Waterfall interpretation

[brucek]

So who knows what waterfall charts are telling me?

I haven't got a clue what I'm supposed to understand from a waterfall.

Below is a response of my main system. It has always had a nice sharp dip at about 27Hz - I've never been able to hear it. I suspect it's too narrow to create much of a problem. It will not respond to filtering, but the interesting part is I can shut the door between my dining room and kitchen and it completely goes away. I like the door open, so the dip remains.....

I did a waterfall of the response shown below and I see the area around 27Hz is a bit funky too.

So, what does the waterfall say to me. If I showed it to some waterfall expert without the response chart, what will they tell me about, or is it just a pretty picture?


servo-waterfall_response.jpg

servo-waterfalls.jpg

brucek

[SteveCallas]

I believe the waterfall shows how much stored energy is in your room at each frequency, kind of like a decay. It should be pretty equivalent to the "Energy Time Frequency Spectrogram" that AV Talk uses (explained on this page). In a room, as opposed to ground plane outside, it's going to be worse, as you get reflections in a room. This is where bass traps come into play - I think many people underestimate how thick they really need to be to have an effect in the subwoofer range though.

I didn't even realize WinISD was capable of measuring that though - I'll have to get around to one of those sometime.

[Otto]

SteveCallas wrote:

I didn't even realize WinISD was capable of measuring that though - I'll have to get around to one of those sometime.

I believe the waterfall plots are coming from REW.

[SteveCallas]

Oops Lol, that's what I meant, REQW, not WinISD.

[Wayne A. Pflughaupt]

brucek,

So, what does the waterfall say to me. If I showed it to some waterfall expert without the response chart, what will they tell me about, or is it just a pretty picture?

This might help, something I posted on a related thread several months ago (here’s the full thread).

I’m not a waterfall expert, but what it’s telling me is that your decay times for the lowest frequencies are really long.

That 27 Hz thing is a real mystery. It’s showing up on the waterfall graph as a peak, but as a null on the regular graph. Can’t say I’ve ever seen that. You might want to give Ethan Winer a shout.

What kind of a room are we talking about? Symmetrical shoe-box type, or something more complex?

Regards,
Wayne

[brucek]

Thanks Wayne, that was a good article you wrote.

Armed with that info, I played with the waterfall a bit.

I see that just like any graph (duh, stupid me), the scaling has to be set, so I matched the log nature and the same axis as the response graph so I was comparing apples and apples.

I also note the waterfall has some smoothing, so I matched it at 1/6th smooth on the response graph.

The number of slices is quite significant in understanding what you're seeing.

Below I have included the response graph with the sharp dip indicated with the cursor and then I keep it there as I show slices of zero, ten, twenty, thirty in the waterfall.

See how it gets hard to determine where the dip is when you get up to thirty? It's hiding at the back where the 3D nature fools you where it starts. By the fact that the dip has a low amplitude, it is overwhelmed by the information on either side as it decays. See in the twenty plot how the dip is finally lost at about the twelfth slice.

The small hump before the dip seems to create the large final long decay hump...

I don't really know how this can help me, but it's interesting anyway.

smoothed.jpg

Zero slices waterfall
waterfall_zero.jpg

Ten slices waterfall
waterfall_ten.jpg

Twenty slices waterfall
waterfall_twentya.jpg

Thirty slices waterfall
waterfall_thirtya.jpg

brucek

[Wayne A. Pflughaupt]

Hey brucek,

Yup, it looks like the decay is overwhelming the notch.

I’ve played with the waterfall graphing a little, but not much. Have you seen anything in the program that indicates a time value? It would seem to be a “must have,” since waterfalls are graphing time domain. I’d sure be interested to see what your graphs is showing us in real time.

By the way, what kind of a room do you have there, symmetrical shoe-box type, or something more complex?

Regards,
Wayne

[brucek]

The room is a typical living room with an L shape dining room and hall off of it. Then a door from the dining room to the kitchen. That's the door that causes the notch. I shut it and it goes away. Funny stuff.... Not really looking to change anything - it's just an interesting topic....

yeah I have the time axis set at the default of 300msec, but it can be set to anything. It takes to 780msec before that large decay stops.... graph below.


waterfall780ms.jpg

brucek

[JohnM]

Wayne A. Pflughaupt wrote:

Yup, it looks like the decay is overwhelming the notch.

It's not really the notch being overwhelmed as such. The deep notch is from cancellation between the direct sound from the speaker and the output of the room's modal response as measured at the listening position. The waterfall shows what happens after the direct sound stops, so in the absence of direct sound the notch rapidly disappears from the plot and what's left is the response of the mode.

Wayne A. Pflughaupt wrote:

I’ve played with the waterfall graphing a little, but not much. Have you seen anything in the program that indicates a time value? It would seem to be a “must have,” since waterfalls are graphing time domain. I’d sure be interested to see what your graphs is showing us in real time.

The times are shown in the trace labels area of the graph, which shows the time of the front-most slice at the top (changes as you move the slice slider) and the total time span at the bottom.

[Wayne A. Pflughaupt]

Thanks for the info, John.

Regards,
Wayne