Path: chuka.playstation.co.uk!news
From: "Michael Enoch" <100413.2514@compuserve.com>
Newsgroups: scee.yaroze.programming.libraries
Subject: Re: Rotation Matrix Functions : The Solution?
Date: 10 Apr 1997 19:58:52 GMT
Organization: PlayStation Net Yaroze (SCEE)
Lines: 35
Message-ID: <01bc45ea$2a00f1a0$c209e8c3@fred>
References: <01bc4074$bd79ee80$d603e8c3@fred> <334A0C83.5959@micronetics.com> <3348D96D.289@interactive.sony.com> <334b2016.21064909@news.playstation.co.uk> <01bc4448$a37ddc80$0b16e8c3@fred> <334BA817.4A3C@interactive.sony.com> <01bc454b$1c8e8fe0$5214e8c3@fred> <334CA29A.539A@interactive.sony.com>
NNTP-Posting-Host: ld16-194.lon.compuserve.com
X-Newsreader: Microsoft Internet News 4.70.1155

> You get this problem because of maths inaccuracies on the PS ( because
> of the 16 bit values ) The normalisation process will oscillate is you
> use it continuously.
> The solution is only to normalise the matrix when you change the
> orientation of your scene/object. Then, when the picture is still you
> won't see any changes occuring.
> 
> Colin.
>

When the picture is still there are no changes as it is, whether I use the
code every frame or only when it moves, the problem is that when it moves,
even with the code in to try to fix it as suggested, the innacuracies in
the matrix all add up and then the object looks wrong, whether it moves or
not.  I hope you understood that, it's rather confusing as I read it back
to myself.  

Here is an example:
	I'm printing out an object's matrix, and I then do a full 360 in the Y
axis.  It starts out as:	4096 0 0
		0 4096 0
		0 0 4096
and it ends up as:
		4096 0 -4
		0 4096 0
		-5 0 4096
I rotated in 5 degree steps (5*(4096/360)).

I think the code I'm using to correct this is quite possibly wrong, if
anyone can see where I'm going wrong, again, I'd be grateful.  The code I
wrote and the description by Stuart are a couple of messages back in this
thread.

	Mike.