Nostalgia for a certain style of design innovation - but ruthlessly simplified

[Evil Twin]

The Kii Three has always, in my perception, been emblematic of the most clever technique and thoughtfulness of Bruno Putzeys.

Impressive... most impressive.





Like the Dutch & Dutch 8C, it proposes to develop a cardioid in room response profile,

Yet, at the same time, I must offer the opinion that it may suffer in some regards from the same level of complexity and tradeoffs inherent in this approach as implemented by the D&D 8C.

This introduction from the Six Moons review can convey key aspects of the configuration, as well as some concerns, yet still miss some more subtle yet important issues that were addressed successfully in early designs such as the Natalie P, which cannot be addressed by this approach.

Wave shaping tech is key to Kii. This fully DSP-controlled active 6-driver compact breaks the rules. Its bass is actually directional. Cardioid aka heart-shaped dispersion is the technical term for its propagation shape. The upshot? Unlike conventional speakers, the Kii Three wants to be more blasé about room interference. It directs most of its sound at the listener, eliminating or at least attenuating side and especially rear radiation with strategic phase-canceling chicanery. This downplays usual room dependencies. It performs superior in far more real-world situations including very close to the wall. For a perfect step response of what is a three-way—reference this for the actual graph—its designers use minor time delay on their frontal drivers. This lets the side and rear woofers catch up.

That trick plus claimed 20Hz response can only do the work because filtering and compensation occur in the digital domain. Analog inputs must convert. By the same token, LF headroom must be lower than mid and high freqs because, to keep its box small, the Kii Three gooses bass gain on its built-in amps to offset natural roll-off. In payment for this, it abolishes your own amplification and D/A conversion. Its functional very high IQ relies on the designers controlling all the parameters, not letting us techno peasants piecemeal it. This eliminates the lego and puzzle crowd who insist on arbitrary system assemblage. Au revoir, amigos. Those who—gasp!—are still with us have notions:

• that Putzeys & Partners also exploit DSP to improve their matching tolerances between drivers
• that they use DSP to enable their 15-click 'contour' knob with its preset EQ curves
• that they use DSP for their 16-click 'boundary' trim pot which adjusts bass compensation for free field, wall and corner placements (the latter two require consecutively less compensation due to boundary gain)
• that their current and voltage 'motion-control' loops are somehow connected to the far earlier motional feedback as mentioned in the 40-year old Philips ad at right.
• that home-theatre users might have latency issues of syncing dialogue to on-screen action given the speaker's digital delay and intensive processing1
• wonder about out-of-standby delays2

The overall result from this level of complexity hints at a design group quite strong in the Force- how much is Dark Side and how much otherwise would be a matter of personal assessment and judgement. But, as with Dutch and Dutch, I would urge caution- "Don't be too proud of this technological terror you've constructed...."

For example, are your DACs comparable to a standalone product such as a TotalDAC D1? A Benchmark DAC3? Or even a Soekris 2541?

And what standard of performance is your ADC? And your master clock?

An interesting technical report is available from Audio Xpress, but curiously, it does not address measured distortion...

And, let's consider driver resonances... one not inconsiderable flaw regarding DSP notch filters is that they have no impact on the generator impedance of the driver at those frequencies, and as a result do NOT reduce resonant amplification of distortion products- a factor taken into account for the Natalie P, and other subsequent designs, but not addressable by DSP. Of course, Bruno's newest association with PuriFi has even relatively recently published an Application note addressing this very same topic.



Given this driver set, and previous design techniques as implemented as long ago as the Natalie P, and as recently as the passive voicing module, is an alternative solution worthy of pursuit? And can it be made worthy of audition as well as detailed technical reporting?

Interesting questions to ponder....

But with some drivers in hand, and others on the way, this will no longer be a purely theoretical question.

One might justifiably advise to be careful not to choke on one's aspirations in this endeavor...

But I have cause to believe that the results will be interesting, most interesting.

To naysayers, I would only offer that I find your lack of faith disturbing...

[Evil Twin]

A few points worth noting for those who may be interested... however few... as for a variety of reasons, this might justifiably be considered a relatively esoteric endeavor, with limited opportunities to duplicate.

• This is being approached as a "value" build, with a focus on the best results possible with a minimal budget, PROVIDED that that end user is willing to make reasonable efforts for room setup and ancillary equipment. As an example, the target development amplifier will be a reasonably priced one, a Denafrips Hestia, not a high current stratospherically priced set of mono blocks. Though, some are available for comparisons.
• The cabinet configuration currently being considered is a single tower, constructed primarily of caramelized bamboo ply and baltic birch ply. The target width is 8-1/2", target height not including base of 42".
• At a minimum, it is recommended to consider the CARDAS placement guide, possibly coupled with the Wilson Audio approach for fine tuning.
• Considering the proximity of the Z axis origins of the woofer, midrange, and tweeter, a Duelund crossover approach is being considered with a center frequency of 800Hz and a bandwidth aleph value of 2.75.

An RMS average of the forward +30 to -30 degree laterally measured previously was used to evaluate a fast Proof Of Concept for the HF section, deemed to be more of a challenge. The result was reasonably satisfactory...




With the early arrival of drivers ordered, there will be some flexibility about starting measurements for the midrange and woofer....

[technodanvan]

Wait, did you start another project?

[Evil Twin]

As I write this measurements of the SBS-160 woofers are in progress.

This is a basic effort to create a simplified system using these "classic" drivers to ascertain what their native performance without active EQ, cardioid cancellation and efficiency losses might provide... and with possibility of a greatly simplified amplification chain.

One might say that this is a combination of nostalgia and curiosity...

[Evil Twin]

Basic measurements are complete for the SBS-160F35AL01. Two test conditions were used-

• A quasi-near field on a large baffle, to simulate the "normal" conditions that should be used in a "factory" measurement
• A sequence of polar measurements in a small baffle cabinet at normal distances, to evaluate polar response and behavior with baffle step

There is reasonable correlation for this measurement with the factory data sheet, except that this measurement with very little smoothing (1/24th) is free of a variety of artifacts that may be due to the acoustics of the Peerless test facility. It is perhaps curious and interesting that the net response curve and bump occurring above the Fs and Fb in the response bears a passing resemblance to the STW-350 - perhaps there is a common designer or design philosophy.

As mounted this way, the response would not be suitable for a typical three way or the proposed Duelund alignment, due to excessive early fall off above 100Hz.


For a "practical" application one must turn to a representative cabinet baffle dimension with measurements at 2M-




Two issues of note:

• the response profile above 100Hz is modified due to the baffle step, and has an initial roll of slope which can fit to the proposed crossover configuration
• With the fairly stiff (3kHz breakup) concave cone, the response tracking from 0 to 40 degrees shows nearly no divergence out to almost 2kHz, which is most satisfactory for the planned three way configuration.

So far, the proposed configuration seems feasible.

[Steve Manning]

Wait, did you start another project?

Danny ...... Jon is suffering from a serious condition called Butterfly Syndrome. Like when a real butterfly that floats by and catches our attention, with Jon it's shiny things like speaker drivers, guitars, routers, etc ..... he just can't help himself! 🤣🤣🤣

[Evil Twin]

Danny ...... Jon is suffering from a serious condition called Butterfly Syndrome. Like when a real butterfly that floats by and catches our attention, with Jon it's shiny things like speaker drivers, guitars, routers, etc ..... he just can't help himself! 🤣🤣🤣

An impressively insightful comment, though rather impertinent! I expect no less from you...

[Steve Manning]

That's Me

[Evil Twin]

Several concepts for the LF configuration have been under consideration and evaluation...

The most promising to date would give cause for titling this project as the Modula Xtreme MkII...

This combines the input tuning capacitor with the passive LF EQ implemented in the same manner as Jonmarsh's voicing filter, passively between preamp and power amplifier, and the first pass does a fairly competent approach to the Duelund LF target, with good bass extension to about 30Hz... a more conservative and realistic value than the Kii Three for 4 6-1/2" drivers. Not using two of them for cardioid shaping helps matters to an additional degree.

[Evil Twin]

It is intuitively clear that for a given LF performance to be modeled, a concomitant enclosure design must be specified first- or have been conceived in the mind of the designer. That has been the case, and these images will convey the fundamentals of the current concept design...








Material choices are relatively basic and focused on the concept of being a "value build", relatively speaking.

• Front and top panels are planned as 3/4" caramelized bamboo ply- materials ordered.
• Side walls are double 3/4" baltic birch ply- on hand in the Imperial storage
• Internal baffles are 3/4" baltic birch ply, pass through holes not yet implemented
• The inlay is black acetal copolymer, from Imperial stocks- Black is back.
• Bamboo panels are 8" wide, overall cabinet width is 11", not including a planned stand. Height of the main cabinet is 42", not including a planned stand.

[Evil Twin]

Incidentally, LF analysis with Unibox 4.08 indicates a single cabinet LF capability of 102 dB @ 30Hz, and 108 dB at 40Hz.

[Evil Twin]

A midrange test enclosure for data capture has been assembled, driver mounted, and data collected for the Peerless FSL-0512R01-08. The results seem quite satisfactory, and in some regards more encouraging than the factory data sheet.


Quasi near field on axis, captured to obtain data which should correlate with large test baffle results in a basically anechoic environment:





And a polar set on 10 degree increments at 2m...





As one would expect from a cone of this size, off axis behavior is consistent and tracks well up to 2kHz, then deviates progressively about that point with the measurement angle. The test enclosure has a width of 11" with 1/2" round overs.

This data is sufficient for crossover concept development, but measurements will be required in the target enclosure design before calculating component values to order.

[Evil Twin]

A basic Proof Of Concept crossover has been formulated using the data measured from test cabinets. The results are within the range of initial expectations, and based on that, proceeding with a project build has been approved by the Chancellor.



This is NOT a final design schematic, merely a POC demonstration.




The target Duelund implementation for this POC is as shown earlier in this topic.

[tktran]

What's the harmonic distortion performance like? Say 94dB @1m? Surely 86 would be a cakewalk...

[Evil Twin]

Individually the drivers look quite reasonable- but at "normal" test levels. But for example, keep in mind, the midrange driver delivers 92 dB with 2.83VRMS. In consideration of that, having HD3 in the -65 to -70 dB range at that SPL over it's operational range is most satisfactory for a "value" build. The DXT sensitivity is in a roughly similar range, but optimizing HD3 and HD2 requires a crossover in place to suppress resonance amplification. Even without that factor being implemented, HD3 has been measured for the raw driver in the -75 dB range between 2500 Hz and 8,000Hz. That should extend much higher with the planned crossover configuration.

The true measure will be with an assembled system and full crossovers.

[Evil Twin]

One other comment- my intuition says that it is long past the time that if Bruno Putzeys is still associated with Kii Audio, that they do a major redesign of their speakers and update to PuriFi drivers and Class D modules. However, that may not be as simple as it seems on the surface... and it could drastically reduce the profitability of the system given likely differences in "OEM wholesale" driver costs.

I mention this because it seems there is a vendor in HK with substantial stock on hand of the Peerless woofer, and volume US pricing is in the range of $20.

[Evil Twin]

A few have asked about the methodology for achieving the low frequency system response... and how the woofers look in "native mode".

In the target cabinet volume of 70L, which is not small for 4 low sensitivity long throw 6.5" woofers, the basic response curve is as shown above with measurements, and if the Duelund target crossover is included, the result is as one would expect...




What is very clear is the relatively high Q 2nd order roll off, caused by the low compliance suspension and relatively low cone mass, and due to system Q, response that is already down about 4dB at the Fb. This is not a system response which would be amenable to a simple first order voicing filter approach, and is why the reflexive mode of the majority of designers might be to reach for a DSP crossover.

A series LF capacitor or an LCR network can be used to modify the low frequency profile, and is useful with drivers within certain parameter ranges, such as employed with the ScanSpeak 10" woofers in the original Modula Xtreme. Adjusting the range of value can produce considerable variation in the resulting basic response curve.

In this case, a value was chosen to enable achieving the desired system response with a simple 5 dB first order filter.

Here is the response with the series cap, but without the passive LF EQ filter (that would be inserted between preamp and power amplifier).



Now one can begin to see how the available "simple" LF "equalization" can adjust the total system response.

[tktran]

The vendor in HK; unclear whether it is a scam website; or legitimate. No mention of it being an authorised distributor for Peerless/Tymphany.

[Evil Twin]

This mirrors my initial reaction to this site also.

[Evil Twin]

Progress step by step...

• The additional caramelized bamboo plywood has been shipped now...
• Further detail work on the front panel is completed, including correcting the location of the lowest woofer, and appending the woofer front panel inserts (acetal now on hand, driver threaded inserts, and stainless M4 screws scheduled to arrive today... (yes, the Imperium builds in Metric)
• POC crossover BOMs in preparation-
• Project specific tooling has been sourced and is in the Imperial workshop.

Detailing of the interior baffles/braces and rear panel will be underway as soon as I can locate a new Sith apprentice...

[Scareurpasenger]

A more reputable source? https://en.toutlehautparleur.com/ful...-5-9-inch.html

[Evil Twin]

THLP is one of my two main vendors in Europe, the other is HiFi Collective. The parts on hand were sourced from THLP, but it is believed they have limited stock which is most likely a buyout from a certain manufacturer...

[Evil Twin]

Kii's new effort would be an interesting concept to consider, except that a passive implementation would have net sensitivity on a part with a Wavecor Ardent...



Now, imagine this configured in a center channel arrangement, and attempting to drive it adequately from a typical home theater receiver who's power ratings are valid only with two channels driven out of 7; this would be an inverted example of the power of the Dark Side of the Force...

[Evil Twin]

An experienced Sith CAD apprentice has been "recruited", though not with the eager and compliant spirit I would usually prefer... however, in these times, needs must, and the results to date seem satisfactory....

Detailing has involved the addition of an inner baffle layer in the front, rear, and top panel baffles, as well as detailing the woofer area braces, and the upper cabinet edges. A rear entry panel for crossover mounting has also been provided.















The majority of crossover components have been ordered from a key European vendor, and a test layout for the tweeter crossover is in preparation, and soon, fabrication.

Progress is slow, measured, but significant, for a secondary priority endeavor.

[JonMarsh]

It is time to advance the effort of this design study, with limited but critical hardware evaluation. To that end, we commence the fabrication of an M-T assembly mimicking the upper section of this proposed system design, to a limited degree. This will also serve the purpose of evaluating a potential M-T combination for the CC Technological Terror.

The concept for this cabinet is a shallower version of the top part of the proposed system:





The construction chosen mimics the planned large cabinet fabrication- for the front panel with a 0.75" Bamboo panel laminated to 0.5" BB ply. Routing was accomplished next, on the larger piece of raw material (12" x 30"), laid out for fabricating two front panels. This allows clamping support for routing which the Final Cut size would make difficult.

After routing in a larger piece, the driver fit is checked, and as expected, found to be satisfactory.





After routing operations are complete, the midrange backside through hole is beveled, then the panels are cut to the final size for assembly.




Side panels are a special fabrication, for this demonstrator evaluation it was decided to use just 1" material instead of 1.5", but a substantial round over is still required.




Assembly operations were conducted using biscuit joiners and the Imperial standard epoxy, BSI 30 minute cure.







Note the use of parchment paper to avoid issues with glue drips and protect bench surfaces and fixtures.


While waiting for epoxy to cure in each assembly step, a layout proposal for the Duelund style high pass tweeter crossover was prepared, and the base board for that was glued up.

[JonMarsh]

Next is cleanup work- sanding, top and bottom round overs, wiring and driver mounting.

[JonMarsh]

Slow work takes time, but eventually it gets done!

[technodanvan]

Looking real nice Jon, are you going to make the bass (base) units as well?

[JonMarsh]

This is just a test box, to confirm previous results and evaluate the vertical window in the crossover.

I could turn these into a real speaker with some of the 6” SB woofers, like the aluminum ceramic I ones, and I did cut pieces to make two cabinets, but only assembling one right now. The wood is on hand for building the full size speaker, but it is a one piece tower, about 43” tall.

[JonMarsh]

Also, a scaled up version will be built, as a test for the PuriFi two way with PR’s for Steve for his neighbors. Not much bigger, but with the PTT6.5X04 woofers, the RSS210PR on the sides, and assuming things work out with measurements on the crossovers, regarding the notch filter behavior, probably DXT’s up tip- I have a nice looking LR3 design in VituixCAD for that systems concept.

[Evil Twin]

While I long for the days of reliable Imperial minions, with fast completion of projects, At least I must acknowledge that when slow work takes time, the resulting quality is most satisfactory... or at least, mostly satisfactory.





Additional POC work has been completed, with several alternative configurations balancing off efficiency and total sensitivity, but now matters are clearly at a point where updated hardware measurements are necessary before proceeding further. Still, the several POC designs to date for crossover configuration all confirm that the driver choices should make the desired end results achievable.

The measurements for the Mid frequency and High frequency components will be ready soon...

[Evil Twin]

It might be worth noting that in this project, more measurements with test cabinets are in the queue prior to construction of the target cabinet designs, even though materials for a pair of those is already in the Imperial holding facility in this local sector.

Materials for an LF array test have been ordered and are expected soon.

A key part of the design path is completion of another detailed examination and documentation of all aspects, and providing for the possibility of HF voicing adjustments. Specific component parameters may require adjustment pending final cabinet+Driver measurements, but the crossover concepts and expected targets have been documented using existing measurements and appear to be fully realizable.

The final target for LF passive EQ has been selected on this basis, and documented within VituixCAD:



This is to be implemented with a simple passive network inserted between preamplifier and power amplifier, and this passive network is what produced the response above. Of course, there are alternatives, such as using a plate amplifier with Linkwitz style shelving EQ and speaker input mode, but this was deemed more complex and expensive than desired for this configuration.



There is a nominally flat system response crossover candidate currently termed RC3...




And there is an HF voiced modification which can be implemented with one or two switches:



Note, the crossover configuration continues at this point to be based on Duelund concepts.

[Evil Twin]

The measurements and analysis of the MT test cabinet were completed today, with few surprises, other than slightly better measurement results, and the opportunity to reduce the number of components in some of the crossover filters.


For reference,


FSL-0512R01G-08 Quasi Near Field





27TBCD_GB-DXT Quasi Near Field




The composite stuffing approach for the Midrange driver now also utilizing denim insulation performed most satisfactorily. And in this cabinet configuration, the 2kHz dip seen in measurements made 4 years ago is not present.


Logistics made progress today also, with the reception of crossover components from the European sector.


Additional orders were placed, including for materials for an LF POC evaluation build, before constructing the planned enclosures.

[Evil Twin]

Potential components for the LF Passive EQ module have been identified from Neutrik components and ordered. This is their "D" module system, and would use a Female XLR at the input end and Male XLR at the output, with a module board slotted in between the sides.

[JonMarsh]

A basic Proof Of Concept crossover has been formulated using the data measured from test cabinets. The results are within the range of initial expectations, and based on that, proceeding with a project build has been approved by the Chancellor.



This is NOT a final design schematic, merely a POC demonstration.




The target Duelund implementation for this POC is as shown earlier in this topic.

How's that old saying go?

"Once I was the learner, but now I am the master!"




Well, tables are turned.


well, decided to work through the same design issues of this project with a new structured design process I've been kicking around in the back of my head, and after applying it just to the midrange of this system, I can comfortably say that at least the midrange crossover in this design will be significantly revamped.




Remember there may only be historical value to watching the sausage get ground... until you get served the final sausage patty, anything can happen.


But I'll also mention the impedance curve is nicer now, too!

[tktran]

Have you taken any HD measurements of the SBS-160F35AL? Can it compete with SEAS L16?

[JonMarsh]

Yup. Possibly on both accounts.





I'd rather it had copper in the gap and didn't have the rise between 300Hz and 1kHz, but it's still under -50 dB. Unlike, say, an Epique E180HE-44, which increases to -30 dB, by my measurements under the same conditions.

[JonMarsh]

OF course, it's probably not fair to mention that there's a 10:1 difference in the price of the Seas versus the FSL-05120R01.

BUT, I hadn't had my coffee yet, and I just re-read and you're interested in the SBS160!!
Oopsie!

[JonMarsh]

Here's the SBS160F35AL01-04... QNF SPL and HD2 + HD3.

Were I to quip a brief joke about this part, it's that, "This is not your father's 6-1/2" mid woofer"! That is to say, it's a bit of a quirky design, not intended for a conventional 6-1/2" two way, (Fs~50Hz, Xmax = 10.5mm, significant roll off slope above 100Hz that is modified by baffle step to be a bit flatter out to ~150Hz) but my intuition says it's a potentially useful part. I say potentially, because all I have so far besides base measurement is design SIM's.

Another project (CC with Coax, for the Chancellor) is at the head of the queue for assembly and testing, though all the crossover parts for this project are on hand, and board layouts done.





The L16 seems to be a well executed more "conventional driver, (more a proper "your father's 6-1/2" woofer) with a 32 Hz Fs (instead of 49 Hz for the SBS160), 7 mm of Xmax, and a smooth response not quite extended enough for a "normal" two way, but possibly a waveguide setup? And higher sensitivity. And with the T/S parameters, well suited to ported or PR configurations.


For the Peerless, HD2 is higher than I'd like to see, but HD3 seems well controlled. Overall behavior is fairly easy to work with, and the price is right- less than 1/4 of the L16 locally. Which is part of what suggests to me using a 4 driver series/parallel array, with it's partial line source giving benefits with regards to minimizing floor bounces irregularities in the frequency response that a single driver creates in a typical room with typical placement. And the partial line source giving the same impedance as one driver, but with a 6 dB boost in sensitivity.

I'm sure that configuration could be effective in the right size enclosure for the L16, but the price would be a bit dear...

This SBS160 response profiles fits well within the desired concept of a Duelund with a center frequency of 800Hz, and an Aleph of 2.75. I do prefer the inverted edge suspension, and it certainly seems to keep suspension contributions in the midrange at a low level.

But then, it's quirky combination of Fs, Mms, Qts, means some special tricks are in order, to get a flat extended response below 100Hz, with the "Xtreme" series capacitor and a shallow LF EQ (done passively).

I'm itching to get the CC coax design tested, so I can get back to this one... with four 180mm woofers, in a sealed configuration and the damping on overshoots that comes with that compared to a reflex design (cross fingers), I'm very curious to measure and listen to how it works and sounds.

YMMV!

😂

[JonMarsh]

Just to tease a bit more, here's the LF crossover build plan:

[Evil Twin]

I am finding it desirable to introduce a different flavor of "rigor" to some driver evaluations, and rein in the "enthusiasm" of my at times over active and recalcitrant apprentice...

In this case, owing to the possibility of alternative HF transducers, one element in this process which is being added to all potential waveguide oriented projects is to prepare a normalized filter network converting the waveguide response not to an arbitrary system response requirement, but to a nominal flat response as one might expect from more conventional tweeters and as used for a standard of comparison, especially for distortion product levels.

This brief post will show how that concept has been applied to the Seas DXT H1499 tweeter, using QNF measurements to approximate an IEC baffle half plane measurement, with a filter to normalize both the response below 20kHz and to "neutralize" the impact of ultrasonic resonance on the SPL curve and resonant amplification of distortion products, if possible. This process has been used for other tweeter evaluations, sometimes only partially, as for direct radiation components such as the Peerless DA25TX-008.

First, the raw QNF response.




Next, the schematic of the dual mode filter- the series cap and parallel LCR to flatten the response level to 1kHz on the left, the ultrasonic notch filter on the right. These filters will be built as a test board for evaluations of the DXT in "normalized" mode against alternative solutions.





The simulated QNF response with these filters...





The measured response, with a focus on achievable distortion levels, will be compared against other potential solutions, including direct radiation tweeters.

Note, it has already been determined that increasing L13 to 1mH from the original calculated value should slightly further flatten the DXT normalized response plot.

[Evil Twin]

A test filter for the Seas DXT has been assembled, and base measurements completed...

L13 was increased to 1.1mH, and C14 was reduced to 2.2uF.


For clarity, the QNF check again...





The SPL response with the "normalization" filter board inserted, and output level increased slightly, with the measurement scale set as before... further optimization of the notch filter design seems possible, but is not needed for this evaluation.




And the measurement with distortion products, which interestingly requires increasing the vertical scale covered in order to capture the HD3 level in the main part of the working range...




Differences in the associated crossover network can interact with the notch filter, it seems- while this version is not fully optimized, it provides fairly clear information regarding the driver potential.

Certainly, based on HD3 levels greater than -70 dB down, at 86 dB output level, it seems reasonable to consider a 2kHz or higher crossover point.