DeeringAmps
I feel like a punching bag!
;~}

bitflipper
Just once I'd like to see somebody publish before-and-after waterfall plots. It'll never happen, though.

mike_mccue
 I am trying to understand how a electronics engineer can *flat line* a room without using some sort of dynamic correction. That is why I keep, optimistically, imagining the use of dynamic convolution.
 
I suspect it is easier to imagine that a flat line response is possible if you are imagining the *wiggly line* is a static or averaged response. In real time the response experienced, in a enclosed space with reflections, is wiggling up and down all over the place over the course of time.
 
If you use a Real Time Analyzer in a room you see how the response is wiggling all over the place through the course of time and the idea of a single drawing of a *wiggly line* begins to seem like an overly simplified abstraction of what is really being experienced at the listening position.
 
 
I think I understand how a FIR filter can use a single impulse response to alter the phase of the output compared to the input. I struggle to imagine how any technology that uses a single impulse can be adroit, and offer the sort of constant corrections, that I suspect are necessary, to approach an ideal such as a flat line graphic represents. I keep imagining that the circumstance demands some sort of dynamic application of a series of rapidly applied corrections.
 

losguy
 
More like DSP, along the lines of Finite Impulse Response (FIR) multipath channel compensators:
http://dsp-book.narod.ru/spra140.pdf
http://www.cplire.ru/rus/informchaoslab/papers/iccsc04ak.pdf
 
A multipath channel is on that has a direct path and other paths (echoes) with different delays and attenuations, where the attenuations may themselves be frequency-dependent.  It translate to the idea of the mirror technique in your room, tracing out the line path that waves take from your speakers to your ear(s).  The compensator uses the information in the main and echo paths to create a filter (a convolution operation) whose coefficients equalize out the echo paths and leave the main path as much by itself as possible. In the second paper above, the multipath channel's impulse response is in Figure 5, and the compensated result is in Figure 7.  The spike is what you want, and the junk is what's left.
 
Techniques like these can actually compensate for echoes that produce complete nulls in the frequency response.  It sounds weird, I know, but it's because they operate in the time domain.  The lit from IKM and MathAudio don't emphasize this, but this has been "the way to do it" for, like 20+ years now.  Perhaps they describe their use of the frequency domain as a type of constraint that the user can apply to the process, to limit over-compensation at certain frequencies, for example.

mike_mccue
Discounting absorption of wall surfaces and presuming constant surface reflectivity characteristics, I'd want to learn more about:
 
Is the impedance of air "linear"?

 
Is RT60 Frequency dependent?

How does varying *musical content* energize a room compared to an all pass sweep or chirp?