The Poly Natalia: A High-Efficiency DIY Loudspeaker
By Dick Olsher

Update (May 25, 2003): The Poly Natalia design is by now over 10 years old, and by my current standards it represents a pretty conventional "old technology" three-way design. I have since the late 90s focused my research in new directions, including CBAE technology and full-range designs (e.g., the BassZilla). The Poly Natalia, as with some of  my other old kit designs, continues to be available online for historical reasons. I would certainly discourage prospective DIYers from building the P.N., as the original midrange and woofer have been discontinued, and besides, if you're looking for a high-efficiency design compatible with low-power tube amplification, you should seriously consider one of the BassZilla designs.

However, if you already own a pair of the P.N., you may be interested in the following alternative crossover network that I developed for the Samadhi Acoustics Natalia and whcih dates back to about 1997. It involves  a total redesign of the original network. I greatly prefer the sound of  this network to  that of the original. The objective was to strive for a totally musical sound, smooth and mellow. Edgless sound, but with the sort of imaging the P.N. is known for.

Update (December 2, 2001): Sometimes last year Audax replaced both the HM210Z0 8-inch woofer and the  HM130ZO 5-inch bass-midrange with upgraded models. As older inventory slowly sold off through distributors, it has become by now virtually impossible to obtain the older drivers used in the original project.

I have confirmed directly with Audax that the new model designations are as follows: the HM210Z12 for the 8-inch woofer and the HM130Z10 for the 5-inch woofer-mid. The changes are minor and involve mainly the voice coil and surround. Specifications may be  downloaded from the Audax website in France ( The design implications for the Poly Natalia bass cabinet are minor, and I offer below an updated bass alignment based on the HM210Z12. However, in the case of the bass-midrange, the new version (with an aluminum voice coil former) is a bit less sensitive by about 2 dB, so expect the  midrange to be slightly laid back relative to the original design.

Regrettably, I will be unable to test and publish an updated version of the P.N. based on  these new drivers. The following bass alignment is strictly based on computer simulations and has not been verified experimentally. However, these type of simulations (LEAP)  have proven to be very reliable in my experience.

Extended Bass Alignment for Poly Natalia Bass Cabinet using 1 pair of Audax HM210Z12 Woofers wired in parallel:

Internal Cabinet Volume: 4.2 cubic feet

 Box Resonance: 32 Hz

Half Bass Power Frequency: 32 Hz

Number of cylindrical Bass Reflex Vents: 2

Vent Diameter: 3.0 inch

Vent Length (measured from outside of baffle): 4 & 3/8 inches

The following is the text of the original article, dating back to 1993.

 Who is Poly Natalia, and what does she have to do with loudspeakers? Let me explain. The name represents a concatenation of sorts linking Polydax and Natalie. Poly (short for  Polydax) is my way of paying homage to  Audax/Polydax's contribution to this project: the High-Definition Aerogel (HD-A) drivers which are a major ingredient in the design. While the second part of the name, "Natalia" -- the Old Country pronunciation of  Natalie -- (in keeping with my custom of naming designs after family members) honors my mother-in-law.

 Poly Natalia was born because of single-ended, low-power tube amplifiers, and the special load requirements that these amps require. In 1993, when I was knee-deep in SE amps, I began to scramble around for suitable loudspeakers  to use in the evaluation. My search criteria included high-efficiency and a benign impedance magnitude. The need for  high efficiency loads is self evident when dealing with 10 watts or so. Just as important, however, is the  impedance the load presents to a SE amp.

 Because of its limited current drive, an elevated impedance magnitude is desirable. The old norm, circa 1950s, of 16 ohm nominal impedance would be ideal not just for SE but  for all tube amps. Unfortunately, 16-ohm drivers (with  the exception of the Lowthers) are no longer being manufactured, and in practice the need to use a pair of woofers in parallel to increase the bass sensitivity, further reduces the impedance magnitude. As I see it, the bottom  line, is a nominal load impedance of 4 ohms, without any significant dips lower than 4 ohms below 500 Hz. Nor can you afford to waste power by mismatching the amp/load. Connecting a 4-ohm load (such as the Poly Natalia) to the  8-ohm output  taps will cost you at least 25% (depending on the amp's internal source impedance) in lost power. In a situation like this, it is imperative to use 4-ohm taps.

 My search for suitable commercial candidates revealed the following status quo. Most audiophile two-way designs spec out out at a 6-ohm nominal impedance magnitude, but  typicaly fall in a sensitivity range of 86 to 88 dB. Most  three-ways offer more bass extension but rarely an increase in sensitivity. The only way to increase sensitivity is to  use horn loading or multiple drivers of ordinary efficiency. My experience with horn-loaded designs has been  disappointing: too many colorations negate their use as a reference  tool. While, on the other hand, the attempt to use multiple mids and tweeters to elevate the ultimate sensitivity of the ensemble runs into problems with  confused imaging and a ragged on-axis frequency response due to interference between drivers. Another problem with the latter approach is what to do about the bass response. It's not a simple matter to  jack up bass efficiency.  The obvious solution of using say a 15-inch woofer in a moderately sized enclosure typically sacrifices bass extension for efficiency. There are plenty of "pro sound" woofers sporting a 95 dB efficiency spec, but of  such high free-air resonance as to make usable response below 50 Hz impossible. And I did want to get sub 30 Hz response.

 So I decided to roll my own 93 dB sensitive 3-way design using an audiophile design strategy. Design criteria included a midband sensitivity of about 94 dB, usable bass response to 25 Hz, and clarity and imaging to rival the  best minimonitor sound money can buy.

Update 12/22/00

1. Over the years the Thiele/Small parameters for the Audax woofer (Model HM210ZO) have drifted sufficiently from the values used originally in 1993, that it makes sense to revisit the bass alignment once again.

(a) I am curently recommending the following bass cabinet design for those desiring a  true extended bass-shelf alignment (EBS). Such an alignment is best suited for medium to small listening rooms where excessive midbass response may be a problem. The EBS  alignemnt achieves lower bass extension at the cost of reduced response in the midbass. Howver, typical room gain below 150 Hz usually fills in the missing bass at the listening  seat. Simulations using the LEAP computer program show that a 4.0 cubic foot bass enclosure gives an ideal EBS frequency response. Note that this is the physical internal volume of the box - no allowance needs to be made for the displacement of the drivers and the vent.  To achieve the required box tuning of 35 Hz using the acoustic stuffing described below, the vent dimensions are: 4.0 inches diameter by 2.25 inches long.

(b) If your listening room is larger than 20 x 20 x 8 feet, or if you simply prefer a fuller bass  balance, consider the following somewhat smaller bass enclosure:  3.2 cubic foot internal physical volume, with a 4-inch diameter vent, 3.25 inches long.

 (c) I have received numerous inquiries about the type and manner in which the bass enclosures are to be filled with acoustic stuffing. The most afforadable material that is  suitable for the job is Dacron polyester fiber fill (i.e., pillow stuffing). Fill the entire interior volume of the bass cabinet (but do not fill the bass reflex vent) to a density of about 1.3  pounds per cubic foot. Method II involves the use of acoustic-grade foam. An acceptable low-cost foam substitute is an "egg crate" foam panel. Cut the foam into small sections, on the order of 6 x 6 inches, and loosely fill the entire interior volume of the bass enclosure. Again, avoid filling the bass-reflex vent.

 (d) The Mid/Tweeter enclosure should also be filled in the manner specified in (c) above. Because the HDA Aerogel midrange (HM130Z0) cone is so thin, it is critical to tame standing wave resonances by filling the entire volume of the enclosure. This is much more effective than simply lining the interior walls with damping material. Another possibility,  suggested by Richard Westenskow, is to replace the closed box with an open-back box or a dipole panel. The benefit of such a configuration is the elimination of cavity  resonances (there is no box!) as well as enhanced midrange ambience. A 2 x 2 foot panel may work well, however, while I have experimented with such an approach in the context of the upcoming BassZilla loudspeaker kit, I have not tried it with the Poly Natalia. Some adjustment of the crossover network may be required to optimize a dipole  midrange. The cost of experimenting along these lines is pretty minimal; hence if you're so inclined, give it a try. I would be glad to entertain any feedback along these lines.

 (e) Many of the questions I receive deal with various proposed modifications to the cabinet geometry. Let me make it clear that as long as the basic interior volumes and  driver spacing along the front panel are adhered to, the shape of the enclosures may be modified to suit individual tastes. For example, If I had to do it over again, I would opt for a single exterior box with an separate upper compartment for the midrange and tweeter. Some of the questions have to do with driver substitutions, most notably alternative tweeters. It is no trivial matter to redesign a loudspeaker to accept a new tweeter. Differences in efficiency, resonant frequency, and harmonic distortion, can result in major  redesign issues. If I am to keep my sanity, I cannot undertake to redesign the Poly Natalia  to accommodate new and wonderful tweeters. Regrettably, such questions and requests to the Audio Mecca will go unheeded.

Driver Selection

Remember the original Dahlia DIY design? Having had good experience with Audax's TPX woofer, I was naturally intrigued by the brand new HD-A driver technology. My  philosophy is to use the best drivers money can buy. It is well known that the residual sonic signature of a particular loudspeaker is determined by the in-band micro   resonances of each of the drivers. So that no matter how cleverly one synthesizes a particular driver complement, being careful to eliminate gross out of band break-up  modes, in the end you're left with the essential flavor of each of the drivers. The ideal driver cone should be lightweight, rigid, and internally well-damped. Damping is a  critical factor, and one that draws some designers back to the old-fashioned paper pulp cone.

Audax's HD-A cone is fabricated as a matrix of acrylic polymer gel in which carbon and Kevlar fibers are embedded using a proprietary process that aligns the fibers along  the polymer chain. The fabrication process allows total control over the contour and weight of the cone and makes it possible to vary the thickness of the cone along its  profile. The end result is a cone that approximates the ideal more closely than anything Audax has  done before. The HD-A series also features a non-resonant, die- cast frame,  an edgewound, flat copper wire voice coil on a kapton former, and a high-loss rubber surround.

 The fact that both a 5.25-inch midrange (HM130ZO) and an 8- inch woofer (HM210ZO) are available, made it possible to reproduce the core of the music using  identical cone materials. To my mind, it's a mistake to transition from a woofer to a midrange using disparate materials such as plastic and metal. As the voicing of each  driver is sufficiently  different, the character of the reproduction will be audibly dissimilar between the lower and upper mids. And the illusion of the speaker speaking with one voice will suffer.

 The tweeter selection process was much more difficult. I went through a slow process of evaluating a grand total of four tweeters before settling on the eventual winner. To  begin  with, I was reluctant to use metal domes, because most of them were getting on my nerves. What I find to be quite appalling is the recent trend toward cheap aluminum  domes as a tweeter of choice in moderate or even expensive systems. Sure, treble transient detailing is improved over the plastic and fabric variety, but at the cost of  increased long-term listener fatigue. Having said all that, I was still persuaded to try the new Focal T125TD, an improved version of the inverted titanium dome tweeter I liked  so  much in conjunction with ORCA's Aria 5 loudspeaker kit. This is a great tweeter, but it did not integrate with the Audax midrange as well as the eventual winner.

 The Poly Natalia started life with the Audax TWO34XO, a 1.25- inch textile dome of good power handling. Despite the fact that it has almost no output above 15 kHz, I was  attracted to it because of its basically natural sound, supported by a clean looking MLSSA waterfall plot. For many weeks, the Audax proved to be a competent tweeter.  Its two main failings, however, wore me down in the long run. It ever so slightly smeared transient detail, obscuring the decay of musical nuances. But more importantly,  a resonance in the upper midrange would flare up when excited by the program material to cast an ugly "neon" glare over the music.

 Next came the Cabasse DOM 4 with its unique construction of carbon-loaded thermoset plastic. The Cabasse worked very well and would have stayed around,  except for a  chance encounter with the the Morel MDT33, hand-treated, soft dome tweeter at the 94 Winter CES. I was so impressed by its wide dynamics coupled with a  most natural voicing, that I immediately ordered a pair for evaluation. An attractive feature for the home constructor is that the MDT33 is sold in matched pairs, so that  response variations between tweeters should not exceed 0.5 dB. Yes, the the Morel did find a home in the Poly Natalia, as it blended very nicely with the HD-A midrange to  extend the remarkable resolution and clarity of the latter into the treble.

Bass Alignment

The bass enclosure was designed with the aid of LinearX Systems' well known  Loudspeaker Measurement System (LMS) and Loudspeaker Enclosure Analysis  Program (LEAP). Because SE amps, especially the feeback-less kind, are notorious for their high internal source impedance and poor damping factors, I opted for an Extended  Bass Shelf Alignment (EBS). The EBS, an inherently over-damped bass-reflex alignment,  offers good bass extension, at the cost of a leaner balance. The bass is  shelved about 2 dB below 200 Hz, but extends to 25 Hz. The basic strategy was to allow the the SE amp  to flesh out the mid and upper bass, without bloating the  mid/upper bass region, thus retaining good pitch definition. You should know that final voicing of the bass registers  was performed with the assistance of Cary Audio Design's  300SE monoblocks. With low-damping factor amps, the bass actually sounds tonally full. On the other hand, with  solid-state amps, the balance is typically on the lean side of reality.

Enclosure Construction

My pair of Poly Natalias is built up in modular fashion using two enclosures: a bass enclosure houses a pair of 8-inch woofers, while the midrange and tweeter are housed  in a smaller enclosure which sits on top of the bass box. The basic idea is to isolate the  midrange from the back pressure of the woofers to avoid acoustical intermodulation  between drivers. Please note that the same objective could be achieved in a  dual-chambered enclosure where the upper compartment is completely sealed from the bottom chamber.

Cabinet details are shown in Figure 1 & Figure 2. The internal volume of the bass box is   2.8 cubic feet (internal dimensions: 26.75"H x 8.75"W x 21"D) with a 4-inch DIA by 8-inch long vent cut from PVC pipe. The box is stuffed to about an 80% fill with  sections of 1"-thick acoustical grade foam. The resultant box tuning is 25 Hz. Should you desire a slightly fuller balance (but with a bit less deep bass extension), try an  internal volume of  2.6 cubic feet with 95% fill and the same vent as before, which gives a box tuning of 30 Hz. Comrade Ron Cox, who built my cabinets and provided the  construction notes, has built Poly Natalia bass cabinets as small as 2.4 cubic feet, which did in fact sound fuller in the bass but not as detailed. So the ultimate choice is yours.

 If you've got even minimal woodworking skills and experience, this should prove to be a fairly easy construction project. Two sheets of 1&1/8" particle board are required for  the  bass enclosures. Panel dimensions are as follows:

6 ea.  8 3/4" x 29" (Front & Back)

6 ea.  8 3/4" x 21" (Top & Bottom)

4 ea.  23.25" x 29" (sides)

 2 ea.  8 3/4" x 2" x 4" (side wall stiffeners)

The front baffle was built up to a thickness of 2.25" by gluing two sections together. A single 2" x 4" stiffener connects the side walls, and is centered on the side panels. It is  glued on the inside and screwed in place from the outside. All joints are butt joints which  are glued and screwed using #6 drywall screws about 2" long. The woofer  cutouts were routed with a router and trammel. It is not necessary to recess the woofer frame for a flush mount.

 The midrange/tweeter enclosures were made from a single sheet of 3/4" particle board (MDF). Internal volume is 0.43 cubic feet, and the enclosure is filled with sections of  acoustical grade foam. The front baffle is slanted at an angle of 25 degrees to time-align the drivers and also to partially radiate the extreme treble toward the ceiling. Such  reflections increase the perceived image height perspective and spaciousness of image outlines. Both the midrange and tweeter need to be recessed onto the front baffle to minimize diffraction effects.

 The drivers were secured with socket-head machine screws and T-nuts, except for the tweeters where wood screws were used. A bead of Mortite ® was used to ensure an air-tight seal between drivers and front baffle.

The Crossover Network

The network schematic is shown in Figure 3. It represents the fruits of an extensive computer-aided design effort using XOPT and LEAP, supplemented as usual by a   tweaking process based on lengthy listening sessions in the reference room.

Note the reversed polarity connection of the midrange. This is necessary to  accommodate the second order electrical filters. No, I haven't given up on Linkwitz-Riley fourth-order networks. There's a time and a place for network  complexity. However, the  nature of these drivers and choice of crossover points (600 Hz and 4 kHz) made possible a much simpler yet elegant approach. The smooth upper  range behavior and speed of the woofers allowed the crossover frequency to be pushed an octave higher than normal and made possible a gentler roll-in for the midrange. The  midrange is allowed to go practically full out. That proved to be an important design parameter. Midrange transparency, already a highlight of the HD-A midrange was thus  allowed full expression.  Alternate networks, although their response looked almost identical on the computer screen, did not allow the midband to breathe in as dramatic a fashion.

 All of my prototype networks were mounted externally to the enclosure, and I recommend this approach for optimum isolation of the network form the harsh  acoustical environment  inside a speaker enclosure. Selection of passive parts is critical to optimizing sound quality. A "parts is parts" mentality will greatly reduce the ultimate  potential of this  speaker. Use only air-core coils and high-quality caps. Please, no electrolytics! The series cap in the tweeter network is especially important, and should  be as a minimum a good-grade metallized polypropylene bypassed with a small polystyrene.

Better yet, give the folks at Sonic Frontiers' The Parts Connection a chance. They now offer a fully assembled crossover board for the Poly Natalia. In fact, the board is  available in two versions. The standard version ($195/pr. parts only) uses Solen metallized poly caps and Perfect Lay air-core inductors. The premium version ($345/pr.  parts only) includes MIT and Solen caps and SOLO's outstanding copper foil air-core inductors. I had a chance to try out Sonic Frontiers' premium networks, and found them  to significantly outdistance my original prototype boards in the areas of soundstage transparency and low-level detail resolution.

 Sound Quality

Because this is a rather expensive kit, which will require a substantial investment of resources and time, I feel that you're entitled to my albeit biased view concerning the  Poly  Natalia's sonic merits. It is one of my all-time favorite dynamic loudspeakers, and one to which I shall return often for critical listening. It is especially gratifying to me that  non-audiophile listeners are as enthused about its sound as much as I am. That for me is the ultimate reality check. These ordinary folks have an intuitive feel for what's hot and   what's not. And when Poly Natalia moves them as much as she does me, I know that this girl is a genuine sonic babe.

She has proven compatible with every SE amp I've been able to mate with her so far --  providing that she's fed from the amp's 4-ohm taps; after all, it's important to squeeze  every watt possible from a 10 watt package. In every case, the amp's unique character  shone through. From the sweetness of Bel Canto Design's Orfeo or that of a Cary  300SE, to the vivid portrayal of space that a Cary 805 is capable of -- this lady tells all. And she does so with a stunning level of midband emotional intensity. Bass lines, as with  the Cary 300SE, were tightly defined. No loss of definition here. Who said that SE amps  have poor bass resolution! For now, this is my reference loudspeaker for SE applications.

A Final Note: This DIY design is presented in the public domain for the benefit of the home constructor. It is not protected in any way and is offered license and  royalty-free for unrestricted use. Because the design may be used by anyone, including a commercial manufacturer, versions of this design may be offered for  sale in the marketplace. In fact, Swans Speaker Systems (RR #1 Mason Road, Charlottetown, Prince Edward Island,  Canada C1A 7J6, Tel.: 902-569-5520, FAX: 902-569-5123) liked the Poly Natalia so much that they now offer a fully  assembled version of this design as the "Allure" Loudspeaker for around $5,500/pr. Dick Olsher receives no royalties or other remuneration from Swans Speaker Systems relating to the Allure.

Sources For Parts

 Drivers, etc.:     Madisound Speaker Components

                              P.O. Box 44283
                             Madison, WI 53744-4283
                             Tel.: (608)831-3433
                             Fax: (608)831-3771

                           ZALYTRON INDUSTRIES CORP.

                               469 JERICHO TURNPIKE, MINEOLA, NY 11501
                             TEL (516) 747-3515
                              FAX (516) 294-1943

 Your comments and experiences regarding this project are welcome.


[Dick Olsher's Audio Mecca Home Page]