Extract single tone information VI: uncertainty of detected phase

[Puc]

Hi,

I'm a student at university, using LabView to make a lab experience with an Elvis II. I took data with the "Extract Single Tone Information" VI and I am now analyze them. I read ELVIS spec. and I work out the uncertainties of "Detected Amplitudes" to be around 2mV, but I can't find how much are the ones on the "Detected Phase" output. Is there a declared error or some standard I can use? Or I need to look at all the code down the VI and guess how much is the error?

Thank you a lot, you're saving my exam.

[Bob_Schor]

When I was a student, I used a sine wave generator to generate low frequency (0.02 to 0.5 Hz) stimuli and recording responses.  I didn't have LabVIEW or its functions, but I did have access to a Card Punch and a computer that understood Fortran.  I knew how to fit functions (like a Sinusoid) to my data, and I knew enough Statistics to know how to compute "Goodness of Fit", to get estimates for the Mean, amplitude of the fit Sinusoid, phase of the fit Sinusoid, and Error Estimates for these three parameters.

I suggest you learn enough signal analysis to understand how to compute the errors of your fit, and how to use that error to estimate the errors of the parameters (Mean, Amplitude, and Phase) of the parameters of your fit.

Bob Schor

[Puc]

Maybe when you were student you had more time to actually learn those things. We had been given few time to analyze data and made a report, since we can indeed use these instruments like LabView even without knowing what they do in every details.
I was just asking for a specific or similar of this instrument "given as it is".

Thank you anyway, but I'm sorry, you're not helping so much by showing what you used to do while being a student.

[Henrik_Volkers]

Some information ist found in the presentation found here

https://forums.ni.com/t5/LabVIEW/Extract-Single-Tone-Information-from-Hann-Spectrum/m-p/139348#M8364...

or

https://forums.ni.com/t5/Signal-Conditioning/frequency-of-sine-waveform-VI/m-p/2638895#M7428

however I suggest to analyse your signal ... noise level, bias drift

since the result depends on number of cycles of the input and noise and other parameters,   

measure in your application the noise floor, add a mathematical known sine (magnitude and phase) and test your application.

Keep in mind that the timestamp of your wfrm is ignored. the tone detection just take the dbl array of data.

In cases where you measure two wfrms another problem arrise: tone detection on two wfrms will come up with two different frequencies (small deviation, but ...)

So I programmed a 4 parameter sine fit (IEEE STD 1057) that matches both wfrms to one frequency. (sync aquisition assumed 😉 ) an found a better performance on noisy signals (on both channels, if one channel is nealrly noise free, another strategy is used). Sorry can't share that vi.

Even better: If your signal source is in sync with your aquisition, you exactly know the frequency and a 3 parameter sine fit is the best I ever tested and used.

EDIT: On the 3 or 4 parameter linear fit, you can find papers that investigate the errors of these methods. But since time and analog resolution and noise are unique for every system, a real uncertainty estination involves own investigations. ... or a generous factor 😄

[Bob_Schor]

---

@Puc wrote:

Maybe when you were student you had more time to actually learn those things. We had been given few time to analyze data and made a report, since we can indeed use these instruments like LabView even without knowing what they do in every details.
I was just asking for a specific or similar of this instrument "given as it is".

Thank you anyway, but I'm sorry, you're not helping so much by showing what you used to do while being a student.

---

So what, in fact, did you do?  Or are you asking (as many on the Forum do, I regret) "Please do my Homework (or Classwork) for me"?  We could provide much more targetted advice if you "show your work", that is, attach (all of) the LabVIEW code (not pictures, but actual VIs) that you are trying to use.

As a suggestion, you can test your algorithm by asking LabVIEW to generate a signal for you, where you can specify a sine wave, set the frequency, amplitude and phase, and add noise if you want.  Look on the Waveform Palette, under "Analog Wfm", "Generation", and choose the Express VI "Simulate Sig".  It may take a few seconds to initialize, but then you can set it up to Simulate a Signal (that you specify).  You can play this signal into whatever VI or functions you are using to analyze it and see if you get the Signal Parameters that you gave to the Generator.  You can also add Noise (I recommend Gaussian, as this is a good model for "noise in the real world") and see how it affects the Signal Parameters your analysis gives to you.  If you are really curious, you can "do an experiment" (and, dare I say it, maybe Learn Something Useful) -- do several runs, varying the Signal-to-Noise (usually Amplitude of Sinusoid / RMS value of Noise).  Do maybe 10 runs at each S/N and keep track of the Gain and Phase measurements.  Note that they will vary from the "true" values (because you added Noise).  Can you see a pattern in the variations of, say, the measured Gain and Phase at a particular Noise Level as you change S/N?  Plot it as a function of S/N.  Notice anything interesting?

Bob Schor