Arranging cameras in an array, or rig, is a simple concept but it takes some careful planning to make the most of the gear available. Especially when trying to get the highest quality capture, using as few cameras as possible, all the while keeping the needs of power, data management and rig weight satisfied.
As manufacturers catch up with demand, simple integrated camera systems will become available at any pricepoint, the same way as what has happened to traditional video cameras. In fact, a number of promising systems are already on the market, or soon to be – a list of some of them can be found at the end of the chapter.
As for rigs, the empty frames to which you can attach cameras, there are a few available to buy, but as there are thousands of different camera shapes and sizes available, their usefulness is often extremely limited. One very notable exception are the Hero360 and Freedom360 rigs, made expressly for GoPro cameras. Much of their success can be attributed to the consistency in design between the gopro models, along with the relatively inexpensive nature of the individual cameras (compared to more advanced imaging equipment).
3D printable rigs for GoPro cameras are also available for free, shared by users on Makerbot’s Thingiverse website. Examples can be downloaded here and here.
The benefits of using a commercially established rig is that the orientation used for each camera along with their layout becomes somewhat standardised – and already there are a number of software packages that reflect this by offer templates for stitching. Where the hard plastic mould of the rig is very robust, it isn’t perfect and is prone to vibrations when the rig is moved. Camera shake is a known and planned for phenomenon in filmmaking, but with non-rigid rigs such as these, the shake is unique to each camera and becomes virtually impossible to remove. To be fair, it is an excellent rig, and one that for many people will be their first as they venture into VR-filmmaking, and it does take a hefty amount of vibration to ruin a shot.
For most other cameras there are far fewer turnkey solutions available. Seeing as there are so few productions being done on a professional level, and owing to the cost involved in renting or buying multiple cameras, the rigs used at the top end are purpose built for each shoot. Designed from the ground up to solve only a single project’s challenges.
Important factors for your rig.
The most important factors for rigging multiple cameras are stability, repeatability and accuracy. Templates created for stitching assume a consistency between the different takes, and as such the rig would need to be robust enough to handle multiple setups in a day, and the odd bump or two – without the cameras shifting. An ideal rig would allow for the camera batteries, memory cards and lenses to be exchanged without altering the arrangement of the cameras in any way.
Where possible, a great deal of trouble can be spared if the cameras can be powered externally, thereby alleviating the need to shut down production when one of the batteries starts running low. More importantly, when a faulty battery in one camera in the array stops working, a section of the footage would record nothing. If the area had no action, it can sometimes be salvages using earlier frames but if there was action, an actor for example, or any kind of camera movement – the entire shot would need to be redone. This is a very real concern with GoPro cameras that have a limited battery life.
The cost of external monitoring might be prohibitive, but weighed against the cost of re-scheduling a shoot due to incorrect cameras or a broken battery the cost becomes trivial. Therefore, when possible the feed from the cameras should be sent to an external monitor, or multiple monitors. Blackmagic design’s decklink cards are more than capable of handling the streams from all but the most demanding camera systems. Having a feed of all the camera views, albeit without being stitched, would also be useful for ensuring that planned action does not stray too close to edges, and that primary or pivotal actions are indeed captured as intended.
Adequately allowing access to the cameras once they are mounted is vital, so that lenses, batteries or memory cards can be replaced during filming. Ensuring a rigid and repeatable mounting point is critical.
If a cameras slips or moves during filming, or is mounted at a different angle – the stitch can fail. As each lens is unique, and even small changes can make a difference to the final picture, always make sure that the same lens is placed on the same camera in the same place on the rig. Simple indicators of camera order such as coloured gaffer tape or labels goes a long way in ensuring the cameras are set up correctly.
Data lines and power lines can be bundled together and, if possible, be hidden as best as is possible. But as we are trying to capture absolutely everything in the scene – more often than not, they would need to be painted out during post production. To make the process easier though, keep cables, battery packs and other gear away from complicated patterns, to make their removal easier.
How to arrange your cameras
It is very hard to sketch out a layout for many cameras that accounts for their orientation and overlap with any amount of certainty. The last thing to you’d want is to have the rig all wired up only to discover the panorama can’t be stitched because of insufficient overlap.
To help illustrate the idea behind a successful rig, let’s break down the problem as a simple visual: The world as a sphere with us in the middle, which at the end of the day is pretty much what we’ll be ending up with, a circular view of the world.
If we imagine the camera as a flashlight in the middle of the sphere, with a beam as wide as the cameras field of view, we can easily see how much we’ll be able to capture and can start to get an idea of how the cameras need to be placed.
As we add more cameras and rotate them around, we can see how they need to be arranged. As the lights intersect and overlap, the light gets stronger, and we know that every edge should have a brighter edge – because without overlap we wouldn’t be able to stitch them.
Translating it into a real example is most easily done using 3D software. Virtual prototyping is massively helpful for planning and visualising multiple camera arrangements. Free software such as Blender, sketch-up or commercial software such as cinema4D, Houdini, Maya and 3DS Max can all be used to prototype without building anything until all the angles have been taken care of.
And because the geometry of the cameras are given a physical shape, it becomes far more intuitive to judge whether an arrangement will be able to accomplish the task of shooting 360.
By using an accurate model of the camera, or even just simple shapes that are to scale as stand-ins, it becomes much easier to plan the layout of the cameras. It is possible to test out different camera combinations and lenses without committing any funds. Furthermore, mounting points for tripods and cable management can all be planned here.
It is even possible to do test renders from the point of view of the camera to know very precisely how much parallax error you will have to deal with and which distances are best for your actors and scenes.
Doing it old School
A step by step for arranging cameras is as follows.
Starting from your lens’ field of view, you can figure out how many cameras you’ll need by dividing 360 by it. For this example I’ll be using a Full frame DSLR set to shoot video, like the canon 6D, with a hypothetical 10mm lens.
The 6D has a sensor size of 36x24mm, but when shooting video it get’s cropped to 36 x 20.25mm. At 10mm, a rectilinear lens, not a fisheye, would have an angle of view of 121.89º horizontally and 90.71º vertically. Calculated using Frank van der Pol’s online FOV calculator and changing the sensor to a 16×9 ratio.
As we can’t use 90% of a camera I’ll round up to 3. The problem is that with three cameras we only have 3 degrees of overlap, or 1º per camera. That would make stitching incredibly hard. So in the interest of better control points, we’ll bump it up to 4 cameras. When using fisheye lenses, 3 cameras often suffice though.
The overlap is much better using 4 cameras, but that only solves the first half and we still need to get the vertical portion of the panorama. Using the same process as before we can calculate the number of images needed knowing that a part of the vertical picture is already being captured by our horizontal row. So we only need to fill in the gap left over.
Our camera has a vertical field of view of 90 degrees which means we only need to fill in an extra 90 degrees above and below to capture the the full sphere.
If we add another camera facing up, and one down, the new cameras slot in like puzzle pieces. Again though, we only have a very small area of overlap. Less than 1 º this time.
We could add another set of cameras to increase the overlap, but that would like increase our rig’s size far beyond what is manageable (or affordable). Often, overlap can be maximised by careful rotation of the cameras.
Given that some of our lenses have more than enough overlap, as in the horizontal row. But not enough for the top and bottom, we can split the difference for the overlap by rotating every second camera in the horizontal row onto its side.
Complete coverage, though there are still areas with limited overlap, especially between the corners of all the cameras.
To make the most of each lenses, you can sometimes optimise the layout further by rotating the rig to align the cameras along their diagonal edges. Each lens will have the greatest angle of view between its two opposing corners (top left and bottom right). Doing this will maximise the chances that action will stay localised inside one camera’s field of view, and therefore, away from the seams. The popular GoPro Hero360 rig uses this technique.