The ABC’s of Line Array Tuning

UPDATE – Aug 27 2010:

An example of a line array tuning using thr ABC method described below – with 19 SIM data screens and some photographs can be found here.

—————————————————-

Every time I teach a seminar it is simply a matter of time before the question arises: How do I tune a line array?  Typically it takes less than 10 minutes of the 4 day seminar to elapse before this comes up. If it were a simple matter, then I would not waste 3.5 days working up the physics of speaker and room interaction to set up the answer to this very important question. I would also not have spent two years writing a book to get to the same answer. So how about we cover it in three paragraphs in a blog? No problem.

The first step is to remove from the table the idea that new physics have been invented, and that new rules apply. This is an adaptation, not a revolution. What is the big difference between tuning a 12 element coupled point-source array that has boxes spread horizontally and vertically (we call those “point-source systems” or “conventional systems” or “old systems”) and a 12 element point-source array that is only a single box wide and is spread vertically (we call these “line arrays”)?  The difference is that the “line array” has a lot more slices of vertical coverage, and a lot fewer slices of horizontal coverage.

The following items still hold true in either case:

1)      We need to define the target coverage area for every element in the array. Does it make sense to you to tune the top boxes that are pointed into the balcony with a mic down on the floor? Not to me – which does not mean I have not seen it done.

2)      Once we define custody areas for the room we can proceed with tuning to make the areas match. By custody I mean: these speakers are assigned to these seats, and these ones are assigned to these seats.

3)      Unique areas require unique tuning. Do you think the same acoustical conditions are found on the first 30 rows on the floor of an arena as the last 30 in the upper deck?  This is absurd. Do you think then that a one-size-fits-all tuning solution makes sense? Not me. That is why we have been dividing up the tuning in such systems by the ABC method for over 25 years. This “old” approach can be applied to the “new” technology to provide uniformity over the space.

4)      The HF response is the most amenable to zonal separation, and the LF response is the least. This is true in both “old” point sources and “new”. Therefore the tuning approach must incorporate the spearatio0n of high and the overlap of lows into its process.

The ABC tuning method (also –known-as Papa Bear, Mama Bear and Baby Bear)

The principal the ABC tuning is to define coverage area for each element. Element A (the longest throw and most dominant subsystem) has a center of coverage and two off-axis edges, Element B has a center and an edge that transitions to element A. This A-B transition is the custody change which we term the “crossover point”. Next we move on to Element C with its area of solo coverage and the crossover to the B element (above) and either D (below) or finally the bottom edge of coverage.

Each element is tuning as a stand-alone system in its area of coverage. The levels for each are set to create the same level at near and far locations. Therefore provision should be made in the design for the A system to have sufficient power capability to go the distance.

After the individuals are tuned we can investigate the transitions A-B and the B-C etc, to see if they are maintaining a uniform response with their A and B soloists. If the overlap between the elements is too high the transition will be too loud and if too wide, they will be too low. Splay angles can be adjusted as needed to minimize the transition errors. Delay can also be added if needed to compensate for path length differences to the crossover point.

The process follows the alphabet. A combines with B first to become AB. Then AB is combined with C, ABC is combined with D and so on.

The equalization will need to be modified as the aggregation progresses. The combination of A+B should have minimal impact on the respective HF responses, since their ranges enjoy maximum isolation in their respective zones. The LF response by contrast, will be a shared resource between all subsystems so we can expect to have to taper the LF response as we add elements.

Practical Application of ABC to line arrays

It would be nuts to sub-divided our 12-element line array into ABCDEFGHIJKL subsystems, even if we had the budget and the time. Instead the practical, repeatable and manageable approach is to break the line into the same sorts of complexity levels as we have managed for 25 years: 3 or 4 levels. So our 12 elements might break down to 3 sets of 4 or 4 sets of 3 for example. Asymmetric quantity grouping could be used as well such as 6-3-3 or 5-4-3 for example. Our A, B and C subsystems are in turn comprised of A1, A2, A3 and A4 and B1-4 etc. This creates what I term a “composite” element, since we are grouping several speakers to together to tune with a common eq and level. The process of subdivision is fairly common practice in line arrays but there is another critical step to making it act like (and tune like) an ABC array: we make each composite element internally symmetric. Why?   

In order for us to know where the center of coverage is for element A it will need to have one. Sounds elementary but its not so simple. If element A is a single speaker the center is on axis to the speaker. If A is comprised of two elements at some splay angle, the center is at the midpoint A1-A2. Still simple. If we add a 3rd element we have A1-A2-A3 and the center will be at A2. Right?  Maybe……………why.

If the A1-A2 splay angle is the same as the A2-A3 splay angle, then symmetry rules apply and the center is found at A2. If A1-A2 is 2 degrees and A2-A3 is 4 degrees where is the center now?  The answer is somewhere between A1 and A3 but now comes the real plot twist: It is different at every frequency. The isolating behavior of the highs will cause the center to move up towards the A1-A2 crossover since there is more overlap combination there than the A2-A3 crossover. The low end, by contrast, will laugh at difference between splay angles of 2 and 4 degrees and will maintain a center position over the 3 speaker spread. 

If we have 12 elements with 11 different angles there are no independently locatable centers that do not shift position over frequency. To tune such a system is the ultimate challenge. Any given mic position may be the center at one frequency and is guaranteed NOT to be the center at others. If you see a peak at 2 kHz this may be the center of the 2kHz beam or maybe it will get louder 10 rows back You won’t know till you move the mic there. Then you will have to do that again for other freqs because they all have different centers.

Conclusion

The ABC approach built from composite symmetric elements ensure that you can place a mic in the A, B and C element areas and that the coverage will transition logically between them. If you can make A, B and C uniform in their respective areas and taper the combined LF response, the transitions will be smooth and predictable, giving you a continuous line of coverage…………. just like we did with the old systems. The result with the modern systems is we get all of the advantages of fine-slicing the coverage (and easy rigging) without losing our ability to tune the array from top to bottom.

This by no means a comprehensive treatise on this subject, but rather the shortest way I could figure out to say it that holds together.

As always, your comments, questions, musings, tried and true methods, better ideas, etc are welcome here.

***************

I am adding some graphics excerpted from the 2nd edition that show the ABC method applied to a theater, opera house and arena. In each case you can see that the A, B and C composite elements are made up of several boxes. e.g. a 20 deg speaker might be 4 boxes at 5 degrees or 5 boxes at 4 degrees etc. There is an icon on the pics that shows where the on axis mic would be placed (the center of the A, B or C speaker) and another indicates where the tranisiton between composite elements is found.

13 Comments on “The ABC’s of Line Array Tuning”

  1. DL Says:

    Hey 6o6,

    Thank you for this. I compared a few different approaches in MAPP, found here: http://www.lundbergsound.com/projects/lao/lao.pdf . Something like this method is in the last 2 pages. I don’t think that there’s any way other than APFs or making the array a whole lot longer to improve consistency below 500 Hz, but this seems to have yielded some pretty good results above 500 Hz.

    Cheers,
    Daniel

  2. 6o6 Says:

    Daniel,
    Soory to be so late in commenting on this. Your studies are interesting. Indeed there is substantive change above 500 Hz. I agree that we would see more effect down low with a longer line – but also, I suspect we are going to see more effect with the tapering if we have more elements – even if they are smaller. This week I will give a go at a similar scenario to yours and see if I can get a more profound effect over frequency. In the meantime – I have some pics here of the ABC method as was applied in the book.
    Thx

    6o6

    • DL Says:

      The difficulty I have encountered in severe tapering of course is the excessive formation of LF nulls near the front. The inconsistency in this example below 500 Hz is much less of a problem in applications with balconies / the top boxes pointed upward. I’ll be interested to see what you find. Thanks.

  3. Goran Says:

    Hi Bob!
    I have a question or two…
    When setting up a line array, lets say we have applied gain tapering we calculated before and we have excellent uniformity in HF, but our LF response has gone wild because we are dealing with large amount of boxes and we have to taper it.. Should tapering be done on processors with gain control or with eq (shelf)? Lets say we divided our line array in ABC sections as you sugested, once we set up A section and add B section, should LF tapering be done on both A and B or just B section..
    Thank you in advance and sorry for bad English.

    With respect,
    Goran

    • 6o6 Says:

      Goran,
      Thank you for writing.
      you say “but our LF response has gone wild because we are dealing with large amount of boxes and we have to taper it.” Typically, if we have a large amount of boxes the LF response is more likely to go tame – than wild. If we have a lot of boxes the low end will become more narrow, more “controlled” , less wild. The big challenge is WHERE does the LF concentrate – and typically this is going to be lower (vertically) than we want. The level tapering predictable shapes the HF because there is some angular isolation (due to directionality). The LF is so overlapped so it is much harder to separate what cabinety is covering what area – they are all covering top to bottom. The level tapering helps to raise the beam up towards the back of the room – that is a start. If you add low cut filters to the lower boxes that can help – but I would suggest to be careful about how much. Even if you cut and cut and cut all day on the lower boxes (C) and (B) , the very close area will still get LF from the upper boxes (A). But if you go too far you may make the LF go too wide and sap away too much power. It is a tool, but use it carefully.
      It is for changes like this that it is critical to measure in multiple locations. When I tune a system like this I have 6-8 mics top to bottom – then I can see if the change is helping only one area – or alot of areas.
      Hope this helps.

      6o6

      • Goran Says:

        Bob, thanks for answer,
        Sometimes I have a problem expressing in English what I thought, it would be a lot easier to write in Croatian 😀 :D..
        Anyway, by “gone wild”, I meant we have LF volume build up, so in order to get “flat” response, should we lower the LF with EQ, or by reducing gain on LF speakers amplifiers?
        But OK, I understand now better… So actually, what you were talking is that LF minimum variance line “lays” under different angle than HF minimum variance line?

        (loud thinking): What if we could add some more boxes in our array on the upper side and lowpass them around say 700 Hz.. That way we would have longer line, and better directivity in LF, but our HF would stay intact…
        Here, I made few simulations of what I was thinkging.
        I put 16 MILOs and “designed” HF coverage with lower 12 boxes, so upper boxes are just LF reinforcement. For simulation, on the left side and over 1kHz I just muted upper 4 boxes.. Btw, I gain tapered lower 12 boxes.

        What do you think about this approach?
        I’m aware of its cons, like problems with array hang height, array weight, price, etc…

        Thank you for your time.
        With respect,
        Goran

  4. Pepe Ferrer Says:

    Hi Bob,

    I discovered your blog, so now even sleep less.

    basically I have some doubts, I understand that separate blocks simetricos helps determine the rank, energy, and its coverage, but you think that it is good to make equalization on each block decisions when we know that LF will work overlapping and its energy will share.
    On the other hand, we can find the XOVER position between a-B, that decrease level change towards less level block crossover point what should we compensate with delay that adjustment?, I understand that no.
    If really we can an identical level for all the coverage, we would not be losing the sound of the center of the stage image since we are modifying the offset to the pink area level.?

    And finally this is possibly a nonsense, but it is difficult to understand for me, that a 3 speaker separate betweem them 1º has coverage of 3º, independent of its nominal coverage is 3º, 12º or 90 °.

    Thank you for everything, and I hope anxious that one day you can return to come to Spain to give a seminar.

    • 6o6 Says:

      Pepe,
      All good points
      1) Shared EQ: If you 3 different big boxes for your PA would you turn them all on at once to start the LF EQ? For me no. It is true that the LF will combine – but the nature of the LF combination will not be clear until we have set the levels for the HF and MF regions. In such cases we have found over the last 25 years that it pays to EQ the LF in stages: 1st tame it as a soloist in the local area – e.g. the far area, the middle area and the near area for and A-B-C array. You are EQing for the box(es) – [the A array might be 3 boxes wide- trapezoidal boxes- or multiple boxes high – line array boxes] and their interaction to the room. Once each part of A,B & C is good to go in its local area – you only have one factor left to deal with – the interaction of A+B+C. This keeps things very clear about what is going on – and potentially how to solve interactions. if we start off with everything going – we cannot tell the room from the array parts and then it becomes very difficult to find a bad driver -or wiring error etc. If we go in steps we can see the progression of interactions and know for sure what is going on.
      2) If your A-B-C is made of single boxes with single, centered HF drivers/horns then delay is a good plan. The displacement and level asymmetry between the boxes can be compensated by delaying at the XOVR. But when the HF is a ribbon – and spread over the height of the box (whether a ribbon series of drivers or a ribbon emulator horn) the displacement at the box-box edge is very different from the displacement box-box center. In such case delay does not help much – A) it is a VERY small number between cabinet edges B) the asymmetric XOVR is very difficult to find exactly because of the spread ribbon sources C) it conflicts with the center-center relationship of A-B (because the XOVR in this case is edge to edge). Bottom line: delay here has potentila to solve very little and screw up alot.
      3) – You asked: If really we can an identical level for all the coverage, we would not be losing the sound of the center of the stage image since we are modifying the offset to the pink area level.? Please rephrase this. I don’t understand it enough to answer.
      4) There are limits to this. This is a big subject but the decisive factor is % angular overlap. Three 90 deg speakers at 1 deg will act like a coupled line source and reduce down to around 30 deg because it is 97% angular overlap. Three 1 degree speakers at 1 degree splay will spread to 3 deg because they are 0% angular overlap. Three 3 deg speakers at 1 deg will stay at 3 degrees combined because they are around 50% angular overlap. the more elements you have, the more certain the outcome will be as the mix of overlap between multiple elements pushes in favor of making the shape of the overall angle spread. I will create some pics to help show this……..soon

      I hope to come to Madrid again as well. I had a great time last year. My suitcase is ready!

      6o6

      • Pepe Ferrer Says:

        Hi Bob,

        Excuse my English, but it’s hard for me sometimes to explain properly.

        In the question, I was thinking about musical theater, where it is important to keep the sound image on stage.

        My question is: if we can maintain the same level in the first row of seats as the last chair in the stands, do we not distort the sound image?.

        Thank you very much for your time and greetings.

      • 6o6 Says:

        Pepe,

        Lo siento, porque mi espanol es muy pobre – mucho peor comparado a su ingles.

        In musical theater it is important to have some sound sources that are low, to aid the localization to the stage. If we only have a high cluster of speakers we will have a lot of image distortion no matter what. In the close areas we get the most sound from the stage – this helps the image , but the angle to a flying PA is the highest – this hurts the image. As we move back the sound from stage is reduced, but the angle to the PA is reduced also , so there is some balance in the effects.
        Frontfill speakers are a must to help the front 3 rows in theaters. Infill speakers, usually 1-2 meters above stage level, can really help the next several rows before the mains take over.
        So please understand – the desired result is even levels front-to-back, but near the front, the even level may be the result of a COMBINATION of main downfill+frontfill or infill. As we move back it will be mains only and then at the very back it might be mains+delay.
        Espero que ahora esta un poco mas claro.

        Gracias para una pregunta interesante!

        seis-zero-seis


  5. Bob,

    Enjoyed reading your blog! A question. I have often worked on tuning line arrays by beginning with a high pass filter on the whole system. I do measurements and listening with HPF @400hz. (I pick the HPF based on the freq where the array looses its directional control) Then, when everything sounds and looks even, I remove the filter and tune the rig as a whole. It has worked for me in the past very well. I often treat point source speakers the same way, checking coverage and doing quite a bit if measurement with HPF in place. I apply the same filter to the ref signal that is fed into SMAART. Any comments?

    • 6o6 Says:

      Thomas,
      I don’t understand. So hopefully you can clarify it for me.
      Since you are measuring a transfer function, and putting the same HPF on both sides of the equation the measurements will look the same whether the filter is in or out. In my experience, if you HPF the reference signal, the transfer function will make it look like your system goes down to DC. For the listening test part I can understand the HPF. It could help reduce the brain’s job of sorting out “what is different as one walks about. For the transfer function measurement I don’t see the advantage.

  6. Goran Says:

    Hi Bob!

    Now that I understand whole story a little bit more, if it’s not a problem, I’d like to ask you if you could clarify a little bit more the use of allpass filter as a tool for LF beam steering in large arrays (you mentioned them in last chapter of the book). How do you select filter’s “corner” (90 degrees phase shift) frequency and to how many boxes do you apply them (I suppose only upmost section of array)? Also do you use just first order filters?

    Thank you in advance and sorry for bad English…
    With respect,
    Goran


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s


%d bloggers like this: