Terahertz Imaging by improving SNR of Terahertz wave

[feroz_eee]

I need to improve signal to noise ratio (SNR) for Terahertz Imaging system. We applied Chopping frequency of 2KHz for generating Step signal as DC component with Terahertz wave’s AC component. I have already applied DFD filtering of labview. After filtering it shows only AC component. Professor said that it should contain DC component also along with AC component. I don’t understand what I should need to do. How can I make Digital filtering of high frequency Terahertz Imaging system?  In another step, I did FFT for this Terahertz wave that contains DC step signal also. In this case, FFT can be also used as filter. But the amplitude of FFT is low in comparison with amplitude of generated DC chopper-based step signal. My question is how DC component can be exhibited after DFD filtering  for Terahertz system? There I took input from ADC and trying simulation for doing filtering using FFT or DFD filtering , even with Adaptive filtering. I don’t know why after filtering, no dc component appeared at DFD filter’s output.

[Bob_Schor]

As a guess (because you failed to tell us anything about the filters other than they were Digital Filters, and failed to show us the code where the Filtering takes place), I would say that if there was a DC component in your signal, you filtered it out.

Here's a suggestion.  Write a tiny test program where you generate a known signal with a DC component (of known value) and other components that might be present in your signal (about which we also know almost nothing).  Add the Filters you are currently using, and look at the output of these filters.  Are the Filters doing what you want them to do?  If not, come back with your simple Test program and ask us "Why not?".

Here is another suggestion -- to the extent possible, try to avoid Express VIs, which tend to "hide crucial details" that we need to understand what you are doing.  Similarly, Dynamic Wires also obscure the data -- I recommend using Waveforms instead.

Bob Schor

[feroz_eee]

Dear Bob Schor,

Thank you very much for your so quick reply. Sorry. Previously I posted only normal FFT that I did. Now I send you one making Step Signal with which I have already added Gaussian White Noise.. After this, I use low pass filter including DFD filtering (Express VI). And you can check the result from here also from the file.. the output after filtering shows nearly same amplitude of Input step signal's amplitude. But with AD converter of THz system, the amplitude after filtering falls seriously with respect to Step signal's amplitude. (The sampling frequency of our AD converter set to 10KHz, No. of samples 1K)

But in my Terahertz system, as step signal comes as DC part, at least even after filtering with our system, I need to see filtered output with appreciable amplitude of signal with the presence of DC component. So, is it possible?? If possible, which filtering component I should use? What type of filter I need to use? I understand you mention that "if there was a DC component in your signal, you filtered it out." To conserve the DC part at final output, what I should need to do?

As in the last part you suggested me to avoid Express VI. So, what type of filtering from LABView I can use?? Please let me know..

[Bob_Schor]

Are you using a special Filtering Toolbox?  I was thinking about the Filters that are on the Filters sub-Palette of the Signal Processing Palette, which include such things as Butterworth, Bessel, Equi-ripple LP, etc.

Your filtering seems to take place in a function called DFD Filtering.vi that was not attached.  It uses the Filter produced by the Express VI Equi-Ripple FIR Lowpass Filter, whose properties I cannot view.

Bob Schor

[feroz_eee]

Yes, Bob. I use FIR filter design using the VIs as part of the Digital Filter Design (DFD) toolkit.  Yes, It uses the Filter produced by the Express VI Equi-Ripple FIR Lowpass Filter.But it seems to be mistake for my case to use such type of DFD filtering.

I send you screen shot of output of my program that I have already sent you.. This is my output. But when I use ADC using NIMAX of Labview, the amplitude at output after filtering falls tremendously.. It is frustrating. 

So, what is your final recommendation? I need to try with  Filters that are on the Filters sub-Palette of the Signal Processing Palette, which include such things as Butterworth, Bessel, Equi-ripple LP, etc. In my case, I can use Low pass filter.. But what's about DC component? Dear Bob, I am doubtful even without DC component's presence I can satisfy my professor. Please let me know, what I should need to do.

Thank you BOB for you so cordial reply.

[Bob_Schor]

My suggestion is to "cheat" (this will probably also work with your fancier filters).  I actually wrote a little VI to test this, and found that if you can "throw away" a bit of the signal at the ends, you can ignore the non-zero mean.  So here's what I did (and I'll post the code, as well, as a LabVIEW 2018 Snippet):

1. Use the Waveform Signal Generator to create Gaussian Noise, N(0, 1).
2. Use the Waveform Scale/Offset to add an offset of 2, so now we have N(2, 1) noise.  This is our "starting Signal".
3. Compute the Mean, then subtract the Mean (by adding an Offset of -Mean) from the Signal.
4. Pass Signal through a Low Pass Butterworth Filter with LP Corner of 20 Hz.
5. Add the Mean back.  Steps 3, 4, 5 are the "cheat".
6. To compare, do the same Low Pass Butterworth Filter of the original N(2, 1) Signal.
7. Plot Steps 2, 3, 4, 5, and 6.  Note that Steps 5 and 6 (the two LP Filter Outputs) agree except at the origin.

Bob Schor 

[feroz_eee]

Dear Bob, I am now trying with my ADC taken as input along with Step Chopping Signal of 2KHz. I shall let you know later whether it is working or not. Thank you very much for your time to time guidance. I am grateful to you