FIR filter with minimum phase - page 4

 
gara:

Thanks for pointing this out, I will definitely try it. Second decomposition variant as I understand it is discrete Fourier transform and there will be no lag, right?

PF is not a filter, but does the same thing (analysis -> synthesis) in the first part. That is, it only does analysis. Not suitable for non-stationary processes.

Everything has to be done with selective or bandpass filters.

==================

There is one more nuance. If you make a bandpass filter for a particular filter by means of the same mapper, the slices will be different. There will be no compensation. There will be additional distortions at the junctions.

Therefore the opposite slice must be done by subtraction. It concerns IIR filters. For FIR filters, everything seems to be possible, but it is easier by subtraction. FIR filters are heavy and do the same thing as IIR filters.

 
Zhunko:

PF is not a filter, but does the same thing (analysis -> synthesis) in the first part. That is, it only does analysis. Not suitable for non-stationary processes.

Everything has to be done with selective or bandpass filters.

==================

There is one more nuance. If you make a bandpass filter for a particular filter by means of one mapper, the slices will be different. There will be no compensation. There will be additional distortions at the junctions.

Therefore the opposite slice must be done by subtraction. It concerns IIR filters. For FIR filters, everything seems to be possible, but it is easier by subtraction. FIR filters are heavy and do the same thing as IIR filters.


I don't know much about bi-filters, but I understand that with the same achh, the bi-filter will have much less delay?
 
gara:

I don't know much about bi-filters, but I understand that with the same achh, the bi-filter will have a much lower delay?
Why?
Here's what... I don't have it handy, but you can find it. About the matrix description of the filters. So you don't have to ask weird questions.
 
gara:

I don't know much about FIR filters, but as I understand it, with the same ahp the FIR filter will have much less delay?

The FIR filter can be made any way you like. The pay-off is the calculation time.

FIR filter parameters are, roughly speaking, tied to each other. For example, you improve one, the other deteriorates.

The most important difference between them is that an IIR filter is a real filter. It is here and now. It has a physical counterpart.

The FIR filter uses accumulated historical data. There is no such thing in nature. Man invented these things. They became possible with the development of computers.

 

Oh, Vadim! Almost philosophical. I almost agree. Go on, please.

===
Don't forget Nyquist. The old man will be offended...

 
Zhunko:

The FIR filter can be made any way you like. The pay-off is the calculation time.

FIR filter parameters are, roughly speaking, tied to each other. For example, you improve one, the other deteriorates.

The most important difference between them is that an IIR filter is a real filter. It is here and now. It has a physical counterpart.

The FIR filter uses accumulated historical data. There is no such thing in nature. Man invented these things. They became possible with the development of computers.


Let's get specific:

In the main graph the green line is the close price, the red line is the arithmetic sum of the signal passed through the bandpass filter with order 128 and the signal passed through the FNF with order 128, in the additional window the red line is the signal through the bandpass filter (we use it in the sum), the light blue line is the same signal with a suppressed constant component, the shift corresponds to half of 128, a question what methods to reduce the lag do you know?

 

Filters without delay do not exist. This conclusion can be reached by considering the following simple example. Suppose a spike is applied to the filter input. We know that the output of the implemented delayed filter starts increasing only after the input spike. If there were a hypothetical non-delayed filter, its output would begin to increase before the input spike. So the principle of causality would be violated. Which is by no means possible in life.

 

It's an empty exercise in words.

What are we all doing here? Trying to predict the future behaviour of the kotir. We are only interested in analysis to the extent that it is useful in predicting that future.

And all these exercises a la DSP are good for picking out the signal that was in the past.

Totally different purposes for which the filters are used.

Yes, filters are applied to the marketplace. But these filters are estimated by other criteria and very few filters are used and there is no need to build new ones, as marketplace problems are in another area.

And so you chew the same thing and then where do you go with it?

 
gara:

Let's be specific:

In the main graph the green line is the price close, the red line is the arithmetic sum of the signal passed through the bandpass filter with order 128 and the signal passed through the FNF with order 128 is the light blue line, in the additional window the red line is the signal through the bandpass filter (we use it in the sum), the light blue line is the same signal with a suppressed constant component, the shift corresponds to half of 128, question what ways of reducing lag do you know?

About the delay Vladimir explained it nicely.

Once again I recommend you to ignore the delay. Use spectrum analysis. I.e. several same filters with different settings.

I agree with Alexander. What do you want from filters? You started with decomposition. I mean analysis. What do you usually do after analysis? Synthesis, perhaps? So, what do you have next?

 
Zhunko:

Vladimir explained the delay beautifully.

Once again, I would recommend that you ignore the delay. Use spectrum analysis. I.e. several identical filters with different settings.

I agree with Alexander. What do you want from filters? You started with decomposition. I mean analysis. What do you usually do after analysis? Synthesis, perhaps? So what do you want to do next?


I want to build a bank of bandpass filters:

https://www.mql4.com/go?http://belisa.org.by/pdf/Publ/Art7_i17.pdf

https://www.mql4.com/go?http://belisa.org.by/pdf/Publ/Art4_i18.pdf

described in sufficient detail with examples of possible use, all seemed very logical, but the author himself, as I now think, somewhat miscalculated without stating a word about the delay, because if in the lowest frequency filter the central frequency of the filter is on the order of 1/MN1 then the delay even in a few samples will be very large, so I think that determining when choosing a filter should be the minimum delay