Sony F3 vs RED?

[Peter Moretti]

Does the BBC broadcast an interlaced or progressive signal? If it's interlaced, it would make more sense why they want 4:2:2 since interlacing halves the vertical resolution.

[Brian Drysdale]

The BBC transmits 1080 interlaced.

[Leonard Levy]

Well maybe I've been too harsh in my attitude towards broadcast specs. If there really is a reason for this I stand corrected - not that anyone at BBC gives a hoot what I think of course.

[Peter Moretti]

IIUC, 4:2:2 was originally designed around the interlaced format. It saves bandwidth over 4:4:4 and makes the horizontal and vertical resolution close to being proportional. B/c interlacing essentially halves the vertical resolution and 4:2:2 sampling halves the horizontal resolution.

Displaying 4:2:2 progressively is something that it wasn't originally designed for. Which doesn't mean it's bad for that purpose. But the improvement over 4:2:0 progressive is surprisingly hard to notice. And it seems odd to value on direction's resolution more than the other's.

From comparisons I've seen using progressive images, the jump from 4:2:0 to 4:4:4 was considerably more visually noticeable than going from 4:2:0 to 4:2:2.

[Leonard Levy]

I had lunch with someone who is often cutting things that are borderline broadcast standards and seems to have been around the block a bit with PBS about it. His take was if you cut and deliver in 4:2:2 unless there are give away shots where the weakness of the original codec is apparent then they don't have any way to tell for sure. They will just be looking for bad video. If you shoot it right you're probably fine. of course he may be wrong and probably has not been dealing with BBC either.

[Graeme Nattress]

Originally Posted by David C. Williams

HDRx will never reach a real 18 stops in any useful way. RED claim 14+ stops in normal mode, yet barely reach 11 in every test I've seen. The current beta version of HDRx is unusable for anything beyond static images anyway.

I spoke with Ted when he was down here, and he said the whole thing was a low priority anyway. I wouldn't hold my breath waiting for it to be a usable mode for anything beyond very controlled situations for a long while yet

The easiest, most reliable way to test DR is to point a camera at a test target with a known wide dynamic range. I'm using the DSC 18 step, 17 stop chart at the moment and it works very well for this, and is clear and easy to read. Simply set the camera so that the brightest step is just clipping, then count the number of distinct steps you can see. Then subtract one to get the number of stops dynamic range. If do this, in a darkened room to avoid any stray light confusing the results, and you can only come up with 11 stops, you are doing something wrong.

HDRx is working well in practise and being actively developed. To say it's a low priority is incorrect. As noted above, Mike Seymour of FXphd has some fantastic examples of what can be achieved with the HDRx, and Davinci were demoing great HDRx support for the forthcoming Resolve 8.

Graeme

[Alister Chapman]

You can make your own DR chart using a large sheet of 0.3ND. O.3 ND is 1 stop, so by building up overlapping layers of 0.3ND each layer is one more stop than the previous layer. Using 18 layers will create an 17 stop range chart. The accuracy will depend on the quality of your ND sheet and any errors are cumulative, so this is not going to be 100% accurate, but it will allow you to compare like for like and deduce which is better or worse. The chart will also need a very even source of illumination from behind such as a lamp bounced off a white reflector.

[Graeme Nattress]

That's a good idea Alister. A stouffer backlit wedge chart is also quite affordable, but only goes up to 13.66 stops, which is no longer enough. Although you could heavily ND one half of it to extend the range somewhat in a similar method to what you describe.

Even illumination is indeed important, as is making sure either the room is dark, or you create a black fabric tunnel between the chart and the camera to ensure stray light doesn't get in. I've used both methods successfully.

Now, it's worth remembering "why a backlit chart" rather than a front-lit chart. Well, the wider the DR of the chart, the easier it is to measure where the camera's range fits, so having a chart with a wider DR than the camera makes life much easier and gives less room for error. A good front lit chart could have a DR of around 5 or 6 stops before you just can't get any more out of the printing process. A backlit chart has basically the practical limit of how bright you can make the light-source without melting the chart (yes, I've melted a chart trying this) and how dark you can make the darkest step, but still let "some" light through. Going beyond 20 stops is rather hard.

Graeme

[Steve Kalle]

Hi Alister,

With the F3 and S-Log, the DR captured from a test chart is a little over 13 stops. BUT, with the compression of the highlights, can you conclude that more than 13 stops is actually being recorded due to the extra 2-3 stops of light that would normally be blown out and another 1 stop of shadows which would be crushed without S-Log?

Maybe Nate can answer my next question: when grading S-Log footage, is it possible to 'expand' the highlights; thus, increasing the appearance of DR? Or has the highlight compression 'compressed' it so much that there is not enough information available?

Thanks

[Brian Drysdale]

Originally Posted by Leonard Levy

. They will just be looking for bad video. If you shoot it right you're probably fine. of course he may be wrong and probably has not been dealing with BBC either.

The guru from the BBC Academy give the impression that they were doing more than just looking at the delivered material during the quality checks. Also, it wasn't just the acquisition codec, but issues from the post work flow as well that they were looking out for.

The BBC commissions many of their HD programmes and they have rejected programmes and insisted on the production company re-shooting them. That's probably not a risk worth taking from the business point of view.

[David C. Williams]

Originally Posted by Graeme Nattress

HDRx is working well in practise and being actively developed. To say it's a low priority is incorrect. As noted above, Mike Seymour of FXphd has some fantastic examples of what can be achieved with the HDRx, and Davinci were demoing great HDRx support for the forthcoming Resolve 8.

Graeme

Just quoting Ted from his Epic tour. That's what he said, low priority in comparison to everything else that is on your collective plates.

Also, referring to Geoff Boyle's test comparing usable stops, testing cameras back to back on the same setup.

[Graeme Nattress]

David, getting HDRx working well was my top priority until I managed to figure it out.

I know Geoff, and I know he likes to do his charts the way he always does his charts, but they're quite useless for measuring DR - for instance, if you take a single chip on the chart, one which just shows clipping and try and track it all the way until it disappears in the shadows, well, it never disappears - it's still there because the range of exposures used (the 7 stop range of the lens, doubled up with a change in exposure) is really not enough to exercise cameras with extended dynamic ranges. On Geoff's test, take the centre mid-grey chip on the chart and track it's exposure. In the first shot it's clipped, in the second one it's not. That is the least clipped chip on the chart that actually clips, thus it can be used to test DR. Down at the bottom that chip never gets dark enough to vanish into the noise floor, hence the whole range was not exercised by that test.

Graeme