Calendar Latest Topics
 
 
 


Reply
  Author   Comment  
eKnudsen

Junior Member
Registered:
Posts: 6
Reply with quote  #1 
I just got a new rig (windows 10, core i5, 16gb ram) set up with the latest version of ML with a single daq, running eye tracking tasks with the eyelink 1000. I migrated my task from an older windows 7 machine running ML version from 2012 with an iScan.

Everything is working great, once I got the configurations switched over to the new card and updated the code for the newer versions of the task specific functions. However, at the end of the trial the ITI is *way longer* than what it's set to be at the main menu (I've got it set to 1000ms, actual ITI is 6000-8000 ms). This never happened on the legacy version on significantly slower hardware. In fact, I set the trial ITI to 500 at the main menu, and during the task I am getting warnings to the effect of "the actual ITI is greater than specified: ~= 750 ms" which comes up, then the eye trajectory gets drawn to the control screen 5-6 seconds later, then the new trial begins. So I think the issue is between end of trial --> draw eye trajectory to control screen somewhere.

Looking at the resource manager, I've got plenty of CPU/RAM/SSD overhead so that doesn't appear to be the bottleneck. Both ML and matlab latency tests were great (<10ms). 

Any thoughts as to what's going on? I'm sampling the eyelink at 500Hz and running the eyelink itself at 500 Hz, but this issue actually is happening on another rig (same computer specs) albeit with an older iscan rather than an eyelink.
0
eKnudsen

Junior Member
Registered:
Posts: 6
Reply with quote  #2 
I'm continuing to troubleshoot here, and the effect persists in stimulation mode. I ran some timers for comparison: around all of my code in the timing file (internal timer) and around the call of the timing function in the main monkeylogic code. There's a huge discrepancy here. Any thoughts would be appreciated...

difference of opinion here.png

0
Jaewon

Administrator
Registered:
Posts: 971
Reply with quote  #3 
What is your ML version? I don't understand what code you updated for "the newer versions of the task specific functions". They have not been changed for a long time.

Are you saying this happens only in your iScan setup, but not in the EyeLink setup?

In the screenshot you posted, where did the message, "Set affinity to realtime", come from? Do you change it yourself?
0
eKnudsen

Junior Member
Registered:
Posts: 6
Reply with quote  #4 
Hey Jaewon, 

I had been running 1.2.66, but yesterday I updated to 1.2.86 (latest on github) and the problem persists (that screenshot is from 1.2.86). I mean we were running the February 2012 version for a long time so really only the input arguments to goodmonkey changed slightly (e.g. "trigger_val" --> "TriggerVal"), so that's all I was referring to there. 

It happens on an iScan setup and the EyeLink setup both, even in simulation mode so I don't necessarily think it's related to AD sampling.

And yeah that was a comment I added in the main monkeylogic routine as a debug to see where in the trial the hang was hitting before I narrowed it down a bit.  
0
Jaewon

Administrator
Registered:
Posts: 971
Reply with quote  #5 
I don't know why the GitHub is linked in the official website, but that version has some performance issue. You should use the Oct 2014 version (which I don't know where to find any more) or try NIMH ML. NIMH ML is still in development to add new features, but should be better than GitHub and I can answer if something doesn't work in it.
0
eKnudsen

Junior Member
Registered:
Posts: 6
Reply with quote  #6 
Ah, good to know. Where can I get the NIMH ML from then? I'd like to stick to the NIMH DAQ toolbox for the single daq feature, but I'm having trouble finding a link that's not on the official page (Oct 2014) or the git.
0
Jaewon

Administrator
Registered:
Posts: 971
Reply with quote  #7 
I am posting a download link here.

https://goo.gl/wuxWg7
0
Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.