Years ago i picked up half of someone's (probably) line array project that never happened.
I have a box of 32 of those 5.25" Foster drivers that Madisound closed out in about 2007. Good response from 100hz to 5khz in an open baffle.
I hesitated on the pile of 32 matching tweeters because they were all embedded in 6x9 frames, and someone else picked those up. But I can get tweeters from Apex Jr.
I am in the process of building a theremin. Art Harrison's revised model 126 6-tube theremin.
The recommended monitor for thereminists is a speaker above and behind the thereminist, with full range response.
So, since i HAVE this box of identical speakers, I think a 5x5 bessel array would look pretty striking on a pole. Maybe with a subwoofer in the base.
Since it's an instrument speaker this is a monaural build, but i have a few 100w+ rated two-channel power amps of fair but not fantastic quality for sound reproduction. So, wiring it as two channels would be ideal for me.
Paul Kembel gives a diagram for a 3x5 array here:
http://www.angelfire.com/sd/paulkemble/soundgb.html
And a 5x5 with "hemispherical sound distribution" is described in the philips document, but i don't understand the notation and can't quite wrap my head around how to wire it up. That document is here, fwiw:
http://www.ee.ic.ac.uk/naylor/LineSourceLoudspeakers/philips091.pdf
Also referenced briefly here:
JUTE on BESSEL
Jute's diagram looks like this:
What gives me a spin is where this is 1, 2, and 4 weightings. There's no 3? How do i get to 4? The Philips doc says something about "easily constructed with 4 ohm and 8 ohm speakers" and the fact that it says "and" instead of "or" suggests that he means a mix, not either/or.
I'm also confused by the question of spacing. I was pretty sure i understood from the stuff I've read that the distance from cone to cone should be 1 cone diameter, but Jute's page describes one array with the spacing "as close as possible".
¯\_(ツ)_/¯
Can anyone fill in the gaps for me?
I have a box of 32 of those 5.25" Foster drivers that Madisound closed out in about 2007. Good response from 100hz to 5khz in an open baffle.
I hesitated on the pile of 32 matching tweeters because they were all embedded in 6x9 frames, and someone else picked those up. But I can get tweeters from Apex Jr.
I am in the process of building a theremin. Art Harrison's revised model 126 6-tube theremin.
The recommended monitor for thereminists is a speaker above and behind the thereminist, with full range response.
So, since i HAVE this box of identical speakers, I think a 5x5 bessel array would look pretty striking on a pole. Maybe with a subwoofer in the base.
Since it's an instrument speaker this is a monaural build, but i have a few 100w+ rated two-channel power amps of fair but not fantastic quality for sound reproduction. So, wiring it as two channels would be ideal for me.
Paul Kembel gives a diagram for a 3x5 array here:
http://www.angelfire.com/sd/paulkemble/soundgb.html
And a 5x5 with "hemispherical sound distribution" is described in the philips document, but i don't understand the notation and can't quite wrap my head around how to wire it up. That document is here, fwiw:
http://www.ee.ic.ac.uk/naylor/LineSourceLoudspeakers/philips091.pdf
Also referenced briefly here:
JUTE on BESSEL
Jute's diagram looks like this:
Code:
1 2 2 -2 1
2 4 4 -4 2
2 4 4 -4 2
-2 -4 -4 4 -2
1 2 2 -2 1What gives me a spin is where this is 1, 2, and 4 weightings. There's no 3? How do i get to 4? The Philips doc says something about "easily constructed with 4 ohm and 8 ohm speakers" and the fact that it says "and" instead of "or" suggests that he means a mix, not either/or.
I'm also confused by the question of spacing. I was pretty sure i understood from the stuff I've read that the distance from cone to cone should be 1 cone diameter, but Jute's page describes one array with the spacing "as close as possible".
¯\_(ツ)_/¯
Can anyone fill in the gaps for me?
I might assume that 2 = -6dB (and 3 might represent -3dB if there were any). I'd assume this based on the 2s giving a region of pressure halfway between the middle and outer region pressures, perhaps for diffraction reduction.
I think the question should be regarding the size of the region. If there was a need, and there were enough drivers then 3s may be justified, and 1s should probably be added also for balance with the 3s.
As it is I'm inclined to assume it is like a roundover.. Any is good, larger is better, much larger only helps at lower frequencies where it tends to matter less.
There is no 3.
The weightings shown represent the amount of power each driver should dissipate.
The "-" signs represent drivers that should be wired up in reverse polarity compared to the ones without the "-".
Use 4 ohm drivers for the "4" weightings. Use 8 ohm drivers for the "2" weightings.
Wire pairs of 8 ohm drivers in series for the "1" weightings.
Alternatively, you could use all 8 ohm drivers. Each "4" driver gets the full drive voltage.
The "2" drivers are wired as 6 series pairs. The 4 "1" drivers are wired in series as a single string.
Both of those arrangements present an inconveniently low impedance to the amplifier.
I can't help on the spacing, but I'd put them as close together as practical.
You might find Don Keele's paper insightful:
http://www.xlrtechs.com/dbkeele.com/PDF/Keele (1989-10 AES Preprint) - Bessel Arrays.pdf
The weightings shown represent the amount of power each driver should dissipate.
The "-" signs represent drivers that should be wired up in reverse polarity compared to the ones without the "-".
Use 4 ohm drivers for the "4" weightings. Use 8 ohm drivers for the "2" weightings.
Wire pairs of 8 ohm drivers in series for the "1" weightings.
Alternatively, you could use all 8 ohm drivers. Each "4" driver gets the full drive voltage.
The "2" drivers are wired as 6 series pairs. The 4 "1" drivers are wired in series as a single string.
Both of those arrangements present an inconveniently low impedance to the amplifier.
I can't help on the spacing, but I'd put them as close together as practical.
You might find Don Keele's paper insightful:
http://www.xlrtechs.com/dbkeele.com/PDF/Keele (1989-10 AES Preprint) - Bessel Arrays.pdf
Last edited:
Eric,
the solution is quite simple. take a look a the following 5 x 5 array using identical speakers:
a1 b1 c1 d1 e1 (l_a)
a2 b2 c2 d2 e2 (l_b)
a3 b3 c3 d3 e3 (l_c)
a4 b4 c4 d4 e4 (l_d)
a5 b5 c5 d5 e5 (l_e)
1. step
Build 5 single horizontal arrays by using the left serial connection scheme (7285803) from fig. 2.
2. step
Each of the the horizontal arrays l_a to l_e can now be connected by using the parallel connection scheme (7285804) from fig. 2.
This serial parallel combination will deliver the proper power distribution keeping the impedance of a single speaker. Alternatively you can start with the parallel connection scheme in step 1 and use the serial connection scheme in step 2. The result will be the same.
Last edited:
Bodo's wiring scheme makes more sense than mine. 🙂
It would give a more practical load to the amplifier.
It would give a more practical load to the amplifier.
My drivers are all the same impedance, so mixing impedances is a non-starter.
Keele's paper sort of lets me down where it says "Only the line array Bessel configurations are analyzed in this study". For the purposes of this thread, I'm interested in panels.
I could make line arrays. I could even design them in CAD and throw some wood or metal on an abrasive waterjet table and cut them out in an hour. For this project, I'm more interested in a panel.
It's quite possible that "as close as possible" but with uniform spacing is what Philips Technical Document 091 was trying to say. Likely, even.
Just to be clear. The conventional speaker array, such as was described by Charles F. Mahler, Jr. in "Hi-Fi Performance from Small Speakers", Audio, December 1959, maximizes both the bass response and the SPL from a large number of small speakers.
However, these panels tend to be increasingly directional, and frequency response suffers as the frequency increases.
I'm interested in bessel panels because
(a) I have 32 identical Foster 5.25" drivers
and
(b) Bessel panels offer radial or hemispherical dispersion and a frequency response essentially identical to the individual driver.
However, these panels tend to be increasingly directional, and frequency response suffers as the frequency increases.
I'm interested in bessel panels because
(a) I have 32 identical Foster 5.25" drivers
and
(b) Bessel panels offer radial or hemispherical dispersion and a frequency response essentially identical to the individual driver.
You have to consider each of the rows as a virtual speaker.
So you start in step 2 with the five virtual speakers l_a, l_b, l_c, l_d and l_e.
Using the parallel network connection you will get the following normalized vertical bessel power distribution:
l_a : 1 (serial connection with l_d)
l_b : 2
l_c : 2
l_d : -2 (polarity reversed)
l_d : 1 (serial connection with l_a)
Each row a1 to e1, a2 to e2, a3 to e3, a4 to e4 and a5 to e5 is already showing the
normalized horinzontal bessel power distribution due to serial connection used in step 1.
a : 1 ( parallel with e) b: 2 c: 2 d: -2 (reversed) e: 1 (parallel with a)
You will get the power distribution for each single speaker by multiplying
the vertical power distribution of all rows with the horizontal power distribution of each row.
Row l_a: 1 * (1 2 2 -2 1) = ( 1 2 2 -2 1 )
Row l_b: 2 * (1 2 2 -2 1) = ( 2 4 4 -4 2 )
Row l_c: 2 * (1 2 2 -2 1) = ( 2 4 4 -4 2 )
Row l_c: -2 * (1 2 2 -2 1) = (-2 -4 -4 4 -2 )
Row l_e: 1 * (1 2 2 -2 1) = ( 1 2 2 -2 1 )
So each row and collumn will show a proper bessel power distribution.
I guess that makes sense?
I went digging for the drivers last night. They are factory cases that hold 12 each, and i remember that one case was incomplete. And that two of the drivers got soldered into another system briefly for testing, so i should have two drivers with solder on the tabs.
Right now i have found two boxes containing 10 and 8 drivers respectively, none with solder on their tabs, so maybe i only have 20 drivers. Or maybe i have a full box of 12 somewhere. I will have to check the UPS labels to see how many boxes were in the shipment from madisound.
I went digging for the drivers last night. They are factory cases that hold 12 each, and i remember that one case was incomplete. And that two of the drivers got soldered into another system briefly for testing, so i should have two drivers with solder on the tabs.
Right now i have found two boxes containing 10 and 8 drivers respectively, none with solder on their tabs, so maybe i only have 20 drivers. Or maybe i have a full box of 12 somewhere. I will have to check the UPS labels to see how many boxes were in the shipment from madisound.
Hi Eric, my thread here will answer all of your questions:
http://www.diyaudio.com/forums/multi-way/301259-improved-array.html
It's a long read, so here's a summary:
The wiring scheme that your using proposes a "hemispherical soundfield."
In 2017 terminology, they are using power tapering to make the beamwidth wider. The "Linus" paper explains how this works. I've also posted a pile of mind numbing simulations that examine whether this works or not.
But ask yourself, do you really CARE about a wide beamwidth? I, you're building a Theremin, who cares if the beamwidth is narrow? You're sitting right in front of the speaker.
A couple of other people who've swayed my opinion are "wesayso" and Monte Kay. Both have extolled the virtues of a straight (unshaded) line.
If you opt for an unshaded line, all of this gets WAY simpler. You can build an unshaded Bessel area without resistors. A shaded Bessel array using a single amplifier is going to require resistors for the shading.
By the way, if you don't have enough drivers, you can do the shading by simply omitting drivers. This is called "spaced shading" and it's used by the seven element Bessel array.
Here's an example:
Let's say that you have a two-row Bessel array with five drivers in each row:
1watt 2watt 2watt -2watt 1watt
1watt 2watt 2watt -2watt 1watt
Does that make sense? Ten drivers?
_____ 2watt 2watt -2watt
2watt 2watt 2watt -2watt 2watt
Now you just made it eight!
See how that works? Same amount of SPL, but with two less drivers, because two are doing the work of one.
This same 'trick' can be used in any shaded array.
http://www.diyaudio.com/forums/multi-way/301259-improved-array.html
It's a long read, so here's a summary:
The wiring scheme that your using proposes a "hemispherical soundfield."
In 2017 terminology, they are using power tapering to make the beamwidth wider. The "Linus" paper explains how this works. I've also posted a pile of mind numbing simulations that examine whether this works or not.
But ask yourself, do you really CARE about a wide beamwidth? I, you're building a Theremin, who cares if the beamwidth is narrow? You're sitting right in front of the speaker.
A couple of other people who've swayed my opinion are "wesayso" and Monte Kay. Both have extolled the virtues of a straight (unshaded) line.
If you opt for an unshaded line, all of this gets WAY simpler. You can build an unshaded Bessel area without resistors. A shaded Bessel array using a single amplifier is going to require resistors for the shading.
By the way, if you don't have enough drivers, you can do the shading by simply omitting drivers. This is called "spaced shading" and it's used by the seven element Bessel array.
Here's an example:
Let's say that you have a two-row Bessel array with five drivers in each row:
1watt 2watt 2watt -2watt 1watt
1watt 2watt 2watt -2watt 1watt
Does that make sense? Ten drivers?
_____ 2watt 2watt -2watt
2watt 2watt 2watt -2watt 2watt
Now you just made it eight!
See how that works? Same amount of SPL, but with two less drivers, because two are doing the work of one.
This same 'trick' can be used in any shaded array.
I had saved a 5x5 Bessel array link but it seems it went off line about a year ago.
The link was: Orange
Sadly nothing to be found there anymore, it had wiring schemes and both measurement and predicted results. Maybe some history can still be found somewhere, I tried using https://archive.org/web/ but sadly no luck there.
The link was: Orange
Sadly nothing to be found there anymore, it had wiring schemes and both measurement and predicted results. Maybe some history can still be found somewhere, I tried using https://archive.org/web/ but sadly no luck there.
Last edited:
I made you a picture that describes my proposal from post #11:
http://www.diyaudio.com/forums/multi-way/301259-improved-array-14.html#post5065262
http://www.diyaudio.com/forums/multi-way/301259-improved-array-14.html#post5065262
It may be advantageous for small gigs or annoying people at the park?
Admittedly the radial pattern may be superior to the hemispheric pattern for those.
Another thought that has crossed my mind is using a shaded array for the tweeters and an unshaded array for a wooflets.
One of the unusual characteristics of the theremin is that the frequency range is limited mostly by the filters in the device itself - it can produce signals well outside of the human hearing range if you want it to. So it's not unreasonable to desire a monitor that is good for all 10 octaves of human hearing.
These Foster 00504 drivers are said to be good from 100hz or so to maybe 5khz. The Apex Jr cheapie tweeters can be crossed over as low as 3khz. There are perhaps some humps in the response of the 00504 that may need ironing out, but in a purpose built, powered instrument monitor, I'm not above using some eq before amplification.
Zobsky used 24 of these wooflets and 100 of the tweeters in a line array build about 10 years ago, described here, shame the images are gone:
http://www.diyaudio.com/forums/multi-way/99960-my-line-array-project-log.html
He said that a dozen of these in a row gets the bass usable down to 40hz, though they struggle at high volume at that frequency.
I am not great at math and haven't dug into the specs of the drivers, but i wonder if he had to use 100 of the tweeters because of the destructive interaction at high frequencies in straight arrays? *shrug*
- Status
- Not open for further replies.