there is none. This can be seen well in the example image. No light reached the sensor during the exposure, but the image still consists of funny colored pixels. This noise is also called dark noise. It is caused by the quiescent current of the individual pixels and the noise of the readout amplifier.
Image noise is a thermal problem. The higher the temperature of the sensor, the stronger the image noise. If I had taken the picture at -20°C, the image would probably be completely black, but I did not test this.
The ambient temperature, the set ISO value and the exposure time have an influence on the surface temperature of the sensor.
In most cases I can’t change the ambient temperature. There are some solutions for active cooling of the sensor, but that would go beyond the scope here. Regardless, these solutions are very expensive and complex to implement.
I can always influence the exposure time of a light painting image. However, one should never get the idea of rushing through light painting in order to shorten the exposure time. This will usually not work. It is always better to work calmly and concentrated and not to think about the exposure time. But what you should do is to optimize your work. I put all the lamps I need in my bags or place them in the right positions so I don’t have to walk around unnecessarily during light painting. I make sure that all batteries are charged and the lighter to ignite the fireworks works, I instruct all participants beforehand exactly what to do, etc..
The Light Painter can exert the greatest influence via the ISO value. The higher the set ISO value, the more the signal from the sensor is amplified. And with that, the image noise is also amplified. The most unfavorable combination is a high ISO value, a high ambient temperature and a long exposure time. The example image above was taken at 20°C ambient temperature. The ISO was 25600, the exposure time 30 seconds. If I had actually taken a picture with these settings it would be unusable. The image noise would negatively affect very large portions of the real, intended image information.
So the Light Painter will always choose the ISO value as low as possible and rather open the aperture wider and/or use a brighter lamp. However, it doesn’t hurt if you know the limits of your own camera. To do this, you can simply take some test shots with the lens cap on. The pain threshold of my Nikon D750 is ISO 3200. Shots with exposure times up to 1 minute at 20°C are definitely still possible with this ISO value.
The manufacturing quality of the sensor also has a big influence on the strength of the image noise. Whether a newer sensor is better than an old one, and whether sensors from camera manufacturer A are basically better than those from manufacturer B, is what we are trying to find out with the following test shots.
One often hears and reads that a higher pixel density, i.e. the number of pixels on a certain area of the sensor, produces more image noise. In my opinion, this effect can be neglected because it is only visible with the same image sections. With the same viewing size of the complete image, the same manufacturing technology of the sensor and the same sensor size, the noise behavior is approximately the same.
Since it is impossible in many situations and with many cameras to completely suppress the image noise during the recording, we will take a look at some possibilities for noise reduction.
One simple way is to make the subject as bright as possible. In this case, the intended information recorded by the sensor is stronger than the unwanted image noise. The image noise is thus “overwritten”, to put it simply.
In light painting, it is common to use the “Live View” function to frame and focus the image. Since the Live View image is captured by the sensor and displayed on the screen, the sensor also heats up.
The display also does not have an efficiency of 100% and thus additionally contributes to the heating of the camera housing. After setting up the camera using Live View, it is recommended to turn it off and give the sensor some time to cool down before starting light painting.
Most cameras use special internal algorithms for noise reduction. In most cases, these can be switched on or off in the menu. Basically, there are two variants for the camera-internal noise reduction. “Noise reduction at high ISO values” only affects the compressed JPEG image, for this reason this function is uninteresting for most light painters. In the vast majority of cases, we will shoot our light painting images in RAW format.
The “Noise Reduction at Long Exposure Times” function (also) affects the RAW image. This function, depending on the camera and the other factors mentioned above, may well provide a better result. With this function, after the actual exposure, another exposure is taken without opening the shutter, i.e. a dark image. All information in this dark image is inevitably unwanted image noise. The information is then “subtracted” from the actual image before it is saved and filled with information from the neighboring pixels. However, this only works up to a certain level of image noise without visible loss of quality. With the exception of some Canon cameras, the exposure time is exactly the same as that of the actual recording. And that’s where we come to the main problem with this function. The impatient light painter has to wait another 10 minutes after the 10 minute performance until he can take a look at the camera’s display and then either repeat the light painting or move on to the next picture. On the other hand, the image noise at shorter exposure times (up to 2 minutes) is not visible in the final result with most cameras, so the function does not bring any improvement. For this reason, the function is almost always switched off on our cameras.
In addition, the subsequent noise reduction, whether in the camera or in post processing on the computer, almost always has a negative impact on other aspects of the image quality (sharpness, contrast range). To be able to use this function sensibly, you should take some test shots with your camera. If the image ends up almost completely noise-free on the memory card with the lens cap on at ISO 100, 20°C ambient temperature and 10 minutes exposure time, you can safely turn off this function. Most light paintings will probably be done within these parameters.
If necessary, noise reduction is also possible in image processing on the computer. Due to the missing dark image, however, this cannot work as effectively as the internal noise reduction of the camera. Of course it is also possible to make the dark image by yourself and to transfer it to the computer (10 light painting images a 10 minutes – 1 dark image a 10 minutes). In the field of astrophotography this method is widely used.
DYNAMIC RANGE, RESOLUTION, SHARPNESS, COLOUR
The Light Painter can have little or no influence on the physical properties mentioned above; these parameters are set by the manufacturer during production of the camera. Only in the subsequent image processing can something be changed. However, only as far as the sensor has recorded information.
A modern sensor from Sony has a larger dynamic range than sensors from Canon or Olympus. The sensor therefore records more information.
The resolution and sharpness of almost all reasonably current cameras are at such a high level that large prints can be made from the images without any problems. The number of pixels hardly plays a role for most requirements. 12 million pixels correspond to an area of approx. 150 x 100 cm when displayed on a monitor with the usual 75ppi. Prints with 300 dpi are possible up to size Din A3. Since the viewing distance increases with larger prints, it is quite possible to print larger formats in lower resolution. Nobody seriously looks at a 3 meter canvas from a distance of 5 cm.
So in light painting, we are only interested in the dynamic range and noise performance of the sensor. All other properties are either not of great concern or can be corrected by simple means on the computer, such as white balance, for example.
HOW WE TESTED?
Two test series were made with all cameras. Both series were carried out in closed rooms at 20°C room temperature. The cameras were not moved during the exposures. Either the lens mount was closed with the matching cap or the lens was covered with the lens cap. All images were captured as RAW at full resolution. The noise reduction function was switched off.
For the first test series, shots were taken with 30 seconds exposure time each with all whole ISO values. In some cases, we will also work with higher ISO values than 100 in Light Painting.
In the second run, shots were taken at ISO 100 with the following exposure times: 30, 60, 90, 120, 150, 180, and 600 seconds. This part is the really interesting one for long exposure and light painting work.
I evaluated the results using the histogram. The further the histogram moves to the right, the more information the sensor has recorded. Since no light hit the sensor during the exposure, all recorded information is image noise and therefore unwanted.
The displayed range is from 0 to 255. 0 means that no information was recorded, 255 that all pixels were excited white at full brightness. I used the arithmetic mean as a criterion for the comparison. The histogram from the example is from the image noise recording from above. 31.2 is an extremely high value for a dark image. With these settings, a normal shot is hardly possible.
For comparison, the calculated average of the Light Painting image from the previous section is 80.8. Even with quite bright images, like the one with the hand above, it is usually not above 100.
Below is the distribution among the individual colors. For the result, however, it is usually completely irrelevant whether the noise has a higher proportion of red, green or blue. If the main color of the light painting is blue, however, a high proportion of blue hotpixels will not disturb as much as the red and green hotpixels. But for the practical work in light painting this does not matter, nobody will analyze his sensor exactly and choose the main color for the light painting afterwards.
Ideally, the calculated average for our dark images is 0, i.e. no image noise at all. In the dark image at 100% magnification, the noise becomes clearly visible from values of approx. 6. When the image noise becomes visible in real images is difficult to predict and above all depends on the black content in the image. As mentioned above, the noise is less visible in bright areas of the image. At values of 12 and higher, the noise is already clearly visible in the small preview image. Thus, in the vast majority of images this will be perceived as disturbing. The noise reduction of the camera will certainly almost always improve the result, but at some point even this will reach its limits. Images with values greater than 25 are probably almost always a case for the trash can. Here are some sample images as 100% crop (click to enlarge): Calculated average from left to right 0 – 1.3 – 5.1 – 10.4 – 31
Meaning of the data in the following tables:
- The “Year” column indicates the year of market launch. Any technical changes without a change in the model designation are not taken into account.
- “Price” is the current street price. For cameras that are no longer in production, the average price for a used example in good condition.
- In the “Mpix” column is the resolution specified by the manufacturer.
- In the other columns is the calculated average for the colors from the histogram. (see above) – 180s and 600s at ISO 100 – ISO 800 and 3200 with 30 seconds exposure time.
- Sensor sizes: FX (35mm format) – 36 x 24 mm – DX 23.6 x 15.8 mm – APS-C 22.5 x 15 mm – MFT 17.3 x 13 mm.
The actual sensor size differs slightly from the above for some models:
The “old” D300s delivers surprisingly good results. I removed the infrared cut filter on the D300 (without S). I can’t say if this has any effect on the image noise. The D300 has a different sensor than the D300s even if the stated resolution is the same. The D300 has exactly 12389760 pixels, while the D300s has a slightly lower resolution of 12212224 pixels. The other Nikon models also do well with long exposures. Moreover, with the D300s and the D750, relatively long times at ISO values up to 3200 are possible without any problems.
|EOS 5D Mark III||2012||1780€||22,3||FX||2,6||2,6||3,9||8,8|
|EOS 5D Mark IV||2016||3000€||30,1||FX||1,2||1,3||2,1||6,6|
* ISO 1600 – the highest ISO value of the 1000D.
The Canon cameras generally have a slightly higher image noise at long exposure times than the comparable Nikon models.
With the EOS 6D at a price of about 1000€, you definitely have a rock-solid working device in FX format with low image noise at long exposure times. Only if you really need the highest quality for large prints, the additional investment of about 700€ for a Nikon D750 is worth it.
The old 5D from 2005 is only of limited use for long exposures and light painting. For the current prices for used copies, you are definitely better off with the Nikon D300s.
The results surprised me a bit here. I had expected more from the Sony cameras. In direct comparison with the Nikon D750, the Sony α7II performs a bit worse. The higher image noise of the Sony is probably due to the slightly worse heat dissipation of the smaller housing, but the gem also costs 500€ less than the Nikon behemoth.
The inexpensive α77 does quite well at ISO 100 and long exposure times. At ISO 3200, however, the thing noises like a waterfall.
The α7RIII has the highest resolution in the test. The values at long exposure times and high ISO values are good. However, the operation of this camera is a disaster when you want to work with long exposure times. There is no “time” mode, the “bulb” mode does not work in “silent shooting” mode, you have to figure it out first. During the exposure, the display backlight was on the whole time. What is this crap? The battery is drained and the display also heats up the housing, so the image noise increases completely unnecessarily. Since the body is smaller compared to the thick DSLR’s, most of the controls are smaller as well. This dial to the right of the display was not my friend at all. Operating it in the dark thus becomes a challenge, at least in my test over a weekend. I wasn’t really sad when I had to give the camera back.
|OM-D E-M10 II||2015||365€||16||MFT||3,6||11,5||2,8||9,1|
The small MFT sensors have a generally high image noise. The E-M1 from 2013 is hardly usable for long exposure times; even at exposure times of 60 seconds at ISO 100, the calculated average is already 9. In a longer light painting performance, the internal noise reduction will also reach its limits. Only the “Live Composite” function, which is appreciated by many light painters, is an argument for buying an Olympus. If you want to work result-oriented and often want to do light painting with exposure times of more than 5 minutes, it is better not to use an Olympus camera.
The lowest ISO value for the RAW format of the Fuji cameras is 200, so the values for 180 and 600 seconds are not recorded at ISO 100 like the other cameras. However, ISO 200 may be too high for some light painting ideas.
According to the test series so far, the recommendation clearly goes to the Nikon D300s. A used Nikon D300s in good condition with no more than 50000 releases is currently available for under 300€, light painting fun for the small purse. The camera is very robust, Nikon gives a warranty on the shutter for 150000 releases. Thus one has surely also with a used copy still for a long time its joy. Our two D300(s) have already been through a lot and work perfectly as on the first day.
Those who prefer to buy a new camera should take a closer look at the Canon EOS 6D and the Nikon D750.
Fujifilm’s X-T1 cuts a good figure, but it’s not exactly a bargain for an APS-C camera.
Actually, I had the plan to take additional comparison shots on the subject of dynamic range. But since the article is already quite detailed, I will include this topic in a separate article. Nobody else will read it 😉
I would like to sincerely thank all those who supported me with this article: Matti Thurley, Dominic Poncé, Gunnar Heilmann, Manuel Paul, Ulrich Günther, Erich Klingenberg, Roland Brei, Martin Wink … I hope I have not forgotten anyone.
Allways good light and little image noise