How to analyse this frequency response measured by labview?

[Lancelot3718]

Hi Coastal

The input sweep sine signal is generated by a piezo-electric actuator, if I put a sweep sine votage on it, it can generate a sweep sine force hit on my cantilever.

Other details I replied to Ravens I put here for you to watch

Hi,Ravens

Sorry, I had to go to sleep last night, you know , it's China.

The target of this vi is to use the sweep sine input to sweep out the frequency characters of the system (my cantilever), and get the response. Then put both the input and output signal (sweep sine excitation and response) into a system identification vi to estimate its transfer function.

The cantilever is not easy to resonate, only occurs at it natural frequences, and the first three orders' natural frequences are 1.6Hz 10Hz 27 Hz (also can be represented by rad/s) .And from the frequency response graph, it is easy to see them, on the left of the yellow cursor.

figure 1: for 0—500Hz input

I do not think the main reason is for the sample time is too short for 0—500Hz, because when I adjust the frequency range to 0—50Hz， it is still the same.  The strange thing is .I the highest frequence input to sitimulate my cantilever is 50Hz, but the measure  vi can still get the response in high frequency range ,like this:

figure 2: for 0—50Hz input

Maybe if we can solve this problem, we can solve all the problems. 

You see in the picure above, when I only use 0—50Hz input to stimulate, the response signal in frequence is supposed to stop by 50Hz, according to the theory about Forcing vibration. But still, there are responses in high frequency range, which is unreasonable.

The most weird thing is all the response signal in frequency flied can be peridicaly divided every 100Hz ，and each period is mirrored at its 50Hz，and the shapes of them are same. It seems that only the first 50Hz is the right result and the frequency range behind is wrong！

It can not be explained by the theory that the ftt should be mirrored at half the sample frequence， if it is，the symmetrical point should be at 500Hz not 50Hz.

[Bob_Schor]

Something is very wrong!  The response to what you say is a 0-50 Hz sweep is virtually identical to the 0-500 sweep response

[Bob_Schor]

Something is very wrong!  The response to what you say is a 0-50 Hz sweep is virtually identical to the 0-500 sweep response -- this seems impossible unless you are not doing what you think you are doing.

I'd like to suggest two additional "proof-of-concept" tests.  First, do one of your experiments (it doesn't matter if you do 0-50 or 0-500), but change your code so that you multiply the values you are sending to your DAC by 0.  This keeps your code intact (in case the problem is in the code), but sends no stimulus to your system, so you should (logically) see nothing as a response.

For the second test, go back to the original code (and actually stimulate the DAC), but disconnect the stimulus (i.e. "cut the wires") as close to the system being stimulated as possible.  This will again not actually stimulate anything (so you should again see a flat response), but if there is electronic cross-talk or feedback in your recording/stimulating system, it could still be present.

I'm not an Electrical Engineer, but this has all the hallmarks of an artifact.  How are your grounds?

Bob Schor

[Lancelot3718]

Hi Bob：

Thanks for your advice，and I follow your words and did some tests~

Here are the results:

First, I used a sine signal of 20 Hz frequence to sitimulate the cantilever and got a response like this:

Figure 1: inupt signal measured by labview

Figure 2: response signal measured by labview

Then, I used a 0 magnitude input to stimulate the system, the input signal had not been measured with values，and the response signal was measured like this:

I know it is weird, but we can find that all the response signal are measured periodicaly with a period of 100Hz，maybe I should call NI's engineers to ask for help.

[CoastalMaineBird]

My next step would be to multiply the MEASURED signal by 0 and see what sort of spectrum you get from that.

In other words, go thru the same motions, but use 0.0 as the input signal and check your processing.

Then try substituting a pure 20 Hz sine wave for the input signal.  Your spectrum should be very pure.