View Single Post
      12-10-2007, 12:46 AM   #198
swamp2
Lieutenant General
swamp2's Avatar
United_States
214
Rep
10,201
Posts

 
Drives: E92 M3
Join Date: Sep 2006
Location: San Diego, CA USA

iTrader: (1)

Quote:
Originally Posted by enigma View Post
Not sure where this thread is going but I am not dusting off the TRS-80 or the Apple IIc.

I have been on both sides of this. The problem with pure simulations is you always end up estimating the effects of components that you don't have sufficient information on to make an accurate model. Good examples are drive train loss, tire grip at launch, shift times, effect of flywheel mass on shifts (next gear lurch), energy wasted as stored rotational energy in the drivetrain (you bleed much of it during shifts), and atmospheric effects.

Unless you have an accurate model for those you are just guessing. The guess are educated based on previous real world result being fed back into the simulation. However as a predictor they are still guesses.
True, but....

These indeed are some of the differences between an OK simulation and a great simulation as well as the difference between using a fairly advanced engineering type tool vs. an out of the box, use it as is type of tool. All of the above variables are able to be taken in to account in CarTest, with either a simple or sometimes a complex model. One should not look at all of the unknowns and conclude that simulation is simply intractable. As much are cars are vastly different; they are in many wasy very common/similar. Much of the value of a good software tool is the appropriateness of its default parameters. This is true with ANY simulation software. I have found that for CarTest, these are pretty darn good, right out of the box. CarTest is noticeably weaker with regards to turbo charged engines and modern low loss/high performance automatic transmissions as well as for any transmission that shifts very quickly (although I learned the hard way how to deal with that one and deal with it with great accuracy). Much of this observed difficulty is likely to NOT be related to the sufficiency of the softwares model and method but related to poor/unknown inputs. Lastly, population of the various parameters is absolutely not always a trial and error or a feedback type of approach, many parameters are without doubt such as tire size and air temperature. Other such as transmission losses can be obtained (although not perfectly) from dyno tests, rolling resistance can be obtained from coast down speed vs. time (subtracting any relevant aerodynamic effects), lastly shift times are now these days often cited direct by manufacturers for automatics and DSG type boxes.

If we remember that the the critical inputs to such a software tool are the power or torque curve, redline, weight, gearing (including tire size) and transmission losses we can do some pretty good absolute simulation and even better RELATIVE simulation without a tremendous effort and without tremendous uncertainly.

PS: The TRS-80 comment was not directed at you in any way!

Last edited by swamp2; 12-10-2007 at 03:57 AM.