Sensor: Omnivision 5647, five megapixel
Lens: 3.6mm F/2.0 fixed-focus
Dimensions: 21.6mm x 25mm x 8.65mm (excluding cable)
Weight: 3g (excluding cable)
Cable Length: 150mm (15-core 1mm pitch ribbon cable)
Connection: Camera Serial Interconnect (CSI)
Maximum Still Resolution: 2,592×1,944 (currently limited to 1,920×1,080)
Maximum Video Resolution: 1,920×1,080 (1080p) 30fps
The Camera Module is the first piece of add-on hardware to come out of the Raspberry Pi Foundation itself – and it addresses a real need in the Pi community. Many users are looking to use the Pi for computer vision projects ranging from near-space balloon launches – in which the Pi provides a live stream of its journey while storing higher-resolution snapshots for when it returns to the ground – to burglar alarms and home automation systems.
The official Camera Module dodges issues encountered using webcams on a Pi – thanks to its esoteric USB implementation – in a couple of clever ways. First, it bypasses the USB part of the Pi altogether by connecting to the previously-unused Camera Serial Interface (CSI) located near the USB socket. Second, it taps into the graphics processing unit (GPU) of the Broadcom BCM2835 chip that powers the Pi – giving it vastly more computational power than devices that use the Pi’s relatively weedy CPU.
The result is a camera that draws surprisingly little power, weighs just 3g, has the footprint of a postage stamp and yet is capable of capturing five megapixel still images and 1080p high-definition video at thirty frames per second. Better still, it does all this for less than £20. Granted, that’s around the same as you’d pay for a Raspberry Pi Model A to which the camera can be connected – but given its capabilities, it’s still a bargain.
The Camera Module arrives, in traditional Foundation fashion, as a bare circuit board with a small, 15cm ribbon cable as the only accessory. Gently inserted into the CSI port on the top of the Pi, the camera is ready to go – once you’ve updated your Raspbian installation. Those who have chosen alternative operating systems for their Pi are, at this point, left out in the cold: currently, it’s Raspbian or nothing.
Even fully updated, the software to drive the camera is in a very early stage. Many functions – such as the ability to record EXIF tags into image files – simply don’t work. Other features work incorrectly: attempting to capture any image above 1,920×1,080, for example, makes the camera take a Full HD cropped image and then resize it rather than capturing the full frame of the sensor – a problem the Foundation’s development team is working to resolve.
Another issue, less likely to be fixed any time soon, comes in how the camera is driven. While webcams are supported on the Pi through Video4Linux, allowing them to interact with various third-party software packages, there’s no such driver for the official camera module. As a result, the only way to drive it is to use two command-line packages – raspistill and raspivid – the source code for which is available.
The software is easily the worst part of the camera, but it’s also the part that will change most rapidly. Even between the writing of this review and its publication, the Foundation has likely fixed several of the bugs mentioned here – and will continue to improve the software.
When used within the constraints of the software, the camera certainly delivers: still images are crisp, providing they don’t hit the Full HD crop limit, and video is impressive – although prone to banding if your power supply isn’t up to scratch. The fixed-focus lens is a slight drawback, with objects any closer than two to three metres appearing blurred – but if you’re willing to hack the camera apart, you can remove the glue from the lens and adjust the focus manually with a twist.
There’s no denying that the Raspberry Pi Camera Module is a bargain, despite costing nearly as much as the Model A itself. The software needs serious work and the lack of Video4Linux support is disappointing, but its small size and high-resolution sensor will likely find it a home in many imaging-related projects.