There are many possible sources of discrepancy from expected:
- noise in the signals - depends what methods is being used to extract magnitude / phase. Look up something called a Transfer Function Analyser for how to do it, but which is less sensitive to noise. I don;'t think that is the case as the noise usually features in the obtained bode plot mainly at high frequencies.
- If your sinewave applied to the system and that measured as it's response, are not properly synchronised - for example if there is a delay in the measurement system - then that will corrupt the bode plot produced.
- the number of sine-waves you are using is important - certain methods are sensitive to having integer number of whole sine wave cycles. The more cycles, the better should be the estimates - it just takes longer to run the test.
- mistake in data capture / algorithm use - given the shape of the response you have I would expect that there is some subtle mistake in how you are doing calculations, so best just check through the data at each step to make sure everything is as you expect.
Consultant Control Engineer
www-isc-ltd.com
