6
$\begingroup$

We are considering setting up a few testing computers using Raspberry Pi 2. There was an interest in this a few years ago, but due to limitations of OpenGL in Raspberry Pi it did not go far. I was wondering if anybody made progress with possibility of running PTB, or PsychoPy, or other experimental software.

Old reference links: Is the Raspberry Pi capable of operating as a stimulus presentation system for experiments?

http://akiraoconnor.org/2013/02/01/using-the-raspberry-pi-to-run-cognitive-psychology-experiments/

$\endgroup$
  • 1
    $\begingroup$ Your best chance at getting a speedy response is to ask Psychopy or OpenSesame developers directly. What really puzzles me is why would anyone want to run experiments on raspberry pi. You can get a second-hand PC that supports OpenGL 2+ (and hence psychopy) for the same price and you get much better performance. $\endgroup$ – matus Oct 3 '15 at 11:48
  • $\begingroup$ Short answer: it's definitely possible (and I think it might have been discussed on www.cogsci.nl), but you almost certainly went be able to get highly accurate response times, or complex animation. $\endgroup$ – Eoin Oct 3 '15 at 21:23
  • $\begingroup$ Aha! osdoc.cogsci.nl/getting-opensesame/raspberry-pi , and cogsci.nl/blog/miscellaneous/… . But as @matus has said, are you sure you really want to do this? $\endgroup$ – Eoin Oct 3 '15 at 22:03
  • $\begingroup$ It would be nice to have a standardized minimal computer system that can be used solely for running experiments. While second hand computers are cheap, you have to deal with slightly different hardware and drivers. $\endgroup$ – Mabu Oct 20 '15 at 15:00
4
$\begingroup$

Speaking for the PsychoPy team, I would expect it to be possible but performance won't be great.

Both PsychoPy and OpenSesame use Python under the hood and this is an interpreted language. That means it isn't super-fast to execute code but we can get away with that by having fast computers (and hardware-accelerated graphics makes a huge difference).

Another option to consider might be a barebones PC like the Intel NUC computers. You can get one of those for £100 (but you have to add memory and hard disk so maybe £180 all in?). It still won't be as fast as a high-spec computer but should outperform a Raspberry Pi

Jon (for full disclosure, I'm the creator of PsychoPy)

$\endgroup$
  • $\begingroup$ Thank you Jon! I am hoping to play with this sometime this year. $\endgroup$ – Mabu Jan 11 '16 at 6:45
1
$\begingroup$

Psychtoolbox-3 maintainer here.

I will probably buy one for some christmas entertainment and see how it does wrt. Psychtoolbox. In the past, with Broadcoms proprietary drivers, i wouldn't have trusted it for a second to provide trustworthy visual timing etc. Also their proprietary drivers only support OpneGL-ES afaik, and that is quite limiting.

However, Eric Anholt, a top open-source graphics driver developer, was hired by Broadcom a year ago to develop a fully open-source graphics stack for RPi / the VideoCore-4 QPI graphics processor, and that stuff is shaping up nicely lately. I was a bit involved in bringup of XOrg support for the xf86-video-modesetting driver of the just recently released XServer 1.18. As Rpi's open source graphics stack uses that driver, i think the user-space support for RPi looks quite decent now. I also did a quick code review of the improvements to the RPi kms display driver in the Linux kernel, and they look as if they might work well enough for somewhat trustworthy display timing. Those stuff is scheduled for Linux 4.5, and maybe would get backported by the RPi foundation to their distribution kernel.

The nice thing about the FOSS driver would be not only increased trustworthyness, but also regular desktop OpenGL support, which would make things easier for PTB and probably also PsychoPy etc.

Of course you won't get the graphics performance or possibly stimulus precision out of a RPi that you'd get from a standard Intel graphics chip, like in a NUC.

So basically i'm cautiously optimistic that the RPi might become at least somewhat useable for at least some not too demanding use cases within the next couple of months or maybe sometimes 2016. But i haven't tried one yet and things always tend to take much longer than one expects and the devil is always in the details, so consider this just a slightly informed opinion, based on some initial code review and skimming of available information.

$\endgroup$
  • $\begingroup$ Thank you Mario. I have been using PTB for a long time now across different platforms, so it will be interesting to see if these can be used for research. $\endgroup$ – Mabu Jan 11 '16 at 6:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.