DI & total room spectra
The EBU did put this in a recommendation R22 (1980), but they made a minor mistake in defining the 'total room spectra' as a flat reverberation time between 200 Hz and 2.5 kHz and a linear diminishing reverberation time between 2.5 kHz and 10 kHz. The minor mistake was that the reverberation was with the speaker system as source, whilst the speaker system had a linearly (v log(freq)) increasing directivity from 2.5 kHz onwards. This was corrected afterwards (EBU 3276) and they now quote a constant reverberation time from 200 Hz on and an operational room response curve that can drop linearly by 0 - 1 dB/oct from 2 kHz. However converting the reverberation time of the initial recommendation and looking at the characteristics of the original monitor speakers, the drop was actually 1.5 dB/oct. (The current recommendation corresponds to the actual practice of having a slight apparent tilt in the high frequencies in broadcast monitoring)
My interpretation of tonal neutrality is a flat response on the listening axis (most cases 15°), a constant operational room response between 400 Hz and 900 Hz and a linearly (v log(freq)) increasing directivity between 2 and 16 kHz. The Harman R&D-ers (JBL, Revel) have researched this and tested listener preferences in the recent past and came to similar conclusions.
Due to the distance of the woofer from the floor and the proximity of walls, the response below 400 Hz is mainly governed by speaker positioning in relation to the walls and the radiation pattern, and thus a similar DI below 400 Hz as beyond 400 Hz is not required. The corner frequency for directivity can be anywhere between 900 Hz and 2 kHz, this is a less critical region for many auditory aspects.
Does this mean that the directivity with a 15 in woofer and a 400 Hz horn starts to increase at a too low frequency?
Well, let's have a look at of my current speaker system having a 10 in woofer crossing over to a 300 Hz Guigue Iwata horn at 600 Hz. The measurements were made in 2007 after the last minor update at the listening position. (I only measure afterwards).
The direct response on the listening axis (15°), first 2.5 ms of the impulse response: fairly flat to 12 kHz and showing a single (floor) reflection.
The first 48 ms: important for perceived coloration.
The total room spectra (1.4 sec):
Although the crossover frequency between the direct radiating woofer and the horn is at 600 Hz, the direct response and the total response are reasonaby flat up to 1.2 kHz. However the total response drops at 2.5 dB/oct instead of the required 1.5 dB/oct.
The speakers sound neutral when the direct response can be heard, but this is not the case when I am in my kitchen and the speakers are out of sight. In that case I hear that the total response drops too fast towards high frequencies.
Can this be changed?
Yes:less toe-in angle: listening axis at 25°, so a small slope is needed towards the high frequencies (e.g. Baxandall tone control) to keep a flat direct response and less slope in the total response. Lynn, this is certainly worth trying after you have finished your speakers, but you should not worry now. A change in toe-in should be accompanied by a change in tone control slope. My experience - also confirmed by my Japanese acquaintances - is that with less toe-in, the speakers need to be at a smaller angle viewed from the listening position. So this also means moving the speakers closer together when experimenting with toe-in. This also changes the distance to the side and front wall and thus the mid frequencies, which gave less good results in my room.
I have considered a dipole and a short horn (Limmer Horn 300) for the mid frequencies. A serious drawback of these solutions is that an extra mid-woofer is needed which can only be placed below the dipole or mid horn. I have used dual woofers in my system and noticed that the bottom woofer needed to be limited to maximum 160 Hz (4th order). This means that the dipole or short horn would need to be able to go down to 150 Hz, which is no easy task. I can live with the compromise that the sound heard when I am preparing dinner is not perceptually flat.
An additional experiment was adding a small dipole (4 in full range in a 5 in wide baffle) 4 inches behind the bass cabinets with the side oriented towards the first wall reflection (zero output in that direction) and absorbing material on the back of the bass cabinet. I added some extra delay in the signal towards these speakers to make sure that the first arrival at the listening position was 8 - 10 ms later than from the main speakers and used a shelving equalizer. I could successfully improve the total response without negative effects on other performance parameters. Since at my normal listening position it made very little difference (a hint of more air), I decided not to keep it in my system.
Siegfried Linkwitz commented that his omnidirectional speakers (Pluto) needed to be closer to the listening position (further from the walls) than his dipoles to have a similar direct/reverberation ratio. Something similar is true for horn speakers: they need to be placed further from the listening position than traditional speakers, much closer to the front wall. This also solves the problem of walking behind the speakers
.
Gary,
I experimented a lot with dipole sub-woofers but never could achieve anything that was better than one closed box sub-woofer in the front room corner; which I could easily improve by adding a second sub-woofer at 3/5 th of the width of the front wall. Still I have not given up on further experimenting with dipole sub-woofers, and I wish you more success than I had so far.
Since Alan Blumlein developed stereophonic sound, there has been been a lot of R&D on this subject for control rooms and listening rooms. From the late 50's, the radio & tv broadcasters (BBC R&D) did set a reference based on many trials and they developed monitoring speakers and control rooms accordingly. Recordings were made and equalized to sound tonally neutral on the broadcast (BBC) reference systems. These became de-facto references for on-axis response and 'total room spectra'. Because subsequent generations of mastering speakers also needed to faithfully reproduce previous recordings, these had to loosely resemble the original reference in on-axis and power response. Recordings sound tonally neutral when reproduced on systems with these characteristics.
The EBU did put this in a recommendation R22 (1980), but they made a minor mistake in defining the 'total room spectra' as a flat reverberation time between 200 Hz and 2.5 kHz and a linear diminishing reverberation time between 2.5 kHz and 10 kHz. The minor mistake was that the reverberation was with the speaker system as source, whilst the speaker system had a linearly (v log(freq)) increasing directivity from 2.5 kHz onwards. This was corrected afterwards (EBU 3276) and they now quote a constant reverberation time from 200 Hz on and an operational room response curve that can drop linearly by 0 - 1 dB/oct from 2 kHz. However converting the reverberation time of the initial recommendation and looking at the characteristics of the original monitor speakers, the drop was actually 1.5 dB/oct. (The current recommendation corresponds to the actual practice of having a slight apparent tilt in the high frequencies in broadcast monitoring)
My interpretation of tonal neutrality is a flat response on the listening axis (most cases 15°), a constant operational room response between 400 Hz and 900 Hz and a linearly (v log(freq)) increasing directivity between 2 and 16 kHz. The Harman R&D-ers (JBL, Revel) have researched this and tested listener preferences in the recent past and came to similar conclusions.
Due to the distance of the woofer from the floor and the proximity of walls, the response below 400 Hz is mainly governed by speaker positioning in relation to the walls and the radiation pattern, and thus a similar DI below 400 Hz as beyond 400 Hz is not required. The corner frequency for directivity can be anywhere between 900 Hz and 2 kHz, this is a less critical region for many auditory aspects.
Does this mean that the directivity with a 15 in woofer and a 400 Hz horn starts to increase at a too low frequency?
Well, let's have a look at of my current speaker system having a 10 in woofer crossing over to a 300 Hz Guigue Iwata horn at 600 Hz. The measurements were made in 2007 after the last minor update at the listening position. (I only measure afterwards).
The direct response on the listening axis (15°), first 2.5 ms of the impulse response: fairly flat to 12 kHz and showing a single (floor) reflection.
The first 48 ms: important for perceived coloration.
The total room spectra (1.4 sec):
Although the crossover frequency between the direct radiating woofer and the horn is at 600 Hz, the direct response and the total response are reasonaby flat up to 1.2 kHz. However the total response drops at 2.5 dB/oct instead of the required 1.5 dB/oct.
The speakers sound neutral when the direct response can be heard, but this is not the case when I am in my kitchen and the speakers are out of sight. In that case I hear that the total response drops too fast towards high frequencies.
Can this be changed?
Yes:less toe-in angle: listening axis at 25°, so a small slope is needed towards the high frequencies (e.g. Baxandall tone control) to keep a flat direct response and less slope in the total response. Lynn, this is certainly worth trying after you have finished your speakers, but you should not worry now. A change in toe-in should be accompanied by a change in tone control slope. My experience - also confirmed by my Japanese acquaintances - is that with less toe-in, the speakers need to be at a smaller angle viewed from the listening position. So this also means moving the speakers closer together when experimenting with toe-in. This also changes the distance to the side and front wall and thus the mid frequencies, which gave less good results in my room.
I have considered a dipole and a short horn (Limmer Horn 300) for the mid frequencies. A serious drawback of these solutions is that an extra mid-woofer is needed which can only be placed below the dipole or mid horn. I have used dual woofers in my system and noticed that the bottom woofer needed to be limited to maximum 160 Hz (4th order). This means that the dipole or short horn would need to be able to go down to 150 Hz, which is no easy task. I can live with the compromise that the sound heard when I am preparing dinner is not perceptually flat.
An additional experiment was adding a small dipole (4 in full range in a 5 in wide baffle) 4 inches behind the bass cabinets with the side oriented towards the first wall reflection (zero output in that direction) and absorbing material on the back of the bass cabinet. I added some extra delay in the signal towards these speakers to make sure that the first arrival at the listening position was 8 - 10 ms later than from the main speakers and used a shelving equalizer. I could successfully improve the total response without negative effects on other performance parameters. Since at my normal listening position it made very little difference (a hint of more air), I decided not to keep it in my system.
Siegfried Linkwitz commented that his omnidirectional speakers (Pluto) needed to be closer to the listening position (further from the walls) than his dipoles to have a similar direct/reverberation ratio. Something similar is true for horn speakers: they need to be placed further from the listening position than traditional speakers, much closer to the front wall. This also solves the problem of walking behind the speakers
.Gary,
I experimented a lot with dipole sub-woofers but never could achieve anything that was better than one closed box sub-woofer in the front room corner; which I could easily improve by adding a second sub-woofer at 3/5 th of the width of the front wall. Still I have not given up on further experimenting with dipole sub-woofers, and I wish you more success than I had so far.
Attachments
Thanks, Peter, for an exceptionally informative post which combines BBC and EBU research with your own personal experience. Good combination.
Gary Dahl's approach (along with other collaborators who have contacted me through email) remains the baseline for my own build. The leading alternative, though, is an Altec/GPA 416 15" driver on a compact dipole (with small side wings) sitting on top of a 15" closed-box woofer, with the closed-box woofer EQ'ed to complement the LF rolloff of the upper driver. (This might well fall in the 160 Hz LPF criteria mentioned by Peter.)
Gary Dahl also had the same experience with dual woofers that you had, Peter. The attempt to use both woofers up to 500 Hz resulted in bass-heavy, slow sound, so the lower woofer ended up doing subwoofer duties only.
It seems possible to combine a 15" OB (upper driver) with a 15" closed-box (lower driver). The upper driver might require a 2nd-order HPF to control excursion, maybe in the 80~120 Hz range, while the lower driver would need a complementary LP characteristic so the net response at the listening position is flat.
If time and phase considerations could be satisfied, there would a two-octave extension in DI compared to a closed-box approach. (The time and phase considerations are not simple, since the transition between the two drivers results in a significant change in polar pattern. Attenuating the backwave from the upper driver with a layer of felt might make things a little bit easier.)
Quick aside on dipoles and bi-poles: The word "bi-pole" is an annoying marketing term that I wish would just go away. If you put another driver on the rear of a loudspeaker, and it is connected in-phase with the front driver, all you get in a near-omni, with a lumpy polar pattern at higher frequencies. People have been making speakers like this decades before some clever marketing guy came up with a cute new marketing phrase for yet-another-omni speaker.
A dipole is quite a different animal. There is a sharp acoustic null at the sides, straight up, and straight down. (This is where the increase in DI comes from.) Putting another driver on the back, and connecting it out-of-phase with the front driver, results in a quasi-dipole at the lowest frequencies (where the wavelengths are larger than the whole speaker). The pattern gets complicated once the wavelengths are the size of the speaker or smaller; there are many nulls that move around with frequency.
Last edited:
Why limit the closed box driver to 15"?........an 18 or even 21 would be more efficient at the task and might better support the aesthetics of the upper woofer's winged enclosure. If you start the HP filter of the upper OB woofer properly, you can use that to your advantage in controlling the dipole peak......which is where the flexibility in the size of the wings comes I handy, allowing you to place the peak right where you want it in order to compliment the sealed drivers electric low pass. There's lots of controls here with properly set dual 2nd order electrical filters to get near exactly the curve desired by varying the corner frequency......better than a single 4th order electrical any day of the week in my experience.
The Fane 15XBN and 18XBN are good candidates, with decent behavior up to 500 Hz and dual spider construction. (If one spider is reversed, there's a reduction of spider nonlinearity, and having two spiders helps prevent side-to-side rocking on long excursions.)
The B&C 15NDL76, B&C 15NW76 and B&C 18SW115 also look interesting.
One merit of a pair of 15" woofers (versus a single 18" driver) is they can be put on the left and right sides of a closed-box enclosure and the mechanical vibration cancelled out. (Particularly if steel or aluminum rods cross-connect the mounting bolts between the drivers.) In the 300 Hz and lower region, the woofers are omnidirectional anyway, so a vibration-canceling driver layout has some advantages.
As for a 700 Hz baffle peak, that leads to a pretty small baffle ... not much room for wings, except maybe to mechanically stabilize the driver and baffle. A 500 Hz baffle peak leaves a bit of room for wings, but then the first null in response falls at 1 kHz, which is inconvenient for the crossover.
A curved felt blanket behind (and to the sides) of the OB driver decreases both the baffle peak, and perhaps more importantly, the depth of the first null in response. Whether a felt blanket several inches behind the driver has an effect on subjective dynamics is a matter of controversy.
The B&C 15NDL76, B&C 15NW76 and B&C 18SW115 also look interesting.
One merit of a pair of 15" woofers (versus a single 18" driver) is they can be put on the left and right sides of a closed-box enclosure and the mechanical vibration cancelled out. (Particularly if steel or aluminum rods cross-connect the mounting bolts between the drivers.) In the 300 Hz and lower region, the woofers are omnidirectional anyway, so a vibration-canceling driver layout has some advantages.
As for a 700 Hz baffle peak, that leads to a pretty small baffle ... not much room for wings, except maybe to mechanically stabilize the driver and baffle. A 500 Hz baffle peak leaves a bit of room for wings, but then the first null in response falls at 1 kHz, which is inconvenient for the crossover.
A curved felt blanket behind (and to the sides) of the OB driver decreases both the baffle peak, and perhaps more importantly, the depth of the first null in response. Whether a felt blanket several inches behind the driver has an effect on subjective dynamics is a matter of controversy.
Last edited:
Returning to the theme of an OB 15" combined with a closed-box 15", the biggest problem with the OB driver is the deep null an octave above the baffle peak. An asymmetric baffle (an "X" or pyramid shape) helps to time-spread the backwave diffracting around the baffle edge, but there's still going to be a comb-filter null in the frequency response. (Time-smearing the diffracted backwave lowers the "Q" of the null, but also makes it broader in frequency. The cancellation dip in response is still there, it just has a different shape.)
If the cancellation dip falls in the crossover region, that's very troublesome, since that causes rapid changes in the phase angle between the OB bass driver and the MF/HF driver. If the null is deep enough, it can't be corrected by any passive network.
Which is why I'm suggesting a simple felt blanket around the rear of the OB driver, spaced perhaps 6" to 12" away from the rear of the cone. It only takes a few dB of attenuation to make the null (and baffle peak) much smaller in magnitude, and thus easier for the crossover to handle.
The frequency-selectivity of the felt, or other sound-absorbing material, works in our favor. It has the most absorption at higher frequencies, exactly the region where the comb-filter nulls are the most troublesome. At lower frequencies, absorption of the backwave is decreased, so the system transitions back to OB behavior. And no cabinet resonances, of course.
If the cancellation dip falls in the crossover region, that's very troublesome, since that causes rapid changes in the phase angle between the OB bass driver and the MF/HF driver. If the null is deep enough, it can't be corrected by any passive network.
Which is why I'm suggesting a simple felt blanket around the rear of the OB driver, spaced perhaps 6" to 12" away from the rear of the cone. It only takes a few dB of attenuation to make the null (and baffle peak) much smaller in magnitude, and thus easier for the crossover to handle.
The frequency-selectivity of the felt, or other sound-absorbing material, works in our favor. It has the most absorption at higher frequencies, exactly the region where the comb-filter nulls are the most troublesome. At lower frequencies, absorption of the backwave is decreased, so the system transitions back to OB behavior. And no cabinet resonances, of course.
Last edited:
There were a few people playing with non-peaking baffles - the results were inconclusive.
Have a look at the Open baffle Nautilus: the Fibonacci solution to edge diffraction thread - it sort of petered out into a discussion about something else altogether
Is this behavior theoretical or have you and Gary tested the blanket method with different material? Either way, sounds like a great solution. Needs a name though.....resistive OB?.....Quasi-anechoic? It would be interesting to see the radiation patterns of such a cabinet in measurements.
The deep null is only present with a round baffle with a point source (radius = 0) in the center.
I suppose that the dipole baffle will not be open to the top (JMLC horn above) or to the bottom (closed box underneath) and will not be round. This already smooths the response above the first peak and the null is not present for angles > 10°. Background material fig 2.10
More importantly: with a large membrane radius > 1/3 x baffle radius, the null is not even present with a round baffle: circular_b2_8a.
Stiff suspensions can have a long break-in time. A demonstration between a 12" speaker that had a 100 hour burn-in (with max linear displacement at resonant frequency) versus one that had toured for 2 months made clear that even the burn-in procedure was not sufficient. That speaker distributor / PA manufacturer's recommendation was that one should not use Cms < 0.2 mm/N unless it could have break-in with very loud music signals.
Oberton also make excellent speakers, but they do not seem to have a US distributor.
If you are planning to use a pair of woofers, you might also consider 12 inch woofers, e.g. 18Sound 12NLW9300.
Lynn, would a cardioid arrangement not be better? A cardioid that crosses over to a (or a pair of) subwoofer driver(s). The cardioid has most of what you describe, as well as ease of placement.
There are also quasi-cardioids where you have a sealed woofer and an OB woofer connected in parallel. That way there might not be a need for a sub.
Deon
Hello,
These are conflicting needs, di of main speakers and reverberant field. I use 2 sets of speakers. An extra pair for the reverberant field. I point them behind me and towards the back wall.
Hook up only the main pair and equalize as normal.
Next, hook up all speakers and apply pink noise and equalize only the 2nd pair for room balance.
The 2nd pair of speakers do not need to have quality sound or flat response.
They only need to be loud.
The widest dispersion is needed, omnis work best
You can even increase there volume beyond the main speakers by up to 10dB to increase the overall perceived sound.
The sound from these speakers should arrive at your ears at least 10mS after the main sound. You may need an electronic delay.
If you use omnis then wire them in reverse polarity to mimic the first reflection which reverses the polarity of the sound.
These are conflicting needs, di of main speakers and reverberant field. I use 2 sets of speakers. An extra pair for the reverberant field. I point them behind me and towards the back wall.
Hook up only the main pair and equalize as normal.
Next, hook up all speakers and apply pink noise and equalize only the 2nd pair for room balance.
The 2nd pair of speakers do not need to have quality sound or flat response.
They only need to be loud.
The widest dispersion is needed, omnis work best
You can even increase there volume beyond the main speakers by up to 10dB to increase the overall perceived sound.
The sound from these speakers should arrive at your ears at least 10mS after the main sound. You may need an electronic delay.
If you use omnis then wire them in reverse polarity to mimic the first reflection which reverses the polarity of the sound.
Last edited:
Myhrrhleine, I agree with most of your post, but in the measurements I've made, reflections preserve polarity, not invert it. It's one of the things that makes them easy to recognize, since they look like low-passed and attenuated versions of the first-arrival impulse.
I want to thank you, and all the others, for the intriguing and interesting posts you've been making. I think you can see why I consider what Gary Dahl, Martin Seddon, and others as the "baseline" version, with a straightforward bass enclosure combined with a MF/HF horn. This is tried-and-true system that's been around for a long time, and it's reasonably easy to obtain flat frequency response as well as good impulse response.
As you can see, my greatest reservation about OB's is the close-in impulse response. No matter what the configuration, the backwave will invariably diffract around the edge of the baffle. It cannot be avoided. It can be diffused and spread out in time (bigger driver relative to baffle, asymmetry, felt absorber on the baffle edge, etc.), but it can't be removed. And the backwave diffracting around to the front is the source of the predicted and measured response irregularities, as well as the ultimate 6 dB/octave rolloff.
If there's a substantial mismatch between the magnitudes of the front and back wave, though, most of the problems go away. The baffle peak gets smaller. The cancellation notch at 2X the baffle peak gets less deep, and if the back wave absorber is broadband, the 6 dB/octave rolloff isn't as steep, either.
As mentioned in post #11601, I tested something similar but with dipoles as a pair of 2nd speakers with the positive polarity towards the wall behind the speakers (which I call the front wall of the room) and the negative polarity towards the main speaker cabinet where it was absorbed. I used an electronic delay to have a 8 - 10ms difference. I fully agree that this allows to improve the measured overall room spectra without negative effect on the sound of the main speakers. However at my main listening position it made very little difference.
Do you mean with 'back wall' the wall behind your listening seat / opposite from the speakers?
Last edited:
Hello Peter
Do you have experience with this driver?
It sure looks interesting on paper.
Yes, This will make very little difference at the on axis listening position.
The last line is interesting. Additional ambience-enhancing speakers, additional channels of power amplifiers, and additional signal-processing to create the 8~10 mSec delay ... and "it made very little difference" in the most-used sitting position. Hmm. Only the measurements were better.
I guess there won't be much additional expense in building the two-box Gary Dahl configuration (as described earlier) with an additional felt-lined prototype OB module for the upper Altec/GPA 15" driver, while selecting the lower 15" (or 18") driver for low-distortion long-excursion performance. Two physical variations that are otherwise similar.
From the perspective of the lower driver, the only real difference between the two variations is the requirement to go up to 60 Hz or 160 Hz. The lower driver either meets the F3 of a closed-box upper driver or acts as a somewhat wider range bass-assist module for an OB system. It's not a difficult requirement either way.
Last edited:
Lynn, my sincere apologies for using your thread to answer this.
I have not heard this driver so far, and I never bought any driver I have not heard.
Drivers do sound very different above the cone edge resonance (350 - 400 Hz for 15 inch paper cone speakers; scale frequency with 1/radius for other sizes).
About the cone edge resonance itself: this is a high Q small amplitude resonance (< 2 dB). This type of resonance is inaudible if it lies below 400 Hz (15" driver or larger), becomes progressively more audible towards 800 Hz and is inaudible between 900 and 1800 Hz. So, this is a non-issue with 15" and 7" - 6" drivers, and a potential problem for 10" drivers as I experienced myself. To make the driver look better on a frequency response chart or on a decay chart, sometimes the cone edge resonance is unnecessarily damped (e.g the rubber edge of the suspension glued to cone outer edge to act as constrained layer damping), but in most cases the complete sound gets damped (no micro-dynamics).
If you are not planning to use this driver above 200 Hz, then it is an acceptable risk to buy one before having heard it.
I have experimented with force cancellation for subwoofers (4 x 10") and it works extremely well. The reason for mentioning using two 12" drivers was that the volume displacement, voice coil displacement, LF distortion etc is the same as for a single 18" driver, but also the total box volume. With two 15" drivers the box volume will be larger. However, the choice in 12" drivers is much smaller.
Last edited:
PeterBracke,
That is a very interesting comment about the placement of the surround on the front or rear of the cone. Have you actually built two similar drivers with the surround on the front and back of the cone to confirm this change in small scale response? I am trying to understand the real physics here as the surround is working as a damper on the edge and also as a termination path to reduce reversion of waves back down the cone after reaching this point. This is one reason that I would avoid a rubber surround as they will return more energy back down the cone due to their hysteresis effects.
That is a very interesting comment about the placement of the surround on the front or rear of the cone. Have you actually built two similar drivers with the surround on the front and back of the cone to confirm this change in small scale response? I am trying to understand the real physics here as the surround is working as a damper on the edge and also as a termination path to reduce reversion of waves back down the cone after reaching this point. This is one reason that I would avoid a rubber surround as they will return more energy back down the cone due to their hysteresis effects.
Hi Pooh, I like your post. At least wave guides give cheap taste of what mid/treble horn can offer. I like Lynns view that simple infinite baffle bass and Mid/Treble works and is proven. Also sympathetic with the variants on this.
Do you favour horns even with plenty of solid state class A oomph?
I still have a 10 cu ft Mahogany solid IB weighing 160 lb with a Fane 18" bass unit I made in 1972 as a sub/mid bass. It may work with the Lynn mid/upper horn so I may try a mid/treble CD on that for fun.
A well set up wave guide system works well with much lower sound levels that some must have, as much as they may want oomph of CD's