I've always felt this way, too, but I am beginning to look at it another way. If the purpose of the multiple spaced measurement mic placements is to supply enough information to create a relatively uniform response in the volume of space that they encompass. Now, if one measures anywhere in that space with any number of positions, the result will probably not replicate the "predicted" result but it should show a significant amelioration when A/B-ed with EQ bypassed. If it doesn't show that, the process failed.
Probably so, Kal.
You know well what I am about to say, of course, probably much better than I. But for others, as I see it, a mike measurement for each speaker at any position will include, inseparably combined, both local frequency response unique to that one mike position plus broader, general room response, which is also part of the measurement results for that speaker channel at other mike positions. Local response peaks/dips in the measurement for one speaker at one mike position are not necessarily smaller than more general ones. The difference is the local ones don't show up for the same speaker at all the multiple mike points. The averaging of response for multiple mike points will therefore be weighted toward the general response issues for that speaker channel in the room and not the individual local ones. The EQ filter determining EQed response for that speaker is then calculated using the average, which has deemphasized the local response issues, though not entirely eliminating them mathematically. But, this should be the aim of the EQ tool, and it should minimize the influence of variable local response based on a single point. Some say no, and prefer EQ tools based on single point measurement. But, I remain more convinced by the multi-point, spatially averaged approach. However, the multi-point, spatially averaged approach creates difficulties and potential inconsistencies between pre-and post calibration measurements.
I am not at all sure how much more uniform in terms of variability from point to point post-calibration response is compared to pre-calibration. Likely so, particularly in the deep bass by reducing room modal issues. But, maybe not, even if on average the response is improved by the EQ process. Some local response issues and their variability may persist as frequency rises.
Smoothing of the frequency response used as input to the filter calculation should also play a role in all this, ideally with less 1/X octave smoothing at lower frequencies and more smoothing at higher frequencies, where chaotic, reflective, narrow band comb filtering becomes more severe. AFAIK, all EQ tools use some frequency smoothing in generating their filters, but I do not know who does use this ideal variable smoothing as a function of frequency. However, most EQ tools, like Dirac, only provide a graph of results with equal smoothing up and down the frequency spectrum, maybe 1/6 octave or so. That may not be sufficient to see issues in the lower frequencies. Amir indicated awhile back that the JBL/Harman EQ tool does provide a response graph with less LF smoothing and more HF smoothing. So, presumably that tool uses that variable smoothing in calculating its filters.
Dirac's specific algorithm for spatial averaging and smoothing is proprietary, and it might possibly include more sophistication. I recall Audyssey touting their "fuzzy logic" method for spatial averaging. It was claimed, though details were hazy, to provide a better picture of the general response for each speaker channel, with further deemphasis of the local response issues in the filter calculation. They also provided no direct method of post-calibration measurement. Nor, does Anthem ARC.
Yes, I would expect the result of a followup measurement test tone sweep to validate some improvement in post calibration response, even if the mike points are inches off of what they were during the calibration. The specific mix of general resposone issues to local ones might well be somewhat different, however, particularly approaching the HF comb filtering region. But, is this more accurate than the predicted response graph? Unless the prediction is a manipulated or overly optimistic fraud - tough to prove - it may have some value as well. But, which is right: the prediction based on the specific measurements feeding the filter calculation, or, if it were available, a followup, similarly smoothed and averaged multi-point measurement but with potential mike position differences, or a single point REW measurement via a different measurement protocol, or none of the above? Are apparent differences between them significant, given they are based on different measurements?
I personally don't want to spin my wheels trying to reconcile or validate different measurements in the messy world of non-anechoic room acoustics. I agree about standing back, letting the EQ tool properly do its measurement thing, and generating its algorithmic idea of improved response. Then, I will subjectively judge the before and after response by listening while easily switching back and forth. Subject to all the caveats about expectation bias, etc., I reach a conclusion, often achieved somewhat more reliably by blind testing. For me, the sonic differences have been sufficiently obvious and beneficial that blind testing yielded no surprises, disappointment or undue agita. And, I found this much more valuable and convincing than comparing before/after graphs, though some useful things like difficult to correct deep bass nulls have been revealed in the predicted response curves.
For over 10 years in different rooms with different systems and with different EQ tools, including those of friends, I am quite sure I have achieved better sound in my room this way, as are my friends in their rooms. I am also convinced that diligent care for and understanding of simple, correct mike measurement techniques during calibration have improved on the calibrated result. Comparing different calculated filter sets in Dirac is a piece of cake. But, YMMV.
Yes, it is clear that the most significant improvements are in the deep bass, both theoretically and subjectively. So, I have no issue with those who prefer to apply EQ only at low frequencies.