Skip to main content

ISO Invariance

Way back in March 2017, Petapixel had a fascinating article on ISO invariance in digital camera sensors. The article details how ISO affects exposure, and how to ascertain the best ISO for low-light work - in particular, for astrophotography. For one of their tests, a Sony A7S was used (Canon 700D and Fuji XT1 for others), and I doubt you can get a better stills low-light performer than that!

Reading the article got me thinking - how would my Micro Four Thirds camera function in a similar test? Bare in mind the Four Thirds sensor is half the size of the full-frame used in the A7S, and 16 megapixels versus 12.2 megapixels in the Sony. So while the sensor is small, the pixel-count is higher, meaning the photosites are smaller in the m43 camera as well.

Anyway, I'm getting into too much detail in the technical comparisons! Let's look at how we test ISO invariance. Take a series of shots in RAW, each at whole-stop ISO and  match the exposure brightness in post across the range.

For the shots, I followed the article's instructions:
  • Shoot at night
  • Shoot in RAW
  • Manual exposure mode
  • Set white balance to daylight
  • Disabled all noise reduction (including Long Exposure NR)
  • Shoot one exposure at each whole stop ISO (200, 400, 800, 1600, 3200, 6400)

The following is using my Lumix GH4 set-up on a tripod in the back garden. Manual-mode, WB set to Daylight, 1 second shutter speed, aperture at @ f/2, and the Voigtlander 17.5mm in front of it all.

I then ISO incremented by full stops, starting with ISO6400. Gradually reducing by full-ISO-stops, but keeping the same shutter speed and aperture.
Oh, and no noise-reduction reduction in camera. RAW files were imported to LR with no sharpening or NR applied, but all exposures normalised. So ISO 3200 image increased by +1EV to get to same exposure as ISO 6400 etc.



Picture info: Lumix GH4, Voigtlander f/0.95 17.5mm, ISO200-6400, f/2, 1 sec
Best viewed full-screen

So, what does it say? Well, to me there is no discernible difference between 6400 and 3200 (+1EV). And I don't think there's too much difference between ISO3200 (+1EV) and ISO1600 (+2EV).
But it gets very noisy after that. ISO400 has a weird colour cast for some reason (bare in mind same white-balance throughout). I think I'd like to reshoot this, just to make sure that the ISO400 result isn't just a blip. 
The only difference between my test and theirs was my starting at ISO6400 as the base. I believe in the Petapixel article, they use ISO3200 as the base, then apply a -1 exposure compensation.

Looking at the image above, I think I'd be happy shooting astrophotography at 1600 or 3200. Anything lower looks like it will increase noise horribly. For low-light shots around town, I'll remain happy shooting at 800 - I don't tend to boost the exposure much for street, just a slight adjustment to shadows (and a lot of that noise can be mitigated through masking).
Labels:

Comments

Post a Comment

Popular posts from this blog

A dark night, and a brace of Voigtlanders

I fell in love with the Voigtlander f/0.95 17.5mm prime as soon as I mounted it on my GH4 a few years back. It properly sprang to life once I moved over to the Lumix G9 - the higher-resolution viewfinder made it much, much easier to check focus, and I feel that the sensor was able to benefit from the lens more than the GH4's was. Due to coronavirus in the UK, I've left the area where I live on very few occasions. I am lucky in that I have the South Downs to the north, and the English Channel to the south - and both are only a five minute walk away for me. Despite coronavirus, I've ensured that I've kept my Photo 52 challenge up to date. This forces me to get out with my camera so that I at least have one image, no matter how rubbish it is, to show for my week. As the nights have grown longer, the Voigtlander 17.5mm and its 42.5mm brother have both joined me on my evening walks, so I thought I'd show you a few shots I grabbed the other evening whilst traipsing ...
Post a Comment
Read more

BOTBs - Panasonic Lumix GH3

Panasonic had realised with the GH1 and GH2 just how much demand there was for high-quality, low-cost mirrorless cameras capable of brilliant video capture. And it was with the 2013-released GH3 that they responded to many of the demands that users had asked for. Panasonic Lumix GH3 January 2014 to December 2016 22500 shots Gone was the multi aspect Panasonic sensor, replaced with a 16 megapixel unit from Sony - the same, it transpires, as used in Olympus's at-the-time recently released OM-D E-M5 . This was a more capable sensor than the one used in the previous model, which helped with better noise handling at higher ISOs and higher-bit-rate video. Gone too was the compact body. Now, the GH3 was a chunkier magnesium alloy affair, and now weather-sealed. It was a huge improvement ergonomically over the GH2 - after shooting airshows for many hours with the GH2, my hand ached, mainly because there wasn't enough room around the handgrip for all fingers. This was no longer th...
Post a Comment
Read more

Lessons from RIAT 2018

Last weekend I got up stupidly early to travel half way across the country to attend an airshow. But what an airshow! The Royal International Air Tattoo 2018! A word of warning - this is going to be a looooong post. Grab yourself a cup of tea and settle down... I haven't been to an airshow, or at least shot aircraft flying since the Goodwood Revival in 2015, where they had a number of warbirds from the Second World War flying. My equipment back then was my trusty GH3 with the first-generation Lumix 100-300mm lens, and my GM5 with Lumix 14-140mm as backup. These performed well as a pair, allowing me to grab take off, landing, and formation shots with the surprisingly-capable GM5, and the longer, single aircraft detail shots with the GH3. At the time, AF-S was used on both cameras, as I didn't trust the reliability of the continuous auto-focus offered by either - bare in mind that at this point, Panasonic had yet to introduce their Depth from Defocus technology to assist wit...
2 comments
Read more