Phase Measurement

[Andy F.]

I noticed that you posted this question twice.  Please see the following thread for an answer.

Frequency Sweeps

Thanks,
Andy F.

[ResearchGuy]

Randall,

Thanks for the information. Will look into the FFT Filter.vi attached in your reply. Another observation made during my data acquisition process regarding the phase algorithm is this: when both i/p and o/p are almost in phase (i.e., when the phase is very close to zero), the algorithm throws out random phase information. By random I mean, the data obtained is not an accurate representation of the actual measurement (by visually looking at the waveforms, one can arrive at a conclusion that the waveforms are "almost" in phase) . To explain further, if the phase is say around +/- 0.5 degree, the measurement shows up anywhere between -2 to 4 degrees. This is so random in nature, that it is really difficult to predict a range. Is it because of the very small phase difference that the phase algorithm's accuracy of measurement is lost? I have attached a Phase.jpg file for your reference. This is the most current version I am using. The mean and standard deviation are to keep track of the accuracy of the phase measurement. By looking at these numbers and the actual phase data, I find out the optimal sampling rate for a given input frequency, setting the number of points to 2000 (constant). Let me know what you think. Thanks!

[ResearchGuy]

Randall,

Here is the Phase.jpg file I forgot to attach in my previous email.

[ResearchGuy]

Randall,

Getting down the basics a little bit, if you don't mind. I couldn't understand the concept of acquiring 'x' no of cycles for 2000 points. How does one go about doing this while acquiring data? Attached is word document file with a range of i/p frequencies and sampling rates I have been using. It also includes the no of cycles required for 2000 points. Can you please explain this concept. Thanks.

[rpursley8]

To get an accurate phase measurement, you must not have any DC offset in your acquisition.  If you acquire a portion of a sinusoid cycle instead of the whole thing, you will have a DC offset in your acquisition (think integrating across a sine wave, if you integrate across a whole cycle, your integral result is zero).  When you compute the phase, the idea is to have one signal present (sine wave).  The presence of the DC component adds another component to the phase measurement (an unwanted component).

Therefore, you need to make sure that you acquire a whole number cycles to make the DC offset zero which is the whole point of this part of the discussion.  If the numbers you posted are the whole list of input and sampling frequencies you want to use, then sampling 10,000 points will give you at least 10 cycles of each input frequency, but any multiple of 2000 points will work for all the combinations.