Ethan Winer wrote:
I've seen negative numbers reported at low frequencies, and I'm pretty sure a second test five minutes later would give different results. Perhaps very different results. I assume the main reason for run-to-run variation is the moving microphone.
The reason for these negative numbers lies in the fact that there is much too little absorption present (< 1m2 added absorption, not physical material).
It's some slogan you mentioned a lot anywhere already to substanciate whatever. It's just a logical physical limitation.
You noticed these negative numbers in these microtraps where they hardly show absorption in the lows.
You compare an empty room with a certain soundfield versus a room where hardly any absorption is added, but whatever added absorption the soundfield will still, even when it's a bit, an altered soundfield.
This falls outside ANY measurement accuracy, hence these values have no sense, and point to physical limitations, not a weak method.
In a reverb room it's of no importance that you add a number of m2 or sft physical material, you have to add a minimum quantity of additional absorption.
There is NO way you can prevent such tolerances with whatever test method.
And it's not even important, since it only occurs in the range where the added absorption material hardly makes a difference with the empty room, hence having very little absorption.
Normally these bands will never be shown in whatever official report.
Since measurements are compared which aren't done at the same time (first empty room, later room with sample), one does assume in the calculation method that the part empty room comprised in the room with sample is still exactly the same as that first measurement.
Since you almost added no other additional absorption, by definition these tolerances will influence that result.
That empty room part in that measurement with sample is too dominant in these numbers. Only a minor mini deviation can result in such seemingly stupid numbers.
As an analogy (poetic license), that's like trying to measure the weigh of a 2 pound package of salt by measuring a truck on a weighbridge, then putting that package of salt on that truck and weighing the truck again on that same weighbridge, hoping the difference represents the weight of that package of salt. It's very well possible that that package of salt calculated as such results in a zero or negative weight also.
No matter what this calculation sais: you measure tolerances of whatever phenomenon (-na), not or hardly the weight of that package of salt.
And it is clear the lower the original weight of that package of salt the larger that mistake expressed in percent, or relative absolute, can become.
But that doesn't matter, the outcome is senseless and recognized as such.
Back to reverb room measurements it will show that effect only in a range where there is hardly any or little absorption to speak of. Hence people don't miss out on useful but lacking data.
Hence that argument is no argument to accept or reject whatever method for practical applications. I think you use it because, without whatever context, it sounds impressive for layman.
The only manner one possibly could minimize this error is in an impedance tube (is very controlled field), where then other limitations occur.
E.g. : small measurement sample (results and depends on properties of small pieces rather than being averaged over more material), sculpted profiled boards = how to measure them (how do you cut e.g. a wedged board?), how to simulate random incidence, how to simulate corner behavior, how to make sure a membrane behaves similar as that same membrane in real live applications?
In fact without these limiting side effects impedance tube measurements are the most accurate method to compare materials.
And you can go accurately down to 20 Hz if you have the right sized and well designed tube (is huge).
I don't see very well how corner absorbtion behavior should be simulated in an impedance tube.
Often people look for absorption numbers of different tiles, not understanding that this has hardly any importance in real live situations, and that this is hardly to measure with room methods. These differences with the room itself are that small that you measure anything except the absorption value of these tiles (no matter which method). And it doesn't matter that you use 72 sft or not.
The added (+ or -) absorption (difference with room and/or room boundaries) is what counts.
If you look at your own density report based on these levels (which were relative, what is wrong about it) you see that there are measurements giving the impression of also showing negative absorption with added absorption. And that's not even for low absorptions.
You can't conclude whatever details (only the obvious) of that density report. You don't know how the exact numbers should look like, since you just have pictures based on a RELATIVE y scale, showing some obvious things and leaving lots of other questionmarks.
And what relates to these 3 measurements in row you asked for/suggested.
I assume you refer to the 3 MiniTraps measurements you did at IBM (spread over some years).
The 1st was very different from the 2nd and 3rd, but the main reason was that you altered the membrane of these MiniTraps significantly, changed the combination of the wool and so on. Hence, nothwithstanding a layman shouldn't notice a visual difference and the name for both was MiniTraps, you were measuring acoustically different products.
The 2nd and the 3rd (with > year difference) were, as per your own words very alike, giving you a reason not to enter your most recent measurements on your site. You yourself said the difference was that minor it wasn't worth to bother.
Hence what people see on your site are your second (2004?), not your most recent measurements (2005/2006?).
There are problems with measuring bass traps in reverb rooms and certainly corner absorption, but I want to put these suggestive comments (unclear what they should substanciate or not) into context.