Just a phase?
As DS-21 mentioned our speakers specifically, I figure it is something I can comment on at least from a designer’s point of view.
If one simulates an arbitrary phase shift with a DSP “all pass” filter or other method and then listens with headphones or speakers, one will conclude that only on some program material is there any real audible difference and at that, it can be hard to say which way is better “with or without” the extra phase rotation. Beyond the purists goal of “wire with gain” or in this case electrical signal in equals acoustic signal out, there are reasons this area matters and why it is so hard to address that it usually isn’t.
Richard Heyser was the first person I know of to write about and also find a way to measure a loudspeakers time delay and then acoustic phase. The acoustic phase is the alignment of the acoustic pressure relative to the input signals phase when all the fixed time delay is removed (at a theoretical zero distance).
In other words, producing a sine wave, the actual pressure at the drivers radiator may be behind at or ahead of the electrical signal, depending on the driver’s acoustic phase. “Minimum phase”, most drivers are that at least down low in frequency, what minimum phase means is that there is a corresponding phase shift for ANY change in amplitude and all the phase wrap caused by time delay has been removed from the display.
While most popular measurement systems use a pseudo random signal to calculate the impulse response and then magnitude and phase, one can measure magnitude and phase and calculate impulse response. Any systems impulse response is set by the systems magnitude and phase response, they are two views of exactly the same thing.
When you add two identical signals together through a pair of resistors, one has simple vector addition and if one of the two were to be reversed, the result is they still add coherently, +1 added to -1 equals 0, cancellation. With the phase at say 120 degrees difference instead of 180, now one plus one equals one and as the differences approaches zero, the result gets closer to 2.
In other words, with coherent addition, the result is the vector sum of the two signals.
With two loudspeaker drivers, this is also how it us at least for subwoofers. With acoustic sources, one has coherent addition up to the point where the two sources are more than about 1 / 4 wavelength apart. With larger spacing, the two sources radiate independently and produce an interference pattern comprised of a series of lobes and nulls. With an interference pattern, if one reversed one of the two sources, the result isn’t cancellation to zero but rather the pattern of lobes and nulls change.
In home hifi, people mostly believe what the magazines tell them and most hifi speakers appear to be partly designed based on expectations as opposed to acoustics or physics as it applies to the subject.
But in commercial sound, the room problems are much worse and all the acoustic problems with loudspeakers get worse, the greater the acoustic power one tries to produce.
The flaws always get louder faster than the desired signal with increasing input level.
For our company, this is where I chose to focus because as an old friend and former hifi reviewer said to me once “there are speakers that go loud and there are speakers that sound good but none do both”.
The real problem here is that the larger the acoustic power you need, the large number of drivers and frequency divisions it takes. That is needed because each time one doubles the distance to the source, the SPL drops by 6dB (1/4th).
You may have noticed that all things equal, there is an inverse correlation between the size of a sound system and the fidelity. You may have noticed that a pair of little fostex FR drivers on a large flat baffle can produce a phenomenal stereo image or that KEF is now making a speaker intended to radiate as a single point source. You may have noticed that in an outdoor venue, if the wind blows, that the sound of a large line array changes wildly or is swishy.
These effects are caused by the interference patterns the speakers radiate (of not.).
For our speakers, the Synergy horn design allows all of the drivers to add coherently into a single acoustic source. This is very important in commercial sound as the object is to make the sound “the same” everywhere in front of the speaker. With the crossover design, the physical position of the drivers and the acoustic coupling between them allowed a crossover that eliminates the “all pass” phase of normal crossovers and the result is some of the speakers can produce a square wave, over a wide band, with little position dependence out front. Signal waveshape in equals acoustic pressure waveshape out. While not thought to be important, is the result of having all the drivers act like a single wide band source.
For the small Fostex drivers, while limited in bandwidth and loudness, they radiate as a single source up to a fairly high frequency. This means that standing in front of just one driver, what reaches each ear is very nearly the same and as a result with one’s eyes closed, it is harder to locate the speakers physical depth using ones ears, the direction sure but not how far away. Most multi-way speakers produce an interference pattern due to the driver spacing etc and so are easy to localize in this kind of test. To localize depth or position in depth with eyes closed, one needs clues, differences between right and left ears. Switch to stereo and the fostex driver produce a strong image anywhere between L and R speakers. An indicator of how much self interference any speaker has is audible when you do the one source, eyes closed test with a human voice. With a “perfect point source speaker” you should not be able to easily hear it’s depth location just the direction, just a voice floating in front of you some distance away.
While the Synergy horns do this too, our business is focused on larger scale sound where this solution makes the audible difference large enough for venue owners to choose an unknown company based on sound quality. Myles, I see your in NY, if you have been to the Madison Square park concerts this season, they have used this kind of speaker instead of the normal line arrays.
To a degree one can “hear” the effect of radiating as a coherent single source in this video (put on headphones), notice how the sound doesn’t change appreciably as he moves around and out to a distance later measured to be 700 feet.
http://www.youtube.com/watch?v=pk54IFD4znw
For a true point source, the spectral balance doesn’t change with distance, only the loudness falls and hf air absorption.. More noticeable is that the wind has hardly any effect as the wind moves the beam around because of the homogeneity of the radiation.
The first public demo of the JH-90 speaker at the Infocom trade show two years ago lead to the BYU stadium installation and Northwestern University, the sound quality of those two jobs brought in 12 other stadium sales, the most recently completed is Lambau field in Greenbay Wi where a large conventional sound system was just pulled out.
The cool part (for me), I get to design and refine most of these in my living room as part of my home hifi, I used to dream about coherence combined with not running out of headroom back when I made electrostatic speakers.
So, phase response in a real speaker (as opposed to simulations of only phase shift) is usually tied to the physical orientation / spacing of the drivers and also tied to the presence of or absence of an interference pattern which can be plainly audible as it is localizable, the speaker is yelling out “here I am” when ideally it’s location in depth should be ambiguous as it is not part of the recorded image .
In the development of the Unity and then Synergy horns (over the last 12 years) I noticed this effect and was puzzled by it. AS they got closer and closer to a single broadband source, it got harder to hear how far away they were with your eyes closed, subjectively the sound got simpler and simpler for lack of a better descriptor..
Best,
Tom Danley
http://www.danleysoundlabs.com/
