Why shooting a feature in HD might be better.

Amadou

Amadou

Carbonite Member
Yes. If you drop an HD file into a Scope 2K DCP, assuming you dont upscale, you will have a 2 to 1 Aspect Ratio with small bars on the side which nobody will notice. You will have saved a fortune by not filming in 4k. With the money saved you can buy a camera that handles 422 10bits in HD. The end result will be as good as that provided by any 4k 422 camera and hour editing will be smooth as cutting trough butter. Your final theatrical Aspect Ratio will be the same as 70mm film. Perhaps that is the way to go. Any comments?
 
If you can get an HD camera with little compression this can all be true..

Arri Alexas and Sony F35 cameras deliver excellent HD/2k and are IMO 'good enough' to make feature films and can even be upscaled to 4k fairly nicely

Most Prosumer HD cameras (fs700, CanonC300mk1 EX1 etc) deliver crappy HD due to thier highly compressed codecs.

The usual route to almost decent 2k is to get a prosumer 4k camera ,like the FS7 or GH5 and downscale the '4k' to 2k. I would not personally consider the FS7 or Gh5 to be true 4k cameras.

So what HD cameras produce good HD?

Arri Alexa
Sony F35
And some oddballs to consider or investigate..
Ikononstop
Silicon imaging Si2k
Black Magic 2.5k
BLack Magic Micro
Canon DSLR with magic lantern
Digital BOlex

To save data you could shoot 4k ursa mini, transcode to HD444 and bin the originals.

--

16.9 - 2.35 anamorphics make best use of a 16.9 imager for a 2;35 delivery as they dont waste pixels on black bars.
 
Last edited:
Excuse me but why would you not consider the FS7 a 4K camera? It does 422 10 bits 600 mb/sec Cinema 4K in XAVC I. Also the FS5 does XAVC L 422 10bits in HD. In what way would you consider these cameras crappy in HD or not 4K for the FS7? Just wondering.
 
I consider the Fs7 when shot at 4k to produce resonable HD.

XAVC is highly compressed..

Rough figures..
A 10bit 4k frame needs (4000px*2000px*3colours*1024levels_of_colour*24framesPersecond) bits per second.
If you do the math you will see how compressed 600mbs is.

It is maybe about the data rate you need to create good HD. (Sony F35 Sr Codec is at least this for HD) Uncompressed HD is 1Gbs (https://westcityfilms.com/wp-content/uploads/2015/03/Filmmakers_Handbook-4th-ed_data_rates.pdf)

If you check this out I zoom to 4k on a 1080 timeline - you can see her jumper is full of compression artifacts.. commercially usable? maybe. true 4k? - no chance :)

 
Last edited:
I don't see an aspect ratio advantage with HD; 4K cameras should also support UHD, which uses the same 16:9 aspect ratio as HD, so there's no advantage to one side over the other. I'm not sure that I follow your other aspect ratio comments. The aspect ratio is 16:9, so are you talking about cropping the top and bottom to fit 2:00:1, or are you talking about pillar boxing the sides?

Editing 4K might not be an issue using a proxy file workflow, but you will need bigger HDDs for archiving the original files.
 
Amadou said:
600 mb per seconds is highly compressed to you?
Click to expand...

At 24fps? No. But the 600mbps statement is facetious, that's only at 60fps. Individually the frames are still quite compressed at that rate.

XAVC Class480, which is 480mbps at 24fps, is more reasonable for a mastering codec.
 
I think what Sam is pointing out is that there are varying levels of "quality" in this industry. TRUE 1080p color and resolution is not always a reality even though cameras of all price points say they are 1080p. If you ever get to see real uncompressed output from full raster imagers the difference compared to what we get after processing and compression is quite large. I remember folks back in the SD days saying true uncompressed output from the best cameras looked like HD.

The arrival of 4K at lower price points basically boils down to giving us "true" 1080p in terms of color and resolution. Can this vary in terms of your needs? Sure, but knowing it exists is good as well.
 
Last edited:
Interesting discussion but I wonder if some of the arguments against delivering 4K are simply attempts to justify still having an HD delivery workflow. :)
I suspect the same will happen in the future with HDR, those who do not deliver HDR will find ways to justify it.

A 4k workflow is simply better and gives a better end quality.
 
So I suppose the way to go is to deliver HD by shooting 4k or UHD. This leads me to my follow up: Does 420 8bit UHD reduced to HD give better image quality results than HD filmed in 422 10 bits?
 
Bassman2003 said:
I remember folks back in the SD days saying true uncompressed output from the best cameras looked like HD.
Click to expand...

Not to sidetrack this thread too much, but I used to say the same. In the very early days of our transition to HD, looking at the straight(non-recorded) video signal coming out of my Betacam onto a broadcast/professional CRT monitor looked as good as transmitted HD on a consumer HD set.
 
Amadou said:
So I suppose the way to go is to deliver HD by shooting 4k or UHD. This leads me to my follow up: Does 420 8bit UHD reduced to HD give better image quality results than HD filmed in 422 10 bits?
Click to expand...

This is not easy to answer - I certainly dont know - Ii guess the answer is once again 'it depends' - in a minimal grade situation with well exposed footage the 420 would probably win, but if the going got a little tougher and the footage was log or needed post correction then the 10bit would come into its own.
 
Run&Gun said:
Not to sidetrack this thread too much, but I used to say the same. In the very early days of our transition to HD, looking at the straight(non-recorded) video signal coming out of my Betacam onto a broadcast/professional CRT monitor looked as good as transmitted HD on a consumer HD set.
Click to expand...

The bloke who started me on the filming road has a big Beta Cam - he was a huge fan too. I think the observations of top end SD matching prosumer HD is universal.

THe numbers I didnt bother to work out shows how much compression is occuring - lots, and one can certainly extrapolate that compression damages an image. Sony say 'visually losless compression' with one hand and then peddle the F65 at 1.2GPS on the other. One can guess which looks better. If visually losless existed then the top end cameras would not burden thier users with data for fun!
 
Amadou said:
So I suppose the way to go is to deliver HD by shooting 4k or UHD. This leads me to my follow up: Does 420 8bit UHD reduced to HD give better image quality results than HD filmed in 422 10 bits?
Click to expand...

This depends a lot upon how the chip is designed and how the camera itself downsamples its images. To be open, true 1080p images are made by oversampling which is having a larger amount of pixels than the end resolution. There are good and bad ways to do this which is camera specific. So one has to test with a given camera to see what is best. In general, this should be about even in a best case scenario.
 
Downscaling can be seen, if done correctly, as a way to alter your chroma sub-sampling...so in theory, 4k 420 can be downscaled into 2k 422. But you are still only working with 255 shades of RGB in 8bit...as opposed to shooting 10bit where you have 1000 shades of RGB. But number of shades per bit is not the same as chroma sub-sampling.

To be quite honest, I'm not sure that this line of query is really super productive in the sense that neither the FS5 nor the FS7 is likely to be the "make it or break it" factor for the OP. A tool is only as good as the people that use it, and the workflows within which it is deployed. Just get the cheapest thing for now and hone your craft. Based on the questions and replies by Amadou, it sounds like experience, study and practice will go much farther than either camera. Of course, that can be said for myself and most others as well.

In other words, it doesn't sound like the OP has enough of a handle on being a Cinematographer that the differences between the two cameras will make themselves know (assuming they will be the cinematographer for your film). Again, I'm just basing my opinion off of their questions and comments, I could be completely wrong. And If I'm correct, that's ok too. Everyone has to learn somehow. I'm still learning too. But I'd say you don't want to over-invest into a learning tool. Script-motivated decisions around lighting, composition, locations, production design and many other factors are going to make or break your film way before the differences between the FS5 and the FS7 come into play.

Either way, always good to see folks pursuing their dream, I wish you all the best in this endeavor.

Sorry if this sounds harsh, I'm not trying to be hostile or discourage anyone, just wanted to be honest...and hopefully helpful.
 
Last edited:
The FS7, C300 II, C200, and other cameras with a 4K Bayer sensor aren't True 4K cameras in the same way 2K sensor cameras weren't True HD cameras (manufacturer marketing exaggerated true/actual captured resolution. Now with "oversampled" sensors they are coming clean). First, Bayer sensors subsample information, so there's quality loss right there. The quality loss is greatest when deBayering is done in-camera. That's why for example, the C200 and C300 II RAW looks more detailed when shooting RAW and deBayering on the computer. I would expect C200 RAWLite and C300 II shooting ProRes RAW (with an Atomos Inferno recorder) to be functionally equivalent.

I did a deep dive analysis to figure out true camera resolution, by studying the math: Nyquist sampling theorem, and then doing computer graphics analysis to really understand what is going on. The end result of the discussion, is that Bayer sensors in practice must be multiplied by .67 to get a good estimate of actual, true resolution after deBayering, before aliasing (important note). Here's the thread: http://www.dvxuser.com/V6/showthread.php?356356-Actual-Camera-Resolution-Before-Aliasing

In certain very limited cases, a 4K sensor can capture 4K resolution (e.g. perfect alignment of detail with sensor: not likely in the real world). However that's sampling at the Nyquist limit, and we must sample at greater than the Nyquist limit to prevent aliasing. My initial estimate was between .5 and .67 (averaging .58), and after further discussion, .67 was the minimum threshold. This .67 factor was validated when discussing these comments privately with ARRI, so camera manufacturers have known about this for a long time, just kept out of marketing until now, where oversampled sensors are now becoming common. To get True 4K we need: 3840/.67 = 5.7K minimum. Thus we have the EVA1 (5.7K), Canon C700 FF (5.9), Sony VENICE (6K), many Reds (>= 6K), and recent Sony DSLRs, such as the A7 III sample a full 6K to produce True 4K. Unless shooting RAW or RGB 444, it's still not True 4K, since color is 1/4 resolution if shooting 420, and 1/2 resolution if shooting 422. RGB 444 is what you get immediately after deBayering, so if the on-camera deBayer is decent and the compression is efficient and light, it can rival RAW quality for many applications. For 4K and above (including 6K), compressed RAW (1 color plane) makes sense vs. compressed RGB 444 (3 color planes). The trade-off is editing performance, since deBayer has to happen in real-time (not as much of an issue when deBayer takes place on the GPU). Also, when targeting 4K RGB from 6K Bayer, since we're oversampling, we don't need to be as aggressive getting every last drop of detail from the 6K Bayer. If desiring to punch in in post, doing max quality (slow, non-real-time) deBayer (transcode to ProRes etc.) can make sense.

Some folks are a bit shocked by how much detail True 4K captures, making comments such as "it's too sharp" or "oversharpened". True 4K is not oversharpened, as that would have artifacts. True 4K has another excellent property: little or no aliasing and moire. If the shot/production requires a softer image, diffusion filters or blurring in post (+ regrain) is always an option. It's always an advantage to capture as much real detail as possible, since lens filters and post work can soften the image. Post sharpening can help with softer source images, however real detail always looks better with less artifacts / digital harshness.
 
You must log in or register to reply here.
Back
Top