Looking for Mr. Cardioid
During the daytime (remember that we start at 11pm) the crew set up 2 measurement mics for us high in the grid. The mics were placed as a front/back pair in the near field of the coupled subwoofer array. The goal was to measure the responses in front and behind on the center axis at symmetric distances. This would allow us to see the cardioid action right there AT the array and help to clarify the mysteries of the day before.
At 11pm we started on it right away. We measured the response of each speaker in front and in back. The mics were not quite equidistant – 16ms (front mic) and 23 ms (rear mic). This translated to around 3 dB so we prorated the data with that in mind. We 1st observed the 3 front-firing drivers, individually and in combination. We found a 1 dB of front/back ratio at 30 Hz but 6 or more by 80 hz. The rear facing drivers did the same – in reverse and we felt ready to put in the parameters we had developed in MAPP (pol reverse, 4ms, -2.5 dB) on-line and ess what would happen. We did. We measured. 1 dB of cardioid action – LESS than just having the boxes all face forward. ????????????
Now we were spooked. We had seen no evidence of cardioid steering in the far field in the room during the day before – now we had none in the near-field. What gives? So we decided to simplify the 5-box rig to the center 3 boxes. Now it became 2 backward and 1 forward (pol rev @ 4 ms). This would steer the OPPOSITE of our design – but MAYBE we could get a measured result that related to the predictions…..pretty please.
Before we measured the combination Vikram abd I looked at the individual parts – amplitude and phase, with all the parameters put in. In phase in front of the two boxes, 180 degrees out in front of the single box. Combined we got the 20 dB ratio we were looking forward. Perfect cardioid steering………….. into the flyspace. Wrong direction, but it was what we expected – that was definite progress. Now we went back to 3-2 ratio and adjusted the level until it was equal in level in the back (2.5 dB down for the rear firing pair – the ORIGINAL predicted number). Now it worked perfectly.
I could not stand to NOT know why it had NOT worked before – so we ran the human error scenarios until we found it. We had previously put the pol reverse on the correct speakers, but the delay on the wrong ones. If you ever want a REALLY REALLY OMNI sub array – let me know, I have the recipe. So NOW it works up in the grid. Maybe our human error was the reason it had not worked in the house the previous day. Back to the far field.
We compared the mic 70 ms away directly in front of the array, with the one at the audience area MOST behind the array. It was TWICE the distance – so it should be 6 db down just by distance. Result: So close to the same level as to be insignicant. Maybe 1-2 dB. On the sides it was about 3 dB down. It was also 3 dB further. The impulse response at the rear revealed something VERY interesting. The second set of arrivals back there were WAY stronger than the 1st. The direct sound wave (earliest arrivals) had been reduced, but the steeing had concentrated the front lobe on a concave GIGANTIC wall of glass where the control booths are – pretty much the only major league reflector in the building. So at the end of the day, we have a cardioid array that reduced the first arrival, but could not stop the second. Our analyzer has a 640ms time window down there so it saw the strong 120ms reflection as integrated with the direct sound. It will be interesting to see how we would perceive the difference between front and back – same level, but much higher direct/reverb ratio in the front. Today the crew will angle the coupled sub array further downward, to reduce the direct sound on the window wall. This will put the cancel lobe more into the flyspace (which can only be a good thing) rather than trying cancel toward the rear of the house. We will test that tomorrow.
The next task was merging the coupled directional configuration with the uncoupled steered subwoofer line array. The levels and delay were set to merge them in the corner (yes I know circles don’t have corners) – the corner was where two parallel lines of 5 subs meet the coupled array. Altogether the subs make the shape of an “n”, with the uncoupled lines on the sides and coupled array at the top.
The latter part of the session was spent in verification mode: checking speakers to make sure they were like the other speakers of the same type. We had already completed the first 1/3 of the circle, now we had to do the remaining 2/3rds. This amounted to: 14 CQ-2’s, 30 UMP-1Ps, and 24 UPJ-1Ps. Each was checked for frequncy response, delay, level, polarity and level. This took about 3-4 hours. We found about 4 poarity reversals, one speaker with a mysterious extra 4 dB of level, a CQ with a lighting instrument partially blocking it, and lots of well matched speakers. This allowed us to turn the system over to Steve and Pierreto try out the Constellation settings they had programmed in.
Tomorrow we have a few mop-up operations: verification of the opposite side main clusters and subwoofers, and 25 little MM4-XP under balcony delays. We will fit these in during times when Constellation tuning does not require the system to be in use. We have been trading time between the teams, prioritizing the SIM tuning to get the Constellation team the data it needed to keep moving forward while we SIM’d other subsystems.