• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Evidence-based Speaker Designs

Ilkless

Major Contributor
Forum Donor
Joined
Jan 26, 2019
Messages
1,757
Likes
3,437
Location
Singapore
Changelog
10/5/20 -
Added ME Geithain, Sausalito Audio/Grimani Systems and GGNTKT entry.
28/2/19 - Placed HEDD entry on main list.
6/2/19 - Expanded the Danley entry and placed it in the main list.
2/2/19 - Expanded the Keele/Dayton CBT entry and placed it in the main list.
27/1/19 - Expanded the NHT, Technics SB-C700, KS Digital and Arendal entries. Added the Gradient Audio Finland entry into the main list.
___________________________________--
There is a "State-of-the-art Loudspeakers" thread on this forum already. A slight semantic distinction from that thread and my intention here has to be made from the outset. The focus is not on absolute pinnacle performance from cost-no-object speakers. Instead, I hope to share about more affordable loudspeaker designs and manufacturers that have a demonstrable commitment to releasing evidence-based designs consistent with acoustical physics and psychoacoustics, even if they do not quite reach the levels of unsurpassed excellence often appearing in that other thread. There is a surprising variety of driver configurations and design choices in the list that follows - much more so than proponents of subjective intuition-led audio reproduction often assert when they speak of a "boring" homogeneity/convergence arising from evidence-based audio.

However, the nature of divergence in the loudspeakers I describe here is, in my view, consistent with what one might call a "reasonable pluralism". I will attempt to point out the different priorities of each design and the relative compromises their designers accept. However, what all these designs share is a spirit of progressiveness, advancement and commitment to making rationally-defensible engineering choices - and are most importantly well-priced. From my perspective, it is actually the conventional "hifi" market that is far more homogeneous in its stagnation of design format - think passive flat-baffle "monkey coffins" without any care to match dispersion.

Note: the list is unsorted, work-in-progress (hence the empty entries) and largely stream-of-consciousness for now. If this thread gains traction, I will probably reformat it with a taxonomy based on driver configuration (2-way, 3-way, coaxial etc.) and expand my descriptions. At least one speaker in every listed manufacturer's line has been independently tested to measure well. Special mention has to be given to German magazine Sound and Recording for world-class measurements for several speakers. They deserve all the support we can give for their contributions. You can do so by purchasing a PDF collating measurements of over 80 speakers (mostly active monitors). You can read more about their measurements in this thread.

Neumann - The German studio legend, now under Sennheiser, has a line of active monitors that have impeccable horizontal dispersion (note: refraining from using the more technical term "directivity" in the interests of making this piece accessible). The designs show a clear lineage tracing back to manufacturer Klein+Hummel, which Sennheiser bought over, rebranding their loudspeakers as Neumann products and releasing new designs under the Neumann name. There is existing material on ASR about the importance of at least smooth horizontal dispersion for accurate (and likely preferable) playback from Toole et al.

The Neumanns are perhaps some of the most conventional speakers you will see in this thread. The entire line uses the multi-way non-coaxial driver configuration that dominates audio. However, they use computationally-optimised waveguides (see here for explanation on what a waveguide is, as well as the importance of seamless horizontal dispersion at least) to ensure a seamless transition in dispersion. Tools like computational fluid dynamics have been used to minimise any sort of anomalous movement of air (think port chuffing and baffle diffraction) within the physical limits intrinsic to the size of each speaker. This explains the complex port shapes and seamless baffles (either cast aluminium or polycarbonate) with no edges. Incredibly advanced construction techniques compared to boring old veneered MDF boxes. All their speakers are excellent, but it is the two babies in the line - the KH80DSP and KH120A - that are conspicuous value leaders for their performance level. The KH80DSP is the latest design. It uses a DSP crossover. It has a 4-inch midwoofer, which basically restricts it to very nearfield or desktop use. However, it is perhaps at the absolute limit of loudspeaker engineering for its size. This is a speaker, that when tested independently at a German acoustics lab, achieves a +/-0.6 dB (you read that right) on-axis response out of the box (scroll down for measurements). It is flat to 60Hz and has a -6dB point at 52Hz - great for the size. Distortion is state-of-the-art, which allows maximum SPL to be above 90dB right down to 60Hz. It has linear phase response even at crossover due to the use of FIR filtering. The crossover works in tandem with the waveguide to achieve absolute seamless and uniform horizontal dispersion. It has vertical dispersion as good as possible for a non-coaxial, non-symmetrical driver placement (aided no doubt by tight centre-to-centre spacing made possible with the tiny woofer). All that for $1000 a pair. It is pricey, but (in my reckoning) commensurate with the amount of R&D. Did I mention the hiss (always a complaint against active monitors) of the KH80DSP is not just inaudible in the typical sense of being masked by room noise? No, it was tested at 0dbA/1m - just about the absolute threshold of human hearing acuity under any circumstances. The KH80 upstages its older, larger brother, the KH120A, which is designed similarly (and measures quite similarly), but is "only" +/-1.2dB on-axis, has a bit more vertical lobing due to the larger midbass driver, uses an analog active crossover and doesn't have FIR filtering. Still a very good value as I've outlined. Hopefully a KH120DSP is in the works.

Genelec - A famous name. Much of their product line is like Neumann's. Smooth, seamless low-diffraction enclosures, smooth dispersion for their non-coaxial loudspeaker designs. Neumann is slightly more accurate. Genelec is sold in more places. Here are polar maps of the most common 2-way Genelecs (provided by the company) compared to the Kali LP-6. However, they have released 'The Ones' - comprising the 8331A, 8341A and 8351A in order of size. This uses coaxial driver loaded in a large waveguide, crossed to double oval-shaped woofers that fire out of slots made by the large waveguide and the rest of the cabinet. There is no baffle in the conventional sense to speak of. The entire front of the speaker is a waveguide. Sound and Recording has measured the 8331A and 8351A. At the bottom of this post, I explain the advantages and disadvantages (have edited it from the original for clarity):

Genelec invented the 8351, a clever active monitor that controls dispersion to an extremely low frequency for its size by using a coaxial driver mounted into a large waveguide, with 2 slots cut into the cabinet, through which 2 oval midwoofers radiate. This allows symmetry in both the horizontal and vertical axis and hence uniform dispersion without lobing issues on both the horizontal and vertical axis. The 3-way design is more compact than conventional 3-way designs used by the likes of KEF or TAD. The 3-way design also reduces strain on the midrange driver and reduces excursion, thereby allowing Genelec to reduce dynamic compression from the midrange upwards vs a 2-way, as well as reduce intermodulation. The slot-loading of the bass drivers and the drivers' crossover with the coaxial work together to control dispersion to much lower levels than a normal multi-way speaker its size. The principle is broadly similar to methods the Kii Three or DD 8C employ to control the wavefront radiated by the speaker, though executed to a less extensive degree.

However, several independent measurements have identified a problem with compression and turbulence in the slots in the critical lower-midrange, reducing the dynamic capability vs a standard 3-way and increasing midrange distortion substantially.

Kali - I don't need to introduce them, there is a thread on this forum already. Their story (top designer, blank slate, budget positioning) is well-known, the design well-optimised for its constraints. I have heard reports of both hiss and no hiss. But the acoustic radiation of the speakers is otherwise beyond any reproach for the price and even much more (see my link early in the post about waveguides and dispersion to understand how to read a polar map). There are some minor response anomalies to consider though, like a slightly-depressed treble.

KS Digital - A highly under-the-radar active monitor manufacturer outside of Germany. Their speakers tick a lot of boxes. Coaxial. Sealed. Active DSP with class AB amplifiers. FIR filtering for linear phase at crossover. Aluminium enclosures (with hardwood trim). German-made even. All that for around KEF LS50 passive pricing. Scarcely believable, right? The model I'm referring to, the C5 Reference, uses a 6.5-inch SEAS Prestige coaxial driver imported from Norway. This entry is special, because it is the C5 Reference inspired this thread. Sound and Recording measurements here.

It has some issues but there is, as I wrote, a broader significance to this product:

Yes, the FR itself is a tad wavier than the best in the pricerange, there's a weird bass boost (easy to cut that especially with parametric EQ), the coax suffers from the massive 10kHz on-axis dip that plagues all but the most advanced coax drivers, distortion may be somewhat higher especially because of Doppler distortion inherent to coaxial mounting of drivers, and headroom might be compromised - but at least they took a shot at it, and engineered some rationally-defensible advancements with much more tech than most speakers its price. In doing so, they offer some verifiable improvements in phase response, crossover sophistication and polar response. The 10kHz anomaly in the polars is entirely because the polars are normalised to the on-axis response, but the dip disappears as one moves off-axis, throwing off the normalisation. Coaxials can also provide smoother vertical dispersion that eliminates vertical acoustic lobing that plagues virtually all non-coaxial speakers. This means it is more tolerant of listening height (though well-designed non-coaxials have a linear vertical window wide enough to cover sitting and standing positions for most listeners outside of desktop/extreme nearfield listening).

This also means that dip is far less audible or consequential than it seems; even the hyper-expensive TAD coaxial speakers have a dip there. The only people to have completely eliminated it are KEF. ELAC, SEAS, TAD, Technics (see below) coaxials all have it to some degree. Designs like this KS that merit consideration (and will pave the way to further improvements) fall under the radar, while the broader market chooses to play around with passive monkey coffins (and for the most part not even fantastic ones like the Philharmonic BMR). The market isn't at all meritocratic.

KSD also has a newer line that load the coaxial tweeter in a rigid waveguide, reminiscent of Presonus and the old Altec coaxials. I have not seen detailed measurements or indeed much information at all about it.

NHT - I think their fairly long history means they should really be in the "usual suspects" section below. However, the brand was in the woods for a short span in the 2000s, before re-emerging with a direct-sale model. They also seem to have a mediocre reputation among pure subjective audiophiles for their ho-hum Chinese OEM provenance, nondescript aesthetic and association with home theatre (all of which do not predict any acoustic property of their speaker design).

Their speakers are highly well-designed across the entire line and (in a welcome departure) sealed. The idea of a well-engineered flat-baffle 3-way is to offer near-seamless dispersion wider than a waveguide speaker. Waveguides narrow dispersion to help achieve a seamless transition between drivers with a large disparity in radiating diameter. The larger driver's dispersion gradually narrows with higher frequency (and shorter wavelength). Sound path length differences between different points of the radiating surface at HF become significant relative to the wavelength of the frequencies in question.

This necessarily entails acoustic interference (due to the addition and subtraction of out-of-phase signals) that accumulate into highly-attenuated SPL at oblique angles compared to on-axis - ie. a narrowing of the net acoustic radiation a.k.a dispersion with increasing frequency. A flat-baffle 3-way with carefully chosen driver diameters splits the difference so the dispersion mismatch is small enough to be almost eliminated (at least on the horizontal plane of the design axis) through optimising driver placement and crossover topology. This is evident in the NHT C3, as well as the Philharmonic BMR.

What of flat-baffle 2-ways? The mismatch in dispersion is very large, and minimising it requires crossing tweeters lower and woofers higher than they'd be comfortable with in virtually all circumstances.

Arendal Sound - Another direct-sale Scandinavian brand, but Arendal hails from Norway instead of Sweden like XTZ. They sell a line of high-SPL speakers, . They provide on-axis and off-axis curves as well as impedance response. Drivers are integrated well with a deep waveguide and good crossover. That sort of waveguide promises narrower dispersion but higher SPL capability (together with what look like beefy woofers), which is good perhaps for rooms with harsh specular early reflections from sidewalls. Unfortunately we don't know what smoothing technique they used for the curve, but the graphs don't seem excessively-smoothed.

Technics - The venerable brand's return to high-performance audio was muted, not least because What Hi-Fi gave a two-star review for their coaxial SB-C700 that has no basis in empirical reality. I will not dignify that site with a direct link on my pose. Stereophile measured them. These are way better designed than the LS50 passive. LS50 has shelved up response (+3dB within the listening window per NRC measurements at Soundstage Network) from 2-5kHz. That's stupidly bright. Its the result of a dispersion mismatch at the crossover frequency - rather ironic because coaxials provide a better basis for smooth-dispersing speakers to be engineered. In contrast, the SB-C700 is totally seamless in dispersion and flat throughout the crossover range. Going by raw driver performance, the Technics is only very slightly behind the Genelec seamless coaxial that covers the midrange up and the KEF Uni-Q. As a complete speaker, the SB-C700 leaves the LS50 (and the other 2-way KEF coaxs without a separate woofer) in the dust. In fact it is to my knowledge, the best implementation of a pure 2-way coaxial, despite using a passive crossover. It extends cleanly (see distortion measurements in the third pic of the first post) right down to its 48Hz tuning frequency, giving up little in bass performance to a non-coaxial midwoofer of similar size.

The "peaks and dips" are mostly artifacts of the normalisation technique as mentioned in the KS Digital entry. Also the midwoofers are much larger (6.5-inch vs 5.25 for LS50), which means more headroom and bass extension. In my opinion their white low-diffraction curved enclosures and flat aluminium honeycomb midwoofer look a damn sight better than the LS50's industrial design too.

Gradient Audio Finland - Read up on Jorma Salmi's contributions to audio engineering, talking about his wide-ranging research publications would be beyond the scope of this thread. Suffice to say he, like Siegfried Linkwitz, was concerned about room-agnostic performance and optimising designs using both in-room measurements and a solid foundation in engineering. To this end, Gradient's top offerings are dedicated to exotic driver configurations that achieve unconventional polar patterns like cardioid, dipole or a mix of both. All their speakers are built in Finland using birch ply. Their entry-level offerings are the Gradient Five standmount and Gradient Six standmount, both of which using SEAS coaxials customised to their specs and passive radiator bass loading. I have no measurements and only an useless subjective report that they did not sound severely coloured in my experience and were enjoyable on and off-axis. Their flagship, the Revolution mixes cardioid (in the midrange below baffle step), dipolar (in the bass) and monopolar (in the treble) radiation. They use dual 12-inch woofers mated to a coaxial acting as MF/HF. The 1997 version of the speaker is exemplary for its era and still good by today's standards. It has been incrementally-revised to keep up with driver and active crossover advances of course. The Revolution now is not that of 1997. The Five and Six use the same coaxial as the current Revolution, except those are used "full-range".

Note that even way back then, the custom coaxial they specced does not have any major on-axis dips below 10kHz, and only one at 16kHz, which is more than can be said for some coaxial drivers today (not that it should be a problem IF it fills in immediately off-axis). The in-room response John Atkinson measured falls within +/-1.3dB at 1/3-octave, one of the best results Atkinson ever measured. Robert E. Greene of The Absolute Sound reports similarly in his own room.

Constant Beamwidth Transducers (Dayton/DB Keele) - There are two CBTs you can buy as manufactured products: the CBT36 kit and the CBT24 kit/complete speaker, both offered by Dayton. The CBT24 is on sale for $1495/pr as of February 2, 2019. For mini-monitor money, you can get a bleeding-edge driver array, albeit limited by driver performance. I would personally love to see someone stick in Vifa TG9s or some other widebander with smoother dispersion and FR. Rick Craig of Selah Audio also does custom CBT work at a much higher price range. The CBT isn't a line array. Despite its looks it is as far as one can get from a line source. The curvature and gradually-decreasing SPL (called "shading") emitted by drivers as one moves up the array come together - as someone described brilliantly on DIYAudio - a "slice" (like an orange) of the wavefront of a perfect point source placed on the ground plane. Project an imaginary line backwards from each driver at the angle at which the driver is pivoted. Do you see the slice of a sphere now? Exceptionally uniform coverage both vertically and horizontally. No floor bounce because the speaker is on the ground plane. Floyd Toole mused that it could be the "perfect" surround loudspeaker due to the uniformity of coverage allowing similar sound over a larger seating area. There is more discussion on another ASR thread.

Note that the CBTs will require subs and ideally DSP to equalise driver response. Don Keele (ex-JBL, designer of the JBL biradials before the CBT array for audio) has an excellent lecture series that examines the properties of CBTs:


Danley - Tom Danley was instrumental to popularising and improving the multiple-entry horn, through the Unity and Synergy horn designs. The designs allow point-source dispersion, without vertical lobing, yet also without the SPL limitations of direct-radiating coaxial drivers. Instead, Danley would load drivers, each firing from a different aperture along the sides of a horn with a single exit point. The spacing between apertures (and hence drivers that were mounted into it) were calculated to approach a quarter wavelength of the crossover frequency. This is to promote optimal driver summation, such that the wavefronts of drivers playing different frequency ranges (woofer, midrange and tweeter) do not destructively interfere with each other within the horn, and instead can be directed outwards smoothly, without major FR discontinuities. Anecdotally, the Synergy horns have drivers so well-integrated that one could stick their head into the horn and still not hear any significant interference between the drivers.

By optimising the driver configuration, crossover and placement, the Danley horns are coaxial, have immense SPL capability and offer very controlled directivity. Moreover the Synergy horns are linear-phase and have relatively flat frequency response. A critique of these horns (regardless of multiple-entry or otherwise) is the diffraction that they introduce due to their relatively sharp edges and straight sides, which have been maligned by researchers such as Earl Geddes as the source of the stereotypical horn "honk" (see page 8 onwards of the linked PDF), even if the FR does not seem to imply that much roughness or brightness.

The alternative per Geddes is a highly-contoured "waveguide" with a minimal amount of sharp edges that focuses purely on having smooth dispersion that gradually fades away outside of its optimal zone. In contrast, horns like the Synergy aim for a dispersion pattern with sharply-defined borders (because of the high-end sound reinforcement applications Danley had in mind, which require ensuring that the Synergies would have a sharply-defined guaranteed-linear coverage within a stipulated area for say, concerts). Think having a blurry, smooth border of constant width versus a border created with a sharp, straight line. A matter of priorities and reasonable tradeoffs between both methods, which captures the spirit of this thread. Here is a comprehensive video of the acoustics behind the Synergy design that might do better than words in explaining how much different it is from the typical loudspeaker we see:


HEDD - ADAM's co-founder starts a new firm. AMT active monitors, all assembled in Berlin (starting to see a German slant here...). More expensive than ADAMs of a similar design made offshore (T5V and T7V). Originally, the only third-party measurements I could find were from Russian sources with measuring conditions I'm not so sure of. I have since found a detailed teardown from Italy, which shows the HEDD Type 05 uses an asymmetric analog active crossover with 6th and 8th-order Butterworth crossover slopes. ICEPower modules (which have detailed AP measurements from B&O datasheets) are indeed used. The midwoofer is beefy too, with a cast frame. FR is fairly wavy, at +/- 2dB, despite the sophisticated crossover, but they have a DSP plugin (HEDD Lineariser) to linearise phase and FR. Arguably, its net effect is similar to that of a DSP crossover without incurring an AD/DA conversion should you be using a computer as source. The only thing this upstream DSP cannot correct is the narrow-band off-axis flare centred at 3.5kHz. Add-on support for either Ravenna or Dante is available through slot-in modules. Clever I think. Trivial to add a card if either protocol takes off, but one isn't left with a white elephant if it doesn't.

ME Geithain - ME Geithain has had a cult following in studio circles for the last 2 decades, with a small following of diehard followers that proclaim it to be at or near the pinnacle of active loudspeaker engineering. The German pro audio manufacturer's offerings are coaxial and active. Models designated with a 'K' suffix are cardioid. They provide datasheets with rudimentary on and off-axis measurements, and Sound And Recording magazine has measured their speakers in further detail before. The cardioid dispersion uses a passive approach, a complex internal cabinet that allows delayed back radiation from the driver to emit from carefully-calculated slots that induce destructive interference. This is a contrast to DSP beamforming with electronics (Kii Three) and hybrid approaches with both DSP and slots (DD 8C). With proprietary coaxial driver assemblies involving MF/HF drivers suspended in front of woofers on driver grilles (as opposed to mounted within), rare cardioid enclosure construction and German manufacturing, MEG speakers are often priced at the top-end of their size class. For instance, the RL-906 workhorse, which is in a similar form factor as the likes of Genelec's 8030 and Neumann's KH120, costs nearly 2,600€ per pair. The cheapest cardioid offering is over 8,000 per pair, utilising a 10" woofer and a 3-tweeter array mounted in front it. Although the cardioid configuration works as advertised (especially on larger mid-field offerings with a wider baffle), considerations include the likelihood of diffraction around the sub-assemblies and relatively ragged response of the cardioid.

Sausalito Audio/Grimani Systems - Audio engineer Manny LaCarrubba designed a novel acoustic lens under his then-company Sausalito Audio Works. It was licensed by B&O and put into several iconic designs, such as the Beolab 5, as well as in the Audi A8. The lens was further developed as the Conic Section Array (CSA) lens under a new company, Sausalito Audio. The CSA be licensed for anyone so inclined. The company has not been shy with measurements, showing Spinorama-style graphs (data is anechoic >500Hz) of CSA designs in various form factors, including the Grimani Systems Alpha. Grimani was co-founded by LaCarrubba with two associates, with an entire lineup of speakers using the CSA targeted at the high-end HT/installation market. Despite the relative obscurity of the CSA design in the stereo market, the Spinorama shows remarkably constant and wide directivity, with HF power response being almost exactly the same as on-axis/listening window, while maintaining a smooth response. Unfortunately, the cost of entry for free-standing CSA-based loudspeakers are expensive; the cheapest is the Grimani Tau, which is a $9000/pair active floorstander with dual 6.5" woofers. However, the Sausalito Audio page shows a proof-of-concept prototype of a 6" woofer + CSA waveguide standmounter with good spins, keeping hopes for a compact and hopefully more affordable standmount alive.

Moreover, Sausalito Audio should also be applauded for releasing spinorama measurements of consumer smart speakers. They released spinorama data of the Apple Homepod and Sonus Play One. Both measure excellently, in an indictment of the conventional hifi market. There is absolutely unfathomable amount of R&D thrown at such speakers to overcome the severe form factor and cost constraints. The economies of scale to manufacture these sophisticated designs cost-effectively are immense. Wilson and Devore need not compare. The natural question to ask is: what if such technology was harnessed in a less compromised form factor? And why hasn't anything like that emerged outside of DD or Kii? But we digress - Sausalito has done the community a great service in providing evidence to raise these questions, and help us hold loudspeaker manufacturers to a higher standard.

GGNTKT - Startup German brand GGNTKT (from the German word "Gegentakt" - "push-pull" - with the vowels removed) released their Model M1, a wide-format 2-way sealed active speaker with DSP-based cardioid radiation (through 2 rear woofers) and a large, computationally-optimised waveguide tweeter that appears to use a variation of the diffraction-blending waveguide geometry popularised by the JBL M2. The GGNTKT M1 shows excellent measured performance, as seen in the detailed spec sheet with polar maps and off-axis curves. Directivity control down to 200Hz is demonstrably achieved through a combination of wide baffle and beamforming DSP with the rear woofers, representing a twist on the progressive style of beamforming speaker design first pioneered by Kii, then DD. Power amplification is achieved with Pascal modules. (thanks to @q3cpmafor the reminder)

<<<<<<<<<<<<<<<<Entries below are work-in-progress>>>>>>>>>>>>>>>>>>>>>>>>>>>>

Merovinger - Custom installer from Germany with a DSP coaxial speaker using an air-motion transformer tweeter. I know the Fluid Audio FPX7 does so too, but the DSP for Merovinger seems more extensive. It is used to linearise both phase and FR, as well as time-align. Need to do more research on these. Intriguing.

ABACUS Electronics - Germany again. Active speakers, preamps, power amps and streamers with built-in Acourate functionality. However, speaker measurements are limited to only one out-of-production model (on German magazine AUDIO). Nonetheless, rudimentary measurement of their proprietary class AB amp modules suggest incredibly low output impedance, that might bode well for overall engineering.

The usual suspects (residual category for evidence-based brands that seem to be well-known, so I'll get to them last or simply leave this list unelaborated. Lots of information available from a Google search anyway) - Salk, Philharmonic, ELAC, Ascend, Selah, Seaton Sound, PSB, B&O, Revel, JBL (LSR and Synthesis), ADAM, Buchardt Audio, Vanatoo
 
Last edited:

Habu

Active Member
Joined
Jan 24, 2019
Messages
219
Likes
533
Location
Montpellier (FRANCE)
Hi,
Congratulations for your post ! Nice work !
Just a remark regarding Neumann, they just released the KH 750 DSP, a 10" closed subwoofer that bring the DSP to the monitors plugged to this subwoofer, so you can use current monitors including KH120A and KH310A, you don't have to wait hypothetical KH120DSP or KH310DSP...

KH 80 DSP : Two way bass reflex monitor with a DSP. Woofer 4", the vertical cabinet is made of composite polycarbonate. (500 USD each)
KH 80 DSP Sound&Recording Test

KH 120 A : Two way bass reflex monitor, Woofer 5.25" : The vertical cabinet is made of aluminium. (700 USD each)
KH 120 A Sound&Recording Test

KH 310 A : Three way closed monitor, Woofer 8.25" : The two horizontal cabinets (L&R) are made of polyurethane front, wooden sleeve, aluminum electronics panel. (2200 USD each)
KH 310 A : Sound&Recording Test
IMG_2126.jpg

Sincerely yours

Habu34
 
Last edited:

SIY

Grand Contributor
Technical Expert
Joined
Apr 6, 2018
Messages
10,383
Likes
24,749
Location
Alfred, NY
Thanks for the post. Could you describe your measurement setup?

One speaker on your list that I've measured is the Kali LP-6. I see some significant differences from your on axis measurement- how was the speaker set up (position and switch settings)?
 
OP
Ilkless

Ilkless

Major Contributor
Forum Donor
Joined
Jan 26, 2019
Messages
1,757
Likes
3,437
Location
Singapore
Thanks for the post. Could you describe your measurement setup?

One speaker on your list that I've measured is the Kali LP-6. I see some significant differences from your on axis measurement- how was the speaker set up (position and switch settings)?

All measurements are third-party. The German measurements are research-grade. The Kali ones specifically were generated from the Common Loudspeaker Format raw data they provided on their site using CLFViewer. Charles Sprinkle got an independent consultancy to measure the Kalis. I was just the first to my knowledge to generate polars and FR from the raw data (which is one of the objects of the CLF) and share it.

i should clarify the CLF measuring standard stipulates full sphere anechoic measurements in 5° angular increments at 1/3-octave.
 
OP
Ilkless

Ilkless

Major Contributor
Forum Donor
Joined
Jan 26, 2019
Messages
1,757
Likes
3,437
Location
Singapore
Where did this plot come from? I want to understand why I see something different than that. Merci!

https://www.kaliaudio.com/lp6-technology-frequency-response

As I said, the raw data Kali provided in CLF, which can be downloaded from their site. The creators of the CLF have also released a viewer (http://www.clfgroup.org/viewer.htm) that allows exporting polars and on-axis FR from the raw data in that format. All measurements are fullsphere anechoic, 1/3-octave, in 5° angular increments under the standard the CLF Group has defined. In the Kali's case, they were measured at 8m. The CLFViewer has very limited export capabilities. There is nothing for me to set to introduce discrepancies on my part. Kali's own graph on their website is using the same data, but they have drawn it outside of CLFViewer with the more common log scale, which CLFViewer doesn't support. (And made aesthetic changes of course).
 

Biblob

Addicted to Fun and Learning
Joined
Sep 13, 2018
Messages
633
Likes
600
This is great, thank you for sharing all this worthfull information.
Regarding the Neumann KH120A, there is a German engineer (again) (Heissmann) that made a DIY design which aims to directly compete with this design. He powered it via a Hypex amp.

He made a direct comparison on his site: comparison DXT-MON with KH120A. It looks promising.

The design is for sale although @JustIntonation pointed out that for the measurements he used slight different box-dimensions (thicker wood) than in the design that he sells.
 
Last edited:

FrantzM

Major Contributor
Forum Donor
Joined
Mar 12, 2016
Messages
4,337
Likes
7,729
WoW!!!

Exellent.

I take notice and strong exception that Wilson Audio doesn't figure in this list .. This has to be an omission :p
 
OP
Ilkless

Ilkless

Major Contributor
Forum Donor
Joined
Jan 26, 2019
Messages
1,757
Likes
3,437
Location
Singapore
This is great, thank you for sharing all this worthfull information.
Regarding the Neumann KH120A, there is a German engineer (again) (Heissmann) that made a DIY design which aims to directly compete with this design. He powered it via a Hypex amp.

He made a direct comparison on his site: comparison DXT-MON with KH120A. It looks promising.

The design is for sale although @JustIntonation pointed out that for the measurements he used slight different box-dimensions (thicker wood) than in the design that he sells.

I am aware of the Heissmann. My own speakers are the SEAS Idunn, which uses the DXT tweeter.

Excellent design, but I'm more impressed by the ER18DXT (imo the best 2-way DXT kit design there is), which mates a larger driver (and its attendant advantages) to the DXT but has equally seamless directivity. The designer stress tested the DXT to 105dB/1m, found it could easily cope even below 2kHz, so he used 1.5kHz second-order LR acoustic slopes. The neat thing is the vertical window is relatively large and smooth, more than a LR4 might've been due to the additional phase rotation and smaller band of overlap causing pronounced vertical lobing.

Edit: Elsewhere, someone summed up the reason I made this thread perfectly:

The market being very backwards relative to deployable technologies available now.
 
Last edited:

Biblob

Addicted to Fun and Learning
Joined
Sep 13, 2018
Messages
633
Likes
600
Excellent design, but I'm more impressed by the ER18DXT

I was not aware of this design. It is already a fairly "old" design, if I may say so. Have there been some recent woofers that measure better and/or fit better with the tweeter? Maybe the fairly recent SB Acoustics SB17NBAC35, measured here (8 ohm) and used in a design by Troels Gravesen here, works better? Or the SB Acoustics MW16P measured here and also used in a build by Gravesen here?

Edit: because of my very limited knowledge about speakers, my questions about woofers will end here, before going to far off topic and polluting this thread.
 
Last edited:
OP
Ilkless

Ilkless

Major Contributor
Forum Donor
Joined
Jan 26, 2019
Messages
1,757
Likes
3,437
Location
Singapore
I was not aware of this design. It is already a fairly "old" design, if I may say so. Have there been some recent woofers that measure better and/or fit better with the tweeter? Maybe the fairly recent SB Acoustics SB17NBAC35, measured here (8 ohm) and used in a design by Troels Gravesen here, works better? Or the SB Acoustics MW16P measured here and also used in a build by Gravesen here?

Edit: because of my very limited knowledge about speakers, my questions about woofers will end here, before going to far off topic and polluting this thread.

Thought experiments for driver pairing can be attractive. However, a couple of caveats: it is only fair to compare drivers of a similar price range. Furthermore, loudspeaker driver product cycles, unlike fully-built commercial loudspeakers, have much longer product cycles, close to a decade give or take a few years, and improvements have been incremental. I'd even say the performance envelope at the state-of-the-art for non-coaxial electrodynamic drivers has barely moved in the last decade since Scanspeak's Revelator (for this discussion, see the 18W/8531)) redefined the mark for smooth FR, low distortion, and deep bass for the size. Innovation has mainly been in getting less compromises at lower prices. Same can't be said for coaxials and other forms of drive unit technology (eg. AMTs, ribbons), which are still improving as we speak.

With that in mind, John Krutke of Zaph Audio measured both the ER18 and SB17NRX (a current-gen SB driver like the one you mentioned). Here are the distortion measurements under similar conditions their ER18 at top, SB17NRX at bottom.

ER18RNX-HD.gif

ER18

SB17NRX35-HD.gif

SB17NRX

Another website that measured both several other SB and this SEAS.

Ok, back to regularly-scheduled programming.
 
OP
Ilkless

Ilkless

Major Contributor
Forum Donor
Joined
Jan 26, 2019
Messages
1,757
Likes
3,437
Location
Singapore
Bingo, that's what I was looking for!

Sorry for double-posting but the high-treble dip, I suspect, isn't a systemic flaw. It could simply be a function of the elliptical waveguide that appears only on a single plane right on-axis. Such local interference effects that exist on a vanishingly thin plane do not accurately reflect the radiated direct sound. This is the reason why Harman uses the spatially-averaged listening window, which I suspect would make things a bit better. An example of the sort of artefacts such a waveguide may introduce dead on-axis is clearly seen in Performa3 F206 spinoramas:

F206%2Bspin%2Bnew.jpg


A more extreme example of the possible invalidity of 0° on-axis measurements can be seen in speakers using Earl Geddes' oblate spheroidal waveguide:

GedLee%20Nathan%20H%20Freq%20Resp%20Plot%20Q1.png


Geddes even argues it is a necessary feature of state-of-the-art minimal diffraction waveguides with constant directivity.
 
Last edited:

SIY

Grand Contributor
Technical Expert
Joined
Apr 6, 2018
Messages
10,383
Likes
24,749
Location
Alfred, NY
Sorry for double-posting but the high-treble dip, I suspect, isn't a systemic flaw. It could simply be a function of the elliptical waveguide that appears only on a single plane right on-axis. Such local interference effects that exist on a vanishingly thin plane do not accurately reflect the radiated direct sound. This is the reason why Harman uses the spatially-averaged listening window, which I suspect would make things a bit better.

I saw a treble dip, but it was higher up in frequency, and there was no overall shelving down of the treble. But my measurement was at 1M, and that makes a difference.
 
OP
Ilkless

Ilkless

Major Contributor
Forum Donor
Joined
Jan 26, 2019
Messages
1,757
Likes
3,437
Location
Singapore
I saw a treble dip, but it was higher up in frequency, and there was no overall shelving down of the treble. But my measurement was at 1M, and that makes a difference.

Thanks. Sorry, but I added some further comments and references to the comment you replied to. I'm not sure of any distance-related mechanisms (air absorption of HF doesn't quite match up) that would explain it. Perhaps microphone FR differences? 1m should be enough for baffle diffraction to emerge fully for speakers on that size (because measuring distance is large relative to baffle dimensions), I'd think. My original point about local interference response still holds though, and I tried to elaborate it with my edits.
 
Last edited:

bigx5murf

Addicted to Fun and Learning
Joined
Nov 6, 2018
Messages
522
Likes
343
Too bad Thiel Audio is no more. Their products definitely fall into this category, especially their earlier designs, which mostly used off the shelf drivers from Audax, SEAS, Vifa, etc. With 40 years of improvements on the same basic concept of phase/time alignment through crossover design and sloped baffles.
 
OP
Ilkless

Ilkless

Major Contributor
Forum Donor
Joined
Jan 26, 2019
Messages
1,757
Likes
3,437
Location
Singapore
Too bad Thiel Audio is no more. Their products definitely fall into this category, especially their earlier designs, which mostly used off the shelf drivers from Audax, SEAS, Vifa, etc. With 40 years of improvements on the same basic concept of phase/time alignment through crossover design and sloped baffles.

I wouldn't put Jim Thiel's Thiel there. While their design choices do have some empirical basis (insofar as verifiable, though not necessarily audible improvements) - namely time and phase alignment, as well as flat, smooth on-axis response - their off-axis response is exceedingly coloured and distorted. There is not the remotest directivity control to speak of. The importance of smooth dispersion with in-room usage is common knowledge on ASR so I scarcely need to belabour the point.

In holding on to the passive first-order XO dogma (instead of infinitely less compromised and more flexible methods of acheiving time+phase alignment like FIR filtering in DSP), I don't regard Thiel speakers as rationally-defensible. The adherence to dogma in spite of severe compromises in other (more) relevant aspects of sound reproduction is a classic example of missing the forest for the trees.

But I would put the now-defunct post-buyout Thiel that were closer to PSB speakers in design methodology on my list.
 

SIY

Grand Contributor
Technical Expert
Joined
Apr 6, 2018
Messages
10,383
Likes
24,749
Location
Alfred, NY
Perhaps microphone FR differences? 1m should be enough for baffle diffraction to emerge fully for speakers on that size (because measuring distance is large relative to baffle dimensions), I'd think.

Presumably, they're using a calibrated mike, and so am I (in my case, a PCB 376A33 condenser mike with NIST-traceable calibration). But 1 meter is not really large in comparison to the baffle dimension, maybe three times greater. In any event, I see differences with 1M versus much larger differences. Unfortunately, everything in my lab is packed up for moving, but once we get to our new place and I'm unpacked, I'll run some measurements and post them to demonstrate what I'm saying.
 
Last edited:
OP
Ilkless

Ilkless

Major Contributor
Forum Donor
Joined
Jan 26, 2019
Messages
1,757
Likes
3,437
Location
Singapore
But 1 meter is not really large in comparison to the baffle dimension, maybe three times greater. In any even, I see differences with 1M versus much larger differences.

Have you considered baffle step losses that are only fully realised >1m? Kind of like a milder version of John Atkinson's infamous bass bump when splicing nearfield measurements to 1m. Would like to see your 1m graphs.

Note to everyone else reading this: I've updated OP. See the changelog for what's changed.
 

SIY

Grand Contributor
Technical Expert
Joined
Apr 6, 2018
Messages
10,383
Likes
24,749
Location
Alfred, NY
I should be up and running again in about 3 weeks.
 
Top Bottom