Friday, May 05, 2023

Low-light performance of Panasonic Lumix DC-S5

This article will examine the low-light performance of the Lumix DC-S5, first by presenting some sensor statistics, then by examining photographs taken at different ISO values. My personal goal is to compare this to the Olympus PEN E-P5 I was previously using. I expect an improvement, but how big will it be?

To start, some numbers

Moving from a Micro-Four-Thirds sensor to a full-frame sensor, I'd expect significant benefits for low-light photography. After all, if the same number of pixels are put into a larger area, each will capture more light. How much more? That depends on several technical factors, tabulated below from info at Digicam Database.

Panasonic Lumix DC-S5 Olympus PEN E-P5 comparison
sensor resolution 6026 x 4017 4627 x 3479
sensor dimensions 35.6 x 23.8 mm 17.3 x 13 mm
sensor area 847.28 mm² 224.90 mm² 3.77x
pixel count 24.2 Megapixels 16.10 Megapixels
pixel area 34.93 µm² 13.99 µm² 1.5x

The Olympus PEN E-P5 was first introduced in 2013 while the Panasonic S5 is 2020 technology. A lot has happened to sensor design and image pipelines in seven years. Many qualities cannot be captured quantitatively. For example, it's not just a matter of how much noise is in an image, but the quality of that noise, which can differ from rather film-like to a less appealing digital look.

Two data points are particularly important, so I've marked them in the comparison column. The sensor on the Lumix is much larger than the Olympus PEN but it also has more pixels. What's important is the area of each photosite. Here we determine that each such pixel will capture 50% more light. 

To continue, some test photos

The only way to understand the photographic significance is... take some photos! So I set up a still life in dim evening light, using a tripod and timed shutter release for stability. I mounted a Pentax 43mm Limited through an adapter, set to an aperture of f/4. I shot in RAW using the Natural photo mode, varying the shutter speed on the camera as needed, to balance the exposure. Try to ignore the differences in colour balance.

Here is a strip of 100% crops for ease of comparison. Be sure to click through to Flickr so you can see this at actual size. You'll also find an album of full-size photos.

ISO test strip

I developed these in Affinity Photo with neutral settings. Normally I'd be trying to make appealing photos, but these I kept flat. 

For the two high ISO cases I decided to present a clean option by setting both luminance and colour noise reduction to 20%. To compensate for softness, I added detail refinement (radius 60%, amount 50%). To be sure, this was only a rough approach to the problem. Normally I would use some moderate detail refinement in all cases. 

Analysis

At ISO 200 the dust and cat hair in the bowl are clearly delineated. The texture of the dry plant matter is well-defined, at least where the objects are in focus. In honesty, it's very difficult for me to distinguish between the jumps in ISO, even up to 4000. 

This result confused me in one respect. The Panasonic S5 has a dual ISO sensor, meaning that two different ISO settings are optimal. The values differ in each picture mode, as follows:

mode low ISO high ISO
Natural 100 640
Cinelike D2/V2 200 1250
HLG 400 2500
V-Log 640 4000

I forgot to shoot exactly at 640 (oops!) but I would expect some indication that 800 ISO offers an improvement over 400. My eyes don't see this, but that's perhaps because all of these low ISOs look great. My experience shooting with the Blackmagic Pocket Cinema Camera 4K is that the two base ISOs on that camera are quite distinct. Hence, shooting near but just under the high base ISO is poor practice. It pays to increase ISO to the higher base, since the resulting image is cleaner. But I don't see that recommendation carrying over to the Panasonic S5.

The second result is the significant reduction in image quality stepping up to ISO 6400. I noticed this right away, so went back to take a shot at ISO 4000. Looking at the 100% crops, it now seems that the issue is one of focus, not noise. The noise itself isn't obvious until ISO 12,800. Even then, is there anything truly wrong with this shot? 

12,800 ISO, developed
ISO 12,800 developed version

At 25,600 the chroma noise becomes unappealing, which is why I decided on the noise reduction option. But this opinion will depend very much on the subject. If I was shooting a band, the noise might be favourable. If I was in monochrome, likewise.

On my Olympus E-P5 I wouldn't increase ISO beyond 1600 unless I had no option. The Panasonic Lumix S5 provides a three stop advantage at all times, with potentially more latitude as necessary. That's a great result!

Addendum

This video contains a noise test similar to the current article, except shot in V-Log. (More about that mode in my next article.) Like me, the author found that there is not a huge advantage stepping up to the high ISO setting, though this does tame the noise in a more visible fashion in V-Log than in the Natural mode I used above. That's because we are already at 4000 ISO, not 640 ISO.

Dual ISO does not fundamentally improve noise because it was designed for a different purpose: to allow the full dynamic rage of the sensor to be used at different light levels. If this happens to improve noise the equivalent of one stop, that is just a happy byproduct.

RELATED POSTS

No comments:

Post a Comment