Discussion on kolkata meet signal

SGM · #51 22nd April 2008, 06:58 PM

Quote:

Originally Posted by kkseal

Ok, first things first

In Sanjay's code (AFL not MS code) insert

a = 22/(7*a);

after 1st line.

This'll solve a basic problem (& regenerate interest in the indi

).

Regards,
Kalyan.

hmm another of those greeks

... this time the Pi fello

so we use another conversion radians/degrees.. interesting

but shouldn't it be a = (22/7) * (a / 180)

anyway use what kalyan says, i have not tested this, he must have tested the code

Quote:

Originally Posted by kkseal

Why bother when the results are so bad?

So it was worth bothering, hopefully the results will change now that we have a fix in place.

Quote:

Originally Posted by kkseal

(Not many will be interested anyway)

Well at last we found something to agree on

rgrds
Sanjay

kkseal · #52 22nd April 2008, 07:24 PM

Right. Code Edited.

I was using Rad values straightaway.
(Currently grappling with ways to deal with EVEN periods

)

Regards,
Kalyan.

SGM · #53 22nd April 2008, 07:46 PM

No sweat, here we go

Code:

_SECTION_BEGIN("Asish_Indicator_01");
a  =  Param("Angle",30,15,45,1);
a  =  (22/7) * (a/180);  

Num = 0;
Den = 0;

n  = Param("Sin_Len",3,1,100,1);

for (i = 1;i < n+1;i++)
{
	Num = Num + sin(n*a) * Ref(C,-n-1) ;
	Den = Den + sin(n*a);
}

j1 = Num / Den;
Plot(IIf(j1 > Ref(j1,-1),-10,10),"j1",IIf(j1 > Ref(j1,-1),colorBlueGrey,colorLime), styleThick);

_SECTION_END();

rkkarnani · #54 22nd April 2008, 08:01 PM

Quote:

Originally Posted by kkseal

Ok, first things first

In Sanjay's code (AFL not MS code) insert

a = (22/7)*(a/180);

after 1st line.

This'll solve a basic problem (& regenerate interest in the indi

).

Regards,
Kalyan.

Any "correction" required in MS code to make it give better results!!!!!

kkseal · #55 22nd April 2008, 08:08 PM

The weighting has to be consistent (triangular) for both even & odd lengths(periods) Here's what i suggest for a 6 period SWMA

Num = (sin(1*a) * P) + (sin(2*a) * Ref(P,-1)) + (sin(3*a) * Ref(P,-2)) + (sin(3*a) * Ref(P,-3)) + (sin(4*a) * Ref(P,-4)) + (sin(5*a) * Ref(P,-5));
Den = sin(a) + sin(2*a) + sin (3*a) +sin(3*a) + sin(4*a) + sin(5*a);

And for 4 period :-
Num = (sin(1*a) * P) + (sin(2*a) * Ref(P,-1)) + (sin(2*a) * Ref(P,-2)) + (sin(3*a) * Ref(P,-3)) ;
Den = sin(a) + sin(2*a) + sin (2*a) +sin(3*a);

For a = 30(degrees) the respective Sine weights would be

For 4 pd : sin30, sin60, sin60, sin30
For 6 pd : sin30, sin60, sin90, sin90, sin60, sin30

Whereas for 5 pd(ODD) it is
sin30, sin60, sin90, sin60, sin30

Correct if wrong.

Regards,
Kalyan.

SGM · #56 22nd April 2008, 08:33 PM

Quote:

Originally Posted by kkseal

The weighting has to be consistent (triangular) for both even & odd lengths(periods) Here's what i suggest for a 6 period SWMA

Hello Kalyan

This is all over the head for me. I never understood the logic behind this indicator. Was just trying to convert it to afl. Its good that you took over the amibroker part of the project.

I think now you and asish should give the members some sort of conceptual inputs / background. (well only if some interest is expressed)

As for me, cycles stuff is way out of line of my current studies. Also I am not currently looking for developing new indicators etc.

Was fun coding/testing. I hope something usable comes out of this project.

Once you stabilize the indicator/signal code, It would interesting to try to build a system around it. If needed, can chip in with coding/testing.

Regards
Sanjay

SGM · #57 22nd April 2008, 08:37 PM

Quote:

Originally Posted by rkkarnani

Any "correction" required in MS code to make it give better results!!!!!

Hello RK

No change is needed in the MS code, the change in afl is required because of a bug / error in the afl code.

Now Kalyan has proposed some changes, that are yet to be coded in ami or MS

Regards
Sanjay

kkseal · #58 22nd April 2008, 08:38 PM

Sanjay, pls see if you can detect the error in this piece of code (possibly in the IF condition)

HTML Code:

pd=6;

Num = 0;
Den = 0;

for(i=1; i<=pd; i++)
{
	if(pd%2==0 AND i==(pd/2+1)) X = sin((i-1)*a);
	else X = sin(i*a);
	Num = Num + X*Ref(P,1-i);
	Den = Den + X;
}

i have taken fixed period=6 for testing. 'a' & the rest of the stuff is same as before.

Regards,
Kalyan.

SGM · #59 22nd April 2008, 08:42 PM

Quote:

For a = 30(degrees) the respective Sine weights would be

For 4 pd : sin30, sin60, sin60, sin30
For 6 pd : sin30, sin60, sin90, sin90, sin60, sin30

Whereas for 5 pd(ODD) it is
sin30, sin60, sin90, sin60, sin30

Hello Kalyan

Coding this should be simple, use a counter to count up (till i = n/2) and then down inside the for loop, use the count instead of i for the angle

rgrds

kkseal · #60 22nd April 2008, 08:51 PM

Yeah got it. Only 'i' will not do for even pds.

Regards,
Kalyan.