Coherence in REW

[John M Caldwell]

Hi all, I'm new to all of this and am trying to understand certain things that are probably basic to others... I did a search for this a nothing popped up so here's my question. Is there a reason REW does not show coherence for a measurement? In other applications like Smaart and SIM I'll see people looking at coherence, albeit, they're generating pink noise in those apps and REW is using a sine sweep which is very different.

1) So, I guess I'm wondering why you don't need to view the coherence when using a sine sweep? I did come across this explanation from 10 years ago in Gear Space,

"The coherence is the ratio of the averaged cross-power spectrum square divided by the products of the averaged power spectra of input and output for a single measurement (capturing one output sequence for one input sequence) the ratio is 1. It is only when multiple captures are available to average that coherence has meaning."

2) I'm already confused by the above statement because I thought the coherence in Smaart or SIM was using a reference loopback signal that was comparing the output's generated source signal to the incoming input signal... Which is ultimately to tell you how much you can trust your measurement. In the above statement, I feel like this person is trying to say that you only need to view coherence across multiple captures of measurements and not for one individual capture... Perhaps I'm misunderstanding?

This person goes on to say,

"Displaying coherence is essential for 2 channel FFT analysis with stimuli that are not guaranteed to have spectral content throughout the measurement range (such as music or random pink noise) as it identifies areas where the output measurement is unreliable due to insufficient stimulus energy; this is not a factor for sweep measurements. It can also identify poor overall S/N, but with sine sweeps that is evident in the extracted impulse response so, again not really an issue..." There's more but I'll stop there.

What I get from this is that because pink noise is lacking certain attributes that a sine sweep has, you have to have a coherence measurement to check yourself when making decisions.

I feel like I'm babbling a little bit at this point so any further explantations of why we don't need coherence in REW with measurements captured using a sine sweep would be welcome. Thank you.

 

[John Mulcahy]

That's a good explanation. Speaking more roughly, when more than one signal block is analysed the coherence measure is an indication of how consistently they relate to the test signal. If every result is identical coherence is 1 everywhere, if there is an area that is dominated by noise (which might be because the excitation signal had little energy in the region, as is often be the case when using a varying test signal like music) the result will differ from block to block and coherence will be reduced. The concept doesn't apply when analysing single captures, and a sweep is synthesised to have the required energy across the region to be measured so the test signal variability problem doesn't arise.

 

[John M Caldwell]

Okay, I think I'm understanding more. Thanks, John.