Guide: Modifying the Canon RF5.2mm VR lens IPD to provide accurate scale

We are working on a project to modify and improve the Canon RF5.2mm dual fisheye lens. This lens is a key element of the Canon EOS VR ecosystem behind most of the high-grade stereoscopic 180° VR content available today. However, the default 60mm IPD gives scaling issues, making the visuals look less realistic and less immersive. Our goal is to alter this IPD to deliver lifelike visuals with accurate scaling. 

Update: We're looking for skilled camera engineers to make Canon RF52 mods in Europe and the States. We will provide the RF52 lens for you to modify and test. High-quality applicants only. Contact ivan@deovr.com if you have the skills to take VR forward!

We will release our modification videos and blueprint files so everyone can create their own version, and help bring top-quality immersive VR to a huge audience. This is what we have learned so far:

First, take a look at footage recorded on the Canon EOS using different IPDs: 

VS

What is the issue with the IPD on the Canon RF5.2 lens? 

Canon originally placed the lenses at 60mm - the lower end of the average IPD, giving the impression of looking out at the world with a smaller head size. Here’s a diagram of the lens elements for the Canon Dual Fisheye Lens. We realized that the IPD can be adjusted as long as the travel distance remains constant. 

With this in mind, we modified a few components to widen the default 62mm camera IPD to 65 and 66mm. The scale is much better with the 66mm modified IPD, as you can see with the videos on the CanonMod channel at DeoVR linked above.

How is the modded lens different to the Canon stock lens? 

You will notice that the difference in scaling is more obvious the closer the subject is to the camera. Our modded lens gives a more natural and appealing look, much more realistic and lifesize - thanks to the improved IPD.  However, the image from the stock lens shows the model as larger and less realistic - which can break immersion in VR. 

To improve the IPD, we needed to modify four parts, the front plate, the main cover, the two prism holders and the inner ring (which is part of the focus ring).

The difficult part in assembling the unit is due to the fact that the four screws that attach the focus ring to the main cover need to be perfectly balanced. The problem is that there is no hard stop on the screws so they can be over/under turned. If they are offset by even a quarter-turn then the focus will be off.

We tested focus using a dolly, through a set range of distances while comparing it to a stock lens. If the focus is off, the unit had to be disassembled to change the balance on the four screws.

Once the focus is equal to the stock lens, the focus on each side must fine-tuned so that the left and right lenses match. Again, this required filming using the dolly again, through a set range of distances.

It takes an entire day to do the modification, plus the lens focus becomes very delicate, and is no longer waterproof. Another issue is that the prisms are glued to the prism holder with epoxy, making it difficult to remove without chipping the prisms.

Right now, two key problems need to be resolved:

1. A foolproof design for the four screws that attach the focus ring to the main cover, in order to get perfect balance.
2. Replacement prisms so that people do not need to remove the stock prism. This way people have the option of easily reverting back to the stock setup. 

Keep checking back for the latest updates on this important project. A lot of time and money has gone into dealing with Canon’s strange design choice, but we hope to have this fixed soon, to help make modern VR as immersive as possible. Join the discussion on the DeoVR Facebook. 

Latest news: Canon has lodged a patent for a MkII fisheye lens with an adjustable IPD which could fix these scaling issues. This adjustable IPD means it can be adjusted to be wider or narrower, depending on the needs of the VR creator. This would mean issues that arise through modding, such as losing waterproofing and warranty, not to mention the time and money required, can be sidestepped. It's important to be aware that this is merely a patent, so there is no set date for when - or even if - it will come to market.