Looking for a 3D camera

[Giroud Francois]

TD10 , TD20 and other consumer level cameras are just useless if you said "pro".
Just because they are not shooting 24p, and because is almost the only way to distribute stereoscopic movies: on blu-ray , encoded in MVC, currently only supported in 1080p24.

The NX3D1 (awfully expensive since it is more or less the same camera than the TD10 for about 3 times the price) is a solution.

[Paul Newman]

Be surprised then, this "nonsense" comes from the Sony website and my TD10 handbook,

Paul :-)

[Pavel Houda]

Everyone can be right in this case. If the farthest point is not much further than the closes point, you can violate the 30x (or 60X or whatever) rule and fix the video in post at the cost of final resolution, or in camera, if the sensors are sufficiently oversized. The viewing screen size matters, and the manufacturer's specs usually take the most favorable case, while some of the feedback deals with more conservative (realistic?) case. Add to it that most of 3D cameras can shoot 2D as well, and the closest focal distance maybe explained. Unfortunately that makes general rules incorrect in certain situations.

[Paul Newman]

I use Ocula from The Foundry, within Nukex -check out what it can do here: OCULA | Features | The Foundry

Even the tools within Edius 3D manage almost everything except depth mapping, which is relatively simple in Nukex.

Paul :-)

[Wolfgang Schmid]

All these rules tend to manage that the disparty does not become to high. But that is no hard limit - it is something that differs in the perception of different humans. We will agree that the stress from 3D should not be to high - but the limit is a little bit different for different people.

It will also depend how far the farpoint is away - if the positive disparty becomes an issue or not.

And it will depend how we adjust the camcorder - with convergence to the object/nearpoint, or without convergence.

So my feeling would be that the 500mm are to small maybe. But 800 mm is to large maybe. But as said, it depends. And no handbook and no calculation tool will help us to overcome the fact that this is something that can be calculated - but that perception will differ for different people. And yes, we should tray to stay conservative here.

[Pavel Houda]

I just placed a test of closeup video on YouTube taken at very close distances at the NAB 2012. Played it on 65" TV, it seems to be o.k. The background would be too diverged on movie screen, but then I didn't try to minimize the functional depth either, so it could be pushed, if taken with that in mind. The clips were taken by the Sony Bloggie 3D. I cannot find the interocular specs anywhere, but I measure it at 20mm. It is strange that the manufacturers do not show the stereobase specs very often. They are probably scared that ignorant buyers will think that smaller is worse. Here is the clip:

Stereoscopic 3D Closeup Test of Sony Bloggie 3D - YouTube

. Of course better cameras and operators could do a better job.

[Mark Rosenzweig]

"Be surprised then, this "nonsense" comes from the Sony website and my TD10 handbook,

Paul :-)"

No, the "nonsense" comes from thinking that closest *focusing* distance and the minimum distance for appropriate 3D are the same. The same TD10 manual also tells you the minimum distance for 3D is 3-4 feet. The TD10 is a 2D camera as well, so closest focusing distance is certainly a relevant spec. :-)

[Alister Chapman]

Originally Posted by Wolfgang Schmid

All these rules tend to manage that the disparty does not become to high. But that is no hard limit

There is a hard limit and it comes at the point when you are forcing your viewers eyes to look apart. This is not natural and almost everyone will find it uncomfortable and disturbing. This point can be calculated provided you know the viewing size. While other than this hard limit there are no fixed limits, most people will find anything with more than 20% total disparity difficult to watch.

The 30x(60x) rule is a guide, it's not a hard and fast rule, but in most typical situations it makes a very good starting point, is easy to understand and easy to implement. If I choose to abide by this rule, then it is rare for my 3D material to not work. If I ignore the rule then much of my material is uncomfortable to watch.

[Wolfgang Schmid]

You should not even come near to 20% - at least not if you define that as % of the monitor width. With the 70 minute rule you end up with 3 or 3.5% of the monitor width, as maxium disparity. And you should stay on the conservative side - it you use 2% or 3% or 3.5% increases the stress, but you are still far away from what you point out.

The best 3D productions that I have seen are with low disparity, shoot with parallel optics, and are adjusted in the postproduction to a low numbers of the negative disparity. What reduces the danger of ghosting.

It is a funny thing: we had the 3Dimensionale the last 3 days here in Vienna, with a lot of presentations. Any my impression as somebody who makes both shooting and editing has been, that in most cases you are able to see in the first few seconds if a production is good or not. Even within the 3D community, there are still some (few) productions where you see that they have not understood the basics of disparity.

[Alister Chapman]

But Wolfgang in response to my first post about the 30x rule you posted that there are no hard rules or absolute limits and everyone 's viewing experience is different.

Now you are saying that we should stick to a maximum of 3.5% monitor width for positive disparity, so which is it? Can we all shoot whatever we want provided it looks OK to you or should we follow the guidelines that have existed for quite some time now and help produce S3D that is viewable by the majority of people?

Yes 20% is extreme, but on a small screen like a youtube window, it can be viewable and I'm talking total disparity, not just positive. Most people won't tolerate anything beyond 20% because of the way our own visual system starts to behave in the real world when we experience more than about 20% of positive disparity. This is normally only reached when we are looking at something very close to our eye's, for example reading a book or if you bring your finger close to your eyes. At this point our own visual system breaks down and when we look at a very close object our brain gives up trying to fuse the background and instead presents us with an unfused double image. This is the extreme limit of what our brain can deal with. I'm NOT however saying that is how to shoot, just pointing out that there are contrary to your original statement there are absolute limits.
All these things can be calculated very accurately, there is actually very little variation between most adults interoccular distance, FL and HFOV and as a result you can figure out the optimum settings provided you also know the exact viewing conditions, but very often we don't know how things will be viewed so generally have to err on the conservative side. Yes, different people may perceive the scene slightly differently, but the calculations allow us to present everyone with the optimum images.

As well as disparity we should also consider roundness. While if you have an extremely shallow scene, you can shoot with a wide interaxial and not end up with excessive disparity, the filmed scene will have exaggerated depth and objects will appear to be deeper than they actually are. Very bad news for actors, balls and any other spherical objects in particular. Of course as you know the reverse is also true, but we tend to be more forgiving of artificially flattened objects than artificially deepened objects as we often see three dimensional objects that appear flat in our every day lives, when they are too far array for our own interoccular to resolve.

The bottom line is that the IA of a camera will limit how close you can sensibly get to your subject, get too close and you may have excessive disparity and/or you subject will have exaggerated depth. The 30x rule is a good starting point for evaluating how close you can get.

Software like Ocula has a place, but you have to consider that it is largely dealing with manipulating flat planes in a 3D space, so if you shoot at one IA and then manipulate the disparity you will end up with some depth distortions, for most amateurs the $10k price of Ocular put's it out of reach and it's better to shoot correctly in the first place rather than try to distort a pair of 2D images to fix a 3D problem.


In case you don't know me, I am the designer of the Genus Hurricane rig, one of the biggest selling professional 3D rigs, with over 100 units in use on everything from features to corporates. I consult for S3D movies, produce S3D cinema commercials and documentaries for 3Net, BSKYB, Nat Geo and many others. I will be one of the instructors at the Seoul International 3D fair and have run over 20 workshops around the world. I do have a pretty good idea of what works.

[Wolfgang Schmid]

Alister, I think that we talk here about indiviudal stress perception. Somebody may feel significnat stress with 3%, another one with 3.5% disparity - and who is right? To my opinion both are right, because it is not a hard limit but a limit driven by perception.

But I agree with you that we are talking about total disparity, I agree with you that divergence would be painfull for everybody, and that the size of the presentation display is important (and that a small display may require even more disparity). And even some other parameters.

In terms of calculation - every calculation will be done based on some basic assumptions. For examle, that the 70 minute rule (or some call it 90 minute rule) is the limit that we assume. You will end up with precise results, but since that is based on a little bit unprecise assumptions the results can be rough a little bit too. The best that somebody can do is to handle that in a conservative way. But for sure we should do some calculations especially if you design what you are doing - but beside that it is also true that there are a lot of very experienced 3D guys who never ever use the calculators.

What is the better way? Well I think it is also a decision about how somebody wants to work, and in a commercial environment like yours you have to protect yourself against risks in a very different way. I really agree about that. There are other guys who run 3D productions for fun, and for the user of a small TD10 it may be less important to use a professional calculator like you will do. Both is fine I think.

In terms of increased IO I really have a little bit mixed feelings. We had that last 3 days the 3Dimensionale here in Vienna, beside some workshops where I have run a presentation about both consumer camcorders and tools like Vegas we have seen a lot of presentations. Great presentations. And for sure some went into the range of hyperstereo, by increasing the IO to such an extend that the end up with a huge depth bracket but also with a signficant amount of liliputism. I am not sure if that is fine for everybody, but we are talking also about an art.

[Alister Chapman]

Hyperstereo does not have to have a large depth bracket. I've done some deliberate Hyperstereo work where the desire has been to create a distorted 3D space, everything from cityscapes to the Northern Lights where the IA was almost 2km, but that is not what is being discussed here.

I rarely use a stereoscopic calculator these days. Probably the only times I use one is when I'm in pre-production and choosing lenses or working out how best to shoot a scene. The rest of the time I shoot to specific sets of guidelines based on the target audience. But a calculator can help someone new appreciate the relationship between IA, FL and disparity and roundness.

[Neil Richards]

I think you're both right, but you're not mentioning the need to take into account the time duration of the 3D you're watching or the depth of the scene you are shooting.

You can break ultra-conservative 1-3% disparity guidelines if the object/s are not in shot for very long, like someone walking past quickly or towards the camera before a cut. This can, and often is, used for dramatic effect.

Background disparity, which may remain in view for some time, does need to be more carefully controlled as this is what is most likely to lead to eyestrain in the longer term.

Some judgement needs to be made as to what you can or cannot get away with depending on the shot and the subject.

Alister shoots storms for a living amongst other things so I suspect he's in the great outdoors a lot shooting to infinity. This will need very careful control of the far disparity and a more conservative approach. Shooting a drama in a front room with nothing more than 10-15m away allows you a different set of trade-offs with the depth budget.

These factors, plus intended screen size, etc all lead to the conclusion that you can't slavishly apply any one single rule - in the end you have to monitor and check and see if it works ok, preferably at the time you shoot. If you can't do that, then err on the side of caution.

One thing I really liked about "Hugo" was that Scorcese broke the "rules" - some of the disparity was enormous by current cinema standards - and audiences didn't run screaming from the cinema. The scene where Sacha Baron Cohen as the Stationmaster leans towards the audience must have been 8-9%, approaching theme park levels, a bit too much in my opinion but it worked and had the intended effect.

I have seen 3D which met the "rules" (e.g. for Sky 3D broadcast) which looked awful and 3D which broke the rules look great. There's some artistic judgement to be exercised. The "guidelines" are your starting point, ad lib from there.

Cheers

Neil
still trying to get it right

[Alister Chapman]

Originally Posted by Neil Richards

These factors, plus intended screen size, etc all lead to the conclusion that you can't slavishly apply any one single rule - in the end you have to monitor and check and see if it works ok, preferably at the time you shoot. If you can't do that, then err on the side of caution.

But only if you are able to monitor with the same size screen as the audience will have. Change the screen size and the parallax/scale ratio changes and the 3D looks very different.

Hugo did break many of the conventional rules, but they did it by having much of the film occurring in negative space. That's fine on a big cinema screen where many of the edge violations go largely unnoticed, but it will be interesting to see how that works on the small screen. Perhaps it will be fine as they will be able to push the space back into positive with the attendant increase in positive disparity which may benefit small screen display.
I agree that the "rules" can be bent and twisted, but tread very carefully. Your example of using extreme positive disparity prior to a cut is normally to be avoided as it creates an issue with the incoming shot and the time it takes the viewers brain to re-map the depth. This is a big cause of 3D fatigue. It might add dramatic effect, but the effect might break the 3D illusion across the cut, which then momentarily distracts from the story telling and instead of helping tell your story it just becomes a poorly executed gimmick.

[Wolfgang Schmid]

I am also keen to see Hugo when the 3D BD is available - I read at Amazon.de that this will be aroung August only. :(

For the point that somebody walkes through the video and generates a strong positiv disparity - I tend to try to correct that in the postpro similar to Alister, by moving that to positive disparty. The only question is if that will increase total disparty beyond the limit that is acceptable.

And yes, 3D must support the story and not try to substitute the story - that is something that never works. For sure not in a professional production, maybe a little bit in consumer videos where the story is more about things like "show us grandma". :)