Phil Bloom's trip to France w/C300

[David Heath]

Originally Posted by Sanjin Svajger

In my opinion the C300 has the best rez, followed by GH2 and then the F3. FS100 has the least rez out of these four cameras......

I will agree with you that at first sight the C300 LOOKS the sharpest - but you have to ask yourself why.

Best absolute resolution can be one reason - but other factors can make an image look sharp, and I'm thinking of contrast and detail settings.

Look closely at the images you posted, and lets just concentrate on the C300 and the F3 for now. Blow up the section showing the stern of the boat saying "Port of London" and look at the two side by side. First impression is yes, the C300 is sharper. Look more closely and it seems the C300 image has more contrast, and FAR more detail enhancement - but does it have any more real resolution? The "London" is definately readable in each case, "port of" just about with difficulty, and the writing to the lower right is illegible in each case.

But the lower right writing in the case of the C300 looks as if it's had a black line drawn round it - a clear sign of too much detail enhancement IMO, whereas the F3 looks set up pretty well.

And that's the problem with "real life" tests, even when just concentrating on a single performance factor. (In this case, resolution.) There are too many variables which can change camera-camera - not just factors intrinsic to the model. Detail settings? Contrast? Lighting? You draw conclusions based on setup - not on the cameras themselves.

And that's why it makes sense to use charts and controlled conditions if you're trying to do any sort of comparative testing, and for resolution the best by far are zone plates. The resulting patterns point to intrinsic resolution alone. Altering detail level, contrast, (even focus, to an extent!) will change the INTENSITY of the pattern - but not it's intrinsic geometry.

None of the above should be seen as criticism of the C300 - far from it. But I'm pretty sure that with a bit of effort at equalising some settings I could make the C300/F3 look very similar - and both better than any of the others shown there.

[Sanjin Svajger]

Yes, I agree. This so called "real world" tests are interesting but they should be more controlled. Otherwise they can give people wrong opinions. I realize that contrast helps alot to the perceived sharpness of the picture - that's why I added pictures with added contrast (albeit done in a crude and quick way...). C300 looks like it could have some digital sharpening there yes. But that text could also have a black outline - which is the case here I think. The F3 and C300 both did a very different job of capturing the blue colour of that boat. The F3 picture has a far more saturated blue compared to the unsaturated blue of the C300 and that's why the black outline probably doesn't show that much in the F3 picture. Maybe it's also down to the difference in the chroma rez? F3 is 4:2:0 and when looking at pictures in a way we are looking at them now such a difference should become apparent - don't you think so?

I just noticed that the GH2 is about 15% tighter compared to other shoots. Mr. Bloom did explain this in his video - just remembered.

[David Heath]

Originally Posted by Sanjin Svajger

Yes, I agree. This so called "real world" tests are interesting but they should be more controlled. Otherwise they can give people wrong opinions.

Exactly so. You think you're comparing two cameras - but you're actually comparing two different lineups!! I could take two identical cameras, line them up differently, and similar comparisons would make them seem further apart than different camera models! That's why I feel charts can tell you so much more than all these "real world" tests - as long as they're used accurately.

Originally Posted by Sanjin Svajger

C300 looks like it could have some digital sharpening there yes. But that text could also have a black outline - which is the case here I think.

No, I disagree, you can see the signs of high levels of detail enhancement (for the C300) in other areas of the picture - for example the gunwale immediately above the writing. I think the F3 is doing a "natural" job of capturing to the limit of it's resolution, the C300 seems to have very comparable res ability - but the detail is adding the "outlining".

Look also at the blue on white writing on the side of the cabin - in each case you can just read the top line, and in each case you can't read the (slightly smaller) bottom line. My conclusion is that in real resolution ability these two cameras are similar (and the best of the bunch) and the differences seen are down to line up.

Originally Posted by Sanjin Svajger

The F3 and C300 both did a very different job of capturing the blue colour of that boat. The F3 picture has a far more saturated blue compared to the unsaturated blue of the C300 and that's why the black outline probably doesn't show that much in the F3 picture. Maybe it's also down to the difference in the chroma rez? F3 is 4:2:0 and when looking at pictures in a way we are looking at them now such a difference should become apparent - don't you think so?

From other images I've seen, 4:2:0 v 4:2:2 by itself doesn't give anything like the differences we're seeing here, so I'm still inclined to think the main difference is amount of detail settings. The trouble is that there are just too many variables between the comparisons to be at all scientific. Look at the FS100 for example - in that case, in your frame grab, the writing on the cabin side is completely burnt out. But is that due to differing exposure? Differing gamma setup? Intrinsic lack of highlight ability? Differing lighting? It's impossible to say.

As far as the C300/F3 colour differences go, then again I'm inclined towards thinking it's a matter of exposure/white balance/lighting/gamma setup etc than colour sampling.

[Sanjin Svajger]

Oh whell... If anything this test atleast showed that the GH2 with its hack is capable of capturing good images. Untill now I wasn't really acquainted with the GH dslr line. So the picture quality or the lack of moire and aliasing suprised me.

What's suprising for me also are the overall "low rez" pictures of all of this new large chip cameras. I expected that the C300 with it's great chip ( no bayer and all that) whould resolve atleast 900 lines of rez but if it's unpar with F3 that would come down to 800 or less maybe...

[Alister Chapman]

You don't want really 900 lines of res or more as that would mean aliasing. The theoretical maximum for a 1920x1080 camcorder is 960 lines and thats with a perfect system. Anything more means aliasing. As camera systems are not perfect with no such thing as a perfect Optical Low Pass Filter, good cameras will have resolution less than around 900 if they also have sensible aliasing performance. If your talking of MTF 50 (which will yield a lower resolution figure than finite resolution) then 800 is about right for a 1920x1080 camcorder.

Contrast and micro contrast are more important than resolution alone. The vast majority of people will perceive a low resolution image with higher contrast as sharper than a higher resolution image with lower contrast. That's why in camera detail correction works so well. It adds a black or white edge to existing edges, by doing this you increase perceived edge contrast, which makes the picture look sharper.

[Brian Drysdale]

I suspect there are too many variables going on in these particular tests. Phil Bloom hasn't really had time to get to know the C300 and there could be a number of factors that could affect how sharp the images look. Differing camera menu set ups can have quite an effect, so perhaps these tests are best regarded as snap shots.

Controlled zone plate tests at least cut down on some variables (including changing location lighting), although the resolution is only one factor in selecting a camera. However, it seems to be the one item that appears to create the most friction.

[Nigel Barker]

Originally Posted by Sanjin Svajger

Oh whell... If anything this test atleast showed that the GH2 with its hack is capable of capturing good images. Untill now I wasn't really acquainted with the GH dslr line. So the picture quality or the lack of moire and aliasing suprised me.

This test doesn't show anything about the hack. These are ungraded static shots that will not show any of the supposed advantages of the hack (better rendering of motion, increased shadow detail, better handling in post for colour grading etc). The hack doesn't increase the resolution of the sensor. Resolution of GH2 video is already outstanding for the price & better than the AF101. The GH2 clips would have looked just the same with stock firmware.

[David Heath]

Originally Posted by Sanjin Svajger

What's suprising for me also are the overall "low rez" pictures of all of this new large chip cameras. I expected that the C300 with it's great chip ( no bayer and all that) whould resolve atleast 900 lines of rez but if it's unpar with F3 that would come down to 800 or less maybe...

Alister's sort of answered it already, but I'd say that the F3 and C300 have more or less as much resolution ability as the 1080 system is capable of handling. I can already hear the cries of "time to move up to a better system then" building up, but for the foreseeable future, at least for home viewing, then home ergonomics, viewing distances, and the abilities of the human eye mean that practically there's not a lot of point in most cases.

It's only a couple of years since a lot of people had a "720's good enough" attitude (including the EBU), which I think was wrong, but generally I'd have to agree that anything over 1080 is probably overkill for most uses. Certainly a law of diminishing returns.

Alister - I'm interested to know if you agree with me regarding too high a detail level on the C300 in these tests? And regarding the feeling that the C300/F3 look similar regarding intrinsic resolution?

I'd also be hesitant about drawing too many conclusions (at least from Phil Bloom's examples) about the GH2 owing to the image magnification effect of the smaller sensor, compared to the s35 cameras. It's making certain features larger - hence more detail can be seen in them - but that's not the same as having the same or better resolution. The only thing it can be easily compared with in these examples is the AF101 (as the same size sensor) and it may be surprising to see just how much better it is than that. (You may like to add AF101/GH2 screengrabs to your examples, Sanjin?)

[Sanjin Svajger]

Originally Posted by David Heath

The only thing it can be easily compared with in these examples is the AF101 (as the same size sensor) and it may be surprising to see just how much better it is than that. (You may like to add AF101/GH2 screengrabs to your examples, Sanjin?)

GH2 appears to have more rez than the AF101 judging from these stills...

[Sanjin Svajger]

Originally Posted by Nigel Barker

The GH2 clips would have looked just the same with stock firmware.

The implementation of the AVC codec in the GH2 is as I understand it not the best. And I higher bitrate should help with the muddiness of the picture should it not?

With the hack, what's the colour rez? Is it 4:2:2?

[John Gilbert]

Were the aperture settings constant with these tests? The GH2 seams to have been shot in diffused lighting while the others in sunshine.

[Alister Chapman]

From what I've seen of the C300 I'd say it has a moderate amount of detail correction and high frequency boost or aperture correction at its default settings. I also think there may be a bit of luma aliasing leading to a sharper look. Having said that, I do think it creates a good looking image, but perhaps the best is still to come once cameras escape into the wild and we can figure out the optimum settings for different types of shoots. Both the F3 and C300 should be very close to the limits of what is possible, resolution wise for 1080 so nothing really between them. It will be other factors which determine which will be the most appropriate for specific jobs. Things like, do you need 10 bit or do you need all-in-one, what lenses do you have or want to use. Is 8 bit at 50Mbps perfect for you broadcast shoot or will 35Mbps suffice, perhaps you really want 10 bit uncompressed.

420 v 422, in these shots you would not be able to tell. Especially as H264 is 420. Show me some zoomed in high saturation interlace footage and it would be different.

David, do you see macro blocking on the C300 on the left hand wall close to the water or is it just the light? Hard to tell, F3 looks cleaner, but the light is different, so hard to say.

Guess that sums it up, it will be the way you light, compose and shoot that will be more important than whether you have an F3 or C300.

[Paul Joy]

Originally Posted by John Gilbert

Were the aperture settings constant with these tests? The GH2 seams to have been shot in diffused lighting while the others in sunshine.

Yes the aperture settings were consistent, the same lens was used for every camera. Where possible built in ND was used to manage exposure and the remaining cameras without ND had shutter speed increased so as not to introduce additional glass at the front of the lens. Unfortunately the weather was very changeable and we only had a few hours of light on a very cold day so we made of it what we could. It would have been more scientific to stay indoors and shoot charts, but we were more interested in getting outside and shooting some real world test shots.

As philip says at the start of the video these were not scientific tests at all, we just had access to all of the cameras so decided to spend a day doing some initial test shots. These res shots were the first done with the C300 for any of us so we were not familiar with the camera at all, as can be seen by our inability to find out how to format the CF cards. If anything I'd say the C300 was at a disadvantage as Phillip owns the rest of the cameras and as such has spent enough time with them to know how to get the best out of them.

The result of our inexperience with the C300 was that we shot the res tests with the C-Log gamma curve but the camera was not in it's Cinema-Lock mode which would have been a better choice as it gives the flattest image possible from the camera. We realised our mistake for the later tests though so hopefully those will represent better the C300's flattest images.

For me these were really just a first pass set of tests to see how the cameras stack up against each other without too much effort being given to push each camera to it's maximum ability. After seeing the results it would be nice to move on the the next step and look at the top four with a bit more detail going into getting the best from each camera.

[Josh Dahlberg]

Originally Posted by Sanjin Svajger

Oh whell... If anything this test atleast showed that the GH2 with its hack is capable of capturing good images. Untill now I wasn't really acquainted with the GH dslr line. So the picture quality or the lack of moire and aliasing suprised me.

It's phenomenal value for money. I recently used a client's unhacked GH2 with kit lens on a shoot alongside my 5DII with 70-200 IS II and was shocked to see the Canon comprehensively outgunned (albeit in daylight). Shocked enough to sell the 5DII and pick up a GH2 and some nice primes. Most of the softness/aliasing/moire deficit with DSLR shooting has disappeared, and as I always shoot two system audio sound is not an issue. Now trying to decide if the gain with the C300 in dynamic range, high ISO performance, codec and monitoring is worth the premium for me.

[David Heath]

Originally Posted by Paul Joy

Yes the aperture settings were consistent, the same lens was used for every camera. Where possible built in ND was used to manage exposure ........

I'm interested in the big differences between the AF101 in normal speed and slo-mo. The slo-mo appears to have what looks mostly like very much worse chromatic aberration compared to the normal speed - visible here as blue haloing around white highlights. That looks far more as if it's lens related than camera related..... and as the slo-mo is later, with the sun obviously having gone in, is it not a case that it was done with the lens aperture opened more widely than the preceding test?

Originally Posted by Paul Joy

As philip says at the start of the video these were not scientific tests at all, we just had access to all of the cameras so decided to spend a day doing some initial test shots. ..........If anything I'd say the C300 was at a disadvantage as Phillip owns the rest of the cameras and as such has spent enough time with them to know how to get the best out of them.

Yes, Philip does say that at the start, emphasises it in fact, but inevitably people are still reading a lot into the tests (too much?), and drawing conclusions about the cameras that are really only down to line-up, lighting etc. And that's why I take "real world" tests with such a large pinch of salt. I'd also say the C300 was at a disadvantage primarily down to too much detail level in the lineup (makes it look too "video") - ironically, that's being interpreted as it seemingly looking best as being "sharpest". It seems like the right conclusion (best of the bunch, with the F3 pretty close) being drawn for completely the wrong reason!