Page 2 of 4
Re: aircraft modelling advice needed
Posted: 08 Aug 2008, 00:45
by Michael davies
Gary, Dave G,
Actually you can beat that poly count by quite a wide margin, I've several models that are over 70,000 and got through ok and perform well, highest so far is 72,000.
Makemdl is a little bit more complex than we might imagine, technically it should be 64K as the compile is based on the BGL structure which is lionked way back to DOS and binary addition. BUT, you can trick it, you do need to be extra extra carefull with your modeling, try to avoid co located vertices, dont use the weld vertices command, be extra extra careful and frugal with your animations and try to keep your parts count as low as possible.
Example, a 64K model with 140 parts may well blow MakeMDL, but a 64K model with 100 parts will sweep through 90% of the time.
Animations are important too, try not to use Euler and stick to linear, the math for linear is much much smaller than the math for Euler which has acceleration and decceleration math in.
Part names are critical as well, the shorter the name the less strain on MakeMDL, also dont leave any spaces, example : my_door_01 is 10 characters, mydoor_01 is only 9 but, my door 01 is actually 14 because a space is written as %20 ( certainly in HTML ) so in compiler language my door 01 is my%20door%2001, ever right clicked an image off the web where the author hasnt used underscores ?, then you'll see plenty of %20 codes to fill in the gaps between the words. I'm not totally sure what MakeMDL inserts, I just know that simply removing spaces between words and making them shorter increases your chance of a successful compile and opens the door for more polys.
MakeMDL seem to work on a bucket theory, a bucket for polys, a bucket for parts, a bucket for textures, a bucket for animations etc etc, if there all full then your limit is around 64K, but if you can empty some of the others then MakeMDL seems to steal space from them and compile larger models.
This is all for Max but I think it'll all work for Gmax too, not sure on FSDS as I've long forgotten how that compiles models.
Hope that helps
Michael
Re: aircraft modelling advice needed
Posted: 08 Aug 2008, 00:52
by Garry Russell
Hi Michael
I am in FSDS so I those setting you mention mean nothing to me/ FSDS now uses MakeMDL
Interesting about the part names :think:
Garry
Re: aircraft modelling advice needed
Posted: 08 Aug 2008, 01:28
by Michael davies
Garry, does FSDS have a setting for different animation timings and scales ?, I'm assuming its the Euler and linear terminology thats un familure ?.
A linear animation is easy to explain, it moves from point A to point B at the same speed, there is no acceleration, instant full speed and instant stop, kind of negates almost all of real world physics, its easy to calculate, one speed, a start and a stop.
Euler is more life like, it also goes from A to B but has an acceleration curve, a steady speed and finally a decceleration curve, thats three lots of math to calculate, landing gear is usually linear as its hydraulic, its an almost instant start constant speed and stop, doors might be Euler.
A good example of Euler at work is LUL tube train doors, they begin to close slowly, reach full speed and then just before closing slow right down to avoid trapping. An auto door closer does the same, slow acceleration, high speed closure but very slow final close to avoid trapping or slaming.
In Max and Gmax ( well certainly in Max ) you can set these curves up to give variable acceleration and decelleration times, but as you migh have already worked out, there expensive in math and that means it has to be coded into the model somewhere and that means more work for MakeMDL and ultimately less space for something else, I've had 57K models fail MakeMDL, simply because the author had Euler set as default animation code, changing them all to linear and it compiled fine.
Best
Michael
Re: aircraft modelling advice needed
Posted: 08 Aug 2008, 02:11
by Garry Russell
HI Michael
Nope
Keyframed so the positions are set relitive....to have something open quick the slow down a larger movement would be set over a small number of frames then the rest over smaller number........like 1-100 might have the halfway set at 80 . That's the only way I know
Not heard of those other types of those curves
Looking at some of the renderings showing frames under mesh it seems max can make a smoother curves with fewer points
Garry
Re: aircraft modelling advice needed
Posted: 08 Aug 2008, 10:13
by Michael davies
Gary,
Yes thats another way to do it, same in Max and Gmax, effectively what you are doing is making your own Euler curve, the door moves at a constant speed so to make a Euler curve you have it move a small distance on say key frame 0-30, then a large distance 30-80, then a small distance 80-100, that would be four key frame pointsin the animation.
An Euler would have only two, start and finish. The less key frame points you have the leaner your animation load is on MakeMDL, although Euler only has two key frames visable there is a lot of math behind the scene for the acceleration and decelleration curves, so your home made Euler using more key frames is actually better for MakeMDL.
Max, yes it does, it has a neat function called auto smooth where by you make a simple box like shape and it multiplies that shape by the factor you set and also blends and smooths it at the same time, you can use one of several preset smooth functions, NURBS being the smoothest, I tend to use a setting of 2 so a cube with six polys will end up at 96 polys, you can clean it up and remove some excess polys in flat areas and retain the extra ones in the curved areas, you can do this by overall part method or go as low as poly method where you select only the polys you want smoothed.
Attached an example of box modeling, it has been cleaned up a little and the overall poly count was about 10K but now sits at 7K, all the curves were generated by Max, al I had control over was the master box nodes which allows you to bend or tweak the shape until all the curves are perfect. It took a couple of hours but the old way would have taken days, maybe a week.
I think you can box model in Gmax as well, it takes some getting used to especially trying to keep the poly count low, box modeling and NURBS are frequently used in serious CGI to smooth out meshes for rendering and thats where you get poly models in the millions, most of it was simple shapes but NURBS was applied to make it ultra smooth.
When I get time I have to teach John all this !, I tried explaining at Duxford but its quite hard to explain LOL
Best
Michael
Re: aircraft modelling advice needed
Posted: 08 Aug 2008, 10:17
by Garry Russell
Cheers Michael
Sounds a neat way of working
Thanks for the info on the curves. I spend a lot of time adding and removing points to try and get smooth curves, ending up with a huge poly count so have yo try to get back the numbers somewhere else..............and of course having to manually re make the polys

It would be greta to have a function
Still FSDS does the job
Garry
Re: aircraft modelling advice needed
Posted: 08 Aug 2008, 10:26
by decapod
Michael davies wrote: I've had 57K models fail MakeMDL, simply because the author had Euler set as default animation code, changing them all to linear and it compiled fine.
Interesting - the fsx XANIM file format and the intermediate XML format used by MakeMDL Have no support for Euler animations - strictly linear, so the animation type should not have made a difference.
I also don't believe the name lengths make much difference, though I would also recommend not using spaces as this upsets the intermediate X file. The %20 conversion is not used. (I have looked at the code inside makeMDL)
I use another tool - 3D canvas pro (
http://www.amabilis.com) for FS9 and FSX but have not released anything yet as the FSX exporter is still in development. It's not free ($70/£35) but it's the one I know and it's very difficult to get modellers to change their tools.
http://decapod-3d.spaces.live.com/blog
Re: aircraft modelling advice needed
Posted: 08 Aug 2008, 10:36
by Michael davies
decapod wrote:Michael davies wrote: I've had 57K models fail MakeMDL, simply because the author had Euler set as default animation code, changing them all to linear and it compiled fine.
Interesting - the fsx XANIM file format and the intermediate XML format used by MakeMDL Have no support for Euler animations - strictly linear, so the animation type should not have made a difference.
I also don't believe the name lengths make much difference, though I would also recommend not using spaces as this upsets the intermediate X file. The %20 conversion is not used. (I have looked at the code inside makeMDL)
I use another tool - 3D canvas pro (
http://www.amabilis.com) for FS9 and FSX but have not released anything yet as the FSX exporter is still in development. It's not free ($70/£35) but it's the one I know and it's very difficult to get modellers to change their tools.
http://decapod-3d.spaces.live.com/blog
Paul, how are you ?, I do bow to your superior knowledge of the inner workings LOL, your right I dont think it adds %20 either ( noted above ) but removing spaces does let MakeMDL compile higher poly models.
Not sure on the Euler, if you use them in Max then they seem to transpose to the sim, at least they appear too ?, maybe using Euler in Max confuses MakeMDL so it does something else and ultimately trip over on poly counts ?.
All I know is that removal of spaces in words, shorter names and non Euler animations will allow a larger poly count to pass through MakeMDL, quite a lot larger than the oft spoken 64K, 67-68K is quite easy and 72K is the highest yet.
Best
Michael
Re: aircraft modelling advice needed
Posted: 08 Aug 2008, 12:40
by Paul K
Thanks for that, Skip.
I keep bemoaning the lack of a truly accurate Duxford scenery for FSX ( with the greatest respect to Alf Denham, who employs stock objects ). A little voice keeps saying 'If you're so worried about it, get off your fat a**e and do it yourself'. I only live 7 miles away...not as if its a great expedition to gather the wherewithall, is it. :roll:
Re: aircraft modelling advice needed
Posted: 08 Aug 2008, 13:25
by DaveG
Thanks for the info Michael. You learn something new every day
