Waterfall with >1/24-octave Smoothing

[Albireo]

Hi,

FP for me! Fairly new to REW, but not new to speaker measurements (own TacT RCS 2.2XP). Quick question: possibly I'm being an idiot, but is there a way to specify >1/24-octave smoothing during waterfall generation? I'm using the waterfall plot for 400Hz-30kHz as a proxy for cumulative spectral delay and the noise makes it a little tough to analyze in that frequency range.

Also is it possible to specify a much longer time range, say up to 5000ms? Finally, I second the earlier request to lower the window limit to 1ms if possible.

Apologies, lots of requests for a first post. Thanks for a wonderful program!

[JohnM]

The trace smoothing only applies to the frequency responses on the Measured and Filter Adjustment tabs. For the waterfall the key is to use a window that is narrow enough for the frequency range being measured, which will have a dramatic smoothing effect of its own at small window settings (e.g. a 1ms window would mean a frequency resolution of 1kHz) so the reduced window limit should address this. Waterfalls at high frequencies using the current 50ms lower window limit include too much of the room's contributions causing a lot of comb filtering.

Can you explain why you want a longer time range? The captured impulse response only extends about 1.75s after the peak, and will usually be well below the noise floor long before that.

[Albireo]

Thank you very much for the info. That is a very reasonable rationale for the 50ms window limit.

As for the longer time range, it is only a desire to see more! On some curiously ill-designed speakers in a poor room even some of upper frequency bands will ring and ring, but this is currently cut off at the current 1s limit.

Out of curiosity, and without knowing anything about the MLSSA system: how does the REW waterfall differ from something like Stereophile's cumulative spectral delay charts?

[JohnM]

Albireo wrote:

As for the longer time range, it is only a desire to see more! On some curiously ill-designed speakers in a poor room even some of upper frequency bands will ring and ring, but this is currently cut off at the current 1s limit.

Whilst not impossible, that is very unlikely. Decay times of many seconds can happen in very large spaces (e.g. cathedrals) but not in domestic-sized spaces, if they appear in measurements the cause is more likely interference of some sort, an instability/oscillation of the amplifier or a measurement artefact than an acoustic resonance. Check the impulse response to see whether what you are seeing actually extends for the time period it appears to, if not you need to look at what settings you are using in the plot that gives this impression.

Albireo wrote:

Out of curiosity, and without knowing anything about the MLSSA system: how does the REW waterfall differ from something like Stereophile's cumulative spectral delay charts?

The difference between waterfall and CSD plots is what happens to the right hand side of the impulse response window. In a CSD plot the window RHS is fixed and the LHS moves towards it, that is why the low freq limit of a CSD increases in later slices. In a waterfall plot the whole window is moved to the right.

Note that CSD and waterfall plots involve a trade-off between time and frequency resolution. For high time resolution you need short windows so that you get a better picture of what is happening in the response as the window moves, but short windows mean poor frequency resolution. Longer windows give better frequency resolution but at the cost of time resolution. There limitations are connected to the near universal use of the FFT to produce frequency responses, but you will have to wait for the next release of REW to see the improvements that are possible with some of the more recent developments in spectral analysis techniques.

[Albireo]

Fantastic, thank you very much for the detailed responses. I'll have a look at my plots tonight.

[atledreier]

John, you tease!