FFT size unexpectedly affecting peak frequency

I was playing around with some exercise from Labview Signal Processing Course Manual (NI, 1997), when I stumbled upon unexpected result. I have included below the block diagram for exercise 3.1 (one sided FFT) from the manual. This demo vi had left the FFT size (default). I am not sure what that default value is. It is not stated in the context help. I tried to add in FFT size. Unfortunately the FFT analysis did not give an accurate answer when I put in FFT size. For example, I have set the sine wave frequency at 100s, Fs 300Hz and No of sine wave samples 500. When the FFT size was set at 256 I get 2 frequency peaks at 51 and 102 Hz. At FFT size of 512, I get one peak at 102 Hz. At FFT size 1024, I get a polymorphic waveform!!!
However when I set the FFT size to -1 then I got back the correct peak at 100Hz. I suppose this is the default FFT size value. I could also get the correct FFT frequency if I set the FFT size (500) at the same value with # no of samples (500).
Is there anyway I could set the FFT size value and yet get the correct FFT output. Thanks.
Solved!
Go to Solution.
Attachments:
3.1onesidedwithnooffft.vi ‏19 KB

Dear 'Dad',
Thanks for the correction. I have made adjustment as you have mentioned. It seems to work but I noticed some new problems.
1)      If the following parameters are used ( sine wave 100Hz, Fs 400Hz, FFTsize 512, Sample Size 512) then everything is fine.
BUT when
2)      When the number of samples was increased to 1000, I ended up with 2 peaks at 100Hz and 300 Hz. (This vi was suppose to be one sided, the negative side of FFT was removed)
3)      Things get even weird when Fs is increased to 500Hz. Then the 2nd peak moves from 300 to 400 Hz. !!!
Attachments:
3.1onesidedwithnooffft2.vi ‏17 KB

Similar Messages

Maximum peak frequency

Hello..
I'm a user of NI product (LABVIEW 8.5).
My problem is, i'm using DAQ VI that read the signal from the hardware and then i transfrom it into frequency domain by using FFT.
In the frequency domain, i get a graph with a set of different frequency. Here is the set of data
1. At freq= 0Hz, magnitude= 23 - this is the offset
2. At freq= 20Hz, magnitude= 5
3. At freq= 35Hz, magnitude= 18
4. At freq= 55Hz, magnitude= 10..
So, which VI should I use so that it will display the frequency of 35Hz which contain the maximum peak frequency of my signal?
TQ

1. It appears that your frequency resolution is 1 Hz. For FFT based systems df = fs/N. To get finer resolution you need more data points (increase N). You can either take more samples, although the insects might not cooperate, or you can pad the data. The FFT.vi will do that internally just by changing the FFT size input to a value larger than the number of samples in the dataset.
Your saved data has 3000 points and you get two outputs (at 8 and 9 Hz) to represent that peak. Changing the size to 15000 results in 7 points in a peak at about 8.6 Hz.
2. Your data looks very strange. It appears that about 40% of the input data is saturated at amplitude = 0.21. This has to distort your results far beyond the DC offset. You cannot do anything about that after the data has been recorded because all the information is lost in the saturated intervals. Please check the gain of your preamplifier.
Lynn

Frequency sweep with peak frequency

Hi guys, I need some help here. I´m using a NI 9215 to do a frequency sweep from 30 to 90 Hz.
So far so good. Now I need a resulting resonance frequency at highest voltage.
I tried a max peak frequency VI but couldn´t get it to work.
Any idea what to do?
mfg, Andi
ok, here is my sweep.vi for better understanding
Message Edited by andinew on 08-11-2009 04:14 AM
Attachments:
sweep.vi ‏216 KB
create_log_frequencies.vi ‏15 KB

Hi Andi,
one thought first off: looking at your "create_log_frequencies.vi", I realized that it generates linear frequency hops, not logarithmic ones...
But on to your question: to find the frequency with the highest amplitude (this will be your resonace, I guess), you will have to look at the entire signal - your "sweep.vi" code iterates the measurement and interpretation five times per second, independantly of the speed of the sine sweep.
My advice would be to generate one waveform containing the whole sweep run, and writing this sweep block for block with a constant sample rate onto the output. You could then read back the response with blocks of the same size as your output, do the single tone measurement on this block and store the resulting frequency and amplitude in an array preserved trough a shift register or simply into two auto-indexing tunnels. The latter is okay as you don't need a "while loop" in the measurement loop (you know how long the analog output loop runs - why not use this information to configure a "for loop").
After your analog input "for loop" has finished, search your amplitude array for the maximum - then use the found element's index to index the frequencies array and return the corresponding resonance frequency.
A good starting point could be the example I wrote a couple of month ago: it runs continuously the same "sweep" (you'd have to adjust the waveform, I know) and reads back the stimulus and the response (wire the output to your probe and parallely to ai0, your response signal will be connected to ai1). You can then insert tone measurements etc. at the marked section...
When trying to open the example, you will be prompted to find a VI called "_Get Terminal Name....vi" - you can find this VI at the following location: C:\Programme\National Instruments\LabVIEW 8.5\examples\DAQmx\_Utility\_Utility.llb
As an afterthought: the "phase difference" you are calculating between every other sample of the analog measurement array is a bit confusing: if oyu need the phase difference between the original signal and the stimulus response, you will need to wire the stimulus to a second input and compare the phases of the two separate signals...
Best regards,
Sebastian
Message Edited by SFK on 08-12-2009 04:12 PM
Attachments:
Multi-Function-Synch AI-AO mit Signalgeneration.vi ‏91 KB

How to detect peak frequencies in the frequency domain?

Hello,
I have built a small test program consisting of two sine waves of variable frequencies, there sum is found and subsequently FFT is performed on it to find the two peak frequencies. I have been trying to somehow display the values of these two frequencies but I just can't find a way.
I searched the forum before posting the question, many of them suggested using the Extract Multiple Tone Information VI which I tried and couldn't get any results.
I have attached my file. Please have a look. If you know any threads on the forum directly related to my problem, please just cite it.
Thanks,
LD
Solved!
Go to Solution.
Attachments:
DTMF Decoder.vi ‏168 KB

Exactly, I just found that out too ... how silly of me.
Thanks very much for replying
LD

Create FFT size

Hi, Can anyone advise me how to create my FFT size from my 1D array say 40926?

There are several articles on the National Instruments website that refer to this subject. I would advise looking into them before you attempt to use the FFT. The FFT is a very powerful tool, but as with most powerful tools if it is not understood the results you get, while appearing useful, may be invalid. I haven't used LV's FFT tools recently, but in the past you could feed any size data (within reason) to the function. If it was a 2^N length an FFT was performed, other sizes would cause it to be evaluated with a DFT, with a sizable increase in execution time. As this was on pretty slow machines when I was doing it (100MHz Pentium) it was pretty critical that the data conform to the "power of two" criteria. One technique, to be used cautiously, is padding with zeros. Doing a search on this website (NI) using "FFT padding" returns a number of articles, i.e.:http://zone.ni.com/devzone/conceptd.nsf/webmain/81227DF4C2952C1A86256CA80053F322, which way the pros and cons of padding.
P.M.
Putnam
Certified LabVIEW Developer
Senior Test Engineer
Currently using LV 6.1-LabVIEW 2012, RT8.5
LabVIEW Champion

Reducing image size without affecting animation

After having trouble rendering, I realized that he images I scanned in are too high resolution even for my HD comp. For my next project, I'll know, but for this project, is there a way to know how to reduce those images appropriately and substitute them so that I won't have to do any new keyframes or something? A way that I can just edit the files in Photoshop and have them update to be lower res in AE without affecting their sizes?

There's generally no easy fix for this problem. Make your life easier by scaling your images in Photoshop by a fixed amount, eg 50%. Then re-scale in After Effects the appropriate amount.
If there is scale-based animation on the layers, you can add a Null to your comp, parent the Null to the layer in question and scale the Null the appropriate amount. You can delete the Null when finished.

Clearing history not reducing catalogue size or affecting speed of LR

I've been looking at ways of speeding up LR and thought to see if clearing 'History' helped.
Weirdly after clearing history from all my images the catalogue is still the same rather large 6.4GB. A restart of LR doesn't change anything either. LR didn't seem to take any time to do this job either which I thought was a bit odd, maybe because it didn't actually do anything.
So I have no history and no change in speed either.
[I backed catalogue up first before testing BTW, so I can always restore history].
OSX 10.8.5 - OS is on an SSD and the LR catalogue is on a separate SSD all by itself in an attempt to speed things up.
Interestingly I decided to clear all previews when starting with new SSD and very briefly LR was a bit snappier. Now I have a mere 4GB of previews and 2.3GB of smart previews and LR is back to Slowsville.

Sorry, should have been more specific. Everything is slow, sadly I could not think of an exception.
Just optimised catalogue after the History deletion and that has now freed up 2.0GB of space and LR certainly seems snappier. Yay!
Optimising has never made any difference before.
But for how long I wonder.

How do you find the frequency of a small sample size and a square wave.

I am collecting data from ten 12v pulses that are seperated pretty uniformly. The pulses make a square wave. I need to know how to input this small sample size and find it frequency. I need this frequency to be output in a readable file. Im not very knowledgeable in labview so if you try and help me please be very descriptive.
I know i need some sort of trigger because of uknown wait times before the sensor will be triggored but other than that im pretty lost. Any help would be appreciated.

What have you tried so far? What parts are giving you trouble? Please be as specific as possible. Post the code you have written.
You ask about frequency measurement and then you are talking about triggering.
If you always have 10 pulses, just measure the time from the beginning of the first pulse to the beginning of the last pulse. The frequency is 10/time.
Break down the problem to smaller parts. Acquire the data. Measure the frequency. Save the result to a file. Get each part working separately, then put them together.
Lynn
Sorry about duplicate post. My internet connection is very slow today.
Message Edited by johnsold on 05-31-2010 12:05 PM

How do i create a list of frequencies +/- a peak output as ascii file?

I have an VI set up that reads in from a Spectrum Analyzer and searches for the peaks and outputs them along with there amplitudes as an ASCII file.
A Macro has been created in Excel that reads the ASCII file and creates a list of 5 frequencies +/- the peak frequencies above a threshold to be tested later. The problem is that it has some frequencies
created more than once.
Is it possible to create this list in the VI that would filter the list and then output it to excel to be printed later?

ChrisKalina wrote:
I have an VI set up that reads in from a Spectrum Analyzer and searches for the peaks and outputs them along with there amplitudes as an ASCII file.
A Macro has been created in Excel that reads the ASCII file and creates a list of 5 frequencies +/- the peak frequencies above a threshold to be tested later. The problem is that it has some frequencies
created more than once.
Is it possible to create this list in the VI that would filter the list and then output it to excel to be printed later?
Of course it is possible to do this. What have you tried so far?
One problem you will have, though, is that LV 5 is such an ancient version, many people on the forums will have trouble being able to open your file, and almost no one will have the means to save back that far. So you may have a hard time getting help.

Frequency Response Function & FFT & Inverse FFT (problem of unit Volts-RMS)

Hello everyone,
I am currently working on a VI in order to compare two analog signals : the first one corresponds to the output signal (my reference) which is sent by my data acquisition card to a shaker and the second one corresponds to the input signal recorded by an accelerometer fixed on the same shaker. The final goal of the VI is to correct the analog output signal by using the analog input recorded signal in order to have the vibrations on the shaker which corresponds to what we really want.
To summary, I have a problem of unit with the Volts-RMS...
So this is my method for the VI :
First, I have to calculate the Frequency Response Function between the two analog signals (output and input). For it, I use the " Frequency Response Function (Real-Im).vi " which returns the complex values of the FRF in Volts-RMS (but I don't want to use this unit).
Then, I want to calculate the FFT of the analog output signal (my reference). There are two different blocs which can be used : " FFT Spectrum (Real-Im).vi " and " FFT.vi ".
The " FFT Spectrum (Real-Im).vi " returns the FFT complex values of the signal in Volts-RMS and the " FFT.vi " returns the FFT complex values in Volts (or say me if I am wrong, thank you). I really would like to use the second one because of the unit.
Then, I divide the FFT just calculated with the Frequency Response Function calculated just before.
For the end, I calculate the inverse FFT of that with the " Inverse FFT.vi " which use the complex values with the same unit than for the " FFT.vi ".
I don't want to use the Volts-RMS unit because I absolutly want to use the blocs " FFT.vi " and " Inverse FFT.vi ".
The problem is that I don't find a bloc which use the same unit for the Frequency Response Function. The " Frequency Response Function (Real-Im).vi " returns only the complex values in Volts-RMS unit. Maybe it is possible to convert it correctly? Or maybe there is an other bloc which can be used in order to calculate the Frequency Response Function with the same init than for the FFT and Inverse FFT ? Because I can't mix everything for the moment...
Thank you for your help,
Best regards,
Sebastien

Hello Preston,
No, I have not use the Sound and Vibration toolkit. I have only used the signal processing toolkit with the two toolboxes " Waveform measurement " and " Transforms ".
But I think that what I have done for the moment in my VI is correct (I have finished the complete VI). But I am not sure of the units (Volts, Volts-RMS...) and I would like to understand.
I have tried with the Sound and Vibration toolkit for the frequency response function (because you say me that it deals with all the unit conversion) and I can obtain the same results than with the " Frequency Response Function.vi " of the toolbox " Waveform measurement ".
But I would like to understand the units (see my previous post please). For example, for the FFT (the result is a complex), why sometimes it is in Volts, sometimes it is in Volts-RMS ? Is it possible to convert it ? How ?
If you want, I can attach on the forum my VI and that will maybe help you to explain me. Maybe it will help other people interested.
And if someone else can give me other precisions or advices about it, do not hesitate.
Thank you for your help,
Sebastien

Correct FFT settings Diadem

Hi,
I am a new user with Diadem (version 10), and use the program to view and analyze EEG data. In order to calculate a powerspectrum from parts (ca. 5 seconds long) of my EEG data (sampled with 5000 Hz) I tested the analysis FFT function in Diadem for one-time signals. Being unexperienced with the FFT details, I have trouble chosing the correct settings:
-type of window (Hanning ?)
-amplitude damping correction (none/periodic/random?)
-FFT functions: should be amplitude I guess. Averaging arithmetic? (I also tried exponential, which did not look too bad, but I guess this puts more stress on the latest calculated parts..). Then calculating the average from amplitude ?(other option is autospectrum)
-time intervals: this of course is linked to my sample size (e.g. 5 sec) and sampling rate (5000 Hz). With my 5 sec of 5000 Hz I believe I can get up to large FFT sizes like 1024 or 2048. But what about the overlap, what are the rules for this? I guess too much overlap will distort the frequency spectrum, but what should be chosen? Reject no values sounds good, but if this is the choice, why?
In the tests with the FFT I did, the 0 Hz peak is dominant. Can one choose the settings in a way that gets around this, or should one put a high-pass filter on the data to get rid of it?
Obviously ignorant, perhaps some expert can give me advice how to make the best powerspectra before I create non-sense. I checked the forum and examples in diadem on this, but I do not seem to find a clear explanation on the options that diadem provides.
Regards
Else

Hi Else,
there is no correct setting for FFT calculations. You can get a lot of
different results and most of them are making sense. On the first view, this
might look very complicated, but most of the time, the settings are only
important if you try to compare your results with the results of other FFT
Analysers or software. If you just need to find out more about your data, you
can try out different settings and look for the results you need.
Here are some answers to your questions.
Your signal is cut of at the begin and the end of the measurement which results
in an error. Most of the window functions are reducing this error. Hanning is
the most common window. You should compare Hanning and Rectangle (no window) to
see which one is giving you the better result. Forget all the other windows if
you don't know them.
Hanning is taking a part of your signal away. In the first place, all
FFT-values are smaller then the results without a window. It is possible to
multiply the results with a correction. If you are looking for a sinus wave
within your signal and you are interested in the amplitude of this sinus, you
should use the periodic correction. In all other cases you should use the
random correction. But the correction is just a factor. If you are not interested
in the absolute value of the amplitudes at all, it doesn't matter.
FFT functions:
It is always the same result, but sometimes squared or multiplied with a
factor.
Amplitude is OK, unless you compare your values with other results.
RMS = Amplitude/square root(2)
Autospectrum = Amplitude^2
Powerspectrum = RMS^2
Averaging is only important if you do more than one FFT calculation. In the
first place, I would do one FFT as large as possible. If your signal is a continuous
signal, you get better results if you take as much values as possible (4096,
8192, ...). Larger FFTs are giving you a much better frequency resolution. Most
of the averaging and overlapping is not useful anymore. We have so much memory
and speed in our computers, that we can do calculations with millions of
values. Today averaging is still very important for 2-channel calculations like
transfer functions on numbers of short measurements.
DIAdem can handle datasets with gaps where the equipment was not able to
deliver any value. "Reject no values" is not important for you,
because you don’t have "no values" in your measurement.
If 0 Hz is the dominant frequency, using a high pass filter is a good way to
get better results.
This are
just some short answers to your questions. Signal processing and FFT is a very
wide area with lots of things you might need to know. If you need more specific
help on analysing your data you can post a set of these data or send it by
mail. It also would help to know what you are trying to find out.regards,
Ulrich Bierwisch
[email protected]

MyRIO FPGA FFT Express VI timing analysis for multiple input mode - trying to perform fft's on a 3-axis accelerometer

Hi Everyone!
Project Background:
I've been working with the myRIO FPGA in an attempt to generate an application capable of sampling a tri-axis accelerometer and performing an fft on each axis. I've successfully developed an application for a single axis, but attempting to duplicate the code to sample the second and third axes in parallel results in an estimated 150% resource utilization for the tiny FPGA's LUT's. Additionally, I'm looking to avoid sequentially processing each accelerometer input using triggers and a single fft block because that reduces my fft update frequency significantly (e.g. I can't calculate another fft for input 1 until I calculate an fft for inputs 2 and 3).
After reading up on the fft vi, I'm thinking that I can use the M-interval input indexes / Continuous output indexes Input/Output Index Pattern mode. My thought is that I can edit the vi to remove any math that "recombines" these three vectors into a single fft, resulting in 3 separate fft's. I'm also hoping that this process requires less time than using the sequential method described above.
The Questions:
1. Has anyone done an fft on three inputs using the myRIO at sampling rates > 20kHz and fft sizes of 1024 or larger? If so, I may just be lacking some proper resource management.
2. Does anyone know where to find timing information on the M-interval input indexes / Continuous output indexes Input/Output Index Pattern mode? The manual only provides timing diagrams for singel channel / single input modes. I don't want to waste my time modifying the vi if it will still take 3x as long (assuming modifying the vi is even a possibility).
Further Information:
I already have an application written that samples the accelerometers at >20kHz and then performs the fft on the main processor, but now I'm looking to see if it is possible to perform all signal processing on the FPGA side. The processor performs decently enough, but the timing is not as consistent as I would like it to be. Lastly, I am aware that the myRIO itself has a built in accelerometer, but I need to mount the accelerometer in an environment where the myRIO would probably be damaged and definitely cannot fit.
Any thoughts are much appreciated! The excessive FPGA compile times for this thing make the old guess and check method less appealing.
-Chris

Hi Chris,
Thanks for posting and the detailed background on the project! To answer some of your questions:
1. The FFT Express VI does use a significant amount of space. The FPGA on the myRIO is somewhat limited space-wise. Your best option may to implement the FFT for 1 channel on the FPGA and the other two on the RT side.
2. I converted the FFT Express VI to a subVI and I am not sure if you can trim too much code from it. The subVI is also very complex so re-working it would be a significant amount of work. I could not find much documentation on M-interval input indexes / Continuous output indexes Input/Output Index Pattern mode timing.
I hope that this helps!
Thanks,
Frank
Application Engineer
National Instruments

Export file size help...output seems way too big

I have a 12 minute movie with audio, video and stills in it.
Currently when I export it the file is 865MB... I need it to be less than 100 with decent quality. It is a tutorial video that will be downloaded from the web.
Can you please give me some direction on how to do this? On what variables to change etc. I know if I exported this out of Camtasia Studio, it would be around that 100 or less bench mark .... problem is I don't know what the settings for the Camtasia preset are.
I am using the following setup:
Quicktime Format
Summary:
Output:
1400x790, 15(fps), quality 100
48000 Hz, Stero, 16 bit
H.264
Source:
Sequece, 01-OverviewSWKit
1920x1080 (1.0), 24fps, Progressive, 00:12:03:12
48000 Hz, Stereo
Settings:
Video Codec: H264
Basica Settings:
Quality 100
Width/height (1400 x 790
Aspect: Square pixels (1.0)
Render at maximum depth - "UN"-checked
Depth: 24 bit
Advanced Settings
Key Frame every 1 frame - "UN" checked
Optimize stills - "UN" checked
Frame Reordering "UN" checked
Bitrate setting:
Limit data rate to xxxx "UN" checked
Use Maximum Render Quality - "UN" Checked
Use Previews "UN" Checked
Use Frame Blending "CHECKED"

Well, with the Apple H.264 encoder in a QuickTime container, you're a bit limited to the number of things you can tweak to decrease final file size. Using a more advanced H.264 encoder, like the free x264, will give access to many of the more "high-tech" bells and whistles that are part of the H.264 standard. One potential would be to tick the "Key frame every # frame" box and try a few values there, usually at multiples of your frame rate. Key frames are "whole" frames that take up much more space in the final file size, and decrease their frequency might help slightly decrease file size, though it will also affect quality.
Short of changing the dimensions of your destination video, decreasing the frame rate, or changing the bitrate of the video and audio streams, you will eventually reach a limit of how much you can push one parameter or another. You might try using H.264 in the Format (not codec) dropdown; this will give you a bit more control over encoding to an H.264 MP4 file. Unless you specifically need a QuickTime MOV, you might find you get better results with the H.264 MP4.
These are of course just the options in Premiere itself. Once you go outside the castle walls, you can find a whole bunch of other options, though they usually come at a steeper learning curve. The big thing to remember is that there is no free lunch when it comes to encoding; you have to give a bit here to gain there, and it's all about setting thresholds of what you're willing to accept.

How to see the peak values in a chart?

I have the following situation: I am applying the FFt to an acceleration signal. In the FFT-peak graph I see different peak frequencies. However I would like to know how much are those peak values and either display them in the chart on top of the peaks or extract them in numerical indicators. Is any of these solutions possible in LabView?
I guess the question stands for any kind of graph or chart where you want to extract and see only certain or peak values. Any suggestions?
Thank you.
Radu

OK. Attached is a VI which allows you to move cursors 0 and 1 to select any section of a graph, and then click a button to get the maximum of that particular section. It also places the third cursor (red) at the location of the maximum it finds. It is all done using property nodes of the graphs.
This VI is based on an exercise of the LabVIEW Basics II course, which I took a long time ago, but the manual of which is still very useful.
Hope this helps.
Alejandro
Attachments:
GetMaxiumumExample.vi ‏81 KB

What is the output on the FFT PSD function?

I am trying to evaluate the FFT PSD function relative to frequency. I expect f(0) = the variance of all signals, but that is not what is given. Does anyone have any information about this. I ran a test array with a 1 Hz sin wave, with 10 samples/second, 120 total data points--the FFT PSD gives f(0)=0.

Hi,
The LabVIEW help file gives the following information concerning this function:
The FFT Power Spectral Density VI completes the following steps to compute power spectral density.
Computes the FFT of time signal.
Forms the power spectral density of time signal.
Averages the current power spectral density with the power spectral densities computed by the VI since the last time the averaging process was restarted.
Returns the averaged power spectral density in power spectral density
The single-channel version of this VI can perform single-channel measurements in both one-shot mode, meaning a single call, and continuous mode, meaning multiple calls with history. The single-channel version can perform multichannel measurements only in one-shot mode. If you want to make multichannel measurements in continuous mode, use the multichannel version of this VI.
The single-channel version of this VI maintains internal state information for a single channel only. Calling the single-channel version to process another channel without using the restart averaging control to clear the history results in an unexpected behavior of this VI. The unexpected behavior results from the VI passing the internal state information from one channel to another.
Note The single-channel version of this VI is primarily intended for continuous processing of a single channel. Do not generalize this behavior to the multichannel case and use the single-channel version in a For Loop to continuously process multiple channels by indexing an array of waveforms. Refer to the examples\measure\maxmpl.llb and the examples\analysis\measxmpl.llb for examples of how to use a multichannel version of a VI like the FFT Power Spectral Density VI in a loop.
Refer to the Noise Waveforms and PS Density VI in the labview\examples\measure\maxmpl.llb for an example of using the FFT Power Spectral Density VI.
Amanda Howard
Americas Services and Support Recruiting Manager
National Instruments

FFT size unexpectedly affecting peak frequency

Similar Messages

Maybe you are looking for