New Raspberry Pi Global Shutter Camera for machine vision and more
Remember those new cameras we launched way back in the dawn of … January? We have another one for you — and this time we’re getting specialised. Today we’re launching something a little different: the Raspberry Pi Global Shutter Camera, available now at $50.

Built around Sony’s 1.6-megapixel IMX296 sensor, the Global Shutter Camera is able to capture rapid motion without introducing rolling shutter artefacts. This makes it a great fit for sports photography, and for machine vision applications, where even small amounts of distortion can seriously degrade inference performance.
Rolling shutters, global shutters
Every camera we’ve released to date, from 2014’s Camera Module 1 to our High Quality Camera and beyond, has used a rolling shutter sensor. These sensors have a two-dimensional array of light-sensitive pixels, which generate an analogue value proportional to the amount of light falling on the pixel during the exposure time; and a row of analogue-to-digital converters (ADCs), which convert these analogue values into digital values which are fed back to your Raspberry Pi.
The row of ADCs is connected to each row of the pixel array in turn, so each row is sampled at a slightly different time. Provided there is no motion in the scene this isn’t a problem, but once things start to move — and particularly if something is moving fast — we start to see rolling shutter artefacts. Linear motion results in compression, stretching, or shearing of the moving object. Rotary motion can create some very odd-looking shapes indeed.
Rolling shutter artefacts in the presence of rotary and linear motion. Credit Cmglee.
Severe rolling shutter artefacts are unsightly, and hard to detect and correct, but even imperceptible artefacts can interfere with the operation of machine vision algorithms. If we want to eliminate them altogether, we need to use a global shutter sensor. This pairs each pixel with an analogue storage element; when the shutter fires, each pixel immediately copies its analogue value into its storage element, from where it can be read and converted at leisure.
The storage element adds complexity and area to each pixel. Global shutter sensors tend to have a lower resolution than rolling shutter sensors of the same size: contrast the 7.9mm, 12-megapixel IMX477 sensor used in the High Quality Camera with the 6.3mm, 1.6-megapixel IMX296.
Enter the Raspberry Pi Global Shutter Camera
Despite the challenges associated with rolling shutter artefacts, our existing cameras are widely used in hobbyist and industrial machine vision applications. And we’ve seen some real ingenuity: to compensate for artefacts when imaging products on a high-speed conveyor belt, one of our industrial customers ended up training their models using pre-sheared input data.


For these applications, a global shutter sensor offers clear advantages. And reduced resolution isn’t a problem, as high-resolution images are generally down-sampled before being fed into machine vision models.
The Raspberry Pi Global Shutter Camera combines the C/CS-mount metalwork of our High Quality Camera with Sony’s IMX296 sensor. It is compatible with the same broad variety of lenses, including the 6mm CS‑mount and 16mm C-mount CGL lenses that we offer through our Approved Reseller partners.
The video below illustrates the benefits of a global shutter in the presence of rapid rotary motion. First we see the output from the High Quality Camera, showing distinctive rolling shutter artefacts, and then we see the artefact-free output from the Global Shutter Camera.
Like all our camera products, you can use the Global Shutter Camera with any Raspberry Pi computer that has a CSI camera connector, and we’ve updated our hardware documentation to include everything you need to know about the new product. You’ll need to update Raspberry Pi OS to use the new camera: sudo apt update
; sudo apt full-upgrade
; sudo reboot
and you’re good to go.
Credits
Like all recent camera products, the Global Shutter Camera hardware was designed by Simon Martin. Naush Patuck, Nick Hollinghurst, David Plowman, Serge Schneider, and Dave Stevenson wrote the software. Alasdair Allan, Simon Martin, David Plowman, Andrew Scheller, and Liz Upton worked on documentation. Austin Su led on sourcing. Jack Willis designed the packaging, and Brian O Halloran (not included with camera) took the photos and video.
We’d like to acknowledge the assistance of Phil Holden and John Conroy at Sony, and of Shenzhen O-HN Optoelectronic.
19 comments
Jump to the comment form
Josh L
What is the fastest “shutter speed” (lowest exposure time) on the new Global Shutter Camera? I feel like this is important information in regards to how well this can stop motion in a scene, since motion blur is unrelated to rolling shutter.
David Plowman
The shortest exposure time is 29 microseconds.
Anders
The documentation for the camera says 30 microseconds.
Raspberry Pi Staff Liz Upton
Check out the documentation, which has everything you’ll need – it’ll go down to 30µs, which should be more than (little) enough for you.
Gordon77
Looks good. Surprised they make a sensor these days that does appear to do full HD (1920*1080).
Gordon77
Should have said doesn’t do HD !!
Nate
HD = 720p (1280×720). From what I’m seeing online, this can do 1440×1080, which is higher than that.
Perhaps you’re thinking of FHD?
Anders
1456 x 1088 pixels on the sensor.
Johan
Does it have a hardware trigger input?
Kelsey Helms
Does it support external trigger for synchronizing multiple cameras?
Simon Martin
Yes the Global Shutter camera hardware does support two synchronised cameras (CM4 I/O board has two camera ports). There is a solder link on the PCB that you have to bridge to define which camera provides the master timing. The software is still in development though so we can only say this function is coming soon.
Jordan
The camera is 1.6 MP. Full HD 1920×1080 is ~2 MP.
Raspberry Pi Staff Liz Upton
Happily for us all, though, HD is 720p, so everybody wins!
Daniel Rozsnyo
Is this sensor support locked by a DRM chip as seen in your other sensor board? If not – would the driver scale for larger resolution models of same line/generation of sensors ?
scruss
> Brian O Halloran
What, as in the guy from “Clerks”?
Raspberry Pi Staff Brian
I’ve been rumbled!
Steve
This looks great…very interesting indeed. I’m guessing the biggest benefits of this will go to industrial customers doing assembly-line scanning at scale, although I’m sure some creative individuals/DIYers will cook up some imaginative applications. Or at least they will once Pi’s are available again on the retail market. Eben, can you give us an update on future availability projections? The last ones I saw were from December.
Carsten Wartmann
I used the HQ cam for some astrophotograpy. I am curious how the bigger pixels would do on the new one!
https://www.heise.de/ratgeber/Maker-Workshop-Pi-und-HQ-Kamera-fuer-Astrofotografie-einrichten-6163212.html?seite=all
Harkant Singh
I was also wondering the same. There is a ZWO with similar resolution and pixel size.