Rational Acoustics



JohnB
May 8th, 2009, 02:31 AM
Hey all.

Just thought I'd point out a recent thread (http://srforums.prosoundweb.com/index.php/mv/msg/44778/0/48/0/) on PSW in which Dave Rat wound up discussing his views on End-Fire Cardiod Arrays vs. Traditional Cardiod Arrays (although his original intent was simply to defend/correct the semantics of one line in his recent blog post (http://www.ratsound.com/cblog/)).

Unfortunately, the core of his argument didn't seem to get communicated well (at least not until much later in the thread), so here's the summary:


First, let's look at the difference in the two approaches:

The End-Fire only uses a difference in time arrival between two sources to create a directional pattern. That is, by delaying the front box(s) to the rear, you create in-time summation on-axis and some time difference off axis. That time difference will be at a maximum 180* off-axis, thus we design the array such that this maximum time difference causes a 180* phase-shift at Fo Hz (where Fo is application-dependent). Now, what happens when we move away from Fo in either direction? Well, the phase shift changes thus, we no longer have complete cancellation. If you take this into account at all points in space, you find that the coverage is cardiod at Fo, but shifts towards omni as we diverge from Fo (in either direction). Thus, we can say the End-Fire array exhibits a smooth frequency response on-axis at the cost of pattern control at low frequencies (and high freq's, though we usually design the x-over point to alleviate that problem).

The Traditional array uses a time difference and a polarity difference to create a directional pattern. By flipping the polarity on the rear box and delaying it to match the front box in time at 180* off-axis, we have cancellation at all frequencies. On axis, however, we now see the front box's signal, plus another signal delayed by [twice the aforementioned delay + (effectively) 180* of phase shift (i.e. a varying delay over freq.)]. Now, the resulting frequency response on-axis is going to look at lot like a comb filter, but since the phase difference seen on-axis varies with frequency, there's an important distinction: the response starts to drop as frequency decreases indefinitely (as opposed to a comb filter, where the frequency response levels off at low frequencies). Thus, we can say the Traditional array exhibits constant pattern control over frequency at the cost of overall frequency response.

Dave's second argument is that of robustness. Let's say a driver fails on a Traditional array; what happens? Well, best-case, it is the rear driver and we lose almost 6dB of sub on-axis and we lose pattern control (sub spills onto the stage). Worst case, however, it is the front driver and the resulting omni-directional signal is polarity-reversed. "Big deal" you may say... "You can't hear a change in polarity," right? Sure, but when you put that Traditional pair in an array with other boxes, you have the potential for some wacky problems. With the End-Fire array, you can only lose the audio from that one box, you can never cancel out audio you didn't originally intend to cancel out.


Ok, that last paragraph was a bit of a paraphrase. In his posts, Dave mentions a specific instance wherein the loss of audio in the front (due to power compression, in his case) actually caused the pattern of the array to reverse itself. Being that that it isn't actually possible for the Traditional array as I am defining it here (that is, no amount of level offset between the two drivers will cause the observed phenomenon), it seems to me that either he is talking about a similar array working on slightly different principles, or the nonlinearities caused by the power compression process some how caused it (in which case it is hard to say for certain which array is better without straight up experimentation).


[Edit: "90* of phase shift" changed to "180*"]

Harry Brill Jr.
May 10th, 2009, 10:06 AM
On axis, however, we now see the front box's signal, plus another signal delayed by [twice the aforementioned delay + (effectively) 90* of phase shift (i.e. a varying delay over freq.)]. Now, the resulting frequency response on-axis is going to look at lot like a comb filter, but since the phase difference seen on-axis varies with frequency, there's an important distinction: the response starts to drop as frequency decreases indefinitely (as opposed to a comb filter, where the frequency response levels off at low frequencies). Thus, we can say the Traditional array exhibits constant pattern control over frequency at the cost of overall frequency response.



I won't comment on Dave's post for now particularly since I haven't read it yet. I was a bit lost by the above statement. Perhaps I'm just misunderstanding.

In this approach the pattern control is much more consistent over the frequency of the bandpass. I have noted a hit on frequency response at the low end. The negative hit is in "feel" or "impact". Also the rear driver is 180 degrees behind the front driver in time at the frequency of interest which is what causes the smearing that steals that powerful impact. Not sure where you got 90. This is what I need clarified.

I have a theory the frequency you are basing your array on it NOT the center of the bandpass since that is where I find I get the best result and haven't really experimented with alternatives with a 2 deep end fire.

I also want to note there are some that call one method end fire and the other cardioid. I believe cardioid is a polar pattern (heart shaped) and end fire is an implementation. I choose to call them end fire method 1 and method 2 until someone shows me an AES adopted standard of naming these arrays.
Otherwise it's confusing.

There is also the method of using 3 or 4 deep sources which is superior to either of the 2 deep methods. Beyond 4 deep there is not real benefit to be had.

JohnB
May 13th, 2009, 01:47 PM
Hi Harry-

First, I apologize for writing 90*. I meant 180, and actually thought I had corrected it before posting, but apparently I was mistaken.


Thus, we can say the Traditional array exhibits constant pattern control over frequency at the cost of overall frequency response.

Second, what you described is precisely the point I was making. The last sentence of the quote you took (repeated above) sums up what you say you've seen in the field: consistent pattern control with a loss in very low frequencies, which would be consistent with a negative hit in "feel" or "impact" (depending where it starts to drop off, of course). I was simply trying to explain what mechanism that comes from.

Lastly, I actually chose to use the terms "End-Fire" and "Traditional" for that very reason. "Cardiod" is just a pattern they are both capable of producing. However, since that naming convention seemed to be throwing several people off (and as method 1 and method 2 can be equally confusing after a long discussion), I chose to define the term "Traditional" to describe that method. Although, personally, I tend use the "Cardiod" term in the field.

Sorry for any confusion. Any other thoughts on what a good naming convention might be?

Ferrit37
May 14th, 2009, 11:40 AM
Hey John,
In my limited experiments (my wife won't let me bring subs into the house :( ) I think that the real issue with "loss of impact" is that all these methods cause the transient response to "smear".
Like all things in life - TANSTAAFL :) (although I'll settle for a cheap breakfast)