REW Window length, FFT size and a few other questions

[David Gilfillan]

Sure much of this has been covered before, but after a search, a few questions I'm not certain about:
What is the relationship between the window length and FFT (or DFT) size in REW?
Can the two be changed independently once an impulse is measured and calculated? (i.e. retain a constant FFT size and change the window length or vice-versa)
Is an impulse zero-padded after the window ends when the window length is smaller than the FFT size? (is this possible)
Can I export magnitude and phase (Bode plot) with a frequency axis and a fixed (known) FFT size?
Can phase be optionally calculated from the window start rather than time 0? This would help with my on-site confusion about arrival time following a review of phase.

Thanks for the fantastic software and all the time that goes into this forum.

 

[John Mulcahy]

What is the relationship between the window length and FFT (or DFT) size in REW?

The FFT will be the next power of 2 greater than the number of windowed samples that gives bin intervals of 0.5 Hz or less, so it is usually 128k for 44.1 kHz or 48 kHz sampling unless the window is very long.

Can the two be changed independently once an impulse is measured and calculated? (i.e. retain a constant FFT size and change the window length or vice-versa)

You can set the windows to whatever you like.

Is an impulse zero-padded after the window ends when the window length is smaller than the FFT size? (is this possible)

Of course.

Can I export magnitude and phase (Bode plot) with a frequency axis and a fixed (known) FFT size?

Sort of. If you deselect the analysis option to allow 96 PPO log spacing measurements will retain linear spacing determined by the FFT length. You can export the measurement as text at its native spacing. If you want the export to have very low frequencies (< 2 Hz) you will need to change the View preference for Frequency axis start to a sufficiently low value. The export will not have entries for DC or Nyquist. Alternatively you can export the windowed IR as a WAV file and process it externally.

Can phase be optionally calculated from the window start rather than time 0? This would help with my on-site confusion about arrival time following a review of phase.

No, but you can apply a timing offset when measuring to put t=0 at your preferred position.

 

[David Gilfillan]

all clear thank you.

Can phase be optionally calculated from the window start rather than time 0? This would help with my on-site confusion about arrival time following a review of phase.

your reply: No, but you can apply a timing offset when measuring to put t=0 at your preferred position.

I guess I'm stuck in the past here - I have to get my head around moving the time axis. I get very confused on site trying to compare phase of loudspeakers (e.g. for active xover alignment) using acoustic measurements and moving t=0 separately from the window. A "phase ref" line associated with the phase response might be useful...

What actually is t=0? How is it placed?

 

[John Mulcahy]

Whenever you use the delay finder in a dual channel FFT system you are shifting t=0. Ideally it is placed where the response starts, so all the delays from processing and time of flight are removed. That way you see the underlying phase response. REW can estimate where that would be using the "Estimate IR delay" button in the SPL & Phase or Impulse graph controls.

When measurements are made with a timing reference the timing of the reference is used to place t=0, that placement can be shifted using the timing offset field on the Measure dialog. That offset might be used when making measurements with a loopback timing reference when the mic is at a known distance from a source, e.g. 1 m speaker response measurements, entering a delay that corresponds to 1 m. REW also has an acoustic timing reference option which uses the time of arrival at the mic of a signal from a reference speaker. That is convenient for aligning speakers as it provides direct timing comparisons with the reference speaker.