Ready Traps

Posted: Wed Nov 14, 2007 4:05 pm

Thanks Ethan! I have been targeting the 80Hz problem with my crude tests, but I will try some more panel moving and waving tonight. Thanks again!

Posted: Thu Nov 22, 2007 6:27 am

Pink noise is an unreliable sound signal unless measurements are repeated several times. Use MLS or sweep for instance. More info at: http://www.acoustics.salford.ac.uk/student_area/bsc3/room_acoustics/MLS.pdf

Posted: Thu Nov 22, 2007 9:07 am

thanks, I have been using fixed frequencies and sweeping through them.

Posted: Thu Nov 22, 2007 3:41 pm

You can use MLS or Sweep signal, should give you the same results for that kind of measurement

Posted: Fri Nov 23, 2007 11:03 am

andrebrito wrote:

You can use MLS or Sweep signal, should give you the same results for that kind of measurement

Swept sine waves are better than noise because the measuring software can use a tracking filter to increase the s/n ratio of the measurement. This is useful especially with nulls, to be sure you measure their true depth.

--Ethan

Posted: Fri Nov 23, 2007 11:23 am

This is great stuff thats so much guys!

Posted: Fri Nov 23, 2007 12:28 pm

MLS and Sweep noises gives you the SAME impulse response if nothing unsual happens ! If you don't believe me check:

http://www.acoustics-engineering.com/dirac/faq.htm#measurement%20and%20analysis%202

"2. What are the differences between impulse responses from MLS and sweep?
MLS and sweep will normally result in the same impulse response, but the methods differ in the effect of system irregularities, such as click noise, system variations during a measurement, distortion in the measurement chain, etc. "

Presuming you are doing your measurements properly and you are in a small room environment they should give you the same results. You stated pink noise so I just recommend other readers that there are better acoustical signals than pink noise nowadays.

Quote:

Swept sine waves are better than noise because the measuring software can use a tracking filter to increase the s/n ratio of the measurement. This is useful especially with nulls, to be sure you measure their true depth.

Posted: Sat Nov 24, 2007 2:14 pm

Andre,

MLS is used as a noise source for convenience - it's a great way to ensure the noise is pure without having to manually select transistors for "good" noise properties. But as I understand it - and I might be wrong! - the actual sequence is not deciphered by software. That is, a tracking filter is not used to increase the s/n ratio as is routinely done with swept sine waves. Do you know of software that interprets MLS and applies a tracking filter? Again, I'm not suggesting that the measurement itself will be less accurate with MLS, just that the s/n is not as good.

--Ethan

Posted: Mon Nov 26, 2007 8:06 am

You need to crossrelate the input an output to get the impulse response when using MLS, so I'm not sure if that is what you meant by "deciphered by software".

Actually MLS has a lower crest factor (close to 1) then a exp. sweeping signal so indeed it has a better Signal-to-noise. The problem with MLS is that it is sensitive to distortion as well as other anomalies during measurements.

This Powerpoint document is very good have a look:

http://iem.at/~majdak/publications/2005_HASSIP_SysId_presentation.pdf

Posted: Mon Nov 26, 2007 8:32 am

MLS is used a lot for TL measurements (and room acoustic) because of its better S/N ratio, certainly when this noise is disturbing background noise. With MLS you expand your dynamic range.

You send out a very specific signal, and measure that very signal back, thereby suppressing unwanted/disturbing background noise not originating from that emited signal.
MLSSA seems to have a solution to synchronize a seperate MLS signal, but normally the generation and measurement of an MLS signal must be done by the same equipment, since it searches on the immission (reception) side to the signal it generated on the emission (generation) side, while ignoring (in as far as possible) other sound.
Normally a good MLS system will warn where it can't distinguish accurate enough its own emitted signal in the measured signal.

If you try to use MLS via a system of e.g. artificial reverb, where that emitted signal is somehow raped MLS doesn't work anymore since it doesn't know what to measure.

When I measured TL at GALAXY, we compared methods.
To get what I got with MLS, I had to use rather complicated methods with noise to suppress that non-related background noise from disturbing these measurements (multi-spectra followed by statistical analysis).

But it depends on what you measure, which method is best to use.
The paper Andre 1st linked clearly refers to Impulse response measurements. You can not use some text from a paper, ignoring the title defining what the paper is about.

I don't see much disadvantage in using pink noise when executing normal pressure measurements. And for me sweeps can involve risks as well in function of too short integration time and less settling time within a room. Hence a signal should be seen in what you want to do with it.

Referring to pink noise as some outdated signal is certainly not correct for me.

Posted: Mon Nov 26, 2007 10:11 am

Hello Eric, how are you doing ?

Yes it was my mistake not refering that quote was about impulse response measurements.

Personally I don't see much advantages in using pink noise when you have MLS or sweep signals nowadays (again refering to IR). You need also more averaging to get reliable results.

Unless in TL measurements it's a bit different, I don't have much experience on measuring those with different types of signal

Posted: Mon Nov 26, 2007 11:06 am

Andre,

andrebrito wrote:

Hello Eric, how are you doing ?

Let's just say: I'm doing

Don't forget that most lab procedures (sound power, RT60, TL) are still executed with pink noise, while MLS is known for long already.

The larger averaging numbers (much more in ASTM than ISO) are not just to increase accuracy of the method but to excite the room in different manners and to measure at different positions (really different predetermined fixed positions or random from rotating beam but with predetermined radius) in function of modal behavior.

Lots of acoustic measurements in the official environmental noise abatement business will be executed with pink noise.

And for "in the field" TL measurements MLS has the disadvantage that you need a cable connection with the measurement equipment on the immission side and the source on the emission side.
Which can be an annoying thing in itself.
I only know one company (Norwegian Electronics) who solved this by sending the signal out via the normal net (you just plug it in on both sides). How they solve that is for clever guys to explain (beyond my area of knowledge).
But it can't be that easy because a specialist electronic guy once, on our request, tried to build something to simulate this, for general use and other brands (as e.g. Larson & Davis), and we never got a final solution.

Eric
.

Posted: Mon Nov 26, 2007 12:57 pm

Thanks for the clarification. Yes I know labs still use pink noise as well as in some official measurements, but most seem to use pink noise since they have old (non digital) equipment. Wasn't also aware of the pratical reasons you have mention.

To get the IR of a room I would prefer to use another type of signal apart from pink noise. With MLS and sweep signals one should also like you said "excite the room in different manners and to measure at different positions".

Posted: Mon Nov 26, 2007 3:35 pm

andrebrito wrote:

This Powerpoint document is very good have a look:

I didn't see anything in there that relates to the point I was trying to make. Smile

I imagine that MLS and sweeps will have a similar s/n ratio if the signals are used raw. But if the sweep is slow, and a tracking filter is employed, that should (as I understand it) give a much greater s/n ratio than MLS. Like 20 dB or more greater. If that PDF document addressed the use of a tracking filter, please point me to the page.

--Ethan

Posted: Mon Nov 26, 2007 4:07 pm

check latest page

it says MLS : highest possible SNR.. crest factor is 1.

This topic sponsored by: