Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Register Log in

extracting time domain for frequency analysis

Status
Not open for further replies.
Z

Zayzoon

Junior Member level 1
Junior Member level 1
Joined
Jun 12, 2010
Messages
15
Helped
0
Reputation
0
Reaction score
0
Trophy points
1,281
Activity points
1,383
Hi,

I have an array of time domain data 'TDD' with multiple frequencies in there. from within that data there is a frequency 'F1' that contains Energy 'E1' that im interested in identifying without ploting the spectrum.

how do I identify the data set within TDD that provides me that 'E1' im interested in. is this reasonable?

Thanks

Zay
 

Zayzoon said:
Hi,

I have an array of time domain data 'TDD' with multiple frequencies in there. from within that data there is a frequency 'F1' that contains Energy 'E1' that im interested in identifying without ploting the spectrum.

how do I identify the data set within TDD that provides me that 'E1' im interested in. is this reasonable?

Thanks

Zay
Click to expand...

Is it a dominant frequency and is the value of F1 known? What about other frequencies? Are they a multiplication of the F1 or are they random?
You can try the linear prediction in time domain. You can also try to filter this freuency.
 

Thanks Cheeryman for allowing me to clarify

F1 is known, there are other frequeincies like noise floor and 60 Hz cycle due to US lighting, and injected signal harmonics. I only want F1.

I have a huge data set from an ADC and I am only interested in the energy caused by signal F1.

is it feasible to extract a portion of the time domain data set to run an FFT with that truncated data set and get the information related to the energy in that frequincy?

Thanks
 

I don't know if I understood the problem correctly, but maybe You can use the Parseval's law i.e. the energy of signal in time and frequency domain are equal. If You specify a region of interest to the neigbour of F1 in frequency domain, You can calculate the value correlated to it's energy.
 

The question isn't clear. Generally, the frequency domain information is contained in the complete set of time domain data. Depending on the nature of the data, e.g. sampling rate compared to signal of interest's frequency and duration, it might be possible to either truncate or decimate the data before performng a FFT.

You would want to give specific informations to phrase a meaningful question.
 

To measure the energy at any frequency without FFT may just filtering of data be used?
Why dont you like bandpass FIR or IIR?
 
  • Like
Reactions: FvM

    FvM

    FvM

    Points: 2
    Helpful Answer Positive Rating
I have a large data set.

I receive 2000 samples per seconds, and i have data with minutes worth of data. for example my least data file has a run with 5 mins that's 600,000 lines of 24 bit data.

I know i dont need to run an fft on the whole data file. all i want is to find energy do to a single frequency without running that large file. i want to truncate that data set into meaningful portions such that when I run the fft there is no data loss.

how do i truncate that data such that no information is lost.

Thanks,

Zay
 

i want to truncate that data set into meaningful portions such that when I run the fft there is no data loss.

how do i truncate that data such that no information is lost.
Click to expand...

Smells like ignorance. Very clearly, there's always information lost when you truncate the data. Nevertheless it may be acceptable according to your accuracy requirements.

Parameters of the signal of interest, particularly noise level and stationarity matter. Why don't you try for small portions of the data and determine variance? Then decide about the required sample size. I presume you know about windowing in FFT.

As Mityan mentioned, a single frequency filter would be the most effective method to use larger sample sizes without much computational effort. You may want to read about Goertzel filter.
 

And if your sampling rate is 2 kHz, and the tone of interest is much less than a half of this value you may perform a decimation to reduce data set.
 

Thats why im asking, im not claiming to know this stuff. no need to insult me FvM.

I understand that I will lose some information. How do i Generate a metric such that i I can say this much loss is acceptable?

Thanks for the Goertzel filter lead. Ill also look into decimating the data set and see what happens per Mityans suggestion.
 

I understand that I will lose some information.
Click to expand...
That's good. I wasn't sure about after your last post where you just continued repeating things.
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top