Shouldn't they be getting progressively closer as frequencies get lower?
IE, the low pass filters introduce a time delay, so the most practical way to compensate for that is to bring the midranges and woofers closer to the listener?
Attachments
Not sure if we are talking about the same thing. If I go from a type 1 (lots of same sized full range units) to a type 2, Then the lower frequency units, the units that are farther towards the end inherently have more recessed voice coils (assuming a flat baffle here) giving more phase shift or group delay. There may then be an argument for compensating for extra group delay of end units.
This is not a requirement though, because most designers have been able to get good blending of non time-aligned tower speakers. The usual trick of increase the network order and then flip the polarity of a driver works just as well here as it does on a one sided tower.
I would suggest people start with low order networks as they seem to work in this application and don't have the broad overlap problems that the non array first order layouts have. (RIP Jim Thiele)
This is not a requirement though, because most designers have been able to get good blending of non time-aligned tower speakers. The usual trick of increase the network order and then flip the polarity of a driver works just as well here as it does on a one sided tower.
I would suggest people start with low order networks as they seem to work in this application and don't have the broad overlap problems that the non array first order layouts have. (RIP Jim Thiele)
True, my assumption is using 2nd or 3rd order filters.
I am no expert on phase, but as I understand it, first order filters have a symmetrical amount of delay. IE, if you have a first order low pass and a first order high pass at the same frequency, the delays cancel each other out.
I generally prefer 2nd, 3rd and 4th order filters because they generally produce superior polar response. But I'll try again with first order filters.
IIRC, your Snell XA used 3rd order filters.
I am no expert on phase, but as I understand it, first order filters have a symmetrical amount of delay. IE, if you have a first order low pass and a first order high pass at the same frequency, the delays cancel each other out.
I generally prefer 2nd, 3rd and 4th order filters because they generally produce superior polar response. But I'll try again with first order filters.
IIRC, your Snell XA used 3rd order filters.
Yes, no problem with higher order filters but they will be more work.
As you increase order the low end of a section's phase bends up and the high end's phase bends down. You must compare each phase curve to the adjacent section's and see if you can either a) get the phase curves to lie on top of each other, or b) get them a near constant 180 degrees apart (at which point you flip the polarity of a unit). Again, the symmetrical array seems to buy some latitude as it is forced to have symmetric polars. Also, there appears to be a tradeoff between "nice" polar curves and classic phase stitching. As long as we aren't cancelling a lot then optimizing polars is preferable.
I don's see Vituix showing phase curves of each section but there is probably a way. You can also surmise phase by how much the elements support each other (add level) as you add each new section.
Or DSP it all.... (Kids today... sheesh)
As you increase order the low end of a section's phase bends up and the high end's phase bends down. You must compare each phase curve to the adjacent section's and see if you can either a) get the phase curves to lie on top of each other, or b) get them a near constant 180 degrees apart (at which point you flip the polarity of a unit). Again, the symmetrical array seems to buy some latitude as it is forced to have symmetric polars. Also, there appears to be a tradeoff between "nice" polar curves and classic phase stitching. As long as we aren't cancelling a lot then optimizing polars is preferable.
I don's see Vituix showing phase curves of each section but there is probably a way. You can also surmise phase by how much the elements support each other (add level) as you add each new section.
Or DSP it all.... (Kids today... sheesh)
Should be enough drivers there.
What always blows my mind is that you can have multiple beams in multiple directions at the same time. I know military antennas are done that way but it still is very cool.
I think Renkus Heinz or some other pro audio company has a steerable array mostly for voice.
I'm impressed by the Starlink beamforming antennas, which can track multiple satellites as they pass overhead. From the teardown photos, it looks like they are using over 500 antenna/amplifier combinations. That's impressive complexity for a product they are selling for around $600.
Yes, and in line-source mode, all four midranges are enabled, and the largest main midrange will have higher efficiency than the three smaller midranges. That's what they called it, the "weighted line-source" system. Is this unusual to the conventional line-array speakers?
Right. AFAIK, the 2030 doesn't use midrange drivers of different size. It just used 4 x 2in domes. It 'simply' gave the option of what it called a 'point source' i.e. with just one of those domes operating and three switched out, or 'weighted' line source where all four of the midrange domes were running, the additional three having a smaller magnet structure -presumably to reduce their sensitivity relative to the primary unit, so they provided a degree of natural output-tapering without actually needing to electrically power-taper.
This is, at the very least, intriguing to me.
According to the brochure, the main midrange driver has a flux density of 1.7 Teslas and a magnetic flux of 1,630,000 nanoWebers. The three auxiliary midranges each have 1.55 Teslas and 1,450,000 nanoWebers.
Can we deduce the degree of tapering from these driver specifications?
No.
As replies go, short & sweet.
But no -you need a lot more information than just two flux density specs. to infer, let alone deduce, anything about any possible output tapering -assuming there was anything particularly significant in the first place. Full driver and crossover data needed, or some fairly comprehensive system measurements -anything else is just guesswork.
As replies go, short & sweet.
But no -you need a lot more information than just two flux density specs. to infer, let alone deduce, anything about any possible output tapering -assuming there was anything particularly significant in the first place. Full driver and crossover data needed, or some fairly comprehensive system measurements -anything else is just guesswork.
Last edited:
As was shown in this thread, if you have enough data on a dome mid you can actually sim it's behaviour and get pretty reliable results.
You'll need the right input data and that may take a while to gather, but programs like Vituixcad can be of great help show what shading can do.
You'll need the right input data and that may take a while to gather, but programs like Vituixcad can be of great help show what shading can do.
Hi David,
I have two questions on the eXpanding array.
1) I simulated such an array using vituixcad and learnt that first order crossovers work the best to achieve constant directivity. Higher order IIR crossovers have too much phase shift to work. Can you please confirm this?
2) Also, if W is the wavelength at crossover frequency and L is the effective length of the array at that frequency then if W/L >> 1 then the array has no directivity. On the other hand if W/L <<1 then lobes appear and polars are not good. So, what is is a good value for W/L?
Thanks and Regards,
WonderfulAudio
I have two questions on the eXpanding array.
1) I simulated such an array using vituixcad and learnt that first order crossovers work the best to achieve constant directivity. Higher order IIR crossovers have too much phase shift to work. Can you please confirm this?
2) Also, if W is the wavelength at crossover frequency and L is the effective length of the array at that frequency then if W/L >> 1 then the array has no directivity. On the other hand if W/L <<1 then lobes appear and polars are not good. So, what is is a good value for W/L?
Thanks and Regards,
WonderfulAudio
I'm sure David will chime in, but in the meantime, the nuts and bolts of the eXpanding array can be fund in one of the AES papers he wrote. I read it about fifteen years back at the library. (I don't have an AES subscription.)
IIRC, the filters were 3rd order.
IMHO, the Horbache Keele paper, available freely online is an evolution of the eXpanding array.
And then someone here on diyaudio took the Horbach Keele design further, by making two dimension instead of one dimensional. I think he called it a "fractal array." I am not aware of one ever being built.
Going two dimensional with the array has quite a few benefits:
1) you increase sensitivity and power handling
2) you can trade sensitivity, power handling, and driver size. IE, instead of using dual 75mm midranges, you can use four 50mm midranges and still get a comparable sensitivity and power handling.
3) you can control the horizontal beamwidth. But this can also become a defect instead of a feature if you're not careful. Most listeners prefer narrow vertical directivity and broad horizontal directivity.
IIRC, the filters were 3rd order.
IMHO, the Horbache Keele paper, available freely online is an evolution of the eXpanding array.
And then someone here on diyaudio took the Horbach Keele design further, by making two dimension instead of one dimensional. I think he called it a "fractal array." I am not aware of one ever being built.
Going two dimensional with the array has quite a few benefits:
1) you increase sensitivity and power handling
2) you can trade sensitivity, power handling, and driver size. IE, instead of using dual 75mm midranges, you can use four 50mm midranges and still get a comparable sensitivity and power handling.
3) you can control the horizontal beamwidth. But this can also become a defect instead of a feature if you're not careful. Most listeners prefer narrow vertical directivity and broad horizontal directivity.