I think You can time-align Your drivers by using different order filters, but this demands Your drivers are pretty flat both for FR and impedance, and without any steep/abrupt phase changes for at least an octave above/below xo freq.
For passive filters, at xo, a 1st order filter will be 45deg delayed, a 2nd order 90deg, 3rd order 135deg and so on. In general, order x 45deg = phase at xo.
By proper choice of filter order, You should be able to compensate for the physical lenghtwise (z) distance between the drivers' acoustic center.
E.g. xo=500Hz, MF->LF=0.172m, then 360deg*0.172m/(344/500Hz)=90deg => You need 90deg extra phase for the driver to be delayed, which translates to use a filter with order (LF-filter-order +2) for the MF. Use filters with Q=0.5 if possible, they sum up flat at xo.
For vertical directivity of line arrays You should check out Horbach-Keele -filters.
For passive filters, at xo, a 1st order filter will be 45deg delayed, a 2nd order 90deg, 3rd order 135deg and so on. In general, order x 45deg = phase at xo.
By proper choice of filter order, You should be able to compensate for the physical lenghtwise (z) distance between the drivers' acoustic center.
E.g. xo=500Hz, MF->LF=0.172m, then 360deg*0.172m/(344/500Hz)=90deg => You need 90deg extra phase for the driver to be delayed, which translates to use a filter with order (LF-filter-order +2) for the MF. Use filters with Q=0.5 if possible, they sum up flat at xo.
For vertical directivity of line arrays You should check out Horbach-Keele -filters.
Sure, so there is no real issue here?
In deciding this, if the factor that is constraining the range of possible choices happens to be lobing and the effect of separation if you cross too high, then the plane of the lobing could be seen as tilted to the side (presuming this is a top down view). The assumption that there would be no delay under these conditions might see the listening position under a null.
Measurement will clear this up.
I agree with this the LF acoustic centers are too far back for time alignment with HF; passive XO would also introduce additional phase offset and non-linearity (especially if higher order) to make coherence nearly impossible.
@profiguy you might pick off some ideas from the myriad coherent minimalist experiments I've posted to the Fullrange Photo Gallery. If your baffle configuration is more like LF/HF\LF i.e. "symmetric-MTM LX flipped on its side" (maybe this is what you intended?) then the acoustic centers can be physically lined up to time-align on-axis (think LX); the XO phase-alignment would require a bit of design/tweak. You might also consider a MEH-like tapped LF arrangement without the horn (see Minmeh vs T/MM vs TM coherent minimalist).
My M.O. for precise time-alignment has been described in various threads e.g. recent Dunlavy.
To preface, please understand I'm not trying to be over argumentative here. Initially, I was actually mainly worried about the shading of LF drivers causing issues. I wasn't even that concerned about the issues cited here.
The crude drawing I posted does have some significant misrepresentation of the actual design. This presents an error in the depth axis of exact placement of individual LF/MF driver VC locations. I'm essentially rotating the LF driver VCs with their center points remaining at the same distance (average) in respect to the MF driver, so in theory at lower frequncies, the LF and MF frequencies, act as if they are at the same depth. The LF driver tilt creates an averaging error, along with the adjacent LF across from it. At lower frequencies, the two LF on their tilted axis will produce a homogenous wavefront that is in exact phase with the MF. At lower frequencies, it would behave the same as both LF drivers mounted on a flat baffle with the baffle surface being on the same plane as the center pivot point of the tilted LF driver arrangement. The only physical error created is minor, produced by the baffle trough from the V shape, but its very negligible.
Again this is all at much lower frequencies where this tilt average discrepancy isnt dominating the situation. If you drew a level line across the LF and MF driver VC axis intersecting MF driver VC plane, it would go through the average pivot point of both LF. Having the 2 LF drivers pulls the average LF phase forward to be effectively on axis with the MF, as if all 3 drivers were mounted on a flat baffle.
Considering this has worked for me before at lower crossover points, where the averaging of VC depth through tilting the LF drivers hasn't affected the FR or directivity, I'm still puzzled why its being deemed as a significantly flawed design premise.
Back in the 1980s, I've had some people from JBL and Fostex tell me it (shouldn't) wouldn't work, but there was one engineer at Fostex Japan who was overseeing the work at ACR who actually backed my findings. He was the only guy who understood all of the details why this idea I had would actually fly. When I finished the driver mock up, it measured almost identical to the flat baffle model at lower cutoff points, just with alot less lobing. There wasn't any significant or goal prohibiting phase error between LF and MF predicted by others, but it affected FR at higher frequencies, which I was staying clear from. Similar I behavior can be observed with curved/folded horn designs, but the averaging error is greatly reduced with strategic WG folding. I guess this has to be mocked up again and measured to be sure.
The crude drawing I posted does have some significant misrepresentation of the actual design. This presents an error in the depth axis of exact placement of individual LF/MF driver VC locations. I'm essentially rotating the LF driver VCs with their center points remaining at the same distance (average) in respect to the MF driver, so in theory at lower frequncies, the LF and MF frequencies, act as if they are at the same depth. The LF driver tilt creates an averaging error, along with the adjacent LF across from it. At lower frequencies, the two LF on their tilted axis will produce a homogenous wavefront that is in exact phase with the MF. At lower frequencies, it would behave the same as both LF drivers mounted on a flat baffle with the baffle surface being on the same plane as the center pivot point of the tilted LF driver arrangement. The only physical error created is minor, produced by the baffle trough from the V shape, but its very negligible.
Again this is all at much lower frequencies where this tilt average discrepancy isnt dominating the situation. If you drew a level line across the LF and MF driver VC axis intersecting MF driver VC plane, it would go through the average pivot point of both LF. Having the 2 LF drivers pulls the average LF phase forward to be effectively on axis with the MF, as if all 3 drivers were mounted on a flat baffle.
Considering this has worked for me before at lower crossover points, where the averaging of VC depth through tilting the LF drivers hasn't affected the FR or directivity, I'm still puzzled why its being deemed as a significantly flawed design premise.
Back in the 1980s, I've had some people from JBL and Fostex tell me it (shouldn't) wouldn't work, but there was one engineer at Fostex Japan who was overseeing the work at ACR who actually backed my findings. He was the only guy who understood all of the details why this idea I had would actually fly. When I finished the driver mock up, it measured almost identical to the flat baffle model at lower cutoff points, just with alot less lobing. There wasn't any significant or goal prohibiting phase error between LF and MF predicted by others, but it affected FR at higher frequencies, which I was staying clear from. Similar I behavior can be observed with curved/folded horn designs, but the averaging error is greatly reduced with strategic WG folding. I guess this has to be mocked up again and measured to be sure.
My other main concern was mid to top end diffraction from the shallow bridge. Some foam baffling could fix this. It just may be too much of a compromise.
I think I understand Your idea, but I'm sure it will not be an easy task to make a passive crossover for this arrangement if You need nice phase, group delay and no lobing. You have CtC-distances vertical, horisontal and in depth, this will be a major challenge with passive xo. Maybe Vituixcad can shed some light on the problem?
What puzzles me is when looking at speakers like from Tekton, these have some novel circular tweeter arrays which could in theory work. The problem with most of these designs is they're on a flat baffle. This causes several issues and limitations. Using all the tweeters to their best advantage requires a DSP based approach. The way it sits is a highly wasteful application of several expensive drivers, which have little advantage over a higher end midrange being able to cover a greater range at lower distortion levels.
The last time I experimented with radial arrays, it showed that anything larger than 4" drivers weren't practical due to limited FR. The last radial array I built was with 6 Peerless 830870 and a Seas T35C002. It sounded very good and was capable of a high dynamic range crossed above 150 hz. The problem it had was it required a long passive delay to the tweeter. I was going to build a concave baffle for it, insetting the tweeter by a few inches to mechanically solve the issue. I may revisit this due to the similar potential issue I'm dealing with on this new design.
The last time I experimented with radial arrays, it showed that anything larger than 4" drivers weren't practical due to limited FR. The last radial array I built was with 6 Peerless 830870 and a Seas T35C002. It sounded very good and was capable of a high dynamic range crossed above 150 hz. The problem it had was it required a long passive delay to the tweeter. I was going to build a concave baffle for it, insetting the tweeter by a few inches to mechanically solve the issue. I may revisit this due to the similar potential issue I'm dealing with on this new design.
Your task would be easier if all listeners was at a single spot 🙂 If I remember correctly, a commercial brand has some kind of directivity control, but making this frequency dependent in two or more axis is by far beyond my level of knowledge. Maybe a simpler solution would be placing the drivers on concentric circles wrt CtC, tweeter in the middle?
I don't know that (yet).. you give too much credit 😉
Engineering is full of compromises that come down to the wire. It can be successful if you know the limitations so you don't give too much in one area, causing another to suffer. Good speakers are often bound this way. Compromise isn't a sign of failure.
That is not a problem, in a view of sufficiently low crossover frequency you have chosen.
Well, that is why I and others objected. With this explanation, it seems the "problem" is far smaller (or non-existent) than we thought.
Center of the voice coil is not acoustic center/origin of the driver! This is a common misconception.
Last edited:
The problem with Tekton design is use of tweeters with full diameter flange. It will work much better without flanges, grouping tweeters in a much tighter circle.
@Sonce Yes, I'm aware the actual acoustic origin plane can be significantly different from one driver to another. The speed of sound in denser solids makes it harder to approximate.
I've measured cone drivers that have acoustic centers as far forward as their surrounds and as far back as magnet backing plates. Deeper cone profiles and larger domes can also show a significant off axis dip. This is similar to some metal domes which require a phase shield to block part of the dome outer edge.
The speed of sound in a metal alloy is so much higher than paper and not as dampened. At higher frequencies, the emitted acoustic signal is partially out of phase across a portion of the dome. The difficulty can be identifying that actual area and use only the exact required shielding. Some people just blindly remove these shields, believing they only hurt the sound.
I've measured cone drivers that have acoustic centers as far forward as their surrounds and as far back as magnet backing plates. Deeper cone profiles and larger domes can also show a significant off axis dip. This is similar to some metal domes which require a phase shield to block part of the dome outer edge.
The speed of sound in a metal alloy is so much higher than paper and not as dampened. At higher frequencies, the emitted acoustic signal is partially out of phase across a portion of the dome. The difficulty can be identifying that actual area and use only the exact required shielding. Some people just blindly remove these shields, believing they only hurt the sound.
@markbakk I was considering that as long as the motor doesn't do weird things at specific frequencies. When you load the motor by mounting it in a baffle or enclosure, the air resistance will possibly induce some noise and even different distortion profiles. I'd have to try this with one driver in revers to figure out how it responds.
One of the goals when mounting the LF drivers in a V configuration is the reduction of CTC with the other LF drivers. A flat arrangement would be roughly 1.4 x more CTC.
One of the goals when mounting the LF drivers in a V configuration is the reduction of CTC with the other LF drivers. A flat arrangement would be roughly 1.4 x more CTC.
@Sonce The acoustic center on the NE180s is roughly at the lowest point of the dust cap to cone joint. I checked this with other 2 way designs using this driver and it does move a little at higher frequencies, although only by a few mm. The Ti VC former is advantageous to have here being it doesn't have much phase lag at higher frequencies due to SOS. Most Ti VC former drivers have very little delay from actual acoustic center of the VC plane.
Acoustic origin/center can not be deeper than the cone/dustcap joint. Acoustic wave in air is forming from the physical object surface pushing air molecules forward and backward, and that physical object is the front (external) surface of loudspeaker cone assembly (with dust cap) - not the voice coil, or any part of the voice coil former (regardless of the voice coil former material). Whole cone assembly, including voice coil former and the voice coil itself, is moving as one complete unison object (ideally) so voice coil can not be acoustic center.
When low-pass filter is used, then there is a delay (phase lag) of the incoming signal, which may be interpreted as deeper position of the woofer acoustic center.
When low-pass filter is used, then there is a delay (phase lag) of the incoming signal, which may be interpreted as deeper position of the woofer acoustic center.
Last edited:
Intrinsic "low-pass filter" of Vifa NE180W has F3 > 10 kHz, so delay is minimal and firmly in the same magnitude order as 22 kHz cutoff of CD music signal. CD and MP3 signals are filtered by very high order "hardcore" low-pass filters.
Last edited:
Edit:
Intrinsic "low-pass filter" of Vifa NE180W has F3 > 10 kHz, so delay is minimal and firmly in the same order of magnitude as 22 kHz cutoff of CD music signal. CD and MP3 signals are filtered by very steep high-order low-pass filters, unlike lower-order natural frequency response roll-off of loudspeakers.
Intrinsic "low-pass filter" of Vifa NE180W has F3 > 10 kHz, so delay is minimal and firmly in the same order of magnitude as 22 kHz cutoff of CD music signal. CD and MP3 signals are filtered by very steep high-order low-pass filters, unlike lower-order natural frequency response roll-off of loudspeakers.