Hello, I am attempting to reproduce the results of a psychological experiment. In the original experiments, messages were delivered -15 to 25 db(A) relative to ambient environment or white noise source using loudspeakers. In my experiments I wish to deliver messages via headphones -15 to 25 db unweighted (RMS average), in an unweighted medium such as white noise. If I keep the same level of -15 to 25 db, essentially keeping the same ratio, should I get the same effect? My desire is not to weight the audio stream to keep the amplification calculations simple. If not, are there algorithms that I may manipulate my digital audio to be within that weighted range. Currently I am analyzing the RMS, less silence, to determine what the audio’s unweighted average. I am using 50ms windows and anything below 44db is considered silence and is not used to factor the RMS. Thank you, Waverly
Weighted vs. non-weighted
Started by ●September 10, 2010
Reply by ●September 11, 20102010-09-11
"WaverlyE" <waverly.edwards@n_o_s_p_a_m.genesys.com> writes:> Hello, I am attempting to reproduce the results of a psychological > experiment. In the original experiments, messages were delivered -15 to 25 > db(A) relative to ambient environment or white noise source using > loudspeakers. In my experiments I wish to deliver messages via headphones > -15 to 25 db unweighted (RMS average), in an unweighted medium such as > white noise. If I keep the same level of -15 to 25 db, essentially keeping > the same ratio, should I get the same effect?First, you seem to be confusing signal level with signal-to-noise ratio (SNR). "dBA" is a measure of absolute power and says nothing about SNR. But to answer your question about dBA, it depends on the spectrum of the input signal. The A weighting curve is a filter that deemphasizes low frequencies, down 10 dB at 100 Hz - see http://en.wikipedia.org/wiki/A-weighting So if the input signal contains significant low-frequency content, the unweighted power will be much higher than the weighted power.> My desire is not to weight the audio stream to keep the amplification > calculations simple.Then your results will likely not be accurate.> If not, are there algorithms that I may manipulate my digital audio to be > within that weighted range.That same article has the s-domain transfer functions for each of the weightings, but that doesn't do you much good with digital data. Converting these to digital filters is not a straight-forward task. There was some discussion about it here a few years back that may help you: http://www.dsprelated.com/showmessage/69319/1.php -- Randy Yates % "How's life on earth? Digital Signal Labs % ... What is it worth?" mailto://yates@ieee.org % 'Mission (A World Record)', http://www.digitalsignallabs.com % *A New World Record*, ELO
Reply by ●September 12, 20102010-09-12
> > http://www.dsprelated.com/showmessage/69319/1.php >--Unfortunately, I have spent the better part of two weeks trying to understand how to turn my unweighted RMS measurements into A-Weighted measurements. I've been good at finding what A-Weighting is and have found formulas for conversion but it is more complex than applying a formula. I haven't found anything that explains how you can get from here to there. The closest thing I have found was too complex in terms of mathematics. I found an article that I was able to glean bits of information to improve my comprehension of how to get to where I need to go. http://web.media.mit.edu/~dlanman/courses/decibel_meter.pdf I don't use MatLab and this was more than I could handle but I *think* I got that using FFT was ideal to the extent that it separated frequencies needed to give to the A-Weighted filter formula for computation. I also got that it is more computationally expensive to use FFT than to use filters. What kind of filters would you use. I am assuming that if you are using multiple filters, that you must combine the results from the filters in some manner. I am lacking knowledge but I haven't given up. I'm hoping for some revelation that will get me on the right track soon. Thank you, W.