Building Troels Gravesen's MUN-17 speakers


In my living room About four years ago I built a pair of Troels's SEAS 3-Way Classic MkII. This is a medium-sized standmount design, with a soft-dome tweeter, a 12cm treated paper midrange and a 22cm woofer in a ported box. The 3WCs were a reasonably straightforward construction project (though I did make things more challenging for myself with my choice of an outer cladding of cherry slats and a leather covering for the baffle). They look rather handsome in the living room, and I have very much enjoyed their smoothness and ease of long-term listening; they are great all-rounders, and are easy for my 300B amps to drive, with a minimum impedance of about four ohms at 800Hz or so, along with a sensitivity of around 90dB/1W/1m. They sound very convincing with small-group jazz and simple music - vocals are clear and tactile, and they are easily transparent enough to reveal clear differences between my Vacuum State SVP and RTP3 preamps.

After a while, though, I started feeling that they sounded slightly confused in the bass, especially with dynamic music, and lacked punch in rock music, to the point where I stopped listening to quite a large tranche of my record collection. It occurred to me that my zero-feedback amplifiers, with their output impedance of an ohm or so, were probably struggling to control the low frequencies, so I couldn't help wondering how they would sound with a beefy solid state amp driving the bass. Inspired by Troels's projects using the Hypex plate amps, I bought a couple of Hypex PSC2.400d amplifier modules. These are designed to be mounted on the back of speaker cabinets, and comprise a pair of 400W Class D amplifiers, along with digital signal processing circuitry allowing them to be used as a standalone active two-way speaker system. I built a pair of boxes for these, adjusted the gain to match that of the 300B amps and connected them to the woofers of my Classics. This immediately improved the bass performance, and the combination went a long way towards removing the boominess and lack of snappiness.

Although my Classics were supremely easy to live with, I had a nagging feeling that they weren't the ultimate in clarity, nor a perfect match to my room. I would never claim that there is a direct equivalence betwen price and performance, but - fifteen years on from my original Neptune speaker project - the three pairs of drivers for the SEAS Classics cost pretty much the same as the slightly more exotic Audax HDA drivers for the Neptune: the SEAS drivers are excellent value for money, but not exactly state of the art, and I did wonder what audible advantages more advanced technology might bring. I also had instinctive reservations about some aspects of the Classics' design:

A little while ago, I set up my Neptune speakers in my system again, this time for a prospective buyer. This confirmed - to my surprise - that substantially more controlled and dynamic bass (for instance Donald Fagen's Morph the Cat on CD) was possible in my room and with the 300B amplifiers than I was getting from the 3-Way Classics.

So, after living with the SEAS Classics for about three years, I started thinking about a new speaker project.

The MUN-17 speaker

Troels's MUN-17 I follow updates on Troels's site regularly, and when this one appeared in mid-2017 it didn't take me long to think "that's it!". The MUN-17 was, as its name denotes, designed for the ScanSpeak room at the Munich 2017 Audio Trade Show. It uses the Scan-Speak 28W/4878-T00 subwoofer driver in a 40-litre closed box, powered by the aforementioned Hypex PSC2.400d plate amplifier, using the onboard digital filter to implement a second-order low-pass filter at 200Hz. The other Scan-Speak drivers, the 18M/8631-T00 midrange and the D2908/714000 beryllium-domed tweeter, are mounted in a separate enclosure with a baffle sloped to time-align the two drivers and shaped to improve the off-axis response, and this section has an entirely passive crossover. The potential for powerful, well-controlled bass from a relatively compact speaker system (it has a footprint of only just over a square foot), as well as providing the potential to tune the bass response through software, made the MUN-17 immediately appealing.

This is what Troels says about the Revelator beryllium-domed tweeter: The new 7140 beryllium dome is an easy tweeter - and it is not. The 7140 Revelator dome has a special frequency response with lot of energy in the 4-8 kHz range, actually declining some 5 dB from 4 kHz to 20 kHz, thus takes some equalisation to reduce upper treble (5000-10000 Hz) and to reduce sibilance. No less than ten different tweeter crossovers were tested before deciding on the one seen below, which is as simple as possible and yet delivers a reasonably flat response, some +/- 2 dB from 3-20 kHz. The easy part was that whatever I tried it almost always sounded good, only different depending on actual frequency response. The 7140 is clean and transparent like few and belongs to the family of top-notch tweeter like ceramics and diamonds and I would say it supersedes ceramics. Here are the impressive measurements he made of the 71400, alongside a couple of other Scan-Speak top tweeters.

And, regarding the 8M/4631T00 midrange driver, he says: This 18M came as a pleasant surprise! It surely rivals my favourite AudioTechnology 18H52 for midrange. It has all the virtues of a high-quality midrange with superb detail, transparency and an overall ear-friendly presentation. With regard to frequency response it's probably the smoothest 6" midrange driver on the market, making crossover work very easy. Troels's description of "astonishing" transparency from the 6" midrange and the beryllium-domed tweeter was the final incentive to start collecting parts for the project. The 28W/4878T00 woofer has a reassuringly weighty magnet, and its linear excursion is an impressive 14mm each way.

Troels suggested the Fusion, a slightly larger 3-way, as a comparable but much more cost-effective alternative to the MUN-17, but I wasn't tempted. I'm sure that it sounds great, but I personally don't like the look of the SBA tweeter in the Fusion (the ScanSpeak D2908/714000 has a much sleeker and less intrusive look than the SBA tweeter, which has a funny-looking belly-button diaphragm and too many screws for my taste) and the large bass enclosure with its two reflex ports threatens to overwhelm my room. The faceted mid-treble baffle of the MUN-17 is carefully designed to manage the off-axis response, where the Fusion simply offsets the tweeter and midrange from the centre line of the flat baffle - just as in my 3-Way Classics. Troels doesn't present any horizontal dispersion measurements for the latter, but I suspect it wouldn't compare all that well with the impressively smooth performance in that respect of the MUN-17. Finally, it still has the stepped baffle of the Classic series, which I wanted to move away from.


Troels places this design in the lineage of his "Three-way Classics", and I can see that it belongs more naturally there than with his recent designs such as the ATS-4 and Illuminator-4, with their multiple box contructions and fore-and-aft adjustable midrange and tweeter enclosures, which show obvious influences from Wilson Audio's top-end speakers like the Alexia. On the other hand, it is clearly from the same stable as his ATS-4 HE, with which it shares the beryllium tweeter and the same Hypex plate amp for the woofer.

Other influences can clearly be seen, though, and not just from the comparably sized Wilson Watt/Puppy system. I read Troels' article on the venerable Snell AIII a while back, and immediately spotted the MUN-17's resemblance to it, despite the Snell's considerably larger dimensions. The overall shape of the boxes, as well as the carefully sculpted midrange and tweeter enclosure to manage diffraction (not to mention the overall aesthetics, with the cutout for the square grille covers over the mid and treble drivers), marked it out as at least an unconscious homage to the earlier design, though Troels denies any direct influence. Here is Larry Greenhill's Stereophile review of the AIII.

Another echo is from the SD Acoustics OBS (coincidentally also designed by a Dane), which is another three-way with a relatively large midrange driver and a similarly chamfered mid-treble section. Aesthetically, it has a fabric-covered grille that extends only partly down the front baffle in a similar way to that of the MUN-17. Back in the early Nineties, the OBS was one of a lineup of speakers I listened to when I was looking to replace my cleverly-designed, but undersized, Shan Shimnas: I was impressed by the airiness and bass of the SD speakers, but in the end settled for the all-round balance and competence of the ProAc Studio Ones I went home with.

Building the MUN-17

Building the boxes

Top box unveneered rear As when I was planning the 3-Way Classics, I toyed with the idea of buying a router, so I could build the cabinets the same way as Troels does, but in the end went back to my proven technique of layering up the baffles, cutting appropriately sized holes in each layer with a jigsaw and finishing with a half-round file. I used standard MDF for the whole of the front part of the top cabinet, rather than the through-coloured black material of the original: for the sake of ease of working, I again made up layers of 6mm and 12mm, instead of the single thickness of 25mm in Troels' original build. I staggered the edges of some of the layers to make it more straightforward to align all the parts before gluing, and this also improved the rigidity of the structure. I started chamfering the front panel with a hand plane, but soon realised this was going to be a lengthy and exhausting task, so I got my electric plane out, then did the fine finishing with the hand plane and sandpaper. The downside of building in layers, along with the 5-degree slant and faceted sides of the baffle, was that this was by a long way the most complex bit of woodwork I had ever attempted!

Top box unveneered front I followed the recommendation to have a removable rear panel, which is a 15mm composite piece of MDF and ply, onto which I installed the two crossover boards, the port and the 4mm terminals. Unlike Troels' prototype, in which the whole rear assembly, including the cladding, was separate from the front half, my speakers had the outer cladding permanently fixed onto the sides of the front parts of the boxes, and the rear panel was seated on 9mm fillets on the insides of the latter, as well as onto the 18mm plywood shelf brace.

Bass cabinet inner front The woofer enclosure was mainly built of layers of 5.5mm ply up to a total thickness of around 22mm, with a single 6mm MDF layer on the front baffle, while I fitted a sheet of 3.5mm ply on the back to allow me to reasonably flush-mount the Hypex plate amplifier. The bracing shelves were also layered up from three sheets of 5.5mm plywood, the centre layer of which fitted snugly into grooves in the innermost layer of the boxes. I made cutouts in the upper braces and holes in the lower ones, matching the sizes and locations in Troels' photos as closely as possible. The pieces ended up fitting together a bit like one of those wooden puzzles you get in upmarket Chrismas crackers, and my hand plane got plenty of use trimming edges to size to the required accuracy. In Troels' MUN-17 bass cabinets, I reckon eight pieces of plywood, not including the compartment for the plate amplifiers; in mine, the count was thirty-four... Before installing the interior damping in the bass boxes, I sealed the joints with frame sealant (normally used for draught-proofing door and window frames). I followed Troels' instructions to the letter for installing the damping felt and polyester foam inside both cabinets, fixing them in with wood glue. Troels doesn't suggest bitumen sheet for damping the panels in this design, and I decided not to apply any - my gut feeling is that it won't have much effect on 20mm or more of plywood. The outer "jackets" were made from two layers of 9mm plywood, giving a total thickness of 18mm. The extra outer layer means that the lower 25cm of the front baffle ended up about 40mm thick, and most of the sides almost 30mm thick - and it feels pretty inert!

Drivers and connections

Bass boxes temporarily connected in my workshop The tweeters and woofers were mounted with hex-head bolts and T-nuts, the latter fixed to the rear of the baffles with epoxy adhesive. For the midrange drivers I used black cross-head wood screws, as the bevelling of the hole behind them didn't leave room for T-nuts to get a grip. To start with I used Lucar spade connectors on the spades on the Hypex modules and on the drivers, to make it easier to connect and disconnect during the construction process, with the option of hard soldering the connections later on.

Top boxes tested on my workbench From the start I planned to install an extra pair of binding posts on the bass cabinet to allow the otherwise unused outputs from the plate amplifiers to be wired up to the mid-treble passive crossover. This gives the freedom to dispense with the external amplifier if I ever get fed up with the hassle and profligate energy consumption of my 300B amps. I ordered a seond pair of the Jantzen nickel-plated connectors: I preferred these to the screw terminals Troels uses in this position, even though these incurred a risk - even if a remote one - of shorting the amplifier outputs. I realised, once the braces and the amp compartments were in place, that locating the binding posts just above the plate amp compartment was the only real choice to allow soldering to the inside ends of the terminals through the woofer holes, once the baffles were permanently glued to the front. I couldn't drill the holes and fix the connectors until I had veneered the boxes, but by that stage I already seriously restricted access to the interior of the enclosures.

The crossover

Rear view Well before I started work on the boxes, I ordered and built the top-level crossover kit (Level 1), with Superior-Z and Alumen-Z capacitors, from Jantzen. I expected to have to match the eight 22uF caps between channels, since their nominal tolerance was 2%, but I think Jantzen must have done the job for me, as these were matched to within better than half a percent.

The crossovers for the tweeter and for the midrange are on separate boards mounted inside the back of the mid-treble cabinet. I built the crossovers on two pieces of 6mm MDF, fixing the parts to the supplied terminal strips with both silicone adhesive and cable ties through the board. I had to cut a larger piece of board for the tweeter crossover than the 200mm one specified on Troels's webpage: that was obviously sized for Superior-Z caps rather than the Alumen-Z parts in the Level 1 kit, and the latter are quite a bit longer. With the big inductors and chunky capacitors, the crossovers are rather heavier than the ones in my 3-Way Classics. I fixed the crossovers onto 6mm plywood battens on the back panel, partly to make room for the cable ties holding the larger parts onto the board, partly to give more depth to anchor the mounting screws, and partly to allow me to route wires behind the boards.

I connected the lower value 2.7ohm series resistor in the tweeter network, rather than the 3.9 ohm choice: I had found with the SEAS 3-Ways that the lower resistance was more to my taste, and I suspect that my room is more absorbent than Troels' more typically Scandinavian wooden floor.


In my living room

I spent several weeks weighing up how I was going to finish the baffles, which in Troels' original MUN-17 speakers are smooth black MDF, finished with wood oil. I don't have the tools to create the smart effect of Troels' version, but wanted to somehow cover my raw MDF in a presentable way. I wondered about using black leather, which I thought looked very attractive on my 3-Way Classics, but it wasn't stretchy enough to conform to the more angular contours of the MUN-17. After a lot of thought and some discussion on the DIYAudio forum, I settled on an ash veneer, which I rubbed down with 180 then 400 grade sandpaper, and before I started applying veneer to the sides and rear of the cabinets I painted the baffle surfaces with three coats of satin black over a grey primer, rubbing down between coats. It turned out that there was quite a difference between brands of satin black paint: I tried both Rust-Oleum and PlastiKote, and the first was far more glossy than I wanted. In the end, three layers of PlastiKote black, then finishing with three layers of clear satin lacquer, gave a satisfying surface finish with just a slight lustre.

I intended from the start to finish the outer plywood cladding and back panels with a hardwood veneer: in the end I chose a nice crown-cut American Oak, using a single length up one side, across the top, and down the other side, matching the figuring across the edges, and mirror-imaging the two speakers. After a couple of false starts, I was pleased to find some oak veneer sheets wide enough to cover the 36cm depth of the boxes in one go, and so managed to avoid having to join up book-matched pairs. As with my first speaker project, I started off using hide glue to fix the veneers, using an iron to flatten, but by the time I got to the crucial final pieces I switched to PVA glue - I found that, although the hide glue went on reasonably well, it made the veneer brittle and prone to chipping and splitting along the grain. With the white glue, I still got the odd chip and wrinkle, but the former were mainly easy to retouch with a smear of glue, and the latter were (surprisingly to me, at least) almost eliminated by brief but firm application of the hot iron. It took a while for me to settle on a finish for the woodwork. I tried out spray satin lacquer, Danish Oil and Liberon Finishing Oil on spare pieces of veneer glued to a sheet of plywood: the latter two turned out to add too rich a colour for my taste, while the spray finish better showed off the natural colour of the light oak. I found four coats of lacquer produced just the right sheen for my tastes. The overall illusion I was aiming for was to suggest that the workings of the speakers were made of black timber, clad in solid oak (compare the actual solid cherry outer jacket of my SEAS Classics!). They turned out well, I thought.

Covering the speaker baffles

Left speaker with cover on Because of regular visits to our house from a toddling grandchild, as well as an elderly resident cat (who, sadly, shook off her mortal coil while I was in the last stages of the project), I thought it best to keep the bass driver covered, along with the upper two drivers. This ended up somewhat similar in style to the covers on Troels's Fusion design, though I had the concept for mine almost complete in my head before Troels updated the Fusion web pages with photos of the covers. I used a single frame to cover both top and bottom enclosures, to give a simpler overall contour, and used the suggested attachment arrangement using paired neodymium magnets in the frame and in the sides of the baffle. Neodymium magnets are fun! I built the frame for the covers with 5.5mm plywood, building it up to three layers where it fitted against the body of the speakers at the rear, and I made cutouts above, to the side, and at the front of the midrange and tweeter section to allow sound to radiate as freely as possible. As I did in my previous speaker projects, I painted the wooden frames with blackboard paint before applying the fabric, to prevent the natural colour of the plywood showing through the black material.

Disappointingly, the fabric from Falcon turned out not to be stretchy enough to cover the top of the frames without folds. I thought about sewing the join at the top edges, as per Troels' suggestion, but I could never quite work out how I could achieve this neatly, given that my sewing ability is several leagues below my woodworking and soldering skills. I lived with the uncovered speakers for the next couple of years, thinking from time to time how I could arrange the fabric on the frames. After a lot of consideration I ended up fixing the fabric around the inside of the top and bottom sides of the frame, and then folding the piece on the top of the frames over itself, which is only noticeable on close examination.

A consequence of extending the covers down over the woofers was that, to maintain the critical dimensions of the mid-treble box the same as in Troels' instructions, the baffle of the lower enclosure ended up a little narrower than in Troels' original, so I extended the depth of the boxes by about 3.5cm to maintain the volume loading the woofer. Actually I prefer the overall aesthetic balance of my build over Troels's - partly because of the slightly slimmer profile, and partly because of the way the single cover envelops all three drivers.


Until I started planning this project, I thought that the only place I was likely to discuss software issues in the context of a speaker building project might be concerning the REW measurement system. When the Hypex PSC2.400d arrived, however, I found I was completely unable to get any sound out of its outputs, and eventually worked out I needed to install the supplied software to set the volume level to anything other than zero. The problem was that Hypex doesn't supply a version of its software for a Mac platform, and a MacBook Air is all I have. After some research I discovered that the Boot Camp utility allowed me to install Windows 10 for free on my laptop, provided I assigned a disk partition to Windows. Once I had done this, I could install and run the Hypex software package without any problem,

The Hypex Filter design program proved very intuitive to use. You can add filter sections and shelves, and although there is a file provided on the MUN-17 webpage with an appropriate parameter set, I found it useful to be able to play with this, especially in adjusting the bass level relative to the passive leg of the crossover.


With the 300B amplifiers

The rest of my system consists of:

I set the speakers up about 25cm from the rear wall and about two metres apart, almost exactly the same placement as my Classic 3-Ways. Ideally they should be further out into the room, and also further from the Ikea Kallax record shelves that are against the wall between them, but the fact that they are in our suburban living room imposes serious restrictions on where they go. They are toed in so I can just see the inside faces of the boxes from my listening seat about three and a half metres from the speakers.

To start with, I set the gain on the Hypex amplifiers to match the levels of the bass and mid-treble sections by ear. I haven't made any direct measurements of the voltage sensitivity of the passive part of the MUN-17s, but the impression is that this is closer to the near-90dB/1W/1m sensitivity of the SEAS Classics than the 87dB or so I have seen in Troels' own measurements. They certainly reveal the high gain of my electronics, as I still have to keep the volume control in the lower third of its range. Because the speakers are perforce quite close to the back wall, I found that the bass boost in the MUN-17 filter settings didn't quite work for me, resulting in a certain amount of boom, so I removed this section, leaving just the second-order low-pass filter, and lifted the volume to compensate.

For the first couple of listening sessions I felt there was something slightly odd about the sound - a hollowness, or a "cupped hands" kind of colouration that gave me a slight headache, and I wondered whether this was part of the "running in" process. All of a sudden it occurred to me that the amplifiers feeding the mid/treble and the bass might not have the same polarity, and switching the plus and minus connections at the back of both the top boxes fixed this at a stroke. Looking at the circuit of my 300B power amps, it was clear that this design inverts the signal, where by default the Hypex modules don't. I toggled the "invert" setting in the Hypex Filter Design dashboard, and returned the connections to the passive crossover to their original.

I couldn't say, hand on heart, that the first impression from the MUN-17s is that they are an easy listen, in the same way as my SEAS Classics are. That's not to say that they are at all offensive to listen to, or unacceptably coloured - far from it! What I mean is that the detail and dynamics that they excavate from a recording makes them quite attention-grabbing, even if the detail is not of the obvious kind. It turns out to be quite hard to pay attention to my laptop while I have a record playing, as I'm constantly distracted by this or that bit of the music. This means that the speakers are very revealing of the human content of the music (even if this might sound a bit "flat-earth". One example of this is Murray Perahia's two CDs of Bach's Keyboard concertos. Although I love these pieces, I had almost dismissed these performances as a little too saccharine, too easy to listen to, but through the MUN-17s I started to appreciate Perahia's superb phrasing and poise, and I found myself drawn into the music far more than I had before.

It's hard to escape the feeling of unlimited power capability with these speakers. In this obvious sense, of course, they sound like big speakers, but - like the best big speakers - in a lot of ways they don't. Along with the dynamic scale and the feeling of unbreakability, there is also real precision and focus. One thing the Hypex amps bring to the table is the way that the soundstage blooms and deepens as I turn the volume up. With the SEAS 3WCs, I could hear the bass start to become soggy as the 300B amplifiers ran out of steam, where with the semi-active MUN-17s the amps only have to drive the mid and treble, and there is no sense of strain whatsoever. As the sound opens up, I hear more detail and presence and everything becomes more real and immediate.

What can I say about the expensive beryllium tweeters? I can't say I heard more "sparkle" or "top-end detail" that wasn't there with the excellent fabric-dome SEAS tweeters in my 3-Way Classics, but I don't think that is where their real advantage lies. I'll try to explain with an analogy. What I like conceptually about a valve phono stage like the one in my RTP3 is the enormous dynamic headroom that a high voltage power supply gives. This, in combination with the wide bandwidth of the cascode input circuit, makes the preamp almost completely benign when fed record surface noise, with its high frequency content and potentially large spikes. This is in contrast with the line level preamplifier I built a few years ago with a volume control based on the DS1802 digital attenuator. This clipped with signals of anything much more than 1.5VRMS and made listening to black disc a hair-raising experience. For me, this is the main strength of a tweeter like the D2908/714000, with its dome resonance of over 30kHz (not to mention its crazy 130W power handling figure), as such a device is not easily excited into non-linear behaviour by transients. This, presumably, is why I find record surface crackles and pops far less intrusive (if not quite inaudible) through the MUN-17s than with any other speakers I have listened to, despite their near-flat on-axis response over the three octaves up to 15kHz.

Listening to an all-Hypex setup

Out of curiosity, I disconnected the cable from the 300B monoblocs and installed short jumper cables from the "high" outputs of the plate amplifier on the back of the bass boxes to the mid-treble unit. I found I had to reduce the level of the high channel from 10dB to 4dB to roughly match the level of the valve amplifier. I expected that I would get a noticeably brighter sound from the all-Hypex drive than with the 300B amplifiers, since Kevin states in his blog that he expectes the Miller capacitance of the 300Bs in this circuit to give a slight high-frequency loss in the latter, but I couldn't hear this.

I had heard good things about the Hypex amplifiers, not least from Troels himself (the ultimate version of his Fusion design uses both channels of the Hypex amplifier), and the better Class D designs are reported generically as having a crystalline clarity. Once I had settled on the right level for the mid-treble section, this turned out to be quite a satisfying arrangement. There was a precision to the sound, a kind of stability to the soundstage and imaging that I didnt get from the combination of 300Bs and Hypex. Music came across in a very lively way, with no sense of fatigue. One example is the way it played the Oscar Peterson trio's Affinity on vinyl, giving plenty of energy and weight to the piano's left hand and to Ray Brown's acoustic bass, but at the same time allowing some sensation of depth as well as plenty of textural detail in Ed Thigpen's drum playing.

Putting the 300B monoblocs back into the system, I realised that what I had missed with the all-NCore amplification was the intangible "magic" of the triodes that brought performers into my living room, and what may be best described as a "luminous" quality to music. I can't define it any better than that - if you have heard a system that does this, you will know what I mean. This experiment also confirmed my previous experiences comparing valve and solid-state amplification: with the Hypex modules powering the mid-treble, there was subtly less differentiation between recordings on vinyl and CD - and in fact took away some of the particular benefits of my excellent vinyl-aying front end, making analogue discs sound more like very good CD recordings.


On-axis at 1m The measurement microphone was a Behringer ECM8000 connected to my MacBook Air via a Focusrite Scarlett 2i2 3rd Gen, with the mic supported on a camera tripod in front of the speaker baffles. I thought about using the Marantz PM44SE integrated amplifier from my garage system to drive the speakers while testing, since this has the "normal" very low output impedance typical of a modern solid-state amplifier with negative feedback, but decided to use the arrangement I normally listen through, with my RTP3 preamp driving the Hypex amplifiers on the woofers and the 300B monoblocks on the mid and treble, all in the usual differential mode. The software was Room Equalisation Wizard (REW v5.20.1), and the figures shown here have 1/6-octave smoothing applied. I measured with both polarities of the connections to the mix-treble unit: the blue curve is clearly the correct arrangement, at least for the on-axis measurement.

My first measurement was at 1m distance on a horizontal axis, level with the point between the mid and treble drivers (right). The response is very smooth above 300Hz, while below that I was basically measuring the room: the big peaks at 42Hz and 100 Hz, as well as the dip at just under 70Hz, are at exactly the same frequencies as the ones I measured in my Neptune speakers in the same room. I do wonder, though, whether the effect of the room modes would be less apparent if I measured the response from both speakers at once, given the asymmetric shape of the living room, but I haven't tried this. The gradual slope down by about 6dB above 2kHz may well be at least partly an artefact of the measuring microphone, as I don't see it in Troels's full-range response, and I don't believe the 300B amp has this marked a treble loss. At 1m distance the mic is also 10-15 degrees below axis, which may contribute to the dropoff. On the other hand, another of the response graphs on Troels's page suggests that this high-frequency tendency is consistent with his build. Closer examination of both Troels's response and of my own suggests that the drop is could be at least partially interpreted as a step at around the upper crossover frequency of 2kHz, so I may try further reducing the series resistance in the high-pass leg of the crossover below 2.7 ohms to see how this affects the measured and subjective performance.

At my normal listening positioni I did a couple of measurements at 0.5m distance which, perhaps unsurprisingly, weren't as smooth as the 1m ones, showing a dip at 3kHz, a little above the crossover frequency. This reminded me of Troels' vertical dispersion graph, which shows a very similar dip for a microphone position substantially below axis.

The main thing I take away from the low-frequency measurements is that the level of the Hypex amps is probably about 5dB too low. This is easy to fix, although I would have to tackle the 42Hz room mode at least for the result to be acceptable with the bass raised by any useful amount. Incidentally, my measurements of the Neptunes showed the same bass shortfall, which in that case was a result of the relatively low real-world sensitivity of the HM210Z0 woofers (which of course I could not correct electronically). At some point I may experiment with the Hypex Filter Design software to see whether I can tackle the main room mode.

I also measured the response at my listening position on the sofa against the opposite wall (left). This had a much more substantial dropoff above 3kHz than the 1-metre curves, which I ascribe partly to the 6-degree tilt of the mid-treble baffle. Besides, at that distance I am about 4m away from the speakers, almost half a metre below the tweeter axis, and surrounded by absorbent furnishings. The fundamental room peak at 40Hz is even more marked, which is consistent with the subjective impression of a slighly boomier bass when I sit there. Interestingly, the effect of switching the polarity of the mid-treble connections is a lot less obvious at this distance, presumably because of the stronger influence of reflections at the microphone position.

Overall I am very happy with the measurements. They confirm my suspicion that the bass level was a little low, as well as reassuring me that the impressive mid-treble response, which shows Troels's skill in crossover design, as well as the high quality of the drivers he chose for this design, is reproduced in my real-world build of this project.


For the reasons I listed above, I expected the MUN-17 to outperform my 3-Way Classics in just about every way. At the same time, I didn't expect any big departures in terms of voicing or character, since Troels' preferences are expressed strongly in all his projects. In the end, I was delighted with their performance - on top of my wishlist was to combine the warmth and immediacy that I hear from high-quality valve systems with the firmness, power and control of a good solid-state amplifier driving a closed-box woofer, and that is exactly what I got. My only misgivings are to do with the room - there is a substantial fundamental mode at 40Hz that has affected every speaker I have used in there, but there is not much I can do to remedy that. Along the same lines, the chipboard-on-joists floor is probably audibly resonant. Finally I would love to hear them out away from the wall, but in our townhouse with a 4m x 5m living room there is very limited potential for that.

Finally, I find my own MUN-17s rather more attractive than Troels' own model. This is partly because of the fine oak veneer on mine, but also because extending the "jacket" down below the woofer gives is a slimmer and more balanced elegance than the original. I am still pleasingly distracted each time I see them in the living room.


For obvious reasons, I am very grateful to Troels Gravesen for his excellent website and for his generous and useful advice on the project.

I have also gained enormous amounts of information and inspiration from reading the Multi-Way Speaker DIY Forum over the last decade or more.

Finally I have to thank Brenda for her patience and for her tolerance of my lengthy absences in the garage, as well has her forebearance in complaining about the dust, the smell of glue and spray paint, and the noise of my power tools.


Troels's pages

Troels's DIY loudspeaker projects site

Troels' advice for choosing one of his projects.

Some "construction tips".

My report on Troels's site on my project.

Another MUN-17 builder! Robert built his pair - ambitiously - as single boxes.

General links

A discussion of the 28W/4878T00 subwoofer on DIYAudio.

Vance Dickason's reviews at Audio Express of the D2908/714000, of the the 4-ohm version of the midrange driver and of the 28W/4878T00.

An interesting DIYAudio thread on measuring the effect of bevelling the baffle edges on off-axis frequency response.

A fascinating comparison between a range of tweeters by Bliesma that differ almost only in their dome material.

Erik Squires' Audiogon comments on the MUN-17 here. Erik also makes the comparison with the Snell AIII.

Timeline for MUN-17 project


Jantzen Audio in Denmark supply crossovers, drivers and other various bits: I ordered the crossover kit from them.

Falcon Acoustics, who supplied the drivers and the grille cloth.

Hypex in the Netherlands.

The driver mounting bolts, spikes and a couple of extra pairs of binding posts were bought from the Hi-Fi Collective.

The T-nuts, blackboard paint and various glues and sealants were bought at Portswood Hardware, a traditional hardware shop in Southampton. This has a range of stock that continues to amaze me!

I bought the timber from the local branch of Wickes. I probably wouldn't source from there again - the MDF was fine (if expensive), but the Chinese-produced plywood was softer and more prone to splinters and warping than I would have liked, and two or three of the barcode labels were so firmly attached that I had to resort to a chisel to scrape off the remnants.

The oak and ash veneers came from the Wood Veneer Hub (highly recommended - helpful and speedy service) and the hide glue from Wood Finishes Direct (no complaints).

Alex Megann, December 2017 - July 2023

Back to DIY page.