FIR filters - page 4
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
The numerical method generator is made in error. The author apparently counts the coefficients correctly, but only uses half of those coefficients.
The signal should be multiplied by the whole impulse response. As a result, the resulting filter has nothing to do with the specified parameters.
I have my complaints about the software too, but there's no need to exaggerate... nothing in common...
nobody forbids cascade filters if damping of one filter is not enough
The numerical method generator is made in error. The author apparently counts the coefficients correctly, but only uses half of those coefficients.
The signal should be multiplied by the whole impulse response. As a result, the resulting filter has nothing to do with the specified parameters.
Firstly, it is 2 times shorter. Therefore, it is sort of "faster". Secondly, the frequency response does not provide the specified suppression.
The code for MQL4 from the "Digital Method Generator" isn't what the author wanted.
Any indicator based on the "Numerical Method Generator" will not filter exactly as the author intended.
The filtering is much worse than expected, but the delay is less because the filter is shorter.
What kind of filtering is needed? I have no idea. But I prefer to understand what I'm doing.
As an example of a FIR filter you can try an LPF based indicator with a Kaiser window.
This approximation allows you to get a lot of suppression. Although in my
my opinion, increasing the delay negates the advantages of filtering.
But it is difficult to cheat nature, though it would be very desirable. The greater the suppression,
the greater the length of the filter and therefore the lag.
I took your generated indicator ED_Raiser_LPF and used it instead of МАшаша in my cluster indicator CL1i_V01.
The figure below has turned out to be much better than the one obtained with MAA.
To get such a price line, as drawn by your indicator, it takes a long time to select the МАшашашаша period, but still it is not as "good" line.
It is assumed that positions are opened and closed at the moment of crossing the line of the indicator with zero line.
Almost all positions are closed with profit.
"As an example of a FIR filter, you can try the LFO based indicator with the Kaiser window" - what is it and where can I read about it?
And I don't understand, is it your indicator or you generated it with the help of the program discussed here?
If you don't mind, tell me how to decrease/increase sensitivity of indicator given by you.....
The sensitivity should be reduced when moving from higher to lower TFs, otherwise it gets very feverish...
It'd be weird if it didn't float.
I wanted first of all to show people that there is a good alternative to the machetes
Agree, all other things being equal, running after the floating spectrum is better armed with a normal filter instead of a machete
why wipers don't think of the running away spectrum I don't know.
after all application of a mashka does not relieve from a problem of non-stationarity
I agree with you.
I searched for "real-time neuro-fuzzy digital filtering" today, but couldn't find anything free...
I agree with you.
I searched today for "real-time neuro-fuzzy digital filtering" but didn't find anything free...
so this free generator is being sold by clowns for a tidy sum http://www.finware.ru/orderdi.html
>> so this free generator is being sold by clowns for a tidy sum http://www.finware.ru/orderdi.html
Yes, I know about this generator. And I was looking for information about creating bandpass filters, but non-linear (neuro-fuzzy).
I also have complaints about the software, but don't exaggerate... nothing in common...
No one forbids cascading filters if the attenuation of one filter is not enough
I meant that you set filtering parameters, i.e. cutoff frequency and attenuation, and get completely different characteristics. And because of this silly mistake a very good match is misleading people. You might as well take arbitrary coefficients and get acceptable results. It would also be a filter. You just won't know its parameters. But if you want to create something more complicated, e.g. a set of bandpass filters, you should know what you're using.
What I meant was that you set the filtering parameters, i.e. cut-off frequency and suppression, but get completely different characteristics. And because of this silly mistake, a very good oscillator misleads people. You might as well take arbitrary coefficients and get acceptable results. It would also be a filter. You just won't know its parameters. But if you want to create something more complex, like a set of bandpass filters, then it is advisable to imagine what you are using.
the suppression and the beats are not clear.
but my amplitude-frequency response measurements show that the resonant frequency of the filter is what you tell the oscillator to use.
there are some other glitches, that need also to be controlled by amplitude-frequency response
I took the ED_Raiser_LPF indicator you generated and used it instead of the MAA in my CL1i_V01 cluster indicator.
I got the picture below, which is much better than the one with the MA.
To get such a price line, as drawn by your indicator, it takes a long time to select the МАшашашаша period, but still it is not as "good" line.
It is assumed that positions are opened and closed at the moment of crossing the line of the indicator with zero line.
Almost all positions are closed with profit.
"As an example of a FIR filter, you can try the FIR indicator with the Kaiser window" - tell me what it is and where to read about it.
And I don't understand, is it your indicator or you generated it with the help of the program discussed here?
If you don't mind, tell me how to decrease/increase sensitivity of your indicator.....
The sensitivity should be reduced when moving from higher to lower TFs, otherwise it is too feisty...
If English is not an issue, I would start by reading this:
The Scientist and Engineer's Guide to Digital Signal Processing
Digital Filters: An Introduction
It does require some basic radio knowledge.
If you find my indicator useful.
We will talk about the parameters in the evening, when I get home from work.
If English is not a problem, I would start by reading this:
The Scientist and Engineer's Guide to Digital Signal Processing
Digital Filters: An Introduction
Although it requires a basic knowledge of radio engineering.
I would be glad if you find my indicator useful.
We'll talk about the parameters in the evening when I get home from work.
Thank you very much for your attention. I have no problems with English, or radio engineering, so I'll read it.
Time is of the essence, so if you could give me a moment to explain
of the indicator you've developed, I'd be most grateful.
Otherwise, it turns out that I've used it without really understanding what it does...
Thank you very much for your attention. I don't have any problems with English or radio engineering, I'll read it.
Time is of the essence, so if you could give me a moment to explain the meaning of
of the indicator you've developed, I'd be very grateful.
Otherwise, it turns out, that I used it without understanding, actually, what it does...
A little about the properties of these filters.
A normal MA has a suppression of about 20dB. In order to improve the suppression, the weighting coefficients are multiplied by some function called a window function.
The Kaiser window allows you to obtain a set suppression value that varies over a wide range. In the calculation, the non-uniformity in the passband and
the suppression in the delay band, but the filter is calculated on the basis of an approximation that is not worse than the required one. You select the worst-case of these conditions.
Another commonly used calculation method is the application of the Parkes-McKelan algorithm (sometimes called the Remez algorithm).
It produces a given non-uniformity in the bandwidth and a given suppression in the delay bandwidth. The calculation requires a fairly large number of iterations and does not always ensure convergence.
I have used the Kaiser window. It is easier to calculate, and the result is comparable in quality to the Remez algorithm.
A little bit about low pass filter parameters.
PassBandBars- bandwidth in Bars.
StopBandBars-The width of the transition zone, i.e. between the bandwidth and the frequency where the required suppression is provided. Also in number of Bars.
StopBandAttenuation- suppression in the attenuation band.
It is not quite correct to measure frequency in Bars, as it is time and not frequency. In reality, frequency is measured in its corresponding time intervals.
F=1/Bars. I.e. at 1 bar the frequency is 1 and this is the sampling frequency. At 2 bars the frequency is 0.5Fd etc.
StopBandBars can be any real number greater than 2.
The filter length (equivalent to the MA period) is not explicitly specified and is calculated based on the specified bands and attenuation.
The greater the StopBandBars or the greater the StopBandAttenuation, the longer the filter is. It lags more and smoothes better.