Attempts at using waterfall graphs with frequency response to set filters

[3ll3d00d]

Hi

I read the recent thread on waterfall graphs (and using them to check that you have set filters accurately to deal with resonances) and thought I'd give it a try myself. The results were somewhat confusing so I thought I'd post them up to see what advice/comments people have.

The room is ~14' * 12' * 10', the frequency response is shown below. The blue line is the raw response and the red is the corrected response. The filters required are fairly straightforward and the REW suggestions work pretty well. The main issue is that peak at ~37Hz which REW suggested a -12dB cut @ 40-7 (IIRC) with bandwidth of 6/60.



The raw waterfall graph just confirms that there is a big hump at 37Hz



However when I generated the waterfall for the corrected response, 2 things stood out....

- the area from 31-45 is still ringing away for a substantial (albeit reduced) period of time
- there is a slightly odd looking ringing at 120Hz



On the 1st issue I attempted various tweaks last night including minor bandwidth adjustments, moving the centre of the filter up and down a few steps and a short attempt at covering the hump with multiple filters. None of those attempts really reduced the decay and some of them increased it substantially (up to nearly 1s in one case) or introduced odd residual effects further down the frequency scale (a ringing from 600ms onto about 900ms down at about 20Hz). As a result I switched back to the suggested filter.

Am I right in thinking that that hump is actually 2 seperate resonances that need to be dealt? 1 at ~33Hz and 1 at ~38Hz and hence I need 2 narrower filters in place of 1 big cut? Scrolling through the slices seems to say that to me, for example this slice = 13 graph.



The 2nd issue (120Hz ringing) I have no idea about

Note that I haven't done any extended listening with these filters, I was doing it as an almost academic exercise to see if I can objectively tame the decay time first and only then live with it for a while.

All thoughts/comments welcome.

Thanks
Matt

[brucek]

Good job Matt. Yeah, I've played with this myself and have come to the conclusion that at a strong resonant frequency, it's hard to stop that energy from hanging around.

WayneP did some tests recently and had me thinking about this. Certainly, when you lower the SPL level of the resonant frequency, you will get much reduced decay and the waterfall definitely looks better. JohnM's claim (author of REW) was that if the filter perfectly matched a resonance, that the decay waterfall would be perfect. And I guess in theory it should be so, keeping in mind that the filter is only really completely valid at the spot where the response was measured. The effect of the resonance in the room can be looked at as a filter, such that if you can exactly match the filter in gain and bandwidth, it should totally remove its effect at the point of measure.

But, if I feed a speaker two frequencies at equal voltage, and one is at a room resonance, then the resonant frequency will produce a much higher SPL level (as evidence in your graph and waterfall at 37Hz). This requires me to reduce the voltage of the resonant frequency to the speaker, so that the resultant SPL in the room is balanced with the non resonant tone. Now both frequencies are producing the same SPL level, but I haven't removed the fact that the resonance is still there by nature of the room. It appears to me that the energy still remains longer in time than the non resonant one.

Either way, your waterfall sure looks a lot better.

I haven't got a clue about the funny little 120Hz thing...

brucek

[Wayne A. Pflughaupt]

3ll3d00d wrote:

All thoughts/comments welcome.

Here’s the link to the discussion brucek mentioned ( http://www.hometheatershack.com/foru...html#post49933 and following).

At the end of the day, an equalizer is a poor tool for addressing modal decay. Kinda like substituting a screwdriver for a hammer, I don’t know why people even try.

Since you asked for thoughts and comments, mine are:

Absorption to address modal decay.
Equalization to address frequency response problems.
Each is a poor substitute for the other.

Regards,
Wayne

[3ll3d00d]

Wayne A. Pflughaupt wrote:

Absorption to address modal decay.
Equalization to address frequency response problems.
Each is a poor substitute for the other.

Regards,
Wayne

Evening

thanks for the replies, that thread is the one that set me down the path of trying this out. Specifically the idea that if you get the filter exactly right then it will be better then getting the smoothest possible curve.

Absorption is a nice idea for a dedicated room but it definitely isn't appropriate for your average English lounge unfortunately so I have to use the screwdriver to do the best I can. I will have another bash at this next time I get a chance and see if using 2 filters at those frequencies does a better job.

Cheers
Matt

[JohnM]

Looks like you actually have 3 resonances in that region. As mentioned in the other thread, dealing with a large resonance often unveils smaller ones that it was masking. The 120Hz feature looks to be another resonance, but one for which your measuring position is at a null - hence the sharp notch at 120Hz in the frequency response. When your main response has died away that 120Hz resonance is still going but has no contribution from your speaker output left to cancel with so its own output becomes visible.

Note that there is still a hint of the 37Hz resonance around slice 7 or 8 or so, which suggests the bandwidth of the 37Hz filter might benefit from a small tweak (checking the frequency of those little peaks will also tell you whether your filter is at the optimum frequency). Bear in mind that if you put in filters to address the resonances either side of the 37Hz peak the correction needed for the 37Hz peak will be smaller (and closer to what that mode is actually contributing). The 37Hz filter is working very well though, you would need really huge acoustic devices to deal with that passively. Absorption would help a lot for the response from about 100Hz up though.