Rational Acoustics

Rasmus Rosenberg
January 22nd, 2012, 10:21 AM
Holla, Folks
So i got a little time off and decided to go look at my venue install. The venue is about 27 mtrs long flat floor wit a 5 box linearray covering the floor. There is a separate system for the balcony that i will not write about. So my "frustration" about the system is that the coverage in the 1 half of the venue is very varying (+/- 12 db difference at the same frequency in the HF) . The system is divided in top 2 boxes on one feed and 3 bottom boxes on one. Same level,dly and Eq on all. After a bunch of measurements i discovered that the top 3 boxes IMO, do a good job HF wise. But when i add the 2 bottom boxes there is quite a lot of combing going on, in the first half of the venue. What i concluded was that the spray angles for the 2 (3) bottom boxes is wrong. But off couse its an install system, so the rigging system is not the tour standard where you can just move a pin to reset an angle. You have to take the whole array down and take every box apart with bolts and nuts. And thats a procedure that takes, very long time and at least two guys. In other words a no go. So i decided to repatch so the 3 top boxes go a feeder and the bottom 2 got there own. Next i tried to reduce the level of the bottom 2 two boxes, that help a little but had to about 12db of level difference so not really an option (I can only adjust the level of the whole box not the HF/Low alone). Next I tried different Eq on top and bottom, but that didn't really pleased me either. After a coffee and cigaret i measured the top 3 boxes alone at a lot of differn't positions and after that the 2 bottom boxes alone and sat down and looked through through the traces. And then it struck me there was an huge area (actually most of the first half of the venue) where it was clear that the phase response was different. SO i tried to dly the top 3 boxes 0,1 ms (the smallest amount on the processor/amp, it was just a tat too much looking at the phase trace 0,07ms would have been better i think) and that changed quite alot. Not only did it sound better, but the measurements showed less variance position to position. I have included a picture the i think shows it quite well. Notice that mic 3 i closer to the array than 1 and 2 (also I did a lot more measurements than the ones showed). So being very skeptical that i had fucked up more than i gained, i spend the rest of the day trying to figure out where I had caused damage, but couldn't find a position that got worse. Meassured and did a lot of listening, with a tablet turning the dly on and off. Yes if I did an average the 3-5 kzh range was lower in level with dly than with out (about 3db), but the variance between positions was minimized and to my "surprise" I could Eq the 3-5 kHz range 2,5 db, back if needed. Top window in the picture is with out dly, bottom is wit dly difference, same mic positions. So guess my question is have you guys tried similar? Knowing its against any manufactures recommendation to dly with in the array, i was quite surprised it worked, or did I miss something?

Arthur Skudra
January 22nd, 2012, 06:29 PM
Just curious, do you have front splays in the bottom boxes of the line array? Been bitten before with that one! :o

Yes, I have done similar tricks to solve physical misalignment problems in the overlap area. I had one case where two point source loudspeakers were overlapping each other in a corner of a rink, and because of rigging limitations, we could not cluster them perfectly, resulting in a physical misalignment of a foot or so between the two. It was enough to cause a very noticeable "hole" in the HF coverage in a bunch of seats in the overlap zone, even with time delay alignment. If you look at the attached file before alignment, you will clearly see the problem with the phase curve above 1,000 Hz. So what we ended up doing was applying an all pass (phase) filter to one of the loudspeakers, and managed to get the two phase curves to line up almost the same (see attached pic after APF alignment). This resulted in a remarkable improvement in the overlap area coverage, the overlap was barely perceptible once this correction was made.

Michael Häck
January 22nd, 2012, 06:51 PM
Hi Arthur,

very good example. Do you have a picture with both speakers on (Corner + Sides) with AP as a result trace? That would be helpful to see the big advantage of your AP-Alignment.

Arthur Skudra
January 22nd, 2012, 07:02 PM
Hi Arthur,

very good example. Do you have a picture with both speakers on (Corner + Sides) with AP as a result trace? That would be helpful to see the big advantage of your AP-Alignment.
Actually the brown trace in the "after" pic is a combined response of both speakers in the area of greatest overlap! The pink trace is the one speaker that was arriving ever so slightly later in time of the two speakers, and is the same as what was shown in the "before" pic. I was only including it in the "after" pic to show everyone what I was hoping to achieve as my "target" response. Pretty close, eh?

Edit: I attached a pic of the DSP file, showing the eq and delay, and all pass filter applied to the side speaker so that it lines up with the corner speaker both in time and phase. Essentially a 1.6 ms delay along with a 1st order all pass at 2000 Hz was applied to the side speaker to make it match up with the corner speaker.

Arthur Skudra
January 22nd, 2012, 07:39 PM
And here is a very rough plan view of the two speakers as they were hung (not my design)! Not the ideal circumstance, they were already installed 40 feet in the air, ice was already in the rink, manlifts were already gone, we couldn't take the speakers down and properly rehang them. Wish I had taken a photograph of them!

Edit: found some photographs! Take a look at image 2556 and you'll see two speakers rather close to each other at the top center of the picture (I hope I'm not violating any copyrights in supplying this link):

Arthur Skudra
January 22nd, 2012, 09:27 PM
A few thoughts of Rasmus' concern with using delay within a line array...

1. Consider that the bottom part of the array, if it's in a "J" shape, is essentially a point source, or several point sources, so you may want to treat this part of the array separately from the rest of the "straight" portion of the line array. Furthermore the top boxes of an array might be treated separate from the main array, particularly if it's a very long throw. Generally I don't consider the first third of the line array's coverage to be a valid measurement point for the "line array" portion of the line array.

2. Depending on how the elements of the array are aimed and how far apart they are splayed, you will see timing differences between the elements of fractions of a millisecond. This is normal. Keep in mind that as you physically adjust the array elements, not only are you changing coverage, but you are making some slight adjustments in time as well. I'd say Rasmus did the right thing here.

3. Any spacing between the elements will cause some gap in the coverage that may be unavoidable and perhaps electronically uncorrectable. This may be caused by front splays, bumpers between differently sized elements, or the maximum splay applied to the bottom of the array to cover the front rows. Time delay and amplitude shading may be helpful here.

4. There is no end to the debate of whether you eq, delay, shade a portion of the array differently from another portion or doing the same processing to the entire array as a whole. I see the benefits in both sides of the argument. I suggest you follow what the manufacturer recommends! I know that one manufacturer recommended for the longest time to process the entire array the same, and then they came out with baby line array, which allows one to process each element individually! It depends!

Rasmus Rosenberg
January 23rd, 2012, 07:14 AM
Hey Arthur,
Thanks for commenting lots of good info. Just to recap, i do think this example is "one of a kind" and do always follow the manufactures recommendation. Lucky its not normal to be so locked in options as I am in that configuration. Normal I would revisit trim height, splay angles, tilt angle of the array etc. One thing you mentioned is that you don't consider the first third of the coverage as a represent for the "line array" part of the line array. I agree and would point out how "wonderful" that with multichannel measurements the transforming of the different zones are "visible". Also trying in treating the bottom part separately, its relatively easy to see if you start to mess up more than you gain. Still can't believe how much info you can get in a short time, with this program.
Great example with the hockey ring! I would recommend any one playing along, at home to try that. Try to align asymmetric: level, splayed, and positioned, speakers (is that the name?). There is AP-filters in the APM crossover plug in, that you can use in AU lab. So it should be quite easy to do a setup with 2 speakers using the internal sound card on your Laptop.

January 23rd, 2012, 09:57 AM
Good stuff guys. All pass filters are somewhat new to me and I had been wondering when you'd use one. So was quite interesting to see a real world example.

Ben Clarke
January 30th, 2012, 05:35 AM
excellent thread - using stuff when things suck - continue.