Is there any difference between time averaging and vector averaging (both using a trigger) to calculate an impulse response?

[Acoustics]

I want to get the impulse response from a linear system (UUT), with noise (deterministic and random) in the output. I generated a deterministic transient signal (sweept-sine) in the input of the system to get its response (trigger on). I did this many times to get an averaged impulse response. I can do this by two ways:
1) Get the time waveform from the input and output, compute the time syncronous averages and then compute the impulse response from IFFT(FFT(averaged output)/FFT(averaged input)).
2)Get the waveforms, calculate the averaged frequency response using vector averaging and then compute the impulse response from IFFT.
Which is the
best choice?

[Randy_H]

Acoustics,

I searched for other information on our web site on this issue and came across KnowledgeBase 1IME9G0O: What Are the Different Averaging Modes for NI-DSA? [link removed; article no longer available]. In this KnowledgeBase, it says that "Averaging is usually performed on measurement results or on individual spectra, but not directly on the time record." It then discusses vector averaging later in the document. From this I have to imply that the vector averaging is better for you.

Randy Hoskin
Applications Engineer
National Instruments

[LocalDSP]

Whether you compute your impulse response from the frequency response of your averaged time signal or from the vector averaged frequency response should not make any difference. This is also valid whether your input and/or output signals are infected with noise.

The attached VI shows that the results are identical (the difference is zero). In this example (LabVIEW 6i or higher) the excitation signal is a repeated multisine infected with noise, the UUT is an IIR filter simulating your system and the output of the UUT is overlayed with noise as well.