[ODE] Re: Torque settings for simulated servo

[mouke at gmx.net]

Well, first, you should try and use your real input values. You can 
achieve quite stable results also with very small/large values, 
especially when using double precision. Just ensure proper setup of 
all parameters.
For example, leave global CFM and ERP at their defaults for the 
beginning, don't use any flags in dCollide you don't really need, and 
ensure a reasonable and constant timestep. 

If you are sure you want to stick with your unit conversion however, 
you just need to: 

1. scale all masses and forces by your factor of choice. Since 
acceleration is force divided by mass, it will stay the same. You 
don't need to scale lengths or even gravity. 

2. since torque is force times distance, scale it by the same factor 
when applying torque directly. Angular acceleration is torque divided 
by inertia, and inertia is proportional to mass, therefore angular 
acceleration will stay the same. 

Hope this helps, 
Michael 

> 
> I'm not sure if this is correct, after all g is acceleration due to gravity 
> and all bodies fall with the same acceleration in a vacuum.  Furthermore 
> the SI units of acceleration are m/s^2.  It seems then that altering the 
> constant for g should only affect units for distance in my simulation.
> 
> As to why I do not use SI units in my simulation, if you check the ODE 
> documentation it recommends that units should be chosen that they lie in 
> the range 0.1 to 10 to ensure the engine's stability.
> 
> That said thanks for the suggestion and does anyone else have a suggestion 
> as to what is going wrong here?
> 
> Garvin.
>