Zaph|Audio - SB12.3 3-Way Tower

I used a 2.07 cubic foot box for I had on hand for initial testing. In that box I got a nearfield response and impedance curve. I used delta compliance to get my Thiele/Small parameters. Given the new parameters, I modeled the the response in that enclosure size to compare model vs reality. It was close enough to move on. 2.07 cubic feet was not a usable enclosure size as evidenced by the peaky low end response, but the box was perfect for obtaining an accurate Vas number with delta compliance. I decided on just under 4 cubic feet as a good tradeoff between enclosure size and low end peaking. The woofer is optimal for a sealed enclosure - vented would have been the size of a refrigerator.

Harmonic distortion was very good in the range that it would be used. Unlike the many smaller woofers I've measured, my test method on this size woofer begins to lose accuracy above 400 Hz. The Le(x) plot shows the real top end limitation of this woofer. I wouldn't use this woofer too high in frequency, and the highish inductance isn't going to let me anyway.

Low end extension in 107 liters (3.78 cu ft) is excellent and that volume works out well for a floor standing system.

CROSSOVER

Crossover schematic

The crossover frequencies are around 220 and 2200 Hz. The woofer-mid crossover is mild asymmetric 2nd order while the mid-tweeter is mild asymmetric 4th order. Both deviations from slope symmetry tilt the main listening lobe in our favor, which compensates for acoustic center differences. Both the tweeter and mid have impedance flattening circuits. The woofer gets by with a single moderate size coil (for a 3-way) mostly because it already has a large inductance built in. There is only a quasi-zobel there to help the inductor hit the target slope. The mid has a mild low end rolloff due to baffle step and gets by with a single cap for a high pass. That single cap hits the target slope only with the help of the LCR trap. That trap on the mid has the largest component values in the crossover and unfortunately is required. Power handling is a driving factor in the selection of crossover frequencies. The SB29 tweeter could easily work down to 1500 Hz in a small TM system, but more is expected of it in a system of this size and a conservative point is selected. The SB15 mids are extremely clean on the top end, so things worked out nicely.

Important: Since the crossover between the woofer and the mid is 2nd order, both the midrange and tweeter are hooked up in reverse polarity with the positive terminals going to ground.

Components were worked out between myself and Madisound, with better quality caps in some locations and large electrolytics in non-critical locations. Potential power handling issues were looked at for every single component. Some resistors are doubled up for power handling, and some larger value caps had to be high power poly. Heavy gauge inductors were used where required. This system can be considered very high output, and component selection had to be held to a higher standard.

MODELED RESPONSE

Modeled response curve on the tweeter axis, 2.5m

Final system response is smooth and extended on both ends with only a few minor issues. The midrange drivers are in MTM format with the tweeters and as such, the tweeter exhibits a slight cavity effect with a dip/peak combo between 3 and 4 kHz. The tweeter has a mild on-axis rise in the top octave that flattens out nicely a few degrees off axis. Anechoic sensitivity is around 91 or 92 dB with the default configuration. The design listening axis is the tweeter, and the design distance is 2.5 meters.

You can see that the wide bandwidth of the midrange helps to smooth out some issues in the other drivers. The woofer has a resonance at 700 Hz but it is down enough to not be an issue. Likewise, the mid's assymetric slightly shallower top end slope smoothes out some of the tweeter's MTM cavity resonance. Polar response is very good, with the only off axis nulls happening due to the midrange/midrange spacing. The vertical lobe is tall enough to not be an issue. Note that the 200, 400 and 600 lobes favor above the axis with nulls below the axis, a good condition for a floor standing tower to minimize floor bounce and lower midrange droop when standing up.

Transfer functions are smooth and predictable. The only reason the low ends of the tweeter and mids have this performance is because of the LRC traps to flatten the impedance peaks. On the mids, I've provided a curve showing what happens without the trap. The trap includes a 7mH inductor, so you'd bet I'd have a very good reason for including such a monster.

MEASURED RESPONSE

Measured response, 2.5 meters gated

This is always a check to make sure the model matches up with reality. This was actually taken at 2.5 meters and was gated to remove the majority of room effects that show up in the bass and lower midrange. I did have to use my ears to make sure I had the relative levels correct between the mids and the woofers.

Nothing too special about the raw in-box measurements except that the woofer was taken with a ground plane measurement, with the entire speaker tilted forward to allow a realistic on axis response. Even then, I had to take it at several different distances and merge them together. Be assured that getting a realistic anechoic curve for a large woofer in a large 3-way is very difficult in anything less than an anechoic chamber. This is probably the single design element that scares people away from large 3-way designs, and for good reason - it's easy to screw up. Or alternatively, it's hard to design a filter for a driver that has room reflections right at the crossover point. You end up with a system that doesn't sound quite right if you try to bring up a floor reflection dip or flatten a reflection induced peak. Note that the mid only has a little baffle step droop and the woofer doesn't have any at all. The woofer's proximity to the floor and natural inductance rolloff counter the minor low end droop from baffle step that this system has. The result is an effectively flat response.

The off axis plots show that it doesn't matter too much if the tweeters are positioned to the inside or the outside. There are some mild differences related to diffraction changes. I personally favored tweeters and mids to the inside. Due to the width of the baffle, these off axis curves had to be taken at the design distance of 2.5 meters for accuracy.

System harmonic distortion is excellent, no issues at all and in fact no sudden transitions in the harmonic levels across the whole bandwidth. Note that this HD plot is not directly comparable to others in my driver tests because of the far field mic postion. It is done only to show trends in distortion and to make sure drivers smoothly transition out of their operating ranges.

System impendance is relatively flat. I had to do a zoomed in curve to show detail. The minimum impedance is 4.3 ohms at 70 Hz, and hovers between 5 and 6 ohms over the rest of the range. This is pretty good for a system that has two 8 ohm mids in parallel. It's a lot more amp-friendly than the average 4-ohm system that often droops below 3 ohms.

OPTIONS

Nothing too special about the midrange and tweeter level adjustments. It's done by changing out a single resistor value, though these are paralelled pairs in this case. This can be done as a fix for batch level variation, or to simply get the system closer to your preference for tonal balance.

The reduced baffle step compensation option brings the midrange and tweeter level up quite a bit, running them full blast without padding resistors. If you need to do this you can, however your amplifier better have some high current capability because the impedance gets quite low in this case. This can barely be considered a baffle step adjustment, given the type of system. It's more of a tonal balance adjustment. The system is 94 dB sensitive at 2.83 volts in this configuration.

The drivers are offset for a couple of reasons, first is for a favorably smooth on-axis response, and second is for simplicity of the midrange enclosure which is built onto the side. The tweeter and midranges can be centered, however the on axis response is no longer as smooth, and the midrange enclosure will require more work and material to build 2 sided in the middle. If you do wish to build centered, make sure the midrange enclosure has exactly the same volume, and face the speakers straight forward to listen about 20 degrees off axis. The ripple response finally flattens out at 20 degrees.

ENCLOSURE

The first thing to mention about this enclosure is that the midrange chamber is critical. It must be that exact volume, and it must be sealed completely airtight. If it's not, the impedance flattening circuit may not work as expected. Additionally, sound pressure from the woofer back wave may affect the mids without a good seal. Test it by slowly and gently pressing in both the mids at once. Hold them in for a few seconds. Let go and they should very gradually move back out. Additionally, manually press the woofer in. The mids should not move out in response. Issues like this could always be a problem in multi-chamber designs, but it becomes more of an issue when you need to nail the midrange impedance peak with a trap.

In case it's not obvious, build mirror image pairs. Aside from that, there is some flexibility. You may build shorter and deeper if you want, as long as the baffle width does not change and the internal volume does not decrease. Remember that the design axis is the tweeter. If you like a leaner and deeper bass, you could increase the volume by 1-2 cubic feet to reduce the effective enclosure Qtc. You'd get a little less in the 50-100 Hz range and a little more in the 20-30 Hz range.

POWER HANDLING

Woofer power handling, 1-256 watts power doubling

This system is ready to handle just about anything you could dump into it, full range without a subwoofer. A 50 watt amp could reach very high output levels, but up to a 250 watt per channel amp could be applicable if the system is used in a very large room or an outrageous sound pressure level is required. Please, value your hearing - that might be fun for a few minutes but more than that could damage your ears. If you use a 250 watt amp and your primary music content includes pipe organs or cannon shots, you might want to keep a finger on that volume control.

Power handling of the mid deserved a quick look. With the filter in place, there is no real worry of running out of excursion. The above plot is for a single woofer. Excursion is decreased even more with the two drivers used in the system, keeping harmonic and intermodulation distortion very low through the critical midrange.

ROOM PLACEMENT

It probably goes without saying, but I'll say it anyway - This system is designed with large rooms in mind. Ideally, the baffles should be at least 4 feet out from the back wall, and 5-6 feet from the side wall. This is not the system to use if you have a 12 foot room with a 60" TV in it.

The speakers do not need to be toed in much, if at all. Anything from 0-5 degrees would be good. I had slightly off axis in mind, considering the tweeter's rise in the top octave. If for some reason you have to go closer to the side walls than the back wall, such as the speakers on the short end of a long room, I recommend toeing the speakers way in to a focal point a couple feet in front of the listener.

SUMMARY

This isn't the largest speaker I've ever designed, but it's easily the best sounding speaker of this size. Some builders may be tempted to pick up some cheaper large diameter drivers for the sole purpose of building something big. This design keeps the emphasis on quality however. Along with the size comes a well balanced 3-way design with high end drivers and clean, smooth, wide band operation. This is the kind of system that won't lose it's composure as you crank it up - dynamics will not compress as much and distortion will remain low. If you've got the room and you would prefer a large, high end 2 channel solution, this could be the system.

Enjoy!

I don't understand how you arrived at system sensitivity, can you go over that?
With a normal smallish 2-way, system sensitivity is generally going to be the woofer anechoic 2.83 V level minus baffle step compensation at the average level through the midrange decade. That simplistic equation begins to not work well when using wider baffles with the bass driver closer to the floor. In that case, boundary reinforcement ends up having a significant effect on sensitivity, and the effect varies per frequency depending on the baffle and driver layout. That is why the best way to measure woofers in a very large 3-way system is with the ground plane method. It then includes the floor plane but without the floor reflections right in the midbass that make crossover design difficult and error prone. Including the floor reflection dip/peak combo in the design often causes a designer to compensate for it in the crossover, generally a bad tendency that will cause audibly non-flat direct response.

This section also deserves clarification on the SoundEasy decibel specification shown on the T/S parameter image. This woofer shows 88 dB there, however people often don't realize that this is an efficiency number based on one watt and is calculated rather than actual. This is a 4.2 ohm DCR woofer, so the 1 watt calculated number is quite a bit lower than the 2.83 V number. The SPL shown on the response graphs is a sensitivity curve based on 2.83 volts and is actual rather than calculated. So indeed this woofer measures pretty close to the manufacturer's specs. The ground plane and baffle width negates most of the baffle step rolloff and allows us a final system sensitivity (anechoic + ground plane) of 92 dB with 2.83 V at one meter.

Why choose MTM format over TMM?
TMM designs are typically used for 2.5 systems where the lower woofer is rolled off to compensate for baffle step loss. With this system however, the baffle is wide enough that most of the low end droop happens below the bandwidth that the drivers are used. As such, the system does not have much baffle step compensation through the midrange. The lower woofer in the TMM would need to be run within the same range as the upper, making it a 2-way mid-tweeter configuration with an effective tweeter to woofer center distance that causes asymmetrical and vertically narrow lobing problems at higher frequencies. MTM does have it's own limitations based on woofer center to center distance, but in this system those trade offs were preferred over a TMM. The trade offs could be easily seen in vertical polar responses I did for each arrangement.

Why choose two 8 ohm mids over one 4 ohm mid?
At the time of design, my selection of SB mids was the SB15NRXC30-8, SB17NRXC35-4 or SB17NRXC35-8. The sensitivity of the 4 ohm SB17 was not enough - I needed even more than the woofer's in-room sensitivity to allow for a range of tweaking and also allow for a bit of series resistance to help get the system impedance out of the gutter. A pair of paralleled 8 ohm SB17's would have done it, but I could reach my needed sensitivity with a pair of 8 ohm SB15's and then benefit from the smaller woofer center to center distance. That in turn allows a higher and more conservative tweeter crossover point. These days people are pushing crossover points very low, and pushing the limits of what a tweeter can handle, but a system like this could potentially be played very loud because there is minimal excursion limitation in the bass range. It's all about a proper balance of power handling - it would not be a good design if one element ran out of power handling long before other elements begin to show strain.

I notice the woofer effectively uses only an inductor for low pass. Would a full 2nd order electrical work better?
Not really, that would give me too steep of a slope. Note that the filter transfer function starts at 6 db for an octave but then transitions to slightly steeper. That's why I called it a "quasi zobel". The final filter's acoustic rolloff is somewhere between first and second order to reach the mild asymmetry needed to compensate for the acoustic center setback in the woofer.

For the midranges, would a full 2nd order electrical high pass filter negate the need for the large impedance trap?
The midrange's monster impedance peak is likely going to be a problem on any system with less than a 4th order electrical high pass at 500 Hz. (guessing, I haven't looked into it in detail) The woofer limitations needed me to cross much lower than that. This is an issue in large 3-ways that designers often neglect, and probably part of the reason so many designers fear 3-ways or fail to get them just right. If the peak were let go, the system would have an audible honk at the midrange's resonance. If the woofer were crossed too high, it's inherent limitation would cause the midbass to sound like mud.

Why don't the SB15's show the same lower treble peak/dip as seen on the manufacturer datasheet?
The SB15 still has a resonance at about 1700 Hz, but it's not bad and somewhat hidden by the diffraction ripple response. It's small enough to be a non-issue and it does not propagate into the non-linear distortion.

Are there concerns about saturation on a cored coil for the woofer low pass? (L22, 4.0 mH)
Not at all. The Madisound Sledgehammer steel laminate cores are power handling monsters. There has always been a stigma about cored inductors but I feel it is generally unfounded. If anyone does indeed have a concern then feel free to pony up a lot of extra money for an inductor upgrade.

I've never heard of ground plane measurement before, what is it?
Very simply, the mic is placed on the floor. This removes the floor bounce reflection and allows accuracy much lower in frequency. The distance to the side walls become the limiting factor. This method works best on a hard floor, but good results can also be had on carpet if you place the mic on a large board. The only ground plane measurement in this design was the lower half of the woofer response, which was merged to a 1 meter on axis curve of the woofer. This combines the frequency ranges where each type of measurement is the most accurate. The mids and tweeter were measured with more conventional techniques.

I understand these are large speakers for large rooms, but doesn't 5-6 feet from a side wall seem extreme?
Not really. The only thing that is extreme is what people want to do without having enough space. My main test room for these was 16 by 18 feet. That's on the smaller side for what these speakers need, but the room opens up into another room with 2 large doorways. One thing about large speakers that often make them require a large room is their bass extension. Smaller speakers can often be designed to roll off enough that there is significantly less output in the range of the primary room node. Large speakers often have response that goes very low before it rolls off. This extended response often emphasizes the primary room node.

The other thing to realize is that large speakers are suited for large rooms because they have high output potential. In a small room where you are listening relatively close to the speakers, there's really no need for a system with 120+ dB output potential. Be realistic about your output needs and be safe about your chosen listening levels. My 12x12x8 "man cave" doesn't need much more than a pair of 5" 2-ways.

All that said, there's tricks to get bigger speakers to work well in smaller rooms. A small room's pressure nodes will be greatly reduced if the room opens into another room. The difference in the low end response between a small room that is sealed compared with having a door open is often shocking. If you have to go closer to the side wall, you can compensate by going further from the back wall. Most of my designs, this one included, have tonal balance adjustments to bring up the midrange and highs - thereby de-emphasizing a boomy low end in a small room. You can always build the enclosure much larger to go for a lower Qtc and it's associated low end rolloff, though I doubt that anyone would be willing to build this monster even bigger. But if you do, go deeper to add a few cubic feet, keeping the baffle the same. And finally, don't be afraid to use EQ, it's not evil. Especially Audessy and the like if your equipment has that functionality and a microphone.

About the only thing I would say that you truly shouldn't do is place these speakers in a corner. The mids and tweeter can't go up in level enough to compensate for that.

I'd like to build this system, but go active instead. Do you have any advice?
You're starting from scratch and you're on your own. But if you want to somewhat replicate this design, your crossover needs to be able to have total control over the filter slopes. You will need to need to accurately replicate the transfer functions from the passive crossover. That's assuming no delay offsets. If you delay the woofer by a time equivalent to about 8", the woofer will need a steeper rolloff and the mid probably a bit shallower. Going active does not really simplify things and you had better be able to measure the final response if you want a truly good sounding active system. You will never get it just right by ear.

Do you have a matching center and surround designs for HT use?
I was contracted by Madisound only for the design of the 3-way towers. There are no plans at this time to do any other designs. If you were to roll your own, a center that could match the output capability of the mains could be a problem anyway, particularly if you need it in a limited height to cram into your HT layout. But if you can forgo high output in the lowest octave of the mains, you could get close with a WMTMW using the SB23NRXS45-8 on the outside and the neo tweeter offset to allow a very close midrange spacing.

What if I wanted to use the MTM portion by itself as a surround speaker or sideways as a center channel?
In either situation, that is so far from optimal that I wouldn't recommend it at all. If you need a small surround speaker however, the DIY world is filled with smallish designs you could use.

Any recommendations for a subwoofer to go with these?
Are you serious? Two SB12.3's in a room is more output and low end than any sane person would need. But if you're not sane and you do seriously want subs, you could probably go with a couple 15's per side in an infinite baffle configuration, crossed over around 50 Hz with an active asymmetric LR4/LR2. Don't underestimate the low end potential of the SB12.3 however. The shallow 2nd order rolloff of the sealed box extends low end usefulness a lot lower than most vented designs. The ridiculous fight for the lowest anechoic F3 isn't going to help anyone much in a normal room. F10 is a little more important.

© Zaph Audio LLC, 11/18/2009

Also visit -Zaph|Audio-