I spent some time on this yesterday. It was interesting to focus on the spectrogram when making adjustments and less so on the SPL response. I started out by delaying the main speakers by 10ms and the sub 0ms. (The distance difference from the main listening point to the sub versus each main speaker is relatively immaterial.) I also looked at a crossover to the sub rather than full range to the main plus LF pass to the sub. I had originally thought that broadening the sources of LF (in effect increasing the number of subs) would lead to the best result and in may in fact still do that if optimising for more than one seating position. (Note the position of the sub in the graphic on page one of this thread.)
In the end I settled on a Butterworth crossover to the sub at 100Hz, slope 24dB (both Hi and Lo). (Previously it was set at 80Hz and the main speakers were running Full with a LP copied to the sub.) I set the phase on the sub to "normal" and began working through delay settings, finally setting on a 6ms delay to the mains. I was using full 5-22kHz sweeps. (For awhile I played an 80Hz tone and walked around the room. This was very eye opening with respect to the impact of room modes. However, it also gave me a headache very rapidly!) The changes in delays manifested mostly in differing decay in the 60-100Hz region.
So to the results. Well I still have a lot of base to absorb at c35-40Hz, it would be nice to soak a little more around 60-100Hz and I have a null at 118Hz. I would be interested in any other thoughts people have.
One thing I found very intriguing was the relationship between SPL response and delays. I was not expecting the two to interact so much. I guess this is combing.
Here are some graphics. All are based on mains and sub running all together (3 speakers) as I was only focused on the low end.
The last graphic shows a comparison of SPL response between the previous settings (sub delayed 8ms and phase reversed, mains running full and LF to sub) and the current ones (mains delayed 6ms, sub 0ms, normal phase, Xover to sub):
I might try several 40Hz Modex Corner Plates to soak up some of the bass there. I'm not sure what I can do about the 120Hz null or other stuff. Roll-on Theta's implementation of Dirac Live...