Welcome, Guest

Author Topic: Timestep function.  (Read 9906 times)

SkyKadet

  • *
  • Posts: 2
Timestep function.
« on: July 25, 2015, 06:07:32 PM »
Hello, I was wondering exactly how the Timestep function works. It seems like sometimes I can increase the speed of the simulation to years per second, and other times only to hours per second or even less. The Timestep icon turns red when I hit the limit. Does this have to do with simulation calculations or something? Does this application limit the Timestep based on computer performance and or physics calculations? I'd like to find know exactly what is going on here. Is there a way to stop this from occurring?

Gordon Freeman

  • *****
  • Posts: 480
Re: Timestep function.
« Reply #1 on: July 25, 2015, 07:29:23 PM »
With an increased timestep, calculations become rougher and less accurate, causing systems to, quite literally, fall apart (which is why there's a limiter). I'm pretty sure Greenleaf has a video demonstrating this.

SkyKadet

  • *
  • Posts: 2
Re: Timestep function.
« Reply #2 on: July 26, 2015, 07:14:58 PM »
Ok, I'll look it up. Thank you!

Jar

  • Developer
  • *****
  • Posts: 732
    • Universe Sandbox
Re: Timestep function.
« Reply #3 on: July 27, 2015, 11:38:26 AM »
Gordon is correct. If it's turning red, that means it cannot reach the target time step without dropping below the accuracy required for each step. The number of calculations it is able to make is dependent on the simulation.

You can adjust the accuracy: click the "sim" button on the bottom toolbar then click the "more" button. The less accurate it is, the higher time step you will be able to run. But this also means a greater risk of the system becoming unstable.

gabriel.dac

  • *****
  • Posts: 517
  • Brazillian dude
Re: Timestep function.
« Reply #4 on: July 28, 2015, 09:08:06 AM »
The timestep function in US1 is still extremely superior than that of US2

linktydraad

  • **
  • Posts: 10
Re: Timestep function.
« Reply #5 on: September 17, 2015, 11:28:16 AM »
The timestep function in US1 is still extremely superior than that of US2

I completely agree

Greenleaf

  • Thomas Grønneløv
  • Development Team
  • *****
  • Posts: 211
Re: Timestep function.
« Reply #6 on: September 24, 2015, 02:28:37 AM »
The timestep function in US1 is still extremely superior than that of US2


I am not really sure how?
Previously, you could turn up the speed and everything would fly off.
Now you can set an acceptable error per step and the simulation will automatically ensure that the steps do not violate the tolerance you have set.


With this kind of physics simulation, you cannot just go as fast as you would like, since every step does take time to compute and a step has limits as to how long it can be before the simulation breaks down due to errors. If you want to go fast and dangerous, you can either set the simulation to "fixed" and not "global adaptive" mode, or you can change the tolerance from its default value.


.. or are you simply saying that the UI makes less sense now and doesn't properly convey what is happening?




Arian

  • *****
  • Posts: 87
Re: Timestep function.
« Reply #7 on: September 24, 2015, 10:47:00 AM »
Actually, it would be nice to maybe have the red timestep icon turn into a prompt to increase the acceptable error rate, so clicking on it would take you to the tolerance settings. Just an idea though.

Greenleaf

  • Thomas Grønneløv
  • Development Team
  • *****
  • Posts: 211
Re: Timestep function.
« Reply #8 on: September 24, 2015, 10:17:52 PM »
Actually, it would be nice to maybe have the red timestep icon turn into a prompt to increase the acceptable error rate, so clicking on it would take you to the tolerance settings. Just an idea though.


Having a popup with some information, the ability to change tolerance and an option to not show that again, that might make sense. I will make a note, because this is something which confused many people.