Originally Posted by jim19611961
when I looked at a filtered ETC/IR @ 500hz and 10ms and the graph says its at -22db, I assume because thats what the data says, that it is right. I dont mean to sound critical, but if what the graph says is wrong, how is it valuable?
The data is right, it is your interpretation of it that is wrong
Applying a filter to a time signal doesn't turn it into a frequency domain signal, it is still a time signal, you are trying to assign frequency domain characteristics to it that it doesn't possess.
The filtered IR plot is primarily about reverberation times, aka RT60 data. An RT60 figure is how long it takes the sound energy to decay by 60 dB, derived from the impulse response. (Partly) because impulse response measurements often don't have 60 dB of dynamic range (they almost never did back when most of the work on RT60 was done) this isn't as simple as just moving along the plot until it drops 60 dB below its peak. The RT60 figures are instead derived by measuring the rate at which the level is decaying over a defined portion of the response (for example, from the point where it is 5 dB below the peak to the point where it is 25 dB below) and then extrapolating that to come up with a figure for the time to drop by 60 dB. The time signal itself jumps back and forth between positive and negative values, so it is hard to see the decay from the time signal directly. Instead, the calculation is done by looking at either the envelope of the time signal or by a process of backward integration of the signal named after Schroeder, who invented the process around 1965.
Typically the RT60 figures are broken down into frequency bands so that treatments can be targeted at those portions of the band where decay time is excessive. Those band figures are obtained by the same process as the overall decay figure, except that the impulse response is passed through an octave or one-third octave filter before the envelope and/or Schroeder integral are calculated. The figures then indicate the average energy decay over the range of frequencies that remain within the filtered impulse response (707 Hz to 1,414 Hz for the octave filter band that is centred at 1 kHz, for example). It's not a hard cutoff of course, filters have slopes outside their pass band rather than cliff edges, but the roll off is fairly fast (3 dB down at 707 Hz and 1,414 kHz, but typically 20 dB down at 500 Hz and 2 kHz for the 1 kHz filter, for example). I repeat that it is still a time signal though, just one with a more limited spectrum.
All that is somewhat irrelevant for what you are trying to do however, which is to examine the variation in frequency content of a signal over time and in particular to see what content it has at the time your engineered reflection arrives. Such time-frequency plots are subject to the uncertainty trade-off I mentioned previously - if you want to be precise about timing you will get very vague data for frequency, being precise about frequency means being vague about time. The spectral decay, waterfall and spectrogram plots are all ways of examining that variation of frequency content over time, with adjustable window widths to provide control over the time-frequency trade-off.