DIAdem

cancel
Showing results for 
Search instead for 
Did you mean: 

Problems with FFT

Solved!
Go to solution

Hi, I'm working in analyze vibration signals through DIAdem, the data is from 4 to 5 million points in three or four hours the problem is when I run the FFT analisys only shows me frequencys up to 300 Hz when I know I measure frequencys of 3,500 Hz.

 

Parameters in the measure:

 

Frequency: 12800 Hz

Number of Samples: 1200

 

 

Parameters in DIAdem:

 

Values: 4,757,810

Time: From 11:24 to 14:12

 

To match my data with the time in DIAdem I divide the number of values between the seconds of the adquisition.

 

Sin título.jpg

 

These were my parameter, I don´t know if they are correct, and well I can´t figure out why it only shows me frequencys up to 300 Hz.

 

Thanks and greetings.

0 Kudos
Message 1 of 7
(5,210 Views)

I'm trying to solve the problem but I'm not very skilled in DIAdem and kept getting frequencies up to 300 Hz, anyone can help me...I´m really stuck.

0 Kudos
Message 2 of 7
(5,159 Views)

HI Rolando-stc,

 

                    I'm having trouble understanding your problem. What do you mean by Number of Samples in Parameters in the measure? I assume that the total number of samples you get is 4,757,810. Do you use your Frequency of 12800Hz to calculate the time step for the Time Channel or just the time interval from 11:24 to 14:12? I think the problem is in the data you are using to generate the time interval, so I want to clarify this information first.

 

Cheers

0 Kudos
Message 3 of 7
(5,129 Views)

By parameters in the measure I mean the parameters in the data adquisition, having a 12800 Hz frequency it means a 6400 Hz measure in FFT, thats why I find so rare that I can only see frequencys up to 300 hz tops.

 

Because I have 4,757,810 data points I need that the time I generate contains the same points to match with my data and I can visualize them correctly.

 

Graf.png

0 Kudos
Message 4 of 7
(5,126 Views)
Solution
Accepted by topic author Rolando-stc

Hi Rolando-stc,

 

            I did some tests and I actually reproduced your problem of getting frequencies much lower than expected with the FFT. The problem is the way you are generating your Time Channel data. ¿Where are you getting your data points from? The thing is that sometimes data comes as waveform data. This means that this data already includes a time channel within the same channel, so you don't need to generate a time channel to calculate the FFT. If your data is just a numeric channel, then you do need to generate a time channel to calculate the FFT. The way to calculate the time channel is taking into account the sampling rate of your data. For example, I have a channel with 325000 points, and the sampling rate was 50000Hz. This means that I have to generate a time channel with a step of 20us (1/50000) in order to get an FFT measure of 25000Hz. If I use a step of 20us with a Time domain selection of Start/Step/Number and 325000 as Number of Values in the Generate Time Channel Window, I get the correct time channel to use with my data to calculate the FFT. In your case, your sampling rate is 12800Hz, meaning your time step should be of 78.125us in order to get an FFT measure of 6400Hz. In your post I noticed that you have a different time step. On the other hand, you say you get a total of 4757810 data points. I don't understand how you get this points with the sampling rate and time interval you mentioned. I suggest that you create a time channel using the correct time step and verifying that the total of data points is correct. You can use a time channel or a numeric channel, it really dosen't matter much since you get the same result for the FFT. I hope this information is useful. Don't forget to post again if you have any questions.

 

Cheers

Message 5 of 7
(5,091 Views)

Thanks for the response Hector and you are totally correct.

 

My main problem was the way that I was saving all the data through labview, I had a problem with the while cycles; instead of saving 1280 samples the program was saving around 400-500 samples, thats why the frequencies were so low.

 

I correct this issue and now I'm getting the 1280 samples, then I created the time channel with the 78.125us steps and diadem gives me the frequencies required.

 

You are my hero...thank you so much!!!

0 Kudos
Message 6 of 7
(5,085 Views)

Hi Rolando-stc,

 

         I'm glad you got everything working fine. Thanks for using NI Forums!

 

Cheers

0 Kudos
Message 7 of 7
(5,081 Views)