When I take frf from experiments with truncation signal, I got only noisy frf. In order to get good frf, I have to use the windowing technique in signal processing technique.
In impulse testing use the exponential window on the response. I don'y know what exactly is your test but if you use also impact/modal hammer for excitation, use the force window on this channel.
In general there is no optimal window for a given measurement. Every applied window triggers a variance/bias trade-off. The smoother the window the wider the main-lobe. As a result the noise is suppressed in the FRF but the frequency mixing due to the wider main-lobe increases the bias. A modern approach which tries to keep the good properties of windowing while controlling the bias is the local polynomial method for FRF measurements.
I don't know how about the potynomial method but in my opinion you are wron writing that
"there is no optimal window for a given measurement". The thing in proper window selection is very often the energy you "throw away". If you use for exammple Hamming window for single impact testing you will cut away almost all information form the trace. In proper applied exponential window the infuence is signifacnatly lower. True is that window always distorts data but you can choose the one which give you what you need with the lowest influence on the original signal.
If you find the problem worth of computation, use an adaptive method to evaluate the error due to a specific window. You can use a (multi) parameter window model, and iteratively find the best one for your kin or class of signal. If this class is completely random, the choice may be of a rectangular window. for example.
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing