Hot pixels in OM1 and OM5 Long Exposure photography
In an earlier post, I compared the performance of the Olympus OM1 to the Nikon Z7 for long exposure (LE) photography. My conclusions from that comparison were correct, but not the whole story. A more nuanced picture has emerged from an enquiry made by a reader of this site. As well as being quite illuminating, it has also led to a major improvement for me in the time it takes to shoot an LE photo.
Noise in long exposure photographs
Noise in digital cameras is the graininess (luminance noise), mottled colour (chrominance noise) or bright dots (hot pixels) that result from the imperfect transformation of incident light on the sensor into a visible image in the camera or on an output device. It is equivalent to the graininess of film in the analogue days.
There are three types of noise that influence digital camera sensors. The first and most significant is shot noise, which is a quantum physics effect caused by the random fluctuations of photons arriving at the sensor. The lower the amount of light, the greater this amount of noise is relative to the output signal. Hence, shot noise is highest in low light situations. It doesn’t vary with time, so images for which there is sufficient light have low noise even on long exposures.
The second source is read noise, which is the compounded errors from the associated electronics used in reading and amplifying the sensor signal. Again, it is influenced by the amount of light rather than time.
Both shot noise and read noise occur in a statistically predictable way, and can be very effectively reduced or eliminated by recent advances in AI post-processing software such as Topaz, Lightroom or On1.
The third form is dark field noise, and specifically a form called dark fixed pattern noise (DFPN). It is caused by inherent differences in individual pixel response due to variations in pixel size, material, or interference with local circuitry. The main things that affect DFPN are temperature and exposure time. So the longer the exposure and the hotter the sensor, the worse the problem. This is the main noise concern for long exposure photography. DFPN shows up as bright “hot pixels” across the image. It cannot be handled by AI in the same way as shot and read noise, and is the bane of LE and astro photographers’ lives.
It should be noted that long exposure hot pixels are different from dead (i.e non-functioning), or “stuck” (i.e permanently on) pixels. These pixel problems are static, and can be eliminated from images by a special function in all Olympus cameras called “pixel mapping”. However this does not work for LE hot Pixels, which are dynamic and appear only in long exposures.
Hot pixels over time
One way to determine if a camera produces a large amount of DFPN hot pixels is to do a dark field test. In this test, a long exposure shot is taken with the lens cap on. The number and brightness of hot pixels can then easily be seen and compared between cameras. In March 2022, I published a post that compared the Olympus OM1 with the full frame Nikon Z7, which showed that contrary to expectations, the OM1 had fewer hot pixels than the Z7.
I was recently contacted by a reader of that article, who was finding a lot more hot pixels in his OM1 than I found in 2022, and many more than he originally saw when he first bought his camera. He asked me to repeat the dark frame tests and see if the hot pixels had increased over two years. I did the tests, and it was immediately clear that this was the case. They had also increased rather more in the OM1 than in the Z7.
More research has shown that this increase is a known factor in digital camera sensors, as referenced in this paper. It is caused by cosmic ray induced defects that begin to appear soon after fabrication. These defects are permanent in nature, and their number increases continuously over the lifetime of the sensor. Moreover, the smaller the pixels, the greater the effect, which explains why my OM1 had more hot pixels than the full frame Z7 after two years.
So it’s cosmic. We just have to deal with it.
Dealing with Long Exposure hot pixels
All modern cameras have a simple but tedious way to deal with fixed pattern hot pixels. Immediately after taking the long exposure shot, another shot is taken for the same amount of time, but as a dark field exposure, so that only the hot spots are seen. This exposure is then subtracted from the first one to eliminate all the hot spots. It’s a great solution except that a 2-minute exposure now takes 4 minutes, and a 6-minute exposure takes twelve – almost unbearable. It is also not a great solution for astro photography where many long exposures need to be taken.
Despite this tedium and the great results of my 2022 testing, I still continued to use this noise reduction (NR) technique for all my LE shots with my Olympus cameras – just to be on the safe side.
Can we manage without noise reduction?
My initial testing didn’t show any difference on my OM1 between the dark frame JPEGs produced in 2022 and 2024. They were both completely clean. It’s taken me a while to understand why this happened. The reason is that I processed the RAW files in Lightroom, and it turns out that, unbeknownst to me, Lightroom automatically removes hot pixels from the raw image. As a result, I saw no difference between the 2022 and 2024 tests. Differences were only seen when I looked at a JPG image, or a raw image viewed in less sophisticated software like FastRawViewer.
The removal of hot pixels by Lightroom is quite remarkable. It prompted me to explore whether other software does the same thing.
Software that removes hot pixels
Here is a comparison of the software I have tried in order of effectiveness:
- On1 PhotoRaw 2024: seems to do the best job – noise/sharpening has to be in use, and there is a further “on/off” switch in the “classic” noise model. The flagship noise model, “No-noise AI” does not have a switch and seems to do about the same job in removing hot spots as classic, but a much better job in removing luminance noise. On1 claim their latest AI noise model (2024.3) has been tuned for astro and “….will be the best for these photos by far”, although I have not tested this yet.
- Lightroom Classic: does an excellent job also, but is always applied automatically, no switch.
- DxO PhotoLab 6: also does quite a good job, and also has a hotspot slider control that lets you see the extent to which hot spots are controlled. Does not remove the most intense hot spots.
On1 for hot pixel removal
For a variety of reasons, I have now moved to On1 PhotoRaw 2024 for my post-processing. Having seen via the dark frame tests what a good job it was doing on hot spots, I did a some real-world testing on my OM1 to see if it worked in complex landscape images as well as pure black ones. I took two outdoor shots, with and without NR, both for 2 minutes. For the no-NR image, hot spots can be clearly seen in the out-of-camera JPG on the left, indicated by the red circles.
However, once processed in On1, the hot spots were removed, and there was no difference to the image with NR applied in-camera, as you can see on the right. Moreover, there was no difference in luminance noise between the noise-reduced, and non noise-reduced images, which was something I was bothered about. I repeated these tests on my OM5 camera and got the same results, except for one stubborn hot spot that I can take out manually.
This is a real game changer for me because it now means that the doubling of the exposure time for NR in-camera processing is not needed. This immensely speeds up and improves the whole LE process. So many times I have completed an LE shot, and the sky turns a beautiful shade, but I can’t take the shot because the camera is still chuntering away doing NR.
Hot spots versus dust spots
One thing I have found in my latest testing: what I thought were stubborn hot spots turned out to be sensor dust. I found this by taking short exposure shots of 1/60s and comparing them to the LE ones. The short exposure shots, of course, have no LE hot spots, but did show a fair amount of sensor dust spots.
Olympus cameras have a high frequency cleaning process for sensor dust, so they are remarkably free of it. As a result, for all my Olympus photographs, I never bother with cleaning up sensor dust – it is quite inconsequential in the overall image. Hot spots are at about the same level of visibility for me. So if I didn’t bother about dust spots, then why be concerned about hot spots (except of course for astro, for which there are separate solutions).
On the other hand, for the Nikon Z7, sensor dust is a massive problem, and I can never get rid of it. It is always there, even right after cleaning the sensor, always accumulates over time, and is always very intrusive in the image. It is far more visible and more difficult to remove than hot pixels. I believe the same is true for Sony (essentially the same sensors). So once again, I am super happy with my Olympus cameras and their balance of features for my kind of photography.