EOSHD Shootout Part 1 – 5D versus GH1 in Low Light

https://vimeo.com/10657409

If possible please watch the test footage on a proper display via DVI or HDMI, not a laptop screen. It makes a HUGE difference in perception.

The results of EOSHD’s low light shootout between the mighty Canon 5D Mark II and the Panasonic GH1 are in! This is a real-life natural light test, this is what happens when you take out the cameras and shoot. There are no test charts. No technical waffle in the video (I left it for this article!). But I have selected the shots quite carefully to show a range of typical low light challenges.

I’ve also made sure that on-camera picture profiles are taken into account. They’re set quite neutral and to match the other camera as much as possible. I also included shots which compare different picture profiles and settings so you can get an idea of their effect on the camera’s footage relative to one-another.

So what do I think about how these cameras perform in comparative test?

Contrast

Well the GH1 is certainly more punchy and contrasty across the image. You can set the picture profile on the 5D Mark II to boost the contrast but the GH1 always seems to be one step ahead. However whilst it’s image does look great and has a lot of impact, it is poorer when it comes to latitude, low contrast shades and details.

Noise and high ISO performance

It’s a widely held belief that in low light the 5D trounces the GH1. But it is simply not the case. In fact the GH1 often matches or produces a punchier image, especially at ISO 800 – and it’s not even much nosier, usually.

However, moise manifests itself differently in different parts of a shot. On skin tones and city skies, noise is most noticeable and here is an area where the GH1 looses to the 5D Mark II.

Because the GH1’s sensor isn’t delivering quite as much colour data at ISO 1600 to the image processor, and because the codec has a lower bit rate, a lot of the GH1’s image subtleties are thrown away in the compression process. It’s not that the sensor isn’t detecting low contrast areas, it’s just that the signal is weaker compared to the large light gathering monster of a chip on the 5D Mark II. To improve the GH1’s low light performance it would have to be a 50-50 improvement in both the sensor and codec. In other words, it’s not just the sensor.

Shadow areas get less excited on the 5D Mark II. They have lower noise and are more baked in. But considering the sensor size difference the GH1 does very well in terms of noise at ISO 800. Where there are subtle variations in shade, or pale low contrast flat areas of colour the 5D’s footage maintains more detail, less noise and banding than on the GH1 – but picture profiles effect this quite a bit. At ISO 1600 the 5D has a clear edge on underexposed low contrast scenes and shadows, but if the scene is relatively bright and slightly overexposed the difference between the cameras is minimal, even at ISO 1600.

The saving grace of the GH1’s more contrasty electronic eye is that you can usually use ISO 800 where ISO 1600 would be required to get a bright punchy image on the 5D Mark II. And as you can see from the video, all of this serves to not really effect the image all that much. The GH1 looks fine, and in some cases punchier with better colours and a less muddy look (though some of that is down to white balance). Quite an achievement against the low light monster then.

Post production

Whilst it’s important to note that this isn’t related to my observations on brightness or contrast above, the other aspect of the test worth mentioning is that on my Macbook Pro, the blacks and low-end of 5D’s raw H264 footage is horribly crushed. Even after converting to ProRes, the shots are much darker than the GH1’s footage, as viewed in Final Cut Pro 7 or Quicktime X, or any other software. Some kind of funky hidden gamma / tone curve thing is being applied and I am amazed it still hasn’t been addressed by Apple.

But the exported footage in XDCAM EX format actually sees the GH1’s footage crushed slightly and the 5D’s footage becomes fine! Why are they handled differently even after being transcoded to exactly the same format – ProRes LT?

Go figure! I think the final output in XDCAM EX format, 1920×1080 24p is pretty much representative of both cameras though, much fairer than ProRes or H264.

Colour

At ISO 800 I have no issues with the GH1’s colour saturation. It’s fantastic.

At ISO 1600 you loose quite a lot of low contrast colour saturation on the GH1. For example if you have a single light source illuminating a red brick wall half way down an alleyway, the 5D will pick up more red, further down the street. The GH1 will show a browner tone, and cut it off to grey or black sooner down the alleyway.

Sometimes the 5D’s colours can look really natural, and sometimes they look a bit weird and neon – especially greens. It’s automatic white balance by the way, is simply horrendous. I was never able to get acceptable results either on auto or manual, under the Taipei street lighting. I have no idea what kind of light they use over here, but it sure as hell turned both cameras very green in the manual white balance K mode, either at the coolest or warmest manual setting. Very odd!! So I had to rely on auto and the 5D’s is absolutely rubbish.

Lenses

The 5D is more demanding. I used the lenses at the same setting on both cameras and the same shutter speed of course. The Zeiss Flektogon 35mm F2.4 lens used on the 5D Mark II really is astounding. Much less flare than the 20mm Lumix F1.7 and the Lumix is also a very good lens!

The Panasonic was set to F2.5 and the Zeiss on the 5D was at F2.4… it was perfectly sharp wide open with minimal chromatic aberration. It’s a contrasty lens, any lack of contrast in the 5D shots is not down to the glass.

The Lumix 20mm on the GH1 is equivalent to a 40mm lens on full frame. The 5D Mark II needed the extra 5mm in wide angle coverage from the Flektogon however, because it’s sensor is 3:2 whilst the GH1’s is native 16:9 – it packs in a bit more than a normal 4/3rds sensor does (like the standard GF1).

Price

You already know the answer.

Usability

A large metal body makes you feel like a pro. But it doesn’t save your back trying to view the camera at a low angle with a screen that doesn’t rotate. It doesn’t help enable AF in video mode or a complete blank-out of live view when one-shot AF is used and it isn’t capable of giving you a built in video EVF and continuous live view. It also adds to the size and weight of the camera. So for the image conscious who would prefer their camera to look less like a toy and more like a tank, the 5D is the best choice. For filmmakers, especially for documentary or run-and-gun, the GH1 wins in the usability stakes hand down, something which hardly anyone ever gives it credit for.

But the world is as it is and I think a professional looking camera is important for professionals – which is why almost all of them use Canon HDSLRs.

The Question Marks

1. Does the 5D have a softer image?

It’s also capable of a sharp one. This is all down to picture profile settings and depth of field. People tend to turn the sharpness down to avoid aliasing (and due to what Tim Smith suggests looks best) but I think this hurts the image more overall. That said, the GH1 certainly resolves more detail from the same lens due to it’s densely packed sensor and crop factor. Take a picture of the moon with both cameras and a 300mm lens and see which wins.

It’s important to view the right area of the image when analysing sharpness. A background object which is further back than the object in focus will always look sharper on the GH1 because the depth of field is deeper on a smaller sensor and shallower on a larger one. That’s why the wall and tree behind the car in shot 6 looks so much sharper on the GH1.

2. Does the GH1’s colour saturation drop when viewed over HDMI?

In-camera footage played back via an HDMI cable from the camera itself really does not look a patch on the 5D. When you transcode everything, combine footage from both cameras together and view it via HDMI from your laptop, they’re equally matched on colour and the GH1 has better contrast. I really do think the HDMI output on the GH1 makes large compromises and it should not be used to benchmark the camera’s performance.

The Winner

In summary, these cameras are more closely matched than I expected. It’s a constant frustration of mine that all the best HDSLR footage is made by the best people almost exclusively with the 5D Mark II. Panasonic should have been much more proactive in pushing the GH1 towards the professional filmmaking community rather than into the hands of consumers, many of whom will just video their cats!

Yet it’s that consumer orientated approach that gives the GH1 some of it’s best features. It’s quicker to shoot with, it handles better, it’s smaller, lighter, more fun to use and you can get a head-on view of the screen from just about any angle when shooting on sticks. It’s video optimised kit lens is ideal for run-and-gun, street filming or documentary style videos / live music videos.

In terms of image quality, the 5D Mark II maintains very little of it’s full frame sensor advantage in video mode. With better video processing the 5D Mark II’s sensor would easily beat the GH1, like it does in RAW stills mode. As it is, the difference is actually quite small. Personally I think the general aesthetics of the GH1’s image are actually more cinematic, but that’s open to subjectivity. The only obvious concrete advantage the 5D Mark II has over the GH1 is it’s low light performance at ISO 1600 and above.

In terms of the Barry Greens (downsides) the GH1 has a lower bit-rate codec which sometimes falls apart altogether and the 5D Mark II has moire. The GH1 has banding, the 5D Mark II has aliasing. The GH1 has mud, the 5D Mark II has a 12 minute shot limit. The list of Barry Greens go on and on. But something odd is happening between one GH1 model and another. Some report that mud is a none-issue, much to the annoyance of people who are absolutely plagued by it. My first GH1 produced codec break-up upon so much as a finger flick of movement, even on sticks. Maybe it missed it’s mum in Japan. My 2nd more recent GH1 hardly ever produces the dreaded ‘mud’ breakup of the image, and hardly ever produces detectable banding. It’s a Japanese model too, so there is surely a factory Q/A issue and variation in electronics performance. This is partly due to the new factory coming online in Japan last September and the later models have a better hit rate and a slightly better performance peak – but the odd dog still was let out and are still around. If you have a bad model, try shooting without the kit lens – this lessens the problem.

The 5D’s renowned aliasing is present in every shot. But I honestly don’t buy Barry Green’s line about the HDSLRs being akin to ‘Digital Bolex’ cameras that resolve a standard def image in terms of resolution and lots of ‘fake resolution’ which is actually just aliasing. Yes the detail can be fatiguing rather than smooth and well resolved, especially on the 5D, but the problem is… whatever the technical test says, the viewer’s eye is king and these cameras look extremely ‘HD’ to me. The truth is that so far in digital video technology, cameras have always produced less detail than the per-pixel maximum, it’s no different with these.

The difference is that the 5D Mark II and GH1 feature extraordinary sensors for video, that deliver an overall image so powerful that even the obvious flaws can’t bring them down. You can mix this footage with RED ONE footage and have a seamless progression, nobody in the audience would even care.

Whether you buy a 5D Mark II or a GH1 is a choice which seems more important than it actually is. I don’t think anybody who watches your finished production will notice the difference quite frankly. The difference is interesting in a technical way, but pretty irrelevant from an artistic point of view.

Wait until we get DIGIC 5 and proper video codecs, then the HDSLRs will really take flight and be completely untouchable by the likes of RED.