Tangential Modes: A Case Study

Posted: Tue Mar 22, 2005 3:25 pm

Great stuff Dave, thanks!

Ethan wrote:

Quote:

The lowest mode is off by almost 10 percent, so I'd be careful not to assume too much at the other frequencies. However, if non-rigid walls can be shown to account for the disparities, it is possible that the disparity becomes less the higher in frequency you look. Though so far nobody has explained the mechanism for this shift other than to blame "construction" generally.

Please do not try and pul things were not written, obliquely into this thread. I clearly gaie a list of half a dozen specific possibilities, nothing of a general nature. Incidently ALL of the reasons/causes could be described as "construction".

Andre

Posted: Tue Mar 22, 2005 3:28 pm

Jeff,

> The common thread between both your (David's and Ethan's) analyses was ETF <

That's why in my very first post I showed a comparison of ETF and sine waves in the same room, to make clear that ETF is probably not in error.

> Remember, too, that we will undoubtedly generate 10^n questions for every n answers we think we have. Wink

<

That was exactly my point with starting all this - to ecourage a discussion of important acoustic principles and hopefully learn more myself. That's also the same motivation for my starting Do room modes even matter? and the rigid fiberglass Density report. And before that starting a discussion of modal versus non-modal peaks and nulls. I have plenty more up my sleeve too. Cool

--Ethan

Posted: Tue Mar 22, 2005 3:33 pm

Andre,

> Please do not try and pul things <

Please stop being so agressive, okay? I'm not here to fight with you or anyone else, and I will not participate if you won't behave like a human being. At John Sayers' site you promised Steve that you would try to be civil, so please knock if off, okay?

--Ethan

Posted: Tue Mar 22, 2005 3:39 pm

David,

> Like Jeff said, using percent is not a good idea. <

I could not disagree more. As I explained earlier, everything in audio is percentages - bandwidth, decibels, frequency response, room mode ratios, distortion, wow and flutter - everything. The only exception I can think of offhand is the lone FFT windows example Jeff mentioned.

--Ethan

Posted: Tue Mar 22, 2005 3:43 pm

David,

> the speaker is located very near a null for the (0,2,0) 115 Hz mode, so it would not be active. <

Since you knew that going in, it would have been useful if you'd run the response a second time with the speaker moved a few feet.

> How do your argument deal with this? <

I'll say this to you too: I am not here to argue! I am here to start interesting discussions and hopefully learn a little too. On the very first page of the initial thread I stated, "Please understand that I'm not taking a position either way! I just want to get to the bottom of this using believeable and repeatable measurements and data."

--Ethan

Posted: Tue Mar 22, 2005 4:34 pm

Ethan Winer wrote:

I am not here to argue!

Me either; I think you're still being prematurely defensive. I'm looking forward to further debate as this is how we learn things.

So how does your argument deal with this?

Posted: Tue Mar 22, 2005 4:44 pm

David,

> I'm looking forward to further debate as this is how we learn things. <

Great, me too!

> I think you're still being prematurely defensive. <

Could be. If you look at the tone of my posts versus the tone of far too many posts by others, you may be able to understand why this is an unfortunate reaction.

> So how does your argument deal with this? <

I have no real argument, only more questions. How close the loudspeaker was to the exact null center is one question, and how much stronger or weaker that mode might have been an inch or two away is another question.

You may have seen a point I've made a few times about deep nulls and placement I once measured. If you find the exact center of a null then move the test mike even a few inches, you can get a 15 to 20 dB increase in level. Nulls are extremely localized, and it takes very little displacement to recover most of the energy.

I guess you could say it's a "percentage" thing. Very Happy

--Ethan

Posted: Tue Mar 22, 2005 9:09 pm

hz vs % vs octaves

The problem with hertz is 18000hz vs 18005hz I can't hear a difference, yet with 20hz vs 25hz there's quite a difference.

Technically we should probably be saying such and such a frequency is "X octaves" off another frequency -- which would give any musician an idea how far it was out. {i.e. difference in octaves = log(f1 / f2) / log(2) }

%, while less accurate, has the advantage of being trivial to calculate.

Posted: Tue Mar 22, 2005 10:51 pm

When talking about pitch perception I'd have to agree, but we are not talking about perception; rather, we are talking about FFT data displayed on a linear scale.

Posted: Wed Mar 23, 2005 1:29 am

David:

Quote:

on a linear scale

I just went back to look at your graphs, and low and behold they are on a linear scale. (It's been a while since I've run ETF5, and I remember choosing a log scale, but apparently I also only did it for the full frequency (20hz to 20000hz) and not for any of the LF graphs)

Posted: Wed Mar 23, 2005 3:40 am

For very low frequencies out ear seems to work more lineair, for full spectrum it is more logaritmic.

Posted: Wed Mar 23, 2005 8:52 am

Ethan,

Ethan Winer wrote:

I could not disagree more. As I explained earlier, everything in audio is percentages - bandwidth, decibels, frequency response, room mode ratios, distortion, wow and flutter - everything. The only exception I can think of offhand is the lone FFT windows example Jeff mentioned.

That's a great observation. I had not thought of that before. Please note that I've prepared a graph of the measured room response from the other thread to reflect this monumental insight:

[url=http://www.auralex.com/savant/Percentages(Ahem).JPG]% graph[/url]

My apologies if I've caused any confusion with my observation of tradition, convention, and my general misguided adherence to the status quo. Embarassed

Posted: Wed Mar 23, 2005 11:59 am

Jeff,

> I've prepared a graph of the measured room response from the other thread to reflect this monumental insight: <

However, I found a very tiny technical error. Really, it's so small I'm almost embarrassed to mention it: Since your graph starts at 0% I believe you need to scale all the higher percentages up by quite a lot. Perhaps Infinity would be the correct amount?

Confused

--Ethan

Posted: Wed Mar 23, 2005 12:04 pm

Gosh, I missed that. Dang. But I thought your standard for graphs was for the ordinate to start at 1 instead of 0? Perhaps the abscissa as well? Either way, I'll get right on it. I'm 97% of the way there...±20% of course... Confused

Posted: Wed Mar 23, 2005 12:08 pm

David,

> we are talking about FFT data displayed on a linear scale. <

Showing frequency response on a linear scale is rare, and I see that as design flaw of ETF. However, a linear scale is good for identifying comb filtering, so perhaps that's what Doug Plumb had in mind with the LF response display.

The reason FFT data comes out as per-Hz bandwidth is because that's just how the underlying code works. Likewise for similar processes like software broadband noise reduction. But that's not how audio and music work inherently. As I've explained twice now, everything in audio is ratios. You can also add to my other examples the fact that musical intervals are ratios too.

--Ethan

This topic sponsored by: