Lumix S1Rii Night Testing

[Kyle Jones]

One of my hopes for this camera was that it would be a solid night performer, allowing me to take advantage of the glass (from Sigma in particular) available for the L-mount along with being able to easily adapt older Canon and Nikon lenses. My first night of testing was at the Hungry Horse Dam and at nearby Lion Lake. I'm sharing photos from the lake first, since the ambient light around the dam made that location "easier" for the camera. Lion Lake was in full darkness with no light on the foreground at all.

I captured five images in each sequence, all with the Lumix 18mm f/1.8 lens (I'd rate this lens as adequate for resolving stars but it's small size and weight make it awfully useful) wide open at f/1.8. Each exposure was for 20 seconds and I only varied the ISO between shots. Exposures were made at ISO 800, 1600, 2000 (I believe this is the 2nd native ISO which is why I included it), 3200, and 6400. I brought the images into Lightroom and did some basic processing, including pushing the shadows to see what detail was there. I wanted to keep the processing somewhat realistic, since I wouldn't want to base any decisions on how things look with the shadows pushed 5 stops. I did not use PureRAW or any of the AI noise reduction tools - just lightroom's older luminance noise reduction set at 20. I then used the exposure slider to balance the overall brightness of the image, with the ISO 800 image at +2, the ISO 6400 image at -1, and the others in between.

Here are the images:

ISO 800

ISO 1600

ISO 2000

ISO 3200

ISO 6400

The first thing that I noticed was that there is a green color cast in the shadows on the left side of the image and a magenta cast on the right. To my eye, the pushed ISO 800 image is definitely noisier than the others. The sweet spot for image quality looks to be between the ISO 1600 and 2000 as I would have expected going in. Strangely enough, although some of the detail has been lost, the ISO 6400 image looks a little cleaner than the ISO 3200. You can see this better in the 100% crops below.

ISO 800

ISO 1600

ISO 2000

ISO 3200

ISO 6400

 

[JimFox]

Interesting testing Kyle, I am glad you did it.

First off before I forget after looking at the 100% images is that ISO 2000 does seem to be the best. I agree that ISO 6400 looks pretty good and better then ISO 3200 which I find surprising. Though there appears to perhaps be a little mushing of the detail at ISO 6400 compared to ISO 3200 so that's why at a glancee ISO 6400 feels better?

I would be interested in seeing 100% crops of the trees either against the sky or the water, though the tree/sky would be more important. The reason for asking is to get something with a sharper edge. The trees are in shadow, but the sky isn't. So it would be interesting to see how sharp that maintains. Though I guess part of that might be the result of the lens and it's IQ compared to the sensor so much. Anyway, just a thought. It may not make sense.

 

[Kyle Jones]

Interesting testing Kyle, I am glad you did it.

First off before I forget after looking at the 100% images is that ISO 2000 does seem to be the best. I agree that ISO 6400 looks pretty good and better then ISO 3200 which I find surprising. Though there appears to perhaps be a little mushing of the detail at ISO 6400 compared to ISO 3200 so that's why at a glancee ISO 6400 feels better?

I would be interested in seeing 100% crops of the trees either against the sky or the water, though the tree/sky would be more important. The reason for asking is to get something with a sharper edge. The trees are in shadow, but the sky isn't. So it would be interesting to see how sharp that maintains. Though I guess part of that might be the result of the lens and it's IQ compared to the sensor so much. Anyway, just a thought. It may not make sense.

Here are some crops at the top of the trees - good idea! I agree that the noise may look better at 6400 vs. 3200 because it is losing some of the detail.

ISO 800

ISO 1600

ISO 2000

ISO 3200

ISO 6400

 

[JimFox]

That's an interesting take with the trees.

I think I would still pick ISO 2000, though ISO 3200 looks good too. ISO6400 is the cleanest but I can see some smudging of details in it.

I wonder if the shadow noise in the ISO 2000 and 3200 would disappear with a Lum NR of between 30 and 40? The sky looks really nice with the Lum NR of 20. I usually run mine between 20 and 30.

 

[Kyle Jones]

I usually use different noise reduction tools. I was just trying to keep it simple for this comparison. I'm pretty sure that any of these except the ISO 800 file would clean up nicely with a little effort

 

[JimFox]

I usually use different noise reduction tools. I was just trying to keep it simple for this comparison. I'm pretty sure that any of these except the ISO 800 file would clean up nicely with a little effort

For my Timelapse's I just stick with the old Luminance NR as it's so much quicker. So my mind went there and forgot about the AI tools.

For my regular Milky Way single image pricessing I have noticed that the AI tools seem to have a tendency to add star detail where there is none, especially on the edges, so I keep the newer AI NR tools to just the ground layer, and stick with Luminance NR for the sky.

That's not 100% hard and fast, but that's my techniques probably 90% of the time. But each time out is different, so I try to stay flexible and use what looks best.

 

[Kyle Jones]

That's not 100% hard and fast, but that's my techniques probably 90% of the time. But each time out is different, so I try to stay flexible and use what looks best.

Exactly - I tend to get better or worse results with each tool on different images. I move between LR's enhance, PureRaw, and Topaz Denoise and see what is better. Sometimes I use different tools on the sky and ground. My favorite these days is still to take multiple images and do a smart object stack for the ground. Then I'll do a little Topaz Denoise (10 at the most, sometimes as little as 5) on the whole image.

I did take 3 copies of each ISO 2000 image so I could try that on these scenes.

 

[Kyle Jones]

I've created a post on my site here if anyone wants to follow along with my testing.

The most recent edit is adding versions of the image shot at ISO 1600 with +0, +1, +2 and +3 stops of exposure in Lightroom to see how everything holds together.
The image holds up OK to +2 stops, although you start seeing a green tint on the left and a purple tint on the right. This gets much more pronounced at +3 stops. Lightroom luminance noise reduction is at 25 for each of these (and all of the images above for that matter).

ISO 1600 image with exposure at +0

ISO 1600 image with exposure at +1

ISO 1600 image with exposure at +2

ISO 1600 image with exposure at +3