Matrix multiplication

[fea12rs]

sorry I know this is unrelated but I've d/l and extracted it.. how do I add it to bloodshed library

[norm b]

I don't use Windows or Bloodshed so I can't help. I did find this though:
http://www.bloodshed.net/faq.html#9

[Duthomhas]

Alas, I apologize, sincerely, truly. My brain must be scrambled or something. I would have sworn one of those subscripts were off last night. And I was a jerk.

:O(

Your matrix calculation looks right.

Uh, alas, I don't have time right now to mess much with your code. Can you wait a few days? (Perhaps someone else will pick it up before then.)

[fea12rs]

Don't worry about it we all make mistakes which is precisely the reason im in this forum.

Hopefully somebody will spot the error I sure as hell cant. The data outputted is so random like a sharp sinusoid up n down etc... I can only think that it may be that every iteration, the matrices dont initialize to their original values for some reason and so making it very random :S ..

Hope to hear back from you soon

[Duthomhas]

Alright, sorry to take so long.

If I understand what you are trying to do correctly, the issue isn't in the multiplication operator, its in your Resultant() iteration.

Comment out line 32 in main.cpp (only multiply by m2 once, before the loop).

Also, on lines 9 and 10 of matrix.cpp you have swapped the signs. It should read

	m12= -(1/n)*sin(k*d);
	m21= n*sin(k*d);

As an aside, the *= operator should be:

MyMatrix& operator *= ( const MyMatrix& fred )
{
  ...
  return *this;
}

Hope this helps.

[edit] Oh, before you run that and say "what?", I modified the parameters a little to get a more sane matrix.

First, in main(), I looped over (k=0; k < 2; k+=0.001). A tenth isn't a very small sample for the wavelength of your graphs.

Second, in Resultant(), I set the 'd' arguments to 10.0 and 5.0, respectively, which is a bit more sane for viewing.

If you want I'll post an image of what I'm seeing.

Hope this helps.

[edit 2] http://home.comcast.net/~michaelthomasgreer/temp/fea12rs-wave-matrix.png

[fea12rs]

Thanks a lot for the reply. Yeah its helped me ive been making progress. I realised for 1 thing i had the wrong expression for R all along which is ironic.

With the line 32 of main.cpp, if i comment that out surely that would mean I only multiply 1 matrix by itself several times, not multiplying them successively like such:

matrix1 = m1; matrix 2= m2;

m1*m2*m1*m2 ...

??

[Duthomhas]

It would help significantly if we could see exactly what your assignment is.

[fea12rs]

By name Its the 1D transfer matrix method but ill try explain in as much detail here.

I start off with 2 matrices, where each matrix has 3 arguments(n,k and d) which are different for each matrix, n and d are fixed.
The 2 matrices are of the form given in the 1st post, I have to multiply them successively - m1*m2*m1.. a certain amount of times and have to do this over a range of k.
Given the final matrix, which is the resultant (mT) of multiplying m1 and m2 together a given number of times, I calculate the reflectivity R which is given by an expression in terms of the final matrix.

The correct expression for R is :

double MyMatrix::calcR()
{
	double a = (m21+m12)*(m21+m12) + (m22-m11)*(m22-m11);
	double b = (m11+m22)*(m11+m22) + (m21-m12)*(m21-m12);
	double c = a/b;
//	std::cout << "a  " << a << "asquared  " << a*a << std::endl;
	return c;

When I plot R vs k I still get a randomly sharp sinusoid type of graph over the whole range where a certain value of k should peak.

[fea12rs]

My apologies I have just changed the variables so that I used the actual values as opposed to scaled down ones, and a get 2 smooth curves which is a lot more satisfying.

I think i've got it solved now. Just being over analyzing when it was something minor.

Thanks again

[Duthomhas]

So you've got it solved?

(I wasn't sure what exactly you were trying to do. Each time you multiply, you are convolving the wave with the multiplicand wave. Which requires some careful observation to see the frequencies combine the way you want.)

In any case, I don't know that much about waves beyond that...

[48T7M4Gy]

If ever there was a case for using a matrix library like that from boost or matrix 11 from Techsoft and then adapting it if you want to invent something useful, this is it.

Normally the assignment operators use 'friend' to gain access to the underlying structure instead of via 'this' depending on the data structure for the matrix class but with all due respect, by not using arrays or vectors the whole scene here is one of rampant spaghetti code.

Once you get it right you can just write A *= whatever;

http://www.techsoftpl.com/matrix/ It's free, no obligation, open source and fairly easy to understand (unlike newmat).

[fea12rs]

Yeah i think i got it. The final graph curves are a bit more jaggedy than expected but on right track.

And kenmort i have no idea what u just said but im sure it made you feel very smart.

[Duthomhas]

No one here is trying to feel smart. He's just suggesting a library you can use to handle all the matrix stuff without having to write it yourself.

@kemort
While he's doing something uncouth with his special-purpose matrix class, it works for the assignment at his level. Further, it doesn't meet the definition of spaghetti code, so play nice, K?

[48T7M4Gy]

@Douas I take your point even though there was no intent on my part.

I know none of us are forced to stay. If fear12rs doesn't know how to design a simple program and can't take a bit of sound advice then he may as well know it and besides, other people may learn something rather than being misled into bad practice.

True, it's not spaghettii code. On reflection I should have called it noodle-code, a sloppy mess.

[Duthomhas]

fea12rs is doing the best he can with what he has learned. No one is born writing perfect code, and for his purposes there's nothing wrong with it.

[48T7M4Gy]

@Doas

If he doesn't want to lift off from the launchpad and fizzle out then that's on fears12.

If you had bothered to take in what I wrote I gave him a pointer to the very functionality he was looking for but it's clear that two overblown egos fighting way over their weight got in the way of it.

Aside from that you only have to make your point once Douas unless of course it's to nag and have the last say, another ego failing sad to say.