I'm honored that another BIG of the line array has joined the conversation
.I was wondering if the idea of expanding vertical directivity beyond +/- 150mm from the sitting position, in my current situation and room, is a good idea.
My "listening room" is now a playground FULL of toys and acoustic obstacles everywhere (first of all a gate that prevents children from reaching the speakers and the TV!).
Perhaps a "laser-beam" speakers could save me from annoying reflections.
In the meantime I read something interesting about and "improved bessel array" (US Patente 20060159289A1) that can solve some off-axis issues. I know that you guys worked a lot on large line arrays so maybe you have never considered investigating this kind of solution because it does not make sense for floor-to-ceiling array.
Any thoughts about this? I can try a Vituixcad simulation to see how it performs.
Thanks,
Manuele
The Bessel array will perform as claimed, directivity wise. Its compromises have to do with dynamic range kind of things. Instead of all drivers working together, some are opposed and some aren't driven at full strength. which reduces maximum SPL. This is a stronger effect in a 5 element Bessel than in a 9 element. Its not that hard to make it Bessel only at mid to HF so as not to lose displacement at LF, e.g. just cross to a sub at the base of the array.
No harm in seeing what it does within Vituixcad. Indeed it doesn't have the same advantages as the regular array.
I know Patrick Bateman has simulated a few Bessel arrays a while back, do you have enough information about the
frequency dependent voltage divider and low-pass filter?
Do I see a division here in two parts? Left part (all +1) being bottom end and right part (Bessel) for top end? Is that
what it's supposed to tell us? That would explain the voltage divider and all-pass filter. Meaning 12.4 and 12.6 do
noting on the top end, 12.5 & 12.8 is inverted from a certain frequency and 12.1 and 12.9 are at half output on top end.
Sim away
I know Patrick Bateman has simulated a few Bessel arrays a while back, do you have enough information about the
frequency dependent voltage divider and low-pass filter?
Do I see a division here in two parts? Left part (all +1) being bottom end and right part (Bessel) for top end? Is that
what it's supposed to tell us? That would explain the voltage divider and all-pass filter. Meaning 12.4 and 12.6 do
noting on the top end, 12.5 & 12.8 is inverted from a certain frequency and 12.1 and 12.9 are at half output on top end.
Sim away
What do you specifically mean with this sentence? With the improved array the lost in sensitivity is much less than a standard bessel array...The off-axis seems really good (at least from the Patent, haven't seen any other measurements of this kind of array).
I'm reading his thread but I'm not sure he simulated the "improved array" topology.
Yes, this is my understanding as well...
Can't wait to see your results! I can't sim now (at work) so you are helping me a lot!
I'm reading his thread but I'm not sure he simulated the "improved array" topology.
Yes, this is my understanding as well...
Can't wait to see your results! I can't sim now (at work) so you are helping me a lot!
Sim away
Took me a while staring at the picture what it should mean
. I didn't read the patent. That's when it hit me that theycombined a Bessel with an ordinary array. I didn't check yet how the wiring works to be honest. It does seem to have
a lot of parallel connections or they just left out the specifics in the diagram.
Curious to see the resulting vertical coverage. As the Bessel used to aim to get the same coverage of a single driver.
That wouldn't have my interest, as that's not what I want from an array.
I don't exactly plan to sim this one... I plan to close up my array this weekend
.
I tried to simulate the "improved bessel array" but, if the crossover schematic is not wrong, I don't see any advantages over previous simulations.
Very low impedance in this configuration (2.5 Ohm minimum). Crossover complication. Directivity worsened.
I think I made some big mistake =)
Manuele
Very low impedance in this configuration (2.5 Ohm minimum). Crossover complication. Directivity worsened.
I think I made some big mistake =)
Manuele
Last edited by a moderator:
No, a Bessel gives about the same coverage as a single driver, see post #86 which is a pure Bessel. The improved part is meant to connect all drivers
with +1 value on bass.
I don't think the (original) schematic is quite right, as that group of 3 will have a different impedance, which would not correspond to the +1 noted.
It's incomplete so to say.
with +1 value on bass.
I don't think the (original) schematic is quite right, as that group of 3 will have a different impedance, which would not correspond to the +1 noted.
It's incomplete so to say.
Looking at the phase plot in your Vituixcad sample @Ethloas, it looks like the inversion you used is doing the exact opposite of what was planned.
I see two drivers that are out of phase at bass frequencies, Yet it is there that they all should be in phase. If you invert those two drivers, what does
that look like? I can see the drivers that are working at -10 dB and those that don't add (are crossed with LR24) above 2 KHz. But the inversion seems
to go wrong.I can't figure out why using 3 drivers in a group isn't visible as a different powered group in the filter graph, becaus these drivers can't
all be receiving equal power as the other groups of 3 drivers, right? (they are all in parallel, that's why)
I'd definitely try to invert D5 and D8 to see what happens.
The inversion is happening too slow to have any benefit. The point where we let the drivers change from 9x parallel to Bessel is showing a break in the pattern.
In short: I don't see much benefit for this at all, most certainly not the 9 drivers in parallel connection creating a very low impedance load for the amp.
I tried several crossover points for all filters but see no clean way out of this...
I see two drivers that are out of phase at bass frequencies, Yet it is there that they all should be in phase. If you invert those two drivers, what does
that look like? I can see the drivers that are working at -10 dB and those that don't add (are crossed with LR24) above 2 KHz. But the inversion seems
to go wrong.
all be receiving equal power as the other groups of 3 drivers, right? (they are all in parallel, that's why)
I'd definitely try to invert D5 and D8 to see what happens.
The inversion is happening too slow to have any benefit. The point where we let the drivers change from 9x parallel to Bessel is showing a break in the pattern.
In short: I don't see much benefit for this at all, most certainly not the 9 drivers in parallel connection creating a very low impedance load for the amp.
I tried several crossover points for all filters but see no clean way out of this...
Last edited:
For the moment I'm getting better results by physically rotating the speakers around their vertical axis with respect to any shading.
One other solution to be tested is most extreme: using 9 amplifiers with DSP on each single speaker! Obviously it will only be a "proof of concept" because it seems not feasible from an economic point of view ...
Manuele
One other solution to be tested is most extreme: using 9 amplifiers with DSP on each single speaker! Obviously it will only be a "proof of concept" because it seems not feasible from an economic point of view ...
Manuele
I remember nc535 mentioning something like that . The option of an amp for each driver.
I'm still a fan of using more drivers, e.g. using a longer array. I'm quite happy with what I've managed to create.
As long as one can make sure it can work with the room.
Having the option of trying out amplifiers is still within reach too. As I've been made aware that not all amplifiers sound alike.
Something I never really thought about until it was demonstrated with my speakers in my room.
I'm glad to have a need for one amplifier to feed my main speakers. Not 9 of them.
. The option of an amp for each driver.I'm still a fan of using more drivers, e.g. using a longer array. I'm quite happy with what I've managed to create.
As long as one can make sure it can work with the room.
Having the option of trying out amplifiers is still within reach too. As I've been made aware that not all amplifiers sound alike.
Something I never really thought about until it was demonstrated with my speakers in my room.
I'm glad to have a need for one amplifier to feed my main speakers. Not 9 of them.
with 9 amps and dsp channels you can do 17 drivers: one for the center driver and 8 pairs of drivers placed symmetrically above and below the center. But then you would need another 9 for the other side. You really need 5 amps per side for 9 drivers per side. If your 3d print capability were to scale up, you could use those 5 amps per side for 5 sets of 5 drivers, as in Wesayso's passive array.
I prefer the active approach despite the cost and made the investment in amps, DACs, miniDSPs, and Jriver long ago. I was happy to pay for the instant gratification of a filter response trim requiring only a few minutes at a keyboard. Once you get those filters trimmed just right for your room, you can think about converting to passive filters. Wesayso developed his filters in simulation but he quality array response measurements at the LP to base those simulations on and he had already treated the room in order to be able to get quality measurements at the LP. If you fall short in either of those areas, you may be in for multiple iterations through filter response.
Another reason he was able to develop in simulation is that he was using floor to ceiling arrays which have exceptional vertical directivity. Your short arrays will show floor and ceiling reflections that will obscure your measurements to some degree. I think for these reasons, its a good idea to prototype with active IIR filters and only later convert to passive.
I prefer the active approach despite the cost and made the investment in amps, DACs, miniDSPs, and Jriver long ago. I was happy to pay for the instant gratification of a filter response trim requiring only a few minutes at a keyboard. Once you get those filters trimmed just right for your room, you can think about converting to passive filters. Wesayso developed his filters in simulation but he quality array response measurements at the LP to base those simulations on and he had already treated the room in order to be able to get quality measurements at the LP. If you fall short in either of those areas, you may be in for multiple iterations through filter response.
Another reason he was able to develop in simulation is that he was using floor to ceiling arrays which have exceptional vertical directivity. Your short arrays will show floor and ceiling reflections that will obscure your measurements to some degree. I think for these reasons, its a good idea to prototype with active IIR filters and only later convert to passive.
@wesayno @nc535 Thanks for your feedback!
As I have already mentioned, my original idea was to print all 25 boxes per side to emulate the success of the array of @wesayno, @fluid, and many others but to leave the speakers without any kind of passive filter, amplifying all with my integrated Marantz NR1200. I would have connected one PC to stream Tidal via EqAPO (to correct the frequency response using measurements at different points in the listening area).
I then discovered that an array of this type needs more "vitamin" and an improved digital filter, so I would have switched to using a nice Purifi or Hyper power amp and DRC-FIR0. All this in the long run
After all the simulations and considerations made in the last few weeks on a mini-array of only 9 elements, I'm realizing that it can do some things in the right way, but certainly the vertical response is not among its strengths.
I am thinking of going back to the original 25-boxes-idea even if this means that, for now, I would use the array in a "limited" version of only 9 elements, and then complete the project as soon as all the boxes are ready. In this case, focusing too much on a possible filter for the mini-array is not very useful.
The mini-array can be installed on a stand to make it a sort of "bookshelf speaker", positioning it in such a way as to have decent vertical coverage starting from a sitting position up to a standing position around the listening distance.
At the end of the day, limited vertical response also guarantees less reflection of the roof and ceiling, and as long as the "sweet vertical zone" is covered, I have no interest in hearing music while jumping. I have other MTM Infinity floor standing speakers that can serve this purpose...
As I have already mentioned, my original idea was to print all 25 boxes per side to emulate the success of the array of @wesayno, @fluid, and many others but to leave the speakers without any kind of passive filter, amplifying all with my integrated Marantz NR1200. I would have connected one PC to stream Tidal via EqAPO (to correct the frequency response using measurements at different points in the listening area).
I then discovered that an array of this type needs more "vitamin" and an improved digital filter, so I would have switched to using a nice Purifi or Hyper power amp and DRC-FIR0. All this in the long run
After all the simulations and considerations made in the last few weeks on a mini-array of only 9 elements, I'm realizing that it can do some things in the right way, but certainly the vertical response is not among its strengths.
I am thinking of going back to the original 25-boxes-idea even if this means that, for now, I would use the array in a "limited" version of only 9 elements, and then complete the project as soon as all the boxes are ready. In this case, focusing too much on a possible filter for the mini-array is not very useful.
The mini-array can be installed on a stand to make it a sort of "bookshelf speaker", positioning it in such a way as to have decent vertical coverage starting from a sitting position up to a standing position around the listening distance.
At the end of the day, limited vertical response also guarantees less reflection of the roof and ceiling, and as long as the "sweet vertical zone" is covered, I have no interest in hearing music while jumping. I have other MTM Infinity floor standing speakers that can serve this purpose...
I played a little bit with some CBT-Straight-style simulations, trying to adopt the same approach of the JBL CBT 50/100 to see if the vertical coverage can improve.
According to this article published by Keele, to have a vertical coverage of at least 25 degrees, I need a "virtual" CBT of at least 39 degrees (64%), with a radius of curvature of about 1m. In order to increase the vertical coverage (-6dB SPL), a smaller radius shall be used.
Following the guidelines of the paper I calculated the delays for each driver pair (using the same numbering as in post # 87):
Actually, I have simulated the CBT-Straight style using the parameters of the crossover (ladder delay network) used in the JBL CBT 50, but this corresponds (according to the specifications published by JBL) in a vertical coverage of about +/- 10 degrees between 1.5 and 16 kHz. No attenuation of outer elements.
The vertical directivity seems a little bit cleaner, even if is still very small and with various dips from 2kHz to 20kHz...
Have any of you ever simulated a CBT with VituixCad? Did you find the same results published by JBL / Keele?
I'll post other "variants" for more vertical coverage asap.
Manuele
According to this article published by Keele, to have a vertical coverage of at least 25 degrees, I need a "virtual" CBT of at least 39 degrees (64%), with a radius of curvature of about 1m. In order to increase the vertical coverage (-6dB SPL), a smaller radius shall be used.
Following the guidelines of the paper I calculated the delays for each driver pair (using the same numbering as in post # 87):
- Drivers D1 & D9: delay of 149uS (~ 51mm)
- Drivers D2 & D8: delay of 83uS (~ 28mm)
- Drivers D3 & D7: delay of 37uS (~ 12mm)
- Drivers D4 & D6: delay of 9uS (~ 3mm)
- Driver D5 (central): no delay.
Actually, I have simulated the CBT-Straight style using the parameters of the crossover (ladder delay network) used in the JBL CBT 50, but this corresponds (according to the specifications published by JBL) in a vertical coverage of about +/- 10 degrees between 1.5 and 16 kHz. No attenuation of outer elements.
The vertical directivity seems a little bit cleaner, even if is still very small and with various dips from 2kHz to 20kHz...
Have any of you ever simulated a CBT with VituixCad? Did you find the same results published by JBL / Keele?
I'll post other "variants" for more vertical coverage asap.
Manuele
Do you mean in terms of a vertical response or in a more general sense?
In theory with short arrays (like mine) using CBT shading should provide more advantages than the effect CBT can have on a real floor to ceiling array, but I wanted to get your feedback as I have no direct experience with CBT...
What intrigues me is the invariance of the frequency response with the distance, so that I can possibly have a coverage of the "whole room" and not just the listening distance (sofa in my case).
The results of the last simulation look better than the others, don't they?
In theory with short arrays (like mine) using CBT shading should provide more advantages than the effect CBT can have on a real floor to ceiling array, but I wanted to get your feedback as I have no direct experience with CBT...
What intrigues me is the invariance of the frequency response with the distance, so that I can possibly have a coverage of the "whole room" and not just the listening distance (sofa in my case).
The results of the last simulation look better than the others, don't they?
I did some simulations here: https://www.diyaudio.com/community/threads/passive-loudspeaker-delay.280664/post-6507079 - 8 driver array with flat and curved baffle, with delays and without.
I did not try short CBT array simulations. I wanted to see if their top end was better than the shaded array I had created.
So I simmed a big CBT with shading. I also simmed a focused array, all out of curiosity. I think it's still there somewhere
in my thread, I didn't even bother to save the models. It's top end wasn't any better than my proposed 25 driver straight
array with frequency shading. So that's all I needed to find out and where my interest in the CBT ended. It doesn't work
that well with 3.5" drivers for tall arrays. A true multi way CBT (small tweeter top end) would be a different animal.
Your result here looks fine, though do try all distances of interest to view the frequency curve and compare to the design
distance. Also check at the height(s) you need. I wanted good/acceptable output when standing and excellent when seated
for more than me alone (different length family members). Also good coverage along my entire couch.
I've got what I wanted, the tonal balance is steady over a large area. large enough to have a pleasing outcome when
moving about and have excellent output in or along the sweet "area" (being the couch).
I've checked this extensively and compared to the unshaded array, which did quite well by itself over a large area once
you mask the parallel planes (damping panels at first reflections). Lived with that setup for years and the frequency
shaded array should be an improvement over the unshaded array on all counts. I can confirm that even without proper
dial in (this has yet to be done). I can tell the difference in sound which is difficult to put into words. I'll wait till I've
done all of my processing to really try and name the differences. I have renovated my room and have a different ceiling,
things like that matter too. So the sound was bound to be different even if I didn't add all those filters. Less "live room"
sound is all I want to say right now.
Sorry for the lengthy reply, tying while thinking has that effect for me... my thread has become that long for a reason (lol).
So I simmed a big CBT with shading. I also simmed a focused array, all out of curiosity. I think it's still there somewhere
in my thread, I didn't even bother to save the models. It's top end wasn't any better than my proposed 25 driver straight
array with frequency shading. So that's all I needed to find out and where my interest in the CBT ended. It doesn't work
that well with 3.5" drivers for tall arrays. A true multi way CBT (small tweeter top end) would be a different animal.
Your result here looks fine, though do try all distances of interest to view the frequency curve and compare to the design
distance. Also check at the height(s) you need. I wanted good/acceptable output when standing and excellent when seated
for more than me alone (different length family members
). Also good coverage along my entire couch.I've got what I wanted, the tonal balance is steady over a large area. large enough to have a pleasing outcome when
moving about and have excellent output in or along the sweet "area" (being the couch).
I've checked this extensively and compared to the unshaded array, which did quite well by itself over a large area once
you mask the parallel planes (damping panels at first reflections). Lived with that setup for years and the frequency
shaded array should be an improvement over the unshaded array on all counts. I can confirm that even without proper
dial in (this has yet to be done). I can tell the difference in sound which is difficult to put into words. I'll wait till I've
done all of my processing to really try and name the differences. I have renovated my room and have a different ceiling,
things like that matter too. So the sound was bound to be different even if I didn't add all those filters. Less "live room"
sound is all I want to say right now.
Sorry for the lengthy reply, tying while thinking has that effect for me... my thread has become that long for a reason (lol).