Building a trading system using digital low-pass filters - page 30

 
jartmailru >> :

I looked again.

Perhaps you know what you are advising - and even use it yourself.

But to me - as a person new to filters - the frequency response/frequency response comb does not give anything.

In the program itself changes to the eye are also unimportant.

So just in case thanks :-).

But everything is clear with input sinusoid.

.

P.S.: the program when the method number 1 is active crashes.

a wrong ragged comb is immediately visible.

correct

ugly

the method no.1 does not break, it just takes a long time to read it.

The first version is based on the digital filtering program created in 1995 by Jake Janovetz
The second version was created using MtxVec 1.51 digital filtering library, which implemented an algorithm described in the literature:
Discrete-time signal processing. Openheim and Schafer, Prentice-Hall, 1989.
Theory and application of digital signal processing, Lawrence R. Rabiner and Bernand Gold. Prentice-Hall, 1975.

 

In one case the offset is there, in the other it is not.

The ridges are identical by eye.

Difference in filter order = 5.

.

.

.

And as for the program... set

KGLP, P1 = 20, D1 = 10, A1 = 60, runout 0.1,

Enable flags: Method1, "Automatic" / "Calculate on the fly".

and use the mouse to change A1 from 60 to 61, from 61 to 60.

That's how it draws nonsense too, and sometimes it drops altogether.

 

it is recommended to use the second method

In any case, the software gives exactly what is asked of it because mathematics is an exact science

that's the peculiarity of FIR filters

Сама по себе импульсная характеристика может быть не очень удобна для изучения свойств ЦФ, поэтому вместо нее обычно используют производные от нее Амплитудно-частотную (АЧХ) и Фазо-частотную характеристики (ФЧХ). Они связаны с импульсной характеристикой преобразованием Фурье. Эта связь опять же взаимно однозначна, т.е. из АЧХ и ФЧХ можно восстановить импульсную характеристику.

 

I built a spectrogram with these filters just for fun


 

Christmas:

Cute.

.

As far as I understand, Y = window shift, X = frequency,

The colour intensity shows spikes in amplitude .

.

Try also building an incremental spectrogram.

Kenny & Goodman's wavelet spectrogram

of the incremental spectrogram looked pretty good.

.

Can you tell me please - what method did you use to build this spectrogram?

KGBP - and a chunk run?

or - as I tried - a multi-band KGLP with subtraction?

 
pumal)))
 

it's a test case

I have made indicators with 8 buffers and 8 colour gradation

I plan to make one indicator without indicator buffers

but with graphic objects (matrix of squares) and gradation colours deep

"amplitude bursts" displayed only for those components that rise above the noise

so what's in the black background is conditionally noise

SZ I did another version... https://forum.mql4.com/ru/6100/page20 bottom post

 

I dug out an article in one of the magazines "Currency Speculator" about optimal LPF in terms of undistorted linear trends, which, as I understand, it was designed to detect (see file below). I wonder what the DSP specialists think about this article?

Files:
ma.zip  905 kb
 

Oh, Neutron, thank you for taking the heat for my article here. I originally tried to write it for dummies, but it didn't work :(

P.S. After January 6 there was no Internet at all, but today I came in for the first time.

 
Mathemat писал(а) >>

Oh, Neutron, thank you for taking the heat for my article here. I first tried to write it for dummies, but it didn't work :(

P.S. After Jan. 6 I had no Internet at all, but today I came for the first time.

Hi Aleksey. Glad to see you.

Are you one of the authors of the article, which I posted in my previous post, or you mean something else?