Fwd: [HCNL] IMPORTANT- Fwd: Psychtoolbox in Windows 10 is not recommended

[Eva Kimel]

Hello,

This is worrying. If anyone has any insights, I'd be very happy to hear them.

Personally, I didn't encounter a situation in which a stimuli wasn't displayed at all, but we do have some timing issues which we are trying to solve.

With kind regards,

Eva Kimel

---------- Forwarded message ---------
From: Carmel Auerbach <carmel....@gmail.com>
Date: Tue, Nov 19, 2019 at 9:38 AM
Subject: Fwd: [HCNL] IMPORTANT- Fwd: Psychtoolbox in Windows 10 is not recommended
To: Eva Kimel <eva.k...@gmail.com>

---------- Forwarded message ---------
From: Leon Deouell <leon.d...@mail.huji.ac.il>
Date: Tue, 19 Nov 2019 at 09:22
Subject: [HCNL] IMPORTANT- Fwd: Psychtoolbox in Windows 10 is not recommended
To: <hcnl...@googlegroups.com>

---------- Forwarded message ---------
From: Melloni, Lucia <lucia....@ae.mpg.de>
Date: Mon, Nov 18, 2019, 11:39 PM
Subject: Fwd: Psychtoolbox in Windows 10 is not recommended
To: neuro-us@aesthetics <neur...@ae.mpg.de>

Dear all,

We have been running into timing issues with windows 10 and psychtoolbox. After some digging out, see recommendations from colleagues (below)

Have you experience sth similar? Are you aware of alternatives to the photodiode?

Any feedback is appreciated!

Many thanks and all the best,

---

Lucia 

Begin forwarded message:

From: Yoav Roll <yoav...@gmail.com>
Date: November 16, 2019 at 5:57:31 PM GMT+2
To: Liad Mudrik <lia...@gmail.com>, "Lepauvre, Alex" <alex.l...@ae.mpg.de>, "Melloni, Lucia" <lucia....@ae.mpg.de>, Michael Pitts <mpi...@reed.edu>
Subject: Psychtoolbox in Windows 10 is not recommended



Hi everyone,

I have made some research around the issue of Psychtoolbox running on Windows 10, asking other member of other labs and reading the official documentation online. I have some important conclusions:

1. Windows 10 is supported by Psychtoolbox, but is not recommended for our type of work. Updates to Windows graphics libraries (the enforcement of the so-called "DWM desktop compositor") beginning in Windows 8, has disabled Psychtoolbox's ability to insure that Stimuli appear when they should - if the appear at all. 

2. If Windows 10 is to be used, the recommendation is to use outside instruments, such as an oscilloscope to measure timing, as the program cannot do it reliably anymore. This is a problem with the operating system (OS). No code-based solution can be made.

3. However, even if we use an oscilloscope to make sure our experiment works in one system at one time, it does not say anything about another system or about another time. While we may assume that using an oscilloscope to make sure timings are proper on a system may indicate that this setup works properly, this assumption cannot be held for other systems. Thus, examining with an oscilloscope cannot assure us the experiment will run properly on all systems.

4. There is only one recommended OS to run experiments in Psychtoolbox - Linux. Windows 7, with a single monitor, is also a possible setup, but Microsoft support will end very soon. MacOS and other versions of Windows are highly discouraged or unsupported. 

5. The use of alternative engines, such as PsychoPy will not solve the issue, as the problem stems from components in an inaccessible part of the OS. Furthermore, PsychoPy shares code with Psychtoolbox. Performance tests of PsychoPy, as reported in on their website and published articles, were all done on Windows 7 and Linux (also on MacOS, which failed to present stimuli with high precision).   

I was worried by how little attention this issue was given in other labs. Some didn't notice it yet. It may be because Windows 7 is still highly used. Furthermore, if Alex wouldn't have seen it, we wouldn't have known either. I believe it should greatly concern all who run such experiments.

What do you think? Did you reach different conclusions?

Best,

Yoav

--
You received this message because you are subscribed to the Google Groups "hcnl_lab" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hcnl-lab+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/hcnl-lab/CACZHGXFR44st3STnk30g-6Uz-ORSwubSVdV-Y5%2BoOaic0hgFdw%40mail.gmail.com.