In addition, as each camera has its own unique IP address, and in the event that you were facing difficulties, a technical lead can jump straight in there and do the tough stuff for you.
Not sure what you mean by that. It has an IP address just like any other computer (DHCP or manually configured), but it's not necessarily unique. So, you can configure your network to allow it access the internet, if you want.
Now, as I see it, the Axiom's real saviour is the 300fps feature. If it can do that at 4K in RAW then that's a real selling point.
The sensor can do 4096x3072 at 300 fps in 10-bit mode, according to the datasheet. As I've showed
here, we can use a LUT to map this data to 8-bit with minimal quality loss, so a raw stream at 4096x3072 300fps 8-bit would require 3.5 GB/s.
The DisplayPort interface maxes out at 25.92 Gbit/s or 3.24 GB/s (
source), so with a small crop, it should be doable.
The problem is... how do you save that image to a storage device?
- Somebody could probably design an external recorder with lots of RAM (expensive), that could save the image to some slow media afterwards.
- I guess it would only be practical if we manage to implement some serious compression in the FPGA (MJPEG?), but that's not something I expect in the next months from my side (it's not a full-time job for me).
The current implementation runs at 30 fps.
- The dynamic range it's better than a Mini Ursa 4,6K? I think you can answer this better than I can read at Google.
There are some promising experiments regarding PLR modes, but I'm not ready with the calibration. Without HDR tricks, the DR of this sensor is not spectacular at all (to me, it looks very similar to the Ursa 4K images I've analyzed earlier in this thread).
My personal opinion:
Currently it's a just a developer kit. If you plan to design a custom interface for it, or to write custom application code, without having to deal with reverse engineering, it's a nice option.
If you just want to use it for filming, and you expect it to be better than 5D3, I'm afraid it's not ready yet. It does record 4K UHD compressed RAW at 30 fps with an external recorder, it has global shutter, it runs Linux, but that's pretty much it. At this stage, you should really not be afraid of the terminal in order to use it. If you can compile ML, you'll have no problems using it.
The main selling point IMO is being able to program it without reverse engineering. This camera is not meant to get stuck at the current capabilities - they promise it will evolve and improve continuously, but that really depends on the open source community around it.
And yes, I also think the price is really high for what it does *right now*, but I'm not sure what I can do about it. Will let the Apertus guys answer this one.