[ekovalsky]

I had my speakers on a long wall setup but decided to try the more usual short wall placement.  The measurements I took were rather surprising. 

The long wall setup had prominent resonances at the predicted frequences and an extremely deep null at the listening position (along center line of long wall, in between speakers) that I cannot explain from room modes or quarter wave cancellation.  The short wall setup had a greater number of modes and nulls, the frequencies of which were generally predicatable from room dimensions and quarter wave cancellations, but they were lower in amplitude. 

Since I do DSP correction I can compensate somewhat for room bass issues.  The correction needed is less for the short wall setup, the deep null in the long wall setup could not be corrected and remained a major suckout.  But I think soundstaging was better with the long wall setup.

The measurements I took were not dissimilar to this graph featured on the Realtraps site, although I only had one extremely deep null at around 55hz.



I was curious why the bass behavior is so different, particularly in my perfectly symmetric room which is only slightly rectangular at 14.5' x 17' with 10' ceiling.  Also any idea what would produce the dramatic null @ 55hz along the center in my long wall setup ?  It was definitely related to position along the long wall, i.e. left and right from the listening position, and not front to back or floor to ceiling distance.  The predicted nodes are 33hz and 39hz, also 57hz for height.  Distances between listening position and speakers and the walls would not suggest a quarter wave cancellation at that frequency.

[bpape]

But seating distance in relation to the length of the room most likely did. 

Bryan

[Ethan Winer]

Distances between listening position and speakers and the walls would not suggest a quarter wave cancellation at that frequency.

Are you sure? The peaks and nulls related to the distance between you and the wall behind you are almost always exactly where they should be. Since you didn't mention the distance from the measuring microphone to the rear wall, I can't confirm or deny your findings.

Also:

> I think soundstaging was better with the long wall setup. <

This is typical because the side walls are farther away. But early reflections are easy to treat with "thin" absorption. Bass peaks and nulls are much more difficult, so I favor whatever placement makes the low end flattest.

--Ethan

[ekovalsky]

Are you sure? The peaks and nulls related to the distance between you and the wall behind you are almost always exactly where they should be. Since you didn't mention the distance from the measuring microphone to the rear wall, I can't confirm or deny your findings.

Pretty sure.  The ~ 55hz null seemed more or less independent of the distance between the microphone and the rear wall, and also height off the floor. It was almost completely dependent on distance from the side walls, i.e. at 8.5' with the microphone centered between the speakers on the long wall.  Quarter wave cancellation at this distance should be 33hz, 66hz, etc.  Distance from mic to the rear wall was about 3.25' so cancellation would around 87hz was expected.  There were some nulls at these frequencies but they were relatively small, at least compared with the big one at 55hz.

The 55hz null should be from cancellation occuring with a boundary about 5.25' away from the mic.  There was nothing at this distance.  My knowledge of acoustics is very limited and I was wondering if oblique or tangential nulls, or other factors I am unaware of, become more relevant with the long wall placement.  Clearly the behavior of a rectangular room changes depending on whether the speakers go on the longer or shorter wall.

[Ethan Winer]

The 55hz null should be from cancellation occuring with a boundary about 5.25' away from the mic.  There was nothing at this distance.

The distance changes with angles. This is one problem when trying to understand 1/4 wavelength interference in a small room. Outdoors with a single large reflection from a building, the 1/4 wavelength distances will be perfectly predictable. Aside from contributions from the ground. After you add four walls and a ceiling, there are so many competing reflections it's hard to know what null is from what reflection. Indeed, the response at any given cubic inch within a room is the sum of the direct sound and many competing reflections.

A few years ago I was preparing an article for EQ magazine about this, and I took the photo and measurement shown below. The Radio Shack SPL meter is 20 inches from a reflecting side wall, and the graph shows the response. Just before I sent in the article I checked the frequencies in the graph with calculations for 20 inches. Oops - they were way off because the speaker was behind the mike and well over to the left. So I moved the speaker to be directly behind the microphone, and then the frequencies were correct.

BTW, this graph shows clearly why having a wall not far behind the listening position is very bad.

--Ethan

[ctviggen]

I think Ekolvasky's software uses MLS, which isn't that good at low frequency ranges.  See, eg, page 57 of the following:

http://www.gedlee.com/downloads/Chapter4.pdf

Where Earl Geddes indicates that these types of analyses are poor under about 200Hz:

The windowing function can have an effect on the low frequency response and this is why the frequencies at or below 200Hz are not considered to be valid. The effect is unpredictable.

A better analysis would be longer-term, frequency dependent analysis such as that performed by the low frequency tests of ETF.  Even Ethan's data below 200Hz is suspect, as that graph was performed by MLS. 

[Ethan Winer]

Bob,

The windowing function can have an effect on the low frequency response and this is why the frequencies at or below 200Hz are not considered to be valid. The effect is unpredictable.

I'm not a math guy, but my understanding is that any of the common signal sources can give highly accurate results if they're resolved highly enough. The advantage of a swept sine wave is it allows the use of a tracking filter to increase the signal to noise ratio. Since the software knows what frequency it's sending at any given moment, it can apply a sharp bandpass filter later when analyzing the data to pass only the current frequency of interest. For example, if you stand near the microphone and breathe too loudly during the test, the breath sound could raise the measured signal level. With a tracking filter the breath sound will be removed and won't influence the results. A good s/n ratio affects mainly nulls - the nulls won't appear as deep as they really are.

--Ethan