The 5D Mark III is already a very capable low light camera. In its factory guise whilst not quite as clean as the Canon C300 or Sony FS100, it is the best DSLR for low light shooting (though the Nikon D5200 puts up a good fight).
But that was before the latest developments with ML Raw.
Has low light improved even further?
The answer is a resounding yes.
The image is cleaner in low light with raw, with even exposures I rated ISO 12,800 coming out perfectly usable. I’m hesitant to say it as a matter of fact but based on my early observations the 5D Mark III seems to have surpassed the C300 as the ‘low light king’.
The power of raw is that you can fully exploit the dynamic range of the image, and in low light that means going into the shadows and giving them a lift, or crushing them back for more contrast and less noise.
First an important note – don’t judge noise from the Vimeo streaming embedded clip. The heavy compression on the website makes it seem MUCH noiser and blockier than the raw actually is. Download the more lightly compressed file or even better take a look at the ISO 1600 shot below. So clean.
RAW (click it to enlarge)
Note on warm cast – that is on purpose, creative choice to match look of scene how my eyes saw it. You can apply cooler white balance to the raw image if you download the DNG.
Here for comparison is the H.264 video mode 1080p (white balance on automatic). Look at the lovely bulb in the old German table light – vanished!
Also look at the poster – half of it is gone.
This is why I am never going to shoot a serious project in H.264 again, until they give us a better image which more closely resembles raw or ProRes.
The standard compressed video mode is totally washed out in terms of colour as well, at ISO 1600. Dynamic range is shot to bits. I’d say only 6 stops usable in it! Why is it so bad? What are Canon doing to the video image to get from the raw sensor feed to this? Compression alone is not responsible. The highlights are burnt – gone – unrecoverable. Any attempt at recovery leads to horrible banding as well. The blacks are crushed too compared to the raw image.
The latitude in the raw shots around the light sources impressed me hugely – in particular watch for the candle flame in the ISO 12,800 H.264 shot.
The whole lantern area is burnt out in the H.264 file but the raw retains everything.
And look at how much more detailed the raw shot is!!
Now actually H.264 is a great codec. The screen grabs you see above were taken from the 24Mbit H.264 Vimeo download file and compressed even further to JPEG for the web. It can handle plenty of image quality despite being compressed, as can a JPEG.
I think serious questions need to be asked of why Canon had to hack the image down so much.
The dynamic range is severely clipped from what it could be, and so is resolution. I don’t understand why compression alone has to hurt the image so much, and it is not due to downscaling or the sensor – as the raw feed is already 1920×1080 (when cropped from the original 3:2). There’s no need for further scaling before it goes to the encoder. No need to chop 4 stops dynamic range off it either.
In that candle shot you’d be forgiven for thinking the exposure isn’t ISO 12,800 if you only notice the retained highlights, which are no longer over exposed – but if you look at the mid-tones (poster on the wall and the blinds) and shadows (sofa area) you will see I have done my best to match the exposure to the H.264 ISO 12,800 shot. To get those lows and mids in the raw shot, you’d have to shoot ISO 8000 to 12,800 on a normal H.264 camera – and your latitude would be tiny, overall dynamic range probably around 6 stops at ISO 12,800.
I notice that noise in the shadows and blacks in general have a magenta cast if you have a warmer white balance and the magenta tint slider to the right but if you put the magenta tint slider back towards green, the warmer yellows go way too green for my liking. This could be a light issue rather than a camera one. Energy saving practical lights do give a green cast. The solution in post was to avoid anything higher than 3200k white balance – this was more keeping with the LED practical light I had for fill lighting in the scene any way.
In the ISO 800 shot I have the table lamp on, I purposefully turned this off for the higher ISO shots to lower the ambient light. A genuine low light test should be done under dim lighting with lots of dark areas, in effect it should be under exposed. Uniformly bright studio lights and stopping the lens down for a normal exposure is not really a good low light test in my view, you need those under exposed tones in the shot!
Better than Blackmagic Cinema Camera?
The 5D Mark III with raw is basically a totally different camera to the factory 5D Mark III and there’s a lot to get our heads around and explore. This is an early test and there’s much more to come. The pace of the Magic Lantern development lately has been astonishing. I have my Blackmagic Cinema Camera sat next to the 5D Mark III and it is going to be very interesting to compare. Already the 5D Mark III has a few advantages on the specs sheet
- Less noisy (in low light)
- Less noisy (no fan!)
- Small interchangeable battery and good run times on one charge considering tiny battery
- Full frame sensor
- 60fps raw (at 720p… In 1080p the data rate is too high for the CF card)
- 22MP stills
- Smaller and lighter form factor
- More physical controls
With that said, it is important I think not to rush to judgement on this as it is such early days. I you are thinking of cancelling your Blackmagic pre-order, I don’t recommend doing so – please wait until this shakes out. Of course Blackmagic will likely still have the resolution advantage on the 4K Production camera and global shutter is still groundbreaking for $4k. Lots to look forward to from them.
I think the 5D Mark Raw is more of an dilemma for Nikon D800 owners looking for video quality. Some of them would have moved to that camera for the gain, and now they will have to come back again big time.