Review and comparison

[Nick Wilcox-Brown]

Granted, the linked image you refer to is more what I would have expected from such a camera, but (perhaps naively) I expect an owner to know how to get the best from their own camera.

It may be interesting to get together with a couple of cameras to do this properly if you have an EX? I can probably get hold of an XF with some notice.

[Alister Chapman]

Your on! Love to do it.

I don't think it's a case of finding a "winner" as both are clearly great cameras, each with strengths and weaknesses. But it would certainly be interesting to explore the limits of both cameras. I've always got a couple of EX's to hand.

[Peter Moretti]

Alister,

You said:

"Alan Roberts tests show both cameras to be at the Nyquist limit for 1920x1080 cameras, so both cameras are producing images of the same resolution. There should be no difference in image sharpness other than differences introduced by differing detail correction settings. The EX's detail settings, while different to those of the XF305 should not be making that big a difference."

While I am very skeptical that the EX-1 was correctly shot (as I mentioned in this and another thread) conclusion you make above is not entirely accurate.

The zone plates do not indicate real use resolution for two reasons. First, they are black and white and hence do show differences in chroma sub-sampling. And indeed, the Canon has 4:2:2 instead of 4:2:0 for the EX-1.

Second, zone plates do not stress the codec the way a real image might. Since codecs do what they do by removing high resolution detail, it is possible that an image with a lot of detail (and/or movement in the case of Long-GOP) will cause one codec to remove more detail than the other.

If these two points were not true, products like Convergent Design's nanoFlash would not have much benefit.

All that said, I do think the EX-1 was not properly shot. And if you guys could come together for a real head-to-head, that would be extremely helpful.

Thanks much.

[David Heath]

Originally Posted by Peter Moretti

Second, zone plates do not stress the codec the way a real image might. Since codecs do what they do by removing high resolution detail, it is possible that an image with a lot of detail (and/or movement in the case of Long-GOP) will cause one codec to remove more detail than the other.

Sorry Peter, the codec is irrelevant in this context. The tests being referred to make no account of codec, they are purely of camera front end. As the first paragraph of the report says:

It records MPEG2-compressed video,but all the tests were made using the HDSDI output, captured uncompressed and analysed in software.

[Peter Moretti]

Right, that's exactly my point. The tests don't take into account codec, but real world shooting w/ both cameras does.

The comparision footage of the car was not captured uncompressed out of each camera's HD-SDI port. So when recording in real life, codec matters. And a camera that has a very good zone plate does not mean you'll always see that same level of resolution in material recorded to the codec the camera uses.

[David Heath]

In this case, I wouldn't expect codec to make any big difference of the type you're talking about. 50Mbs XDCAM-HD has been described as "quasi-transparent" in the EBU tests, and whilst 35Mbs is not fully approved, the discrepancies are only likely to show up in heavy post work, and/or subsequent recompressions. I wouldn't expect the differences between the 35 and 50Mbs codecs to be visible on raw footage under normal viewing conditions, given the same source.

I believe Alister himself has done tests using the same camera front end recording the same scene to both codecs at the same time (internally and via a nanoFlash). From memory, whilst the 50Mbs codec is better, the difference is subtle and barely visible under normal conditions.

[Alister Chapman]

The resolution in the images shown, or at least the apparent sharpness, is primarily related to the luma channel. The difference between 4:2:0 and 4:2:2 only really shows up in well saturated colours and dosn't as a rule make the overall image look softer or sharper. I've seen hours and hours of side by side 4:2:0 35Mb/s and 4:2:2 50Mb/s (and even 100 and 220Mb/s) footage as I have been using NanoFlashes on my cameras for almost a year. In most cases it's hard to see any difference at all, certainly not the difference we are seeing here. It's only after grading or heavy post production that the difference really start to show and even then they are rarely as dramatic as this. When they do show up it's rarely a softening of the image but increased artefacts, mosquito noise and blocking. You only have to look at the link to the frame grab I posted and compare that to the car image to see a significant difference.

It really looks to me as though the back focus was off on the EX.

[Peter Moretti]

Alister, you have a lot more experience than I do, so feel the need to tread lightly when I disagree w/ you, LOL.

But even looking at posts in the CD forum, there are quite a lot of frame grabs that show significant visible differences in resolution between the native EX-1 footage and higher bit rate nanoFlash recordings.

I also had the opportunity to talk with one of the designers of Sony's F35 and he said he refers to codecs as the "low pass filters" in digital cameras. For example, if a lens aberration created the visual degradation that lower bit rates codecs cause, the lens would be roundly criticized.

All that said, once again, I do believe the EX1 was not properly set in the comparison. But I do think codec plays a bigger role than you're giving it. Maybe I'm splitting hairs.

[Brian Woods]

Originally Posted by Nick Wilcox-Brown

Granted, the linked image you refer to is more what I would have expected from such a camera, but (perhaps naively) I expect an owner to know how to get the best from their own camera.

It may be interesting to get together with a couple of cameras to do this properly if you have an EX? I can probably get hold of an XF with some notice.

If this was ten years ago when I lived in Bristol, I would love to get together with you guys on this... Oh, well - a day late and a dollar short (again!) ;)

[Alister Chapman]

Some things to consider. The Canon is 1920x1080 4:2:2 while the EX is 1920x1080 4:2:0 so the EX is recording around 33% less picture information. 35Mb/s is about 40% lower bit rate than 50Mb/s. While it's not quite as simple as I have implied here, the difference in compression ratios between the two is not hugely different. With such a small difference in compression ratio the actual implementation of the encoder and the hardware and software decoders will be as important as the bit rate itself. Yes the Canon will be slightly less compressed which is a good thing, but unless there are large amounts of motion within the image, at this level, you really aren't going to see a visual luma resolution difference in the majority of shots. The luma channel is the key channel when it comes to the apparent sharpness of an image. 4:2:2 is desireable, especially with interlace material, but the difference between 4:2:0 and 4:2:2 in progressive material is hard to discern.

4:2:2 50Mb/s is good to have, no doubt about that, but it won't be a day and night difference. If you look at the CD forums you'll see that to get a useful improvement in the image you need to go to the 100Mb/s level, a two fold increase in the bitrate.

[Nick Wilcox-Brown]

Originally Posted by Brian Woods

If this was ten years ago when I lived in Bristol, I would love to get together with you guys on this... Oh, well - a day late and a dollar short (again!) ;)

Sorry too Brian. I have to get hold of a camera - like Hen's teeth right now, but we'll publish it all when done

Nick.