Making a lego camera: Apertus engineers consider modular design
Earlier this week Apertus team made a bit of a splash in the filmmakers community with their announcement of an open modules concept for their upcoming Axiom 4K camera which is also open hardware. We spoke to Apertus's Sebastian Pichelhofer about some of the most important industry's concerns.
The idea behind the open modules concept is to split the Axiom camera into upgradable units that can be replaced as technology moves on, and be able to attach completely new units such as gyro sensor etc. — basically, wherever the fancies will take the industry in 5 or 10 years time.
The specs for Axiom so far look like this:
- 4K resolution;
- Super 35mm sensor;
- Global Shutter;
- 15 stops of dynamic range;
- Cinema DNG Raw Recording;
- Price "well below $10,000".
On the matter of the modular design the team says:
Modularity is of course nothing new: analogue film and photographic (medium format) cameras have been built around systems using this principle for more than half a century.
We want to stress that Axiom will not be outdated anytime soon — it will be built from the ground up as a system that will evolve over time — as we rely on FPGA based designs rather than fixed “set-in-stone systems”.
We and the entire community will be able to deliver new interfaces, codecs, firmware and significant degrees of newer functionalities all via a simple software download to the camera.
It is our intention to use this to create a very powerful ecosystem around Axiom, providing long term support alongside the development of new features.
Engineers from the Apertus team carefully put a disclamer that so far it's only a concept, and things might go a different way in the end. But the idea of a "lego camera" is certainly tempting in the market that badly needs an injection of frugality.
The general trend with open solutions however is that people tend to treat them extremely critically and demand perfection, while being more relaxed towards deficiences of products by well-established software and hardware vendors.
Naturally, the industry's media response to the announcement so far is along the lines of "if the price is competitive, you better make the final product really compelling". But there are a few more questions that need to be asked. So we spoke to Sebastian Pichelhofer, leader of the Apertus project.
Sebastian, do you think it's realistic to create an interface between modules that would last long enough (decades?) and stay backwards compatible even with further revisions?
Definitely, we are looking into a standard for the connector between the modules similar to the FMC connector which is very mature and widely available.
FMC has a high pin count version with 400 IO lines and a low pin count version with 160 lines. So if we enter a thought experiment, each differential pair pin couple can theoretically go up to 2 Gbit/s (per FMC specifications), and we plan to use 2 of these kind of connectors: one at the top and one at the bottom of each module, since this placement is also great for timing, if the processing chip on the open module board sits exactly in the middle.
If we do the math (greatly simplified), with the 400 pins FMC version there would be 800Gbit/s of total throughput available, not accounting for reserved pins, overhead, electrical limitations, etc. I think it shows that there will be plenty of room for future sensor/module upgrades as 4K 16:9 at 4096x2304 resolution with 12 bit at 25 frames per second results in a raw datarate of 2.7 Gbit/s.
The rest is software, and we will make sure that the data transmission protocol is as simple and at the same time as flexible as it can be.
I am not sure, whether it will establish itself as a global industry standard for decades, but we are confident it would have the potential — if others realize that as well... We will see.
Even though this work is still at pre-design stage, do you have any kind of understanding, what kind of modules you'd most likely be shipping as default part of Axiom? Or would you be providing some sort of "build your own Axiom" service for each customer?
What modules we will be able to develop from the very beginning, and what 3rd party module developers will come up with then, will greatly depend on the success of the crowdfunding campaign. But in the end there is no default configuration — that's the beauty of the open modules concept. Everyone can get a camera tailored to his/her way of working and to the kind of projects he/she does.
Even the camera head alone without any modules will already be usable with an external 3G-SDI recorder for Full HD productions, if you don't need 4K or raw. I guess that for a big percentage of projects/customers this would already be sufficient.
What needs to be done, exactly, to proceed further with this concept?
Right now we need to evaluate connectors, dimensions, electronic characteristics of parts, EMI emissions & solutions for shielding, mechanical properties, fixtures, etc. and that way get closer to final specifications for the open modules "standard definition". And at the same time — mature the concept and the presentation.
What do you think is the biggest challenge the project is currently facing? Getting industry's attention? Finding active contributors?
Finding skilled active contributors is definitely a big challenge, but the crowdfunding campaign is meant to solve that problem by enabling us to buy us the worktime we need aka pay the people to get the essentials parts of the camera done in time.
That doesn't mean we will stop development in the open and accepting contributions. It just means we will have a fixed team of contributors dedicating a minimum of a certain time to the project.
So the biggest challenge will definitely be to run crowdfunding successfully.
The feedback we received so far on the plans for Axiom and the Open Modules concept have been very good, and a lot of people support us in what we do. But there is a big difference between pressing a like button and handing us a big sum of money in the crowdfunding campaign. So we will see if we will actually be able to get the money.
At the moment Axiom design is at the alpha stage. The team recently received first PCBs for the prototype:
The next step is to create a memory interface for reading/writing to and from DRAM, storing and retrieving the image data from inside the FPGA. Stay tuned for more news.