Certain none
minimum-phase "issues" seen in the in-room measurements should be best left alone.
Why? Well, in terms of just EQ, if you measure elsewhere besides a single point in space or MLP, the response will not necessarily be any better -- sometimes
much worse.
Sorry, but I'm going to be adding lots more images into this thread!
Bass-managed left and right front mains (LR) as well as surrounds (LsRs) as measured in the center MLP of my couch:
There is obviously an "issue" around 130 Hz -- esp. when applying a lot of
frequency dependent windowing. Should this be fixed via EQ?
Phase and Time Alignment and EQ
Phase Responses in Detail
*Notice the abrupt phase kinks around 130 Hz? That's caused by the boundaries and placement of the speakers in the room.
Impulse Responses
The room not only mangles the in-room response of the speakers badly, but also that of the subwoofer. The
IR start time of the sub is set at around 2 ms in relation to the fronts and surrounds.
Step Responses
Mixed phase EQ --
additional FIR only added to surrounds -- makes the responses more
alike overall. The subwoofer's maximum phase peak is delayed by about 15 ms. *Notice some of the waviness in the last graph between fronts and surrounds above 2.5 ms equalized each other out. The last magnitude SPL graph traces of the BM fronts (red) and surrounds (brown) at the bottom-most graph posted at the end correspond with the step response very nicely.
Envelope Time Curve
Again, more similar and
controlled post EQ.
Spectral Decay
Having the subwoofer augment the "satellite" speakers' very imperfect amplitude response in the bass is crucial. We see the bass is quite smooth and have no "major" visible (and
audible) dips for the most part.
Group Delay
It should be obvious from the
LR only and
LsRs only graphs that the GD anomaly should not be force-fixed via EQ. FIR filters to correct this narrow and varying region requires very steep/high Q filters.
Wavelet Spectrograms
This is about as good as it will ever get for now. No significant visible signs of pre-causal ripples under t=0 as well. Both fronts (internal DSP) and surrounds (external) have some form of FIR filtering applied.
Multiple microphone positions set in the room -- RTA pink noise -- both "steady" and MMM
*last graph traces are all
MMM and not "steady" (with pseudo-surround mode enabled)
Ignore the peak below 20 Hz. The room is sealed and pressurization in the room is sometimes exaggerated with RTA.
The last measurement set is with "pseudo-surround" DSP enabled. This essentially matrixes the LR inputs -- and decorrelates it as well -- playing the delayed signal 20 ms afterwards for an artificially induced pseudo-surround effect.
The MMM measurements are much more stable and evenly smooth.
I had to manually apply a 90 Hz linear phase HPF to the surrounds in this DSP mode, as well as get rid of the bass mixing below in order for there not to be any additional interference between channels in the bass.
FDW 15 cycles
It should be clear from the
multi-mic measurements taken in
more than one spot in the room that specific anomalies in the response should not be treated in isolation. EQ boosts and cuts should be applied with caution. Taking care of a single seat maximally with EQ is fine, but you have to know that there is always going to be a trade-off somewhere.