Lets assume we have four 15" drivers each with 90 dB / 2.83 V sensitivity, and we connect them in parallell. That should give an increase of 12 dB in voltage sensitivity, which is 102 dB / 2.83V
Now to the interesting point: is there any acoustic interaction between the drivers that gives us any more increase in sensitivity? I've seen claims from a speaker designer that we will get up to 3 dB additional increase in this case because of driver interaction.
Is this true, not true, or someway in between?
Now to the interesting point: is there any acoustic interaction between the drivers that gives us any more increase in sensitivity? I've seen claims from a speaker designer that we will get up to 3 dB additional increase in this case because of driver interaction.
Is this true, not true, or someway in between?
It can be true.. but it isn't broad-band. Nor is it a set db figure.
Vance Dickson has this in his Loudspeaker Cookbook when using 4+ loudspeakers, often freq.s between 300 and 1.5 will receive some additional gain. An increase in drivers results in an increase in gain.
There are of course other factors like driver seperation distance, driver diameter vs. wavelength, etc...
Member
Joined 2003
Doubling drivers (1 to 2) will gain 3db by doubling current/power delivery, 2 to 4 will gain another 3db, again by re-doubling the current/power. If there is close acoustic coupling, 1 to 2 will gain up to an additional 3db though increased radiation resistance, 2 to 4 can gain another 3db.
So, four drivers can deliver up to 12db through a combination of power delivery and acoustic coupling. Don't recall the driver spacing for full coupling, but I believe it is at/under a wavelength...think it is covered by Olsen etc.
So, four drivers can deliver up to 12db through a combination of power delivery and acoustic coupling. Don't recall the driver spacing for full coupling, but I believe it is at/under a wavelength...think it is covered by Olsen etc.
Would it, by any chance, be when the speaker stops being a point source and becomes a plane/line source? These tend to have much greater sound projection than a single driver speaker, when they're both at the same level at 1m.
I'm no expert, but it hasn't been mentioned yet, so I figure I should put it into the melting pot.
I'm no expert, but it hasn't been mentioned yet, so I figure I should put it into the melting pot.
Simply doubling the "baffle area" by positioning a second driver alongside the first will give you up to 1.5 dB. See my simulation in the diagram to the right (stig5.gif). Dotted line is for the added baffle.
Putting four drivers in a square could give +6dB - with still only one driver working. See how the on-axis response (upper line) is more affected than the power response - which was to be expected.
Another effect caught my attention: I simulated four drivers in the corners of a square with 2 m side length. Upper red line is the combined on-axis response, blue line same for a single driver, lower red line is the power response (second from left Stig3.gif).
(third from left Stig4.gif) is the same, but now with all drivers grouped closely together at 40 cm distance. On-axis response would not change in free space, but the power response would be raised by almost 5 dB at 100 Hz. So it would be a good idea to have those H-frames not too wide apart. 😉
Putting four drivers in a square could give +6dB - with still only one driver working. See how the on-axis response (upper line) is more affected than the power response - which was to be expected.
Another effect caught my attention: I simulated four drivers in the corners of a square with 2 m side length. Upper red line is the combined on-axis response, blue line same for a single driver, lower red line is the power response (second from left Stig3.gif).
(third from left Stig4.gif) is the same, but now with all drivers grouped closely together at 40 cm distance. On-axis response would not change in free space, but the power response would be raised by almost 5 dB at 100 Hz. So it would be a good idea to have those H-frames not too wide apart. 😉
Attachments
When discussing multiple drivers systems it is important to separate out the differences between efficiency and radiated SPL. For a multiple driver system where the drivers are acoustically close (much less than a wave length), assuming that each driver individually radiates the same SPL as if it were isolated, then the SPL is just the vector sum of the SPL form each driver. Being acoustically close means all drivers operate in phase so the SPL would increase by 6dB for two driver, 9.5dB for 3 drivers and 12dB for 4 drivers. This assumes also that the multiple driver system, is also small in dimensions compared to the wave length. (In other words, the radiation pattern remains spherical).
When looking at efficiency it is a simple matter of summing the input power. To keep the radiated SPL of each individual driver constant at the single driver level, the power delivered into the 2 driver system increases by 3dB, the three driver system by 4.77 dB and the 4 driver system by 6 dB. The the efficiency of the 2, 3 and 4 driver system increase by 3, 4.77 and 6dB as well. Or you cam look at it as the two driver system produces the same SPL as the single driver with 3 dB less power. The 3 driver system, the same SPL with 4.77dB less power and the 4 driver system the same SPL with 6dB less power.
Lastly, we should not confuse the effect of the baffle size on the SPL. It will produce the results the Rudolf indicates, but this is a result of the change in where the baffle related 2Pi tp 4Pi transition occurs, not changes is efficiency of the drivers.
When looking at efficiency it is a simple matter of summing the input power. To keep the radiated SPL of each individual driver constant at the single driver level, the power delivered into the 2 driver system increases by 3dB, the three driver system by 4.77 dB and the 4 driver system by 6 dB. The the efficiency of the 2, 3 and 4 driver system increase by 3, 4.77 and 6dB as well. Or you cam look at it as the two driver system produces the same SPL as the single driver with 3 dB less power. The 3 driver system, the same SPL with 4.77dB less power and the 4 driver system the same SPL with 6dB less power.
Lastly, we should not confuse the effect of the baffle size on the SPL. It will produce the results the Rudolf indicates, but this is a result of the change in where the baffle related 2Pi tp 4Pi transition occurs, not changes is efficiency of the drivers.
There are inter-driver effects, but they would best be described as 2nd order for 4 drivers. In a sonar torpedo head there were 256 individual transducers. At one location the inter-driver loading actually went negative at a certain frequency and blew up this transducer. Hence, in arrays of that complexity the effects can be significant, but with only four, its not really going to matter. That makes all the analysis above correct to first order.
After a nights sleep over the problem:
If you have some 15'' drivers to spend and we are talking of <200-300 Hz, you don't want to concentrate those drivers at one (or two) places. You want to distribute them WRT room modes. The advantage will be much greater than any possible gain for the clustered drivers.
BTW: Does anyone know, whether we can locate the below 200 Hz part of music (not sine waves!) for itself, when simultaniously hearing the part above 200 Hz from another location in the room? I'm not interested in individual observations, but scientific studies only. Do they exist?
Rudolf
If you have some 15'' drivers to spend and we are talking of <200-300 Hz, you don't want to concentrate those drivers at one (or two) places. You want to distribute them WRT room modes. The advantage will be much greater than any possible gain for the clustered drivers.
BTW: Does anyone know, whether we can locate the below 200 Hz part of music (not sine waves!) for itself, when simultaniously hearing the part above 200 Hz from another location in the room? I'm not interested in individual observations, but scientific studies only. Do they exist?
Rudolf
What if the 'individual' is yourself? 😀
---------------
I'd like to recommend a perfect man (and his awsome system) for this experiment:
http://www.diyaudio.com/forums/multi-way/90804-large-midrange-ob-scott-g-40.html#post1953221
However the re-config of such system must be very difficult....
---------------
I'd like to recommend a perfect man (and his awsome system) for this experiment:
http://www.diyaudio.com/forums/multi-way/90804-large-midrange-ob-scott-g-40.html#post1953221
However the re-config of such system must be very difficult....
See Jim Griffin's line array paper.
http://www.audioroundtable.com/misc/nflawp.pdf
In addition feel free to model in:
Vertical Polar Response: Line Array
The nice thing about arrays is they can work with the room when long enough.
http://www.audioroundtable.com/misc/nflawp.pdf
In addition feel free to model in:
Vertical Polar Response: Line Array
The nice thing about arrays is they can work with the room when long enough.
Last edited:
The relationship for multiple drivers only applies for relatively small numbers of drivers. Efficiency does not double each time the number of drivers is doubled without end. If it did we could exceed 100% efficiency.
Fair point.
Maybe the efficiency fails to increase when you've exceeded a certain number of wavelengths (Elliott seems to think 10 wavelengths to be a significant figure)
And the output now sees an infinite plane, rather than something with an edge. His article on PA systems (line array section) seems to conform to this.
PA Systems
Maybe the efficiency fails to increase when you've exceeded a certain number of wavelengths (Elliott seems to think 10 wavelengths to be a significant figure)
And the output now sees an infinite plane, rather than something with an edge. His article on PA systems (line array section) seems to conform to this.
PA Systems
Pure scientific perhaps not, but logically, based on the wavelengths involved, localization must fail below some frequency. In tests that we did in vehicles, this frequency was about 150-200 Hz. In small rooms I would suggest 100-120 Hz is a clear demarcation below which no localization is possible.
I completetly concur with the distributed subs as I have expounded on many occasions.
From experince it is possible to localize a sub IF it generates sounds above its intended range. These can come from a number of sources - noise, nonlinearities. This makes distortion an issue not because we hear it or are offended by the distortion products (which are heavily masked at these LFs) but we can localize on them and this defeats the positive effects of distributing the sources.
This is the point that I made before. With enough drivers the 2nd order inter-driver effects come into play. The drivers will start to load one another (called Mutual Coupling) as the efficiency goes up and this load will get greater and greater with more drivers. The individual drivers will start to decrease their output because of the increased load thus always keeping the efficiency below some level, usually 50%, but it could go higher. It can never get anywhere near 100%.
Personally, I don't like the word efficiency thrown around even used in the line array paper. Efficiency is typically presented in percentage since it is a ratio of what goes in vs what comes out. Speaker efficiency is always a very low number because of the impedance mismatch.
I have to agree with Geddes findings on cars, room modes dominate in the midbass region making localization harder higher in frequency. It depends on room size, the larger the room the lower the frequency, the smaller the room the higher the frequency localization can occur.
Use this room mode calculator to understand.
http://www.bobgolds.com/Mode/RoomModes.htm
In larger rooms >40Hz
http://www.filmaker.com/papers/RM-2SW_AES119NYC.pdf
I think there is a point however our hearing cannot resolve location. We simply have not evolved this way. In nature (out in the open)we hear or feel low frequency reflected from the ground (thunder, stampedes...etc) and the rest is bone conduction. This brings to point if objects close by make noise with low frequency excitation (pictures shaking on the wall), it will most likely fool you into thinking that is where the sound comes from over-riding room modes as the domiant factor.
I have to agree with Geddes findings on cars, room modes dominate in the midbass region making localization harder higher in frequency. It depends on room size, the larger the room the lower the frequency, the smaller the room the higher the frequency localization can occur.
Use this room mode calculator to understand.
http://www.bobgolds.com/Mode/RoomModes.htm
In larger rooms >40Hz
http://www.filmaker.com/papers/RM-2SW_AES119NYC.pdf
I think there is a point however our hearing cannot resolve location. We simply have not evolved this way. In nature (out in the open)we hear or feel low frequency reflected from the ground (thunder, stampedes...etc) and the rest is bone conduction. This brings to point if objects close by make noise with low frequency excitation (pictures shaking on the wall), it will most likely fool you into thinking that is where the sound comes from over-riding room modes as the domiant factor.
Last edited:
While room modes certainly degrade the issue, I believe the bigger effect is our hearing. At 100 Hz the wavelength is ten feet. Our ears are spaced by a few inches. The SPL at the two ears is virtually identical. On what basis could our ears even begin to identify a sound direction? The modal situation pushes the situation higher in frequency for smaller rooms, but even in free space there has to be a limit to what we can resolve directionally.
This is relating to what acoustically close means. Acoustically close usually means much less than a wave length. 10 wave lengths is acoustically far apart. When the sources are far apart the SPL form each source in basically uncorrelated and there is no gain in efficiency. The radiated power is 3dB greater than a single source and the input power is also 3dB greater. However, the SPL becomes a strong function of position reaching +6dB over a single driver at some positions to cancellation at others.
This figure shows the power radiated by two omni directional sources 4' apart (282 Hz).
An externally hosted image should be here but it was not working when we last tested it.
At low frequency the power is 6dB over a single source. At higher frequencies (above 3k Hz) it is only 3dB over a single source. In between you can see the way things behave in the transition region.
Ear spacing is what gives us the ability to locate in the lateral direction anything below 1.6khz using timing differences. There are lots of frequencies below 1.6Khz that are greater than this spacing. How are you going to place a limit on LF localization based on that one measurement that separates the upper limit of using timing differences to locate? I think unequal pressures cause us to want to turn our head or find the location that is least unpleasant for out of phase signals and this means turning our heads until we are at 0deg azimuth with the source since that would equalize the differences.
Free space is not reality so I'm not really interested in what we hear in a fictious place and I know the room is always going to dominate, at the very least you cannot remove the ground. I still content though that VLF is going to shake something at reasonable output levels and give away a real or false location first, then room modes, and the final limitation is our hearing/brain resolution.
Last edited:
I would probably respond if I understood the question.
But:
1) in a steady state there aren't really timing differences, there are phase differences and at LF these phase differences at the ears are negligable, as are the amplitude differences.
2) the sound field at LF is basically always steady state because it takes at least four or five periods of the sound to even begin to detect its frequency and by this time the ear is in steady state.
Only the high frequency transients would be detectable directionally, but these have been LP'd out of the sub signals.
- Status
- Not open for further replies.