Power spectral density

Hello,
I cannot seem to change the "desired units" of the SVT Power Spectral Density.vi in the sound and vibration toolkit. It is still possible to switch between EU^2/Hz and EU/sqrt(Hz) on the Power Spectral Density function?
Regards,
Raoul Enning

Hello Beat,
thanks for your help in this matter, I really do appreciate it. First and foremost let me mention that I am working with SVT version 3.1- maybe quite antiquated; i un- and reinstalled it just to make sure that no one here in my lab had fooled with it. On that note, my spectrum units.vi looks somewhat different to yours (see attachment). The relevant portion of the naming of the y-axis variables comes in the "svl_Modify Unit Label for Spectrum Type and Spectral Density.vi". If we look in this VI (also attached) we see that the larger outer conditional, which defines if the units are X^2 / Hz or X / sqrt(Hz), has itself a condition that only allows free movement when the input is linear. When the input is in dB, the conditional is always false, which amounts to forced labeling the Y-axis with X / sqrt(Hz). I actually see the same conditional in the VI you posted.
In addition to this, the data doesnt change either. This is because of the sv_Linear to dB.vi (also in attachment). when switching between EU^2/Hz and EU/sqrt(Hz), the "input signal" data switches between either 1 set of data, or that same set of data squared. so far so good. However, the calculation that is done here when EU/sqrt(Hz) is selected (i.e. magnitude) is 20log(Vin/Vref) which is equivalen to 10log(Vin^2/Vref^2). So, this returns us to power and not magnitude, converting the data right back to how it is when EU^2/Hz. Now I can imagine that programmers have done this to maintain the definition of the dB in electrical engineering, but I will argue that when I select EU/sqrt(Hz) (this cannot be called power spectral density anymore) I want the units in Volts, and not in Volts squared.
Regardless of the semantics of the data computation, the units reported on the y-axis are always wrong when dB is active. Following the computation of the data in sv_Linear to dB.vi, they should always be EU^2/Hz, and not the other way around.
In order to fix this, I am compelled to convert the 20log(Vin/Vref) computation to 10log(Vin/Vref) and to remove the "linear" condition in the svl_Modify Unit Label for Spectrum Type and Spectral Density.vi.
What do you think of this idea? Am I completely wrong in my thoughts?
Regards,
Raoul Enning
Attachments:
spectrum units.JPG ‏67 KB
unit label.JPG ‏77 KB
Linear to dB.JPG ‏120 KB

Similar Messages

Cross power spectral density

Hello
I can not figure out how to use "Cross Power Spectral Density" (CPSD)
in Labview. I know that Labview has something about
"Cross Power Spectrum" but I need Cross Power Spectral
Density, because I am going to measure the coherence which
needs CPSD as far as I know.
Thaks for your help.

The following example code should be helpful. Please take a look, thanks!
Analyzing Two Input Signals Using the Cross Power Spectrum VI

Power Spectral Density.vi

Can someone tell me what the parameter "ENBW" stands for and what it really means in Window Parameters.vi (in Scaled Time Window.vi) (used for Power Spectral Density.vi). See attached.
Is it "envelope bandwidth?" I checked some of my signal processing books but came up with no definite answer.
Thanks,
Don
Attachments:
enbw.png ‏283 KB

It's the equivalent noise bandwidth of the window.
From the help file (LV8) on Scaled Time Domain Window:
"You can use eq noise BW to divide a sum of individual power
spectra or to compute the power in a given frequency span. "
and more from:
Scaling Smoothing Windows
Applying a smoothing window to a time-domain signal multiplies the
time-domain signal by the length of the smoothing window and introduces
distortion effects due to the smoothing window. The smoothing window changes the
overall amplitude of the signal. When applying multiple smoothing windows to the
same signal, scaling each smoothing window by dividing the windowed array by the
coherent gain of the window results in each window yielding the same spectrum
amplitude result within the accuracy constraints of the window. The following
plots are the result of applying scaled smoothing windows to the time-domain
signal.
An FFT is equivalent to a set of parallel filters with each filter
having a bandwidth equal to Δf. Because of the spreading effect of a
smoothing window, the smoothing window increases the effective bandwidth of an
FFT bin by an amount known as the equivalent noise-power bandwidth (ENBW) of the smoothing
window. The power of a given frequency peak equals the sum of the adjacent
frequency bins around the peak increased by a scaling factor equal to the ENBW of the smoothing
window. You must take the scaling factor into account when you perform
computations based on the power
spectrum.
The table below lists the scaling factor, also known as coherent
gain, the ENBW, and
the worst-case peak amplitude accuracy caused by off-center components for
several popular smoothing windows.
Window
Scaling Factor (Coherent Gain)
ENBW
Worst-Case Amplitude Error (dB)
Uniform (none)
1.00
1.00
3.92
Hanning
0.50
1.50
1.42
Hamming
0.54
1.36
1.75
Blackman-Harris
0.42
1.71
1.13
Exact Blackman
0.43
1.69
1.15
Blackman
0.42
1.73
1.10
Flat Top
0.22
3.77
<0.01
Here's another discussion on it:
http://forums.ni.com/ni/board/message?board.id=230&message.id=2637&requireLogin=False
Message Edited by Jonnie 5 on 04-06-2006 10:55 AM
2006 Ultimate LabVIEW G-eek.

Plot Power spectral density of a signal from myDAQ.

Hello,
I need to plot the power spectrum of a signal acquired using myDAQ. I have already been able to acquire and plot the input signal. The example given in Labview for power spectrum uses a pre-defined sine wave and replacing the sine wave by the input signal does not get the work done..
Please Help...
Thanks
Solved!
Go to Solution.
Attachments:
example psd.vi ‏117 KB

Why have the second loop? Looks like you should have what you want with just the first loop. Here's a cleaned up version of your code. I only removed the second loop and then did some slight rearranging to make it easier to read.
There are only two ways to tell somebody thanks: Kudos and Marked Solutions
Unofficial Forum Rules and Guidelines
Attachments:
example psd.zip ‏70 KB

Recording data for power spectral density

Hello
I need to record the power spectrum for a duration of 2 weeks for 2 hours per day. I can get the averaged power spectrum for the 2 hour duration per day as it is embedded in the Power Spectrum Function. Is there a possiblity of saving the data to some Excel file and later calculating the averaged power spectrum for the whole 2-week duration..
Please Help. Thank you.

I was save the data into a TDMS file. The API is fairly simple and there is a free add on to view the data in Excel.
You can later also read the TDMS file and do whatever averaging, etc you want to do with the data.
There are only two ways to tell somebody thanks: Kudos and Marked Solutions
Unofficial Forum Rules and Guidelines

How to calcualte Cross power spectral denstiy in Labveiw??

Hi all,
I unable to figure out how to calculate the cross power spectral density(cpsd) in labview ??? I'm able to calculate crosspower by using
Cross Spectrum (Real-Im).VI but i need cpsd using labveiw functions.
i know that, we can get that through script node function cpsd(x,y) but i shouldn't use that it is my functional requirement.
And one more, Is this right?? psd=power spectrum/ df. where df is frequency resolution.
could any one please help me...
Thanks and regards
Giridhar

Hi girikavali,
does this help you? cross power spectral density and Analyzing Two Input Signals Using the Cross Power Spectrum VI.
Note: The "cross power spectral density" should mean the same as "cross power spectrum", says Wikipedia - CPS/CPSD.
The term "density" seems correct to me, because it is devided by the number of the points, see Cross Power Spectrum. If this is not the case, then please supply a mathematical definition, so one could know.
Cheers,

Determining spectral density of poisson process (carson theorem)

Has anyone ever determined code for working out the spectral density (or
equivalent) of a Poisson process, based on Carson theorem (or a variant
thereof)? It doesn't have to be in Labview, but that would be the easiest (or VB).Message Edited by Jo on 05-01-2005 11:34 PM

Power Spectrum Density conversion to Time Series Data

Hi,
This may seem an odd request but is there a way to convert power spectrum density data back to the time series data that generated it in the first place. I have lost the original time series data but still have the PSD and need the time series to do other analysis.
Thanks,
Rhys Williams

Hate to be the bearer of bad news, but there are an infinite number of time series that will generate a given PSD. You lose all phase information upon taking the PSD. For this reason I almost always save time domain data, or at least complex FFT values.

Power spectrum density question: One sided convert

Hi, there
I am using the PSD function. In order to get the RMS noise, I integrated the PSD data and doubled the result because I have two sides in positive frequency range and negative frequency range. Then my result is 1.414 (or Sqrt(2)) times higher than my RMS data measured in time domain.
Then I looked into the PSD fuction and found out that in the function of "Convert to on-sided spectrum", the amplitude was doubled. But it is not meantioned in the help file.
My thinking is: either change the help file of PSD function and mention that it is a doubled one-sided output or do not double the amplitude when convert the spectrum from two-sided to one-sided output.
best regards
Chengquan Li
CQ

Dr. Chengquan Li,
I know sometimes the help can be hard to navigate, but I think I found the help documentation that address this issue. It can be found here: Power Spectrum.
Under the heading "Converting a Two-Sided Power Spectrum to a Single-Sided Power Spectrum", you can find the following:
"A two-sided power spectrum displays half the energy at the positive frequency and half the energy at the negative frequency. Therefore, to convert a two-sided spectrum to a single-sided spectrum, you discard the second half of the array and multiply every point except for DC by two, as shown in the following equations."
This is why you noticed that the block diagram of "Convert to One-Sided Spectrum" had the output doubled. I hope this helps clear up the issue.
Regards,
Elizabeth K.
National Instruments | Applications Engineer | www.ni.com/support

What's the unit label of y-axis of PSD when using Spectral Measurement Express VI?

Hi:
What's the unit label of y-axis of PSD when using Spectral Measurement Express VI?
Selected Measurement : Power spectral density
Result : dB
No Averaging..
Is dB/Hz the magnitude label of PSD output? Thx!

There could be an issue the the Express VI converting your waveform data into the dreaded Dynamic Data Type, which is just causing more and more memory to be used.
So what I would recommend is use the DAQmx Configure Logging VI to configure DAQmx to stream the data as you read straight to a TDMS file. There are various other properties you can set to make the logging go across files. You just set up the logging before running your task.
There are only two ways to tell somebody thanks: Kudos and Marked Solutions
Unofficial Forum Rules and Guidelines

Spectral Measurements - Y axis units?

All;
When I choose to use the RMS magnitude selection on the spectral measurements dialog attached below.... If I am taking angular velocity measurements in degrees/second, what are the units of magnitude on the y axis of the subsequent frequency domain result? Additionally, what are the units if I select any of the other selections? Thanks!!
-Nick-
Attachments:
Capture.JPG ‏66 KB

When the selection is rms the output units are the same as the input units. Power Spectrum units are the square of the input units and power spectral density is in units of (input units)^2/Hz. This information is in the detailed help for the VI.
Lynn

How does the window selected on the spectral measurements express vi affect the output?

I'm using the spectral measurements express vi to find the power spectral density of a fairly broadband signal. I've already discovered a bug in the express vi; it doesn't divide the power spectrum by the equivalent noise bandwith for the window as it should. However, even after taking this into account by using the scaled time window vi to divide this factor out, the PSD is still changing with the window. How can I get the PSD to be constant with varying windows??

I would like you to prove the "bug" that you're describing by creating an example that shows the calculated and the expected results. By opening the code of the "Spectral Measurements" Express VI you can dig into the heart of the calculation, and there (in the "ma_FFT Power Spectrum and PS Density no State.vi") you will see that the Power Spectral Density indeed IS calculated by using the formula PSD = (Power Spectrum/(df*Noise Power BW of Window)). See the attached PSD.jpg that shows where the calculation happens.
The Power Spectral Density is supposed to be dependant of the chosen Window. Please read chapter 5 "Smoothing Windows" in the "La
bVIEW Analysis Consepts" manual.
Best regards,
Philip C.
National Instruments
- Philip Courtois, Thinkbot Solutions
Attachments:
PSD.JPG ‏139 KB

Optimizing Power Spectrum Performance

I'm continuously acquiring data from two channels on a PXIe-5122 at a rate of 50 MS/s (100 MS/s total). I would like to compute the power spectral density of each channel while acquiring the data. I'm currently using the FFT Power Spectrum and PSD VI. I benchmarked the performance of the FFT Power Spectrum and PSD VI on its own and was processing at a rate of about 3 MS/s per channel.
Is there a more efficient VI should I should be using to calculate the PSD?
Is calculating the PSD at these rates even feasible? If not, any suggestions for alternatives? (Would an LV FPGA be an option?)
Thanks!
geoff
Geoffrey Schmit
Fermi National Accelerator Laborary

Hello Geoff,
This is Andrew Brown, an Applications Engineer from National Instruments. The FFT Power Spectrum and PSD VI you mentioned is the recommended method of determining the PSD of data channels. Calculating the PSD given your sampling rates (100 MS/s) would definitely be much faster using the LabVIEW process. However, I was unable to find a specific benchmark.
Some additional resources on using LabVIEW FPGA are located at NI LabVIEW FPGA and in the LabVIEW FPGA Help for FFT. Please let me know if you have related questions on this issue. Thanks, and have a great day!
Regards,
Andrew Brown
Applications Engineer
National Instruments

WAV power spectrum

I am conducting a sound measurement test using a IEPE microphone and the NI USB 9233. I used the SVXMPL_WAV Recorder (DAQmx).vi to record a Wav file of the sound and then the SVXMPL.WAV Power Spectrum.vi to produce a power spectrum of the wave file. I had a concern that scaling was occurring because the spectrum seemed louder than expected. I read in another thread that no scaling occurs in the vi. Next I used SVMPL_FFT Subset (DAQmx).vi to produce a power spectrum over the same frequency range (20 to 20,000 Hz). In general, the two method agree well except that the peaks in the spectrum are at lower frequencies for the WAV Power spectrum vi than that obtained for the SVMPL_FFT vi. I was wondering if any one could please explain why this is occurring?
Regards,
Bill

Hello Bill!
Thanks for your posts back and the information. So what I have done is I ran approx 1 KHz signal through a series of different programs to see if I could reproduce anything you were seeing. I also used the program you posted as well. However when I run the WAV Recorder (DAQmx) vi and then put that wave file into the WAV Power SPectrum I see the signal at about 1KHz like I expect. Then I aslo use your program to see if I can get the same results with the same function generated signal. I always se my signal in the same matter and the frequency components line up accordingly. See the pictures attached below. Thanks for posting your code and it seems like I am unable to reproduce what you are seeing. Did you make any changes to the VI's in your code? If you could perhaps post the wav files that you are producing to see what I can make of them? Take a look at the screen shots that I was able to produce.
Cheers!
Corby_B
http://www.ni.com/support
Attachments:
Power Spectral Density.jpg ‏225 KB
WAV Power Spec.jpg ‏202 KB
Power Spectrum.jpg ‏142 KB

Cut off some data from a set of data

I have a set of data our of from power spectral density measurement.
I want to curve fitting with the data.
But I want to cut off below a certain frequency. For instance, the PSD
data
contains {(1,10),(2,3),(3,4),(4,3),...} and I want to fit from 3 Hz
{(3,4),(4,3)...}.
I tried high pass filter but it does not change. I tried several ways such
as
get time stamp and find the index for the element and remove the array
using
'Array'tools. It cuts off the data sets that I wanted to remove but it
does
not remove the frequency values. So it works like {(1,4),(2,3)....}
instead of
{(3,4),(2,3)..}. Please let me know how I can make this work.
Thanks.
Attachments:
new-PSD-factor-v3.vi ‏522 KB

Hey Arrow,
I think I understand your question better now. You'll want to use the "delete from array.vi". It can be found by going to all functions>>Array. You'll need to input the array to be edited, which index to begin deleting from and how many elements to delete. From your application it sounds like you'll want to start from element zero and delete the number of elements that corresponds to 0 through 4 Hz.
In addition, you'll need to create an array that represents the frequencies you want to plot (4 1000 Hz or your upper limit). There'll need to be one frequency on your x axis for each element in your PSD array. Then bundle the two arrays together and plot them on an XY plot. You can use the "bundle" vi to accomplish this. You can find examples using clusters from help>>find examples. Then navigate to Fundamentals>>Arrays and Clusters.
Hope this helps out.
Chris Cilino
National Instruments
LabVIEW Product Marketing Manager
Certified LabVIEW Architect

Power spectral density

Similar Messages

Maybe you are looking for