Joint Time Frequency and Order Analysis Tool Outputs

[dralistersmith]

Dear All

I am currently using the Joint Time Frequency and Order Analysis tool to produce plots of a waveform. I am using DIAdem 2015, both 32- and 64-bit, and have used the latest versions of this script, including the beta version for 64-bit.

I am trying to analyse the waveform below. It has been filtered to between 10kHz and 100kHz. When I perform the JTFA, I get the plots below. The first JTFA plot is of an 11-second window when the waveform activity is minimal, the second plot is of an 11-second window when the waveform activity is high.

My key question is: what are the units of amplitude, and how is it being calculated? The values presented by the JTFA are very small compared to what they appear to be in the original waveform. Note that I use 'Ampl.Peak' as the amplitude type (the second image below shows the settings I have used for these particular plots, but I have tried numerous other settings too).

Secondly, the amplitude range goes up to the colour red, but the images exhibit colours below light blue only. Is this because so few points in the image are red that they are not visible?

If I set ‘Amplitude’ to ‘dB’ instead of ‘NoRef’ then the issue is the opposite - colours in the image appear to be only red to yellow, with no green or blue.

Is there any way to adjust the amplitude colour range manually?

Many thanks for any advice you can provide.

Best regards

Alister 

[CourtneyAlgar1]

Hi Alister,

Firstly I have a few questions, you have filtered the signal between 10kHz and 100kHz, how do you know the frequency lies between these values? and what are you expecting to see from your JTFA spectrum plots as they are configured? 

JTFA plots as you can see in 'Capture3.JPG' plot Time against Frequency with the Z axis or Amplitude being frequency magnitude or intensity. 

What do your graphs look like if you increase the period of time displayed on the Y axis? The below article you may benefit from reading, it contains a lot of good information about setting up both JTFA and Order Analysis.

https://knowledge.ni.com/KnowledgeArticleDetails?id=kA03q000000x14vCAA&l=en-US

You might also find reading the white paper on JTFA useful, 

https://www.ni.com/en/shop/labview/joint-time-frequency-analysis--jtfa--overview.html

Thanks 

Courtney

[dralistersmith]

Hi Courtney

Many thanks for your response. I have read http://www.ni.com/tutorial/3549/en/ and http://www.ni.com/white-paper/3548/en/, which are very useful, but I cannot seem to find the answers to my questions there, hence this thread.

You state that the Z-axis is showing 'frequency magnitude or intensity', please could you elaborate on this?

When we look at the time series, we see how the amplitude varies with time. When we do an FFT (of a specific time window), we see how the amplitude varies with frequency. I'm aiming to see how the amplitude varies in both time and frequency domains, in a single plot. 

The monitoring system used to measure this waveform (shown in 'Capture2.jpeg') had band-pass filters (both 'hard', i.e. in the components, and 'soft', i.e. in the acquisition program) to attenuate frequencies outside of the 10kHz-100kHz range, for many reasons. Moreover, I applied another 10kHz-100kHz filter to the waveform data in DIAdem. So, because increased waveform activity was measured when a stimulus was introduced (between t~3450 and t~3525), and this waveform only included frequencies within 10kHz-100kHz, we know that the frequency content is within 10kHz-100kHz. 

Regards

Alister

[Brad_Turpin]

Hi Alister,

If you look inside the Joint Time Frequency Order Analysis folder, in and among all those various files, you will find a dozen or so *.TDR files.  These get loaded when you create a graph in REPORT with this tool.  You can load the corresponding *.TDR file you want to edit into REPORT, change the color scheme for the 3D coloring, then resave that *.TDR file.  The color scheme will be saved separately in each *.TDR file.  Thereafter each time you click on the button in the tool dialog, it will load and use your edited *.TDR file.  You might want to keep a backup *.TDR file of the original.

Brad Turpin

DIAdem Product Support Engineer

National Instruments

[dralistersmith]

Excellent, many thanks Brad - this is very helpful!

Are you also able to help with my other question?: When using 'Ampl.Peak' as the amplitude type,  what are the units of amplitude, and how is it being calculated? I assume the units are volts? But the values presented by the JTFA are very small compared to what they appear to be in the original waveform. Does the script perform RMS, resample? In order to be able to report the outputs of the analysis, I need to explain how the amplitude has been calculated and what the units are.

Many thanks for your help.

Alister

[Brad_Turpin]

HI Alister,

The short answer is that the 'Ampl.Peak' method returns the largest peak value associated with each particular frequency bin, so it's in the same unit as the original data channel.  But FFT peaks are much more useful as relative indicators of frequency prevalence.  The absolute peak value will change based on the number of data points you acquire and the sampling rate you acquire with, since these together determine all the bin sizes.

Here's what R&D had to say to your question:

Amplitude Peak calculates the peak value for all frequencies. If a pure sine signal is used,
the peak value at the frequency of the sine should be in the range of the size of the sine
and all other values are about zero. Depending on the window function the peak value can be
smaller and the energy of the signal is distributed over several FFT-lines.

The time-data of the customer looks more like a random signal with spikes. In this case,
the signal contains a wide range of frequencies and the energy is spread over a large part
of the spectrum. Usually, the values are smaller in this case. The two results in the first
post (Capture2.JPG) are showing the results at different times. The first one is at a point
with low amplitudes. In this part is a constant frequency at something over 40,000 Hz with
a constant peak value. The second picture, at a point with high amplitudes, looks like a much
more random signal.

It is difficult to see much more in this kind of pictures. In this case it would help to look
at single FFT-calculations from time data at this points.

Brad Turpin

DIAdem Product Support Engineer

National Instruments