Waterfall "Window" Parameter - please explain it to me - Home Theater Forum and Systems - HomeTheaterShack.com

Old 05-10-11, 08:52 AM Thread Starter
Shackster

Join Date: Oct 2010
Posts: 44
Waterfall "Window" Parameter - please explain it to me

Hi all

I've RTM and had a look through a few posts but I'm still confused as to what the "Window (ms)" parameter in the Controls section of the Waterfall graph does. I just don't get it! I wondered if anyone could explain it to me or link me to a thorough explanation?

Cheers

Max

Old 05-10-11, 10:40 AM
REW Author

John

Join Date: Apr 2006
Location: UK
Posts: 6,297
Re: Waterfall "Window" Parameter - please explain it to me

Frequency response plots are generated by taking an impulse response, applying a window (sometimes called a gate) to it to select a portion of the response for analysis and then calculating the frequency content of that windowed portion (using a Fast Fourier Transform or FFT). When looking at a whole room response the window tends to be fairly long, up to a second or so, to capture both the direct contribution from the source and the later arrivals reflected from the surfaces of the room. When trying to isolate the direct sound the window needs to be short, to exclude the contributions from reflections.

In a waterfall we are interested in how the frequency content of the impulse response is changing over time, so that persistent features (room resonances typically) can be identified. To produce the waterfall a smaller window is used than when looking at the whole response, for example 300ms. The first windowed section includes the impulse response peak, and so includes all the direct sound and the earlier contributions of the room. Analysing that first section produces the first (rearmost) slice of the waterfall. The window is then shifted along the impulse a little and a fresh calculation of the frequency response is carried out to produce the second slice. This is repeated to produce all the slices that make up the waterfall (the Slice control lets you move between the slices), moving the window over a total range set by the time range control.

When choosing the window width for a waterfall (or spectral decay, which is a 2D view of the same thing) a compromise has to be reached between the frequency resolution (how well we can identify different frequencies in the slices) and the time resolution (how well we can see how quickly things change over time). Long windows give high frequency resolution but, because you in essence see the average level of each frequency within the time span of the window, lower time resolution. Short windows improve the time resolution, but give lower frequency resolution. For room responses and analysing modal resonances good results are usually obtained by choosing a window width of around 300 - 600ms and choosing the time range to be about the same as the window width, for longer window widths a longer time range helps to show the evolution of the response over time.
JohnM is online now
Old 05-10-11, 05:49 PM Thread Starter
Shackster

Join Date: Oct 2010
Posts: 44
Re: Waterfall "Window" Parameter - please explain it to me

Hi John

I really appreciate the time you took to write that explanation. Unfortunately I'm as thick as two short planks and still don't get it! I'm sorry and don't mean to awkward. I'm trying to wrap my head around it but it just won't work for me.

Perhaps if I illustrate my confusion with an example..... Here's three waterfalls of the same measurement:

Time 300/Window 300:

Time 500/Window 500:

Time 500/Window 300:

Now, just for examples sake, let's look at 500Hz:

Pic 1: (300/300). The last data for 500Hz is 52.5dB at 140ms

Pic 2: (500/500). The last data for 500Hz is 52.5dB at 215ms

Pic 3: (500/300). The last data for 500Hz is 50dB at 131ms

And no doubt the more changes I made to those two parameters, the more different results I would get.

That's where I'm mostly confused (aside from still not begin able to grasp what it does!). In this practical example, changing those parameters changes the reading for a frequency and it's ringing. But it is the same measurement in all three pics. So, if I were to base decisions on acoustic treatments on these readings, my decision would be different depending on which I chose to go by.

Please note that I'm saying all of this just to try to give an insight into why I don't understand and nothing more! Hopefully it will help someone to help me!

Many thanks

Max
Old 05-11-11, 04:28 AM
REW Author

John

Join Date: Apr 2006
Location: UK
Posts: 6,297
Re: Waterfall "Window" Parameter - please explain it to me

Note that the data doesn't actually stop at the floor of the waterfall, it is just that the values are below that level - if you shift the waterfall upwards to reveal lower levels you will see them.

The utility of the waterfall is in assessing the uniformity of decays. An ideal (unattainable) waterfall would have everything decaying at the same rate from an initially flat response, with no outcrops. The waterfall is good at showing the parts that are decaying more slowly and which need attention, such as the resonance in your plots just below 40Hz. The better the treatment is working out, the more uniform the decay becomes. The absolute levels (50dB here, 70dB there, whatever) do not convey useful information unless you are comparing two measurements from the same position with the same window and time span, for example to assess whether a tuned trap has helped speed up the decay of a resonance or whether broadband absorption has helped bring more uniformity to the mid band decays. The information content of a waterfall is in its shape, but you need to be consistent in the parameter settings to have a basis for comparison between measurements.
JohnM is online now
Old 05-11-11, 05:10 AM Thread Starter
Shackster

Join Date: Oct 2010
Posts: 44
Re: Waterfall "Window" Parameter - please explain it to me

Thanks John. The need to be consistent across measurements makes total sense, as does the usage of the WF for identifying decay and the effects of LF bass trapping.

As for the function of the "window" setting, it still baffles me a bit, but I'm going to read through again, have another play with them, and see if the penny drops. I did also post this query over on GS - here's the link:

http://www.gearslutz.com/board/studi...xplain-me.html

As I end up asking on that thread, and bearing in mind what you've said in your last post, is there such thing as a recommended setting for window and time for my type of circumstances?

Cheers

Max
Old 05-11-11, 05:35 AM
REW Author

John

Join Date: Apr 2006
Location: UK
Posts: 6,297
Re: Waterfall "Window" Parameter - please explain it to me

For general treatment evaluation the 300ms/300ms REW defaults work fine. For more specific tuned traps using a longer window and correspondingly longer time span can offer more accurate identification of specific problem frequencies, but those kind of investigations are better done via the EQ window using the modal resonance analysis tools.

To understand the windowing process it may help to think about a slightly different scenario. Say you set up to record the sound of a triangle struck once. The signal you capture will be pretty much the impulse response of the instrument. You load the signal into your waveform editor and take a clip of the first 300ms of the signal, with the content before and after that forced to zero. You then take a look at the spectrum of that clip to see what frequencies it contains. Then take another clip, but starting and ending 10ms (say) later than the first and look at the spectrum of that section. Then do another clip, starting 20ms in. That is exactly the same process as the waterfall plot uses. Increasing the window is the same as taking a longer clip. Increasing the time range is the same as moving the start of each clip by a bigger step.
JohnM is online now
Old 05-13-11, 02:55 AM Thread Starter
Shackster

Join Date: Oct 2010
Posts: 44
Re: Waterfall "Window" Parameter - please explain it to me

Hi

John - that last explanation was really helpful in explaining what is going on, so thanks for that.

However...... Lying in bed last night - much to the disgust of the missus - another thought occurred to me.

I've read several times that an even-ish decay over the frequency spectrum of around 300ms is a good aim for a small control room. I've been looking at my waterfalls with this in mind. But as I get a better idea of how changing the parameter changes the presentation of the data, it seems more prudent than ever to choose an appropriate time and window. Otherwise, I might end up making adjustments to the room to achieve 300ms, which - when viewed with a different time/window - is no longer 300ms.

So while I fully appreciate how keeping the parameters consistent when making comparisons is a good idea so that things remain relative to each other, when assessing for a desired level of decay wouldn't this be less relevant?

If so, would there be a recommended setting in this example (aim = 300ms decay). Or would the waterfall no longer be the best tool in this situation (and if not, what would be)?

Cheers

Max
Old 05-13-11, 07:55 AM
REW Author

John

Join Date: Apr 2006
Location: UK
Posts: 6,297
Re: Waterfall "Window" Parameter - please explain it to me

The RT60 plot, whilst not strictly applicable to (acoustically) small rooms, is the easiest way to get a view of the uniformity of decay as an absolute figure.
JohnM is online now

 Bookmarks

 Tags "window" , explain , parameter , waterfall

Message:
Options

## Register Now

Random Question
Random Question #2

User Name:
OR

## Log-in

Human Verification

In order to verify that you are a human and not a spam bot, please enter the answer into the following box below based on the instructions contained in the graphic.