Russell,
Copy that opinion on the sweet sixteen, though, after a 32 cone EnABL treatment, it was at least musical at low levels.
The Nestorovic and Speaker Lab speakers were a bit unusual. Mille used a normal low pass filter for the smaller diameter woofer, at about 1K for an 8 inch, and then used his patented low low pass filter to turn off the larger woofer as it approached 50 Hz or so, gradually becoming a passive radiator. This system produced prodigious amounts of superbly colorful music at the bottom of the spectrum. Also required an incredibly robust smaller driver.
I know, I built them at his Lab for a number of years and got to participate in his design process while pretending to be his R&D engineer. I was that, but this mans fluency in combining test data with corrective listening procedures was just astounding to watch. I felt like a kid in a candy store, every day, for 5 years.
He does have a patent on this crossover so you may be able to dredge it up. He also has a number of other patents from his work at McIntosh.
Bud
Copy that opinion on the sweet sixteen, though, after a 32 cone EnABL treatment, it was at least musical at low levels.
The Nestorovic and Speaker Lab speakers were a bit unusual. Mille used a normal low pass filter for the smaller diameter woofer, at about 1K for an 8 inch, and then used his patented low low pass filter to turn off the larger woofer as it approached 50 Hz or so, gradually becoming a passive radiator. This system produced prodigious amounts of superbly colorful music at the bottom of the spectrum. Also required an incredibly robust smaller driver.
I know, I built them at his Lab for a number of years and got to participate in his design process while pretending to be his R&D engineer. I was that, but this mans fluency in combining test data with corrective listening procedures was just astounding to watch. I felt like a kid in a candy store, every day, for 5 years.
He does have a patent on this crossover so you may be able to dredge it up. He also has a number of other patents from his work at McIntosh.
Bud
Lynn Olson said:
This seems like a great approach. Get the middle right, that's where the music is. Build up and out from there. The midrange is the heart of the music - you only fill in the rest of the spectrum to compliment the mids. Getting that good midrange is the challenge.
Very much the approach taken by Lukasz Fikus on his Lampizator Site.
Without wishing to criticise Lukasz' designs, there are two aspects puzzling me.
Yes the main (fullrange) driver is the most important, but Lukasz wires his out of phase with the supporting (bass-treble) drivers, which means that the predominant range responses of those supporting drivers are going to be out phase.
Also Lukasz tends to mount his tweeter below the fullrange, as opposed to what I think of as giving it 'air' above.
Yes the main (fullrange) driver is the most important, but Lukasz wires his out of phase with the supporting (bass-treble) drivers, which means that the predominant range responses of those supporting drivers are going to be out phase.
Also Lukasz tends to mount his tweeter below the fullrange, as opposed to what I think of as giving it 'air' above.
Are the woofers attached to the front baffle? It is an interesting way to do a dipole speaker. More interesting than the standard big baffle, attached to a stand on its bottom.
I can see that the black pieces can be replaced with acrylic, and the woofers are attached to a separate, thin metal frame. It will be similar to an I-beam used in construction, but with two pieces of plastic connected by a piece of metal.
I did that also, and it seems fine. I suppose there is a shorter path to the edge at the top that way but is that really an advantage? The side distance is the same... There's a lot of disagreement as to whether a small baffle realy gives "more air "
Or is that just the way you are used to seeing things? I did it to keep the tweeter at proper height in my speakers and because tweeter and mid are mounted in a single open baffle on top of a boxed bass speaker. By putting the tweeter below the mid, I was able to raise the dipole mid away from the horizontal surface of the box. Since the tweeter only fires to the front, it isn't affected by the proximity of the box below. I could have mounted the tweeter in the bass box but I wanted to be able to isolate the baffle from the box with rubber feet. I could have made the box shorter, but it was getting pretty deep and wide already. Sometimes it's fun to try other ways to do stuff.
I know Lukasz used to follow established ideas more rigorously, but seems to have "broken free" from over thinking this stuff. As choky says, he's on an other plane now!
Now with Lynn pointing out that the drivers are to be mounted off center in the baffle, I suppose I'd be more tempted to mount it above as it would be close to both the side and top. I am also influenced by how people normally do things...
Maybe Lynn has a more specific reason why having the tweeter in the middle might not be optimal. Might depend on you crossover freq. Mine crosses from bass to mid at about 200hz so I imagine I have some room to play around.
panomaniac said:
Zen Mod said:
his spks are just from another realm
you can't realy criticize anything.......mebbe even xover schmtc is incorrect
but - with that approach,usual hair splitting is less important than in "usual realm"
What a great site!!! I approve! Lukasz is a serious engineer - his CD mods are pretty much exactly what I'd do. Not mentioned are how the miserable op-amps found in just about every CD player - and I include the Sony SACD-1 - are too slow by a factor 20 to 100.
I am not joking. Looking at what comes out of the DAC chip, my friend Matt Kamna and I measured a flat comb spectra extending out to 20 MHz, and only descending into the noise at 50 MHz. That works to a slew rate over 1000V/uSec, far above the 13V/uSec of the ubiquitous 5532/5534. If the op-amp slews, it is 100% distortion during the slew interval. Even though this is lowpass-filtered (rather badly by the op-amp), the slewing distorts the amount of energy that should be present in the sample, creating a lot of HF distortion that is program-dependent.
Feeding the DAC output through a transformer that is electrically screened between primary and secondary, then straight into a tube grid, sidesteps the slewing, and also filters the RFI quite effectively. The output from the DAC is actually pretty good so long as it isn't slewed by an op-amp.
----------------
As for driver polarity, there are ways to keep all drivers in the same polarity, so long as you're willing to accept a 360-degree phase rotation at the crossover frequency. That's what the Ariel does, with an electrical 2nd-order network and an acoustic 4th-order crossover.
So if your constraints are mounting all the drivers on the same baffle (which puts the tweeter acoustically ahead of the woofer), and you want all the drivers in the same electrical polarity, increase the slope of the HP and LP network. The ratio of the HP/LP slopes sets the phase angle between the drivers, which in turn controls the vertical polar pattern.
I prefer an in-phase relation, with less than 10 degrees spread between the drivers. I find this inter-driver phase angle much more important than the broad-brush horizontal polar pattern - again, because the inter-driver phase controls the vertical polar pattern. Most important of all, small ripples on the skirts of the rolloffs shift the inter-driver phase, making the vertical polar pattern shift abruptly with frequency, which is very undesirable.
To continue the discussion of the previous page, after you select the widerange driver to meet your tastes (hemp or paper cone, Alnico or Neo magnets), you start the crossover design by coming up with a smooth (and I mean textbook-smooth) rolloff curve for the widerange driver. The driver will tell you where it wants to cross over - follow its guidance, not an arbitrary formula of driver size vs recommended crossover frequency.
You want to listen to the lowpass-filtered widerange driver with pink-noise (most important) and music before you add the tweeter. You want silky-smooth sound, rolled-off, but silky-smooth nonetheless, and most of all, musically valid and pleasant to listen to.
After you get a good-sounding widerange driver with a smooth rolloff, then you add the tweeter and its network - as before, designing a HP network with smooth curves, and trimming the crossover frequency to get a desirable in-phase relation between the drivers. Since inter-driver phase is difficult to measure at high frequencies, I temporarily reverse-phase the HF driver and measure the depth of the null. The null in the Ariel is 25 to 30 dB deep, indicating 3-degree phase match.
It takes a fair bit of adjusting and trimming to get really smooth transitions between the drivers - since the widerange driver and the tweeter are working at the edge of their ranges (that's why we need a crossover, remember?), we have to move things up and down in order to find a smooth region. Crossover smoothness translates to a vertical polar pattern that doesn't go through sharp transitions with small changes in frequency.
This can be confirmed by turning the system on its side, and auditioning with pink noise from a distance of one or two feet. It should like one driver, not two - if it sounds like two, you have phase-angle problems, and you need to re-visit the crossover.
I am another one who heard the Emeralds in Denver; I also was highly impressed by the overall presentation. Many people who were bowled over by the sound, while attributing it to Emerald Physics' implementation, I believe were in fact hearing the gestalt attributable to open baffle dipole sound itself. I asked three people, and in fact it was there first exposure to it. To be honest, I was impressed by just how similar it was in the midband and bass presentation to my own DarkStar. More similar than different.
I have ALOT of faith in Ciare's current offerings, and they have been putting out new designs like sparks off the ol' grindstone, and the Italians in general. For anyone interested in following Emerald's lead with Nd 8, might want to know that Ciare also offers an 8 inch and a 6.5 Nd wideband driver buried in their Home Line. The 8 looks as though it employs the same basic basket and motor, and could not but be a great driver. Its plot indicates to me that it could use some active contouring, which to me as well is fine, as I consider DSP pretty much mandatory, to get the most out of multi-driver open baffle rigs. Flattened out, it shows an honest ~96db all the way out. I like the 6.5 model even more, for a scaled-down condo rig, and I am about to order a pair from our friend "Validator," Thorsten Weber, at www.boxen-baustelle.de one of the good guys.
I also like that new Eminence Nd DeltaLite. The Kappa looks even better for $14 more. Now you can dive into a simple 2 way rig with state of the art drivers with Nd magnets. I also have great faith in Eminence, and their published FR plots. They have zero to hide. I would appreciate any more discussion on the idea of using the DeltaLite as a wideband driver; a low-rider rig, with just this and a compression driver on top. Using boundary gain with this one, you could add cheap bass on your way out of Costco. There are some outstanding choices right now, it would seem.
I have ALOT of faith in Ciare's current offerings, and they have been putting out new designs like sparks off the ol' grindstone, and the Italians in general. For anyone interested in following Emerald's lead with Nd 8, might want to know that Ciare also offers an 8 inch and a 6.5 Nd wideband driver buried in their Home Line. The 8 looks as though it employs the same basic basket and motor, and could not but be a great driver. Its plot indicates to me that it could use some active contouring, which to me as well is fine, as I consider DSP pretty much mandatory, to get the most out of multi-driver open baffle rigs. Flattened out, it shows an honest ~96db all the way out. I like the 6.5 model even more, for a scaled-down condo rig, and I am about to order a pair from our friend "Validator," Thorsten Weber, at www.boxen-baustelle.de one of the good guys.
I also like that new Eminence Nd DeltaLite. The Kappa looks even better for $14 more. Now you can dive into a simple 2 way rig with state of the art drivers with Nd magnets. I also have great faith in Eminence, and their published FR plots. They have zero to hide. I would appreciate any more discussion on the idea of using the DeltaLite as a wideband driver; a low-rider rig, with just this and a compression driver on top. Using boundary gain with this one, you could add cheap bass on your way out of Costco. There are some outstanding choices right now, it would seem.
15"Deltalite ll:
http://www.eminence.com/pdf/deltaliteII2515.pdf
15" Kappalite
http://www.eminence.com/pdf/deltaliteII2515.pdf
15" Kappalite LF
http://www.eminence.com/pdf/kappalite-3015lf.pdf
If you look at the curves, you can see that the Deltalite is exceptional in the upper frequencies. not only are they smooth there, when they do drop off, there are no sawteeth and it's STEEP. You could probably run this driver with no crossover filters at all. Possibly a filter to reduce the rising rate. I did that with another 15" driver that had this same kind of response , and was very pleased with the final result.
The Kappas are fine-maybe better if you are crossing them at a low frequency, but the upper part of their curves are pretty typical Eminence..
http://www.eminence.com/pdf/deltaliteII2515.pdf
15" Kappalite
http://www.eminence.com/pdf/deltaliteII2515.pdf
15" Kappalite LF
http://www.eminence.com/pdf/kappalite-3015lf.pdf
If you look at the curves, you can see that the Deltalite is exceptional in the upper frequencies. not only are they smooth there, when they do drop off, there are no sawteeth and it's STEEP. You could probably run this driver with no crossover filters at all. Possibly a filter to reduce the rising rate. I did that with another 15" driver that had this same kind of response , and was very pleased with the final result.
The Kappas are fine-maybe better if you are crossing them at a low frequency, but the upper part of their curves are pretty typical Eminence..
missing colors / dissimilar woofers
Hi
Well, I'd rather think so!
With a fairly smooth FR loudspeakers do not behave like TV or computer monitors where there are definitely not all colours available depending on make (DLP, LCD, Plasma...) and in addition to that there are also might be some missing spectral lines form the illumination source.
That there are sounds that can't be reproduced (accurately) is obvious for anybody. Thinking of it that there are "colors" that are missing is a nice concept but I don't think this holds true. At least unless we don't suffer from finite resolution of upcoming digital speakers. Anything could equally be explained with coloration and masking effects IF we had a better understanding about that.
While continuing my research about diminishing resonance effects in loudspeaker cabinets WITHOUT the use of damping material, I found it most interesting to first have a closer look at the effects that come up with different constructions.
Note this is what we are used to see. Sharp high Q resonances fading away equally.
Note that the resonance at 300 Hz starts lower but falls much less steep than the resonance at 180 Hz.
Here the resonance at 300 Hz builds up and also shifts its frequency down to some degree
Here we can observe kind of frequency swapping between 150 Hz and 180 Hz and kind of oscillation at very low frequency
Here we can see decay in steps
Above gives a glimpse on how complex sound coloration might get. Sure, all measurements I presented were from speakers in cabinets and open baffle speakers are different.
On a smaller scale, I'll bet, same strange things happen with the speaker itself.
On an larger scale, we have to take room response also into account and a typical listening environment can always be seen as if sitting inside a loudspeaker cabinet. This is most obvious for those who like to build their speakers into the walls.
(By the way, my Avatar basically tells you the same story, showing the outspread of a single frequency wave from two loudspeakers into a simplified listening room.)
Coming back to overlapping frequency range with multiple speakers:
there was an experiment I was told of with one pair of very neutral sounding speakers against a pair of speakers where they throw a lot of resonating things into – and guess what – the multiple resonating pair was subjectively judged as the better one.
So at the end using different speakers over a broad range might balance also nicely.
Greetings
Michael
Hi
Well, I'd rather think so!
With a fairly smooth FR loudspeakers do not behave like TV or computer monitors where there are definitely not all colours available depending on make (DLP, LCD, Plasma...) and in addition to that there are also might be some missing spectral lines form the illumination source.
That there are sounds that can't be reproduced (accurately) is obvious for anybody. Thinking of it that there are "colors" that are missing is a nice concept but I don't think this holds true. At least unless we don't suffer from finite resolution of upcoming digital speakers. Anything could equally be explained with coloration and masking effects IF we had a better understanding about that.
While continuing my research about diminishing resonance effects in loudspeaker cabinets WITHOUT the use of damping material, I found it most interesting to first have a closer look at the effects that come up with different constructions.
Note this is what we are used to see. Sharp high Q resonances fading away equally.
Note that the resonance at 300 Hz starts lower but falls much less steep than the resonance at 180 Hz.
Here the resonance at 300 Hz builds up and also shifts its frequency down to some degree
Here we can observe kind of frequency swapping between 150 Hz and 180 Hz and kind of oscillation at very low frequency
Here we can see decay in steps
Above gives a glimpse on how complex sound coloration might get. Sure, all measurements I presented were from speakers in cabinets and open baffle speakers are different.
On a smaller scale, I'll bet, same strange things happen with the speaker itself.
On an larger scale, we have to take room response also into account and a typical listening environment can always be seen as if sitting inside a loudspeaker cabinet. This is most obvious for those who like to build their speakers into the walls.
(By the way, my Avatar basically tells you the same story, showing the outspread of a single frequency wave from two loudspeakers into a simplified listening room.)
Coming back to overlapping frequency range with multiple speakers:
there was an experiment I was told of with one pair of very neutral sounding speakers against a pair of speakers where they throw a lot of resonating things into – and guess what – the multiple resonating pair was subjectively judged as the better one.
So at the end using different speakers over a broad range might balance also nicely.
Greetings
Michael
Driver Resonances
I strongly suspect the CSD measurements are actually showing cabinet modes, not the drivers themselves. I'll bet you two pixels
that if the drivers were removed from the cabinets, mounted on a large flat baffle, and measured near-field (microphone 2 cm from cone), those low-frequency high-Q modes would all disappear.
The lowest-frequency departure from flat, resonance-free piston-band operation I've measured in a bass driver were spider resonances, typically around 500 Hz or higher. As it is, spider resonances are a design defect, and shouldn't be present in a monitor-quality loudspeaker. Spider resonances are sometimes disclosed by very small ripples in the impedance curve - and translate to severe, high-Q resonances in the acoustic realm. When the prosound manufacturers make obscure references to "silicon" damped spiders, they're trying to tell us the spider is damped with a silicone-rubber compound.
Resonances in a cone driver below 500 Hz hints at a gross design defect - the cone is flexing in a region where it should be rigid, some part is loose and buzzing, etc. I would be quite concerned by any driver resonance much below 1 kHz - it just ain't natural.
Conversely, we should ALWAYS expect to see severe cabinet resonances from 1 kHz on down. That's consistent with quarter and half-wave internal cabinet dimensions, and the fact that cabinet filling and damping materials on the cabinet walls have only modest effectiveness at lower frequencies. The region between 100 and 1 kHz is where cabinet colorations are measurably and audibly most severe - and coincidentally, where cone drivers are nominally in "piston mode" where they should be intrinsically flat.
In other words, cabinets are at their worst at the same frequencies where cone drivers are at their best! This is why I suggest getting the drivers out of the cabinets and measuring on a LARGE flat baffle in free air, or even measuring nearfield hanging from a string. Unless drivers have gone dramatically downhill from my Audionics days in the 1970's, I didn't see any narrowband resonances much below 700 Hz. Once you get below the spider resonances, drivers behave pretty close to Theile/Small models.
Now if drivers have degraded over the last several decades, I'd like to know. Last time I spoke with Linkwitz, he did say there were some famous, widely-used audiophile drivers with outright defective spiders.
------------------
Some very impressive links were sent to me by a reader - the webpage of David Griesinger. He makes a very strong case for separate stereo subwoofers that are positioned on the sides of the listener in this paper. Highly recommended reading!
The comments of audibility of the velocity wave is appropriate to this discussion - remember, measurement microphones are only sensitive to pressure, and as the articles show, equalizing for only the pressure term can be a mistake. Dipoles generate a strong velocity term close to the speaker, unlike a monopole, where it takes a LF room mode to create a velocity term (close to and right at the pressure null).
By the way, suggestions for "other" drivers like the PHY, Supravox, Fertin, and the Eminence Deltalites are very much appreciated, showing all of us what's out there. If a notch filter can be avoided, that's good!
As for mixing and matching, seriously, I don't know at this point. It's only an instinct that it would work well - in the bass region only. It's not something I'd try at higher frequencies, where crossover design, inter-driver phase relationships, and vertical and horizontal lobing become much more critical and hard to design. That's part of the reason the system has only one "real" crossover, around 1.5 to 2.5 kHz, depending on the properties of the widerange driver and the HF ribbon or horn tweeter.
Everywhere else, the drivers are close to phase-synchronous - and one reason I'm keeping a separate, isolated tweeter, and the possibility of a 5 to 10-degree slanted baffle, in the mix. Crossovers are much easier to design if the arrival times from the drivers are reasonably close together.
mige0 said:
I strongly suspect the CSD measurements are actually showing cabinet modes, not the drivers themselves. I'll bet you two pixels
that if the drivers were removed from the cabinets, mounted on a large flat baffle, and measured near-field (microphone 2 cm from cone), those low-frequency high-Q modes would all disappear.The lowest-frequency departure from flat, resonance-free piston-band operation I've measured in a bass driver were spider resonances, typically around 500 Hz or higher. As it is, spider resonances are a design defect, and shouldn't be present in a monitor-quality loudspeaker. Spider resonances are sometimes disclosed by very small ripples in the impedance curve - and translate to severe, high-Q resonances in the acoustic realm. When the prosound manufacturers make obscure references to "silicon" damped spiders, they're trying to tell us the spider is damped with a silicone-rubber compound.
Resonances in a cone driver below 500 Hz hints at a gross design defect - the cone is flexing in a region where it should be rigid, some part is loose and buzzing, etc. I would be quite concerned by any driver resonance much below 1 kHz - it just ain't natural.
Conversely, we should ALWAYS expect to see severe cabinet resonances from 1 kHz on down. That's consistent with quarter and half-wave internal cabinet dimensions, and the fact that cabinet filling and damping materials on the cabinet walls have only modest effectiveness at lower frequencies. The region between 100 and 1 kHz is where cabinet colorations are measurably and audibly most severe - and coincidentally, where cone drivers are nominally in "piston mode" where they should be intrinsically flat.
In other words, cabinets are at their worst at the same frequencies where cone drivers are at their best! This is why I suggest getting the drivers out of the cabinets and measuring on a LARGE flat baffle in free air, or even measuring nearfield hanging from a string. Unless drivers have gone dramatically downhill from my Audionics days in the 1970's, I didn't see any narrowband resonances much below 700 Hz. Once you get below the spider resonances, drivers behave pretty close to Theile/Small models.
Now if drivers have degraded over the last several decades, I'd like to know. Last time I spoke with Linkwitz, he did say there were some famous, widely-used audiophile drivers with outright defective spiders.
------------------
Some very impressive links were sent to me by a reader - the webpage of David Griesinger. He makes a very strong case for separate stereo subwoofers that are positioned on the sides of the listener in this paper. Highly recommended reading!
The comments of audibility of the velocity wave is appropriate to this discussion - remember, measurement microphones are only sensitive to pressure, and as the articles show, equalizing for only the pressure term can be a mistake. Dipoles generate a strong velocity term close to the speaker, unlike a monopole, where it takes a LF room mode to create a velocity term (close to and right at the pressure null).
By the way, suggestions for "other" drivers like the PHY, Supravox, Fertin, and the Eminence Deltalites are very much appreciated, showing all of us what's out there. If a notch filter can be avoided, that's good!
As for mixing and matching, seriously, I don't know at this point. It's only an instinct that it would work well - in the bass region only. It's not something I'd try at higher frequencies, where crossover design, inter-driver phase relationships, and vertical and horizontal lobing become much more critical and hard to design. That's part of the reason the system has only one "real" crossover, around 1.5 to 2.5 kHz, depending on the properties of the widerange driver and the HF ribbon or horn tweeter.
Everywhere else, the drivers are close to phase-synchronous - and one reason I'm keeping a separate, isolated tweeter, and the possibility of a 5 to 10-degree slanted baffle, in the mix. Crossovers are much easier to design if the arrival times from the drivers are reasonably close together.
Latest Thoughts
This time I didn't draw the support structures - there are lots of ways to do this, from the steam-punk look of brass cylinders and adjustment thumbscrews to the modern-day black-painted space-frame. My suggestion for the tweeter is hanging it on threads, so it is isolated from the vibrations of the rest of the assembly. I think being able to adjust the tweeter front-to-back, as well as rotating it, might be a desirable option - this lets you fine-tune the system without wrestling with the main driver array.
The delay of the horn compression driver, or ribbon tweeter, is approximately correct - about three inches behind the voice coil of the widerange driver. So if a ribbon is used, yes, it'll be in about the same place as the compression driver drawn here. The prettier look of aligning the faceplate of the tweeter with the front panel means more work on the crossover and accepting time and frequency response that isn't quite as good - your choice.
There is a possibility of a OEM version of 100 dB/metre ribbon specifically designed for studio monitors, with a suggested 1.5~2 kHz crossover. Stay tuned for more news on this front.
I'm still leaning towards the 18Sound driver array, with the 12NDA520 for the 12" driver, and the 15NMB420 for the midbass and bass drivers. The Eminence Deltalite-II 2515 looks entirely suitable as well.
It's interesting many of these drivers have a minor (3 dB or less) peak at 2.2 kHz. If you don't want to use a notch filter, they could be crossed over with a low-Q (0.5 or less) lowpass filter somewhere around 1.5 to 2 kHz - the net acoustic result of the minor peak at 2.2 kHz and the low-Q 2nd-order filter would result in an acoustic 3rd-order Butterworth lowpass filter around 2.2 kHz.
This time I didn't draw the support structures - there are lots of ways to do this, from the steam-punk look of brass cylinders and adjustment thumbscrews to the modern-day black-painted space-frame. My suggestion for the tweeter is hanging it on threads, so it is isolated from the vibrations of the rest of the assembly. I think being able to adjust the tweeter front-to-back, as well as rotating it, might be a desirable option - this lets you fine-tune the system without wrestling with the main driver array.
The delay of the horn compression driver, or ribbon tweeter, is approximately correct - about three inches behind the voice coil of the widerange driver. So if a ribbon is used, yes, it'll be in about the same place as the compression driver drawn here. The prettier look of aligning the faceplate of the tweeter with the front panel means more work on the crossover and accepting time and frequency response that isn't quite as good - your choice.
There is a possibility of a OEM version of 100 dB/metre ribbon specifically designed for studio monitors, with a suggested 1.5~2 kHz crossover. Stay tuned for more news on this front.
I'm still leaning towards the 18Sound driver array, with the 12NDA520 for the 12" driver, and the 15NMB420 for the midbass and bass drivers. The Eminence Deltalite-II 2515 looks entirely suitable as well.
It's interesting many of these drivers have a minor (3 dB or less) peak at 2.2 kHz. If you don't want to use a notch filter, they could be crossed over with a low-Q (0.5 or less) lowpass filter somewhere around 1.5 to 2 kHz - the net acoustic result of the minor peak at 2.2 kHz and the low-Q 2nd-order filter would result in an acoustic 3rd-order Butterworth lowpass filter around 2.2 kHz.
Attachments
Hi
You won!
Of course all measurements show cabinet behaviour not speaker behaviour.
Don't get me wrong, omitting the cabinet seems to be the best way to get rid of cabinet coloration, making cabinets small compared to the wavelengths involved is an other possibility especially suitable for subwoofers.
Showing the backside of a coin though, can bring up the real juicy stuff!
The mechanisms demonstrated like:
- non linear decay
- frequency shift and interaction
- Q blurring
- small interval echo
basically apply to any more complex resonating system, is it a cabinet, a speaker or a listening room.
The point here was, that there is plenty of coloration going on with audio playback at different scales but I didn't find any evidence for the concept of missing sound colors so far.
Greetings
Michael
You won!
Of course all measurements show cabinet behaviour not speaker behaviour.
Don't get me wrong, omitting the cabinet seems to be the best way to get rid of cabinet coloration, making cabinets small compared to the wavelengths involved is an other possibility especially suitable for subwoofers.
Showing the backside of a coin though, can bring up the real juicy stuff!
The mechanisms demonstrated like:
- non linear decay
- frequency shift and interaction
- Q blurring
- small interval echo
basically apply to any more complex resonating system, is it a cabinet, a speaker or a listening room.
The point here was, that there is plenty of coloration going on with audio playback at different scales but I didn't find any evidence for the concept of missing sound colors so far.
Greetings
Michael
mige0 said:
I couldn't agree more. Loudspeakers do many dreadful things to the signal, especially long-term high-Q storage effects - one of the more bizarre effects I discovered in that MLSSA series of tests were resonances that swept in frequency as they decayed, like chirp radar!
Room resonances can be resolved by the hearing system, while cabinet resonances follow the direct sound so quickly they can't be separated out. The first in a long succession of room reflections typically arrives 5~7 mSec after the direct sound, while cabinet resonances start within a millisecond. This falls within the interval used by the location-sensing mechanism, and also degrades the impression of timbre - especially vocals, which not surprisingly sound they are coming from a box. They are!
By partitioning the spectrum, we can use devices in the region where they perform the best. Subwoofers, whether monopoles or W-box dipoles/cardioids, operate with an enclosure a small fraction of a wavelength. This is good; all the box has to do is be rigid, and not generate high-order decay harmonics in the several-hundred Hz region.
If we avoid the region where boxes go into characteristic modes (as a function of their size), the drivers can operate in the piston region, and deliver inherently flat response. The 1/f characteristic of dipoles can be equalized, or compensated by bringing in additional drivers to keep efficiency constant.
The delayed rear-wave coming around the edge of the baffle can be smoothed out by using asymmetric shapes, Karlson slots, or open-mesh resistive grilles to smooth out the delayed-arrival signal. Rather than guessing as is traditional for OB, MLS can be used to inspect the delayed-arrival signal and confirm the smoothing techniques are working. True, these are some of the same techniques used inside boxes, but it has the merit that we're dealing with one signal coming around the edge, not hundreds of internal reflections.
At higher frequencies, we shift technologies again, and select tweeters that with low time storage - to match the performance of the OB drivers. A real hard-core purist might insist on dipole electrostats or planars in the 700 Hz to 5 kHz region, just to avoid cone-driver resonances, but that would discard the high dynamic range (and efficiency) of prosound drivers, and would certainly add to system complexity. Harmonizing the sound of cone drivers to electrostats and ribbon tweeters sounds like an awful job.
The trick is to find tweeters with very high dynamic range and low energy storage, two things that don't usually go together. The candidates are compression drivers + advanced profile horns, modern versions of the AMT, and studio-monitor-grade ribbons.
What's so interesting about OB is that this is largely uncharted territory for loudspeaker design: Siegfried Linkwitz was one of the first to discover what it can do for modern systems, but there's lots more to territory to explore. It's amusing is that Americans are so late to the table on this one - I was surprised at the 2004 ETF that OB was so well established in Europe, while it's really radical here.
Russell Dawkins said:
fiacono said:
Russel, I'm surprised he liked forward-facing treble in the first place. True, it's the "traditional" sound from conventional speakers that we're used to, but SL was trying to get away from that and move towards something more like real life. Just because we disagree with Amar Bose's 901 concept doesn't mean there isn't something worthwhile in there.
In my own experience, the challenge with rear tweeters is rear sound sneaking around towards the front, and degrading coherence from the front driver. This is no problem with a true radial tweeter (Walsh, MBL, etc.) which radiates from one point (OK, one radiating surface), but is awkward with multiple radiators that are spaced apart. One solution to this is hiding the rear tweeter behind the rear of the speaker, and shaping the highpass filter so not as much sound diffracts around the front.
When I tried this with my first speaker for Audionics in 1975, I wasn't completely happy with the sound - it was more spacious, true, but there was an artificial, pasted-on quality to the HF dimensionality. But that was a conventional forward-facing speaker, a really big transmission line with four drivers and four-way crossover.
With a dipole with a generous amount of rear-directed energy over the full spectrum, a rear tweeter is a different matter. In that event, you want the spectrum coming out of the back of the speaker to have a similar character to the front - maybe not identical, but similar. So you now have the fun task of measuring the speaker over the front listening angle (a 60-degree sweep), at the sides (what's the null look like?), and across the back (is the spectrum continuous, or are there holes?).
Exact one-to-one spectral matching is less important than gross discontinuities - big peaks and holes at certain angles. I'd look for the most trouble at +45, +135, -135, and -45 degrees, where the the drivers are reaching the edge of their polar pattern, and the sound from the front and back are meeting, resulting in a partial null. This is where the spectrum is going to be the roughest, and time response most distorted.
I suspect it all comes down to quality of implementation. Dissimilar front and rear tweeters, with different sonic characters, is asking for trouble. HF response shaping for the rear tweeter is fairly sensitive, since you're controlling the polar pattern in the horizontal plane with the phase relationship between front and rear.
Frank, there are several ways to go with this concept - none of them wrong, just different. First, pick any widerange driver you like, 8, 10, or 12 inches. Second, choose the bass-assist drivers - one or two, 12 or 15-inch, or a combination of 12 and 15-inch.
If the same amplifier is going to drive the entire front panel, a pair of 16-ohm Tone Tubby's, either 12 or 15-inch, fills the high-Q Midbass and Bass requirement. All of the cone drivers have a cascade crossover, starting with a 12 dB/octave lowpass at 2 kHz. This is connected directly to the widerange driver at the top of the baffle. (The Widerange driver has an inductance-correcting Zobel network so it presents a resistive load to the 12 dB/oct lowpass and the other drivers.)
After going to the widerange driver, the signal flows to a tapped lowpass inductor (transformer-grade iron-core) tunable between 160 and 250 Hz. The Midbass driver is connected to the other side of the inductor, and also has a Zobel network to compensate for its inductance.
The signal continues to another tapped lowpass inductor, tunable between 80 and 125 Hz. The Bass driver is connected to the other side of this inductor, and has its own Zobel network.
All three drivers, working at parallel at the lowest frequencies, present a nominal 4-ohm impedance, and are in-phase with each other. The overall slope of the network gradually steepens for each driver in the cascade, but the Q is very low, since the poles are spread quite far apart.
If 16-ohm drivers are not available, and/or the Qts is in the more conventional 0.22 to 0.4 range, then multi-amping and equalization is a good idea. The high-quality amplifier is reserved for the widerange driver and its tweeter (which are crossed-over passively). Any equalization of the widerange + tweeter is done by the passive crossover, not active EQ, partly because I strongly recommend drivers that don't need much equalization in the first place.
Things are different once you get to 250 Hz and below. The room response merges with the what the speaker does - in fact, the room cannot be distinguished by ear, or far-field measurement, from the speaker. This is where individual L and R channel parametric EQ to remove peaks created by room modes is a good idea - but trying to "fill in" room nulls is a bad idea, creating massive distortion and overloading the drivers and the amplifier to no good effect.
Equalization is most useful in the range covered by the midbass and bass drivers - 80 to 250 Hz. I'd still recommend connecting the Midbass and Bass drivers in parallel, as in the example above, with the Bass drivers additionally lowpass-filtered by a tunable 80 to 125 Hz inductor. This controls the overlap between the drivers, and allows response and phase shaping in this critical region.
For the region from 60~80 Hz and lower, stereo subwoofers, located as far apart as possible, or even at the sides of the listening room, are a good idea. I feel it is quite legitimate to put the stereo subs on each side of the listening chair or couch, for example - it decreases power requirements, and increases the tactile sensation. If the subwoofers are conventional monopoles, choose drivers with the lowest distortion possible, to match the dynamics and low-distortion character of the rest of the system.
In the multi-amping example, this is a good application for obsolete Dolby Surround home theater receivers, or Class D or T amps. I'd reserve the quality amplification for the 200 Hz-on-up range.