Home > Linkwitz Speaker Build: Thinking Outside the Box

Some notes and observations made while building a Linkwitz dipole speaker system. The opinions expressed are my own; use at your own risk.
Please do not copy and repost these photographs or link directly to them from my server.

I'd seen the Linkwitz site many times over the years and always had a sideways glance at the Orion pages. As a long-time user of near-field two way speaker systems I did not get how these strange things with speakers mounted on baffles with no boxes could possibly produce decent sound. When a friend mentioned them in mid-2013 and pointed out about a dipole's radiation pattern and the fact that they create a null at the sides I had to look again.

Spending as much as or more than the cost of a pair of speakers on acoustical treatment of a listening room always seemed like an uncomfortable reality with conventional loudspeakers. With a dipole, not only does a null at the sides reduce the need for acoustic treatment of the side walls but the dipole radiation also reduces the need for controlling room reverberation times. See the Linkwitz site for an in-depth explanation of why this is so.

I ordered the Orion construction plans in September of 2013 but put everything on the back burner for a while. At the start of the new year I decided to move forward with the first stage of the project - completing the ASP boards.

The Rev.03072011 orion-4 PCB has a few oddities that are pretty hard to accept for someone who enjoys fooling around with PCB design. The orphaned copper pour in the middle of the U8 footprint, tracks that extend slightly beyond the final pad (pin 3 of U8, the four vias near C38), the smaller hole size for pin one of every IC footprint, the maddening thermal pads where the tracks always seem to cross to one or the other side of the target pad (see C38, C40) are some of the things I find difficult not to focus upon.

But more bothersome than the cosmetics are the component legends that sit mostly within the footprint of the devices themselves. Once you mount a component good luck finding out where it lies in the schematic without referring to the documented overlay. A board where the component names cannot be quickly identified creates extra work particularly when troubleshooting.

Double Vision

A side effect of having the component designations within the component footprint is that it makes it more likely that your designation will get obscured by the pads of the component footprint. In this case, the component in the above image appears to have the same designation (C206) as the one in the previous picture. But since the screen layer cannot be printed in the same area as a component pad it is very difficult to tell. Surely there cannot be two C206s on the same board?

In fact there are two C206 designations on the Rev.03072011 board. The one that is distinct near U5 is not to be used as many of the component footprints surrounding that IC. The one that has an obscured designation should be filled with a 47nF capacitor. This is mentioned in the Linkwitz support pages.

Change of Plans

After completing one ASP I started looking seriously at purchasing the drivers. Between the higher cost for Orion drivers and upon reading further about the Linkwitz LX521 I decided to shift gears and build the newer system instead.

It's fascinating how initially I was put off by the appearance of the newer design and how quickly these objections were quashed by a few hundred dollars. Not that I was ever that excited about the appearance of the Orion either!

Here's a table I created for changing an existing Orion ASP to an LX521:

Value Location
Jump R48, R308, R540
1.1k R37
1.62k Remove R217, Move to R309
3.48k R31
4.22k R214
2.37k R40, R217
1.33k R53
1.96k R69
2.87k R33, R49
5.62k R51, R55
14.7k R65
21.5k R304
19.6k R57, R58, R208, R211, R221, R222
38.3k R210, R213
31.6k R54, R302
147k R66, R67
4.7nF C306
10nF C35, C36, C48, C49, C50, C51, C52, C53, C54, C55
33nF C61 Remove C73
47nF Remove C400, Add C43, C47, C73, C208, C210
120nF C64
220nF C63 Remove C58
Note: There is a post buried within the resources forum that points out that the value of R40 on the ASP is 2.37k, not 2.87k as indicated in the bill of materials supplied with the documentation. The value is correctly marked on the schematic and mentioned on the Owner Support page.

Redesigned ASP Board

Rather than just complain about the PCB I decided to redesign it. The new board is designed to fit inside an enclosure also housing four power amplifiers - one box for each side.

The power amplifier boards are fairly conventional bipolar transistor emitter follower output designs. The majority of components are surface mount on the bottom side with conventional electrolytics (for reliability and ease of replacement) and other through hole devices on top. Thermal and DC protection are onboard while a global AC detection and power on delay will control all four amplifiers.

Thinking that, realistically, the required eight power amplifiers would take some time to build, troubleshoot and put in a suitable enclosure I ordered two more Crown XLS 1500 Power Amplifiers. The plan was to hook the woofers of each side in series making a reasonable 8 ohm load. The remaining two amplifiers would handle the ASP outputs for the passively crossed-over midrange drivers and the tweeters.

But as luck would have it, not long after ordering the Crowns the opportunity to get a Yamaha MX-A5000 11 channel amplifier presented itself (only three extra wasted channels or just one extra channel if I later decided to cross over the mid section actively). The prototype boards came in faster than expected - great service from Gold Phoenix. The prototype amplifier modules worked perfectly aside from a minor issue with the thermal protection circuit. So I quickly went from no solutions for amplification to having a glut of them.

24 October 2015: New measurements above substitute Caddock thick film 8 Ohm 100W resistor load in place of imported wirewound type.

The ASP board had a few mistakes that - when repaired - proved to produce an ASP with identical characteristics to the official design. Since many seem to be going down the DSP path it may seem a strange pursuit but I've created a new ASP revision that fixes the earlier mistakes and adds some features. Since I will, for the medium term, be using one of the manufactured amplifier solutions I've also put the RCA jacks on the new PCB for convenience. The Neutrik combo input jack remained but with a footprint for an optional RCA input and a plug that allows both to be used (though not simultaneously, of course).

Original ASP sits on top of two redesigned ASPs.

Original ASP is the yellow line while the green is the redesigned ASP.

A last minute addition was the shelving low pass filter that was eliminated from the LX521 design by Mr. Linkwitz on the 12th of July 2014. A header was included on the board to allow a front panel switch to be fitted. Unfortunately it was connected incorrectly and did not function as expected (or at all!). One cut track and a short jumper restored its operation (above).

I do agree with some comments that the inclusion of the shelving low pass filter makes some (many?) recordings more tolerable at the top end. Whether the speaker system should include compensation for poor mixing/mastering is another question.

Madisound Flat Pack

While I enjoy woodworking the lack of suitable raw materials in Australia made the Madisound flat-pack very attractive. Shipping was expensive but a quick calculation proved that getting the tools and material needed to build the cabinets myself would outstrip the price of the flat-pack and shipping.

The raw Baltic birch had fuzzy edges caused presumably by the CNC tool that cut the various shapes.

A random orbital sander fitted with 400 grit took the fuzzies off while bringing the surface to a smooth finish. The plan is to use a solvent based polyurathane finish to enhance the natural look of the Baltic birch plywood.

Some small voids were visible on the edges of a few pieces. Easily filled with some water-based wood filler.

Other marks will end up hidden by virtue of their location. In the case of the woofer bridge the scratches will end up on the inside of the support between the bass cabinet and won't be visible.

These marks, however, will be difficult to disguise.

Some de-lamination has occurred on the bottom of the baffle. Repaired by flowing some wood glue in the gap and clamping it.

One of the woofer bridge tops was missing from the original two-box shipment. Madisound sent a replacement that arrived in time for the second-last finishing session.

I used my trusty Malaysian HVLP spray gun to apply the polyurathane finish. If you need proof that insects rule the world, try getting a blemish-free coat on a sanded and cleaned piece of wood.

After three coats of finish I began assembly and immediately regretted that I had not masked off the bonding areas. While the idea of finishing before assembly was valid, getting finish where the pieces were to be joined was not a good idea. Instead, I sanded these areas as best I could to promote good bonding.

Following this the pieces were aligned and drilled for gluing and screwing together.

While the pre-drilled holes were beautifully countersunk the screws provided were pan head and not ideally suited to the angled profile of the countersunk holes. I purchased the correct screws at the hardware store and kept the nice pan head black screws for another project.

The two woofer baffles meet at a 90 degree angle but are secured only by the side edges, leaving a potential for vibration. I chose to use a biscuit joint here to secure this area.

The baffles were fitted into the mortises in the side panels to ensure the correct alignment and then glued and clamped into place.

Due to the length of the 'mortise' relative to the 'tenon' formed by the machined edges there is a considerable amount of slop where the fitted piece can slide within the mortise. Part D in the woofer assembly is supported by a rectangular piece of wood from behind but I chose to glue and clamp it to the top of the woofer assembly.

Factory Sound in South Melbourne had 8 conductor speaker cable in stock. One less thing that I'll have to order.

I wanted to hide the speaker wires to the degree possible using the stock Madisound baffle. Some scrap MDF was cut to the approximate midrange driver hole size.

Baffle in position to check the spacing.

The slot cutting bit is fixed at a 14mm depth so the jig is necessary to ensure that the slot won't interfere with the mounting bolts for the drivers. The depth is a guess since I don't have the drivers in hand as yet.

Even with a very slow approach the bit causes some burning of the wood. Unfortunately my router does not have a built-in variable speed control.

However, the method is a good first step toward hiding the wires.

The data sheet for the U22REX/P-SL does not show the diameter of the mounting holes. With the drivers in hand it is now possible to mark the holes. The mounting hole diameter missing from the SEAS data sheet appears to be 208mm.

The information on the data sheet for the MU10RB-SL for an 89mm diameter of the mounting holes appears accurate. The screws will clearly have to miss the tweeter wires that will pass around the perimeter.

A long 1/8" drill bit directed toward the rear of the front baffle allows the tweeter wire to pass though the routed groove to the tweeter compartment.

A manual chuck was required to drill the tunnel for the wires through the midrange channels due to lack of space.

The 1/8" long bit was used again to drill a tunnel from the lower midrange to crossover area.

I had some solid-core PVC covered 20 AWG wire from a tube amplifier build left over, so I gave it a gentle twist and routed it through the groove. So far so good.

Using non-acetic acid curing RTV silicone to secure the wires.

Smoothed with a finger, the silicone fills the channel well.

Front tweeter baffle is positioned over mounting hole with top parallel to the overall baffle.

Area is masked and cut against tweeter baffle's complex bottom shape while sides and top use the straight edge of the masking tape itself.

Regrettably varnished surface sanded to promote adhesion.

PVA glue spread on mounting surface.

Baffle clamped.

Excess glue is easy to remove by pulling up the mask. The tweeter baffles must be attached to the main baffle before mounting and wiring the tweeters. Doing so does not interfere with wiring the tweeters (something that took some time to work out in my befuddled brain).

Mentioned by others but not, to my knowledge, previously shown - a jig knocked out of scrap plywood to hold the bass cabinet.

The jig holds the cabinet in a way that allows the woofers to lie flat. This is probably essential if you are mounting the drivers on your own as it is difficult to both hold the driver and secure it with screws at the same time.

When installing the upper woofer but with the cabinet in this kind of jig, note that when the instructions say to hold the woofer horizontally you will need to hold it parallel to the top surface of the cabinet with the cone facing the top panel. You don't want to try to put the driver in magnet first. As you push it towards its destination the edge of the rubber boot surrounding the magnet will prevent it from entering the hole. Work the rubber boot through the hole gently and the driver will drop into place. It is a tight fit.

Orient the T nuts in such a way that the spikes form a triangle facing away from the driver.

Otherwise the spikes can tear out the wood or even fail to pierce the wood at all.

Madisound thoughtfully provided quick disconnects for the bass and midrange drivers. Despite what has been written on the Orion/Pluto/LX521 Users Group by some, using the correct crimping tool will ensure reliable, long lasting yet removable connections. Notice that with a proper tool such as this Molex crimper (#63811-1000) it creates two curled cylinders that grip the wire securely instead of simply crushing them like the crimping portion of your basic $2.99 multi-function tool. An Engineer PA-20 or PA-21 would probably be an even better tool than the Molex. Making crimped joints properly takes practice - see here for instructions on how to do them correctly.

Unable to resist piecing it together with rubber bands, bits of string and tape, I threw together the whole system and plugged it into the MX-A5000. Astounding - the impact of cone drivers with the effortless delivery of a Magnepan dipole. The best analogy I can come up with is that the system is like putting on a new pair of glasses. Things that looked good look even better while flaws stand out in stark contrast. I have heard the future of sound reproduction and the first true advance in the art of speaker design in a long while.

After my initial excitement I noticed an imaging bias toward the left speaker and a mildly strident upper midrange. The cause revealed itself when passing the test signals through the system. Alternating pink noise showed that there was a marked difference in the frequency response of the two channels.

Upon investigation it appeared that the tweeters in the right channel were not working. I carefully prised them from the baffle and tested continuity. The rear tweeter was fine while the front tweeter measured open circuit. Since they are wired in series neither of them was working. Since they are crossed over quite high in frequency it wasn't dramatically noticeable as it would have been in a small two-way.

While I did test the left hand side tweeters for polarity as instructed in the construction plans I decided to trust the markings on the right side tweeters and simply wired them as described. Knowing that they had not been hooked up incorrectly or powered through the wrong amplifier output I was at a loss as to why it had failed.

The MX-A5000 amplifier has a unique circuit that by design excludes the first emitter follower stage from the feedback path. Differences between the input transistor's gain cause a current imbalance that creates a small voltage that is amplified by the remainder of the power stage. The resulting voltage is servoed out by an op amp, however if the difference in the beta of the input buffers is large enough it can exceed the capabilities of the servo to correct this offset voltage. But measurement of the tweeter channel showed a voltage of 0.001V - below the several millivolts typical of an ordinary amplifier.

Since there did not seem to be any electrical reason why the tweeter had failed I had to look for a physical cause. After removing the plastic retaining tabs from the back of the tweeter I held the threaded bushing with pliers and carefully removed the magnet structure.

The end of the voice coil winding is secured by a dab of stiff adhesive. It appears as though the wire had been bent right at the point at which it exited the glue, causing it to weaken. Perhaps it had initially remained intact long enough to pass inspection but failed either in transport or upon first operation.

By using a soldering iron to soften the glue I was able to free enough of the wire so it could be stripped of enamel. With the terminal end stripped the two wires were positioned at right angles to each other and soldered. The re-assembled tweeter seemed to work fine. I would assume that this was an isolated case and not typical of SEAS tweeters.

A few drops of Zap-A-Gap to fix the magnet assembly back in place.

Passive Crossover Mounting

Details in the LX521 Contruction Manual regarding passive crossover mounting within the mid/tweeter assembly base are surprisingly vague. Bill Schneider's page has one of the more clever ways of mounting these components securely and in an aesthetically pleasing way.

The usual round inductor can really only fit in a vertical orientation; making use of its mounting hole is difficult within the confines of parallel boards between which the inductor fits with little room to spare. Bill uses a nylon tie to secure it to the sled prior to sliding it into place.

I wanted to make use of the hole in the inductor's bobbin to screw it to a crossover container but what I had envisioned was something I really did not want to create more than one of by hand. Remembering back to when my father used to cast various items using resins I began to explore that avenue. If I only had to make one master that could be molded and cast that would satisfy my sensibilities.

From the LX521 FAQ: "Q6 - Are you done with designing the LX521?
A6 - Yes."

Midway through my efforts of passive crossover mounting the LX521.4 was announced. I was ready to persevere with the original design especially after my earlier work on the new ASP board. So going DSP was not a decision I made lightly.

But once the miniDSP 4x10 Hd arrived and I loaded the configuration files I began to feel I had made the right decision. Converting the nearly-completed LX521 to an LX521.4 was painless compared to the effort I'd made in getting my version of the ASP and passive crossover built and within a couple of hours I was listening to a markedly better system. For the first time I find myself listening to the music and the recording rather than to the speakers. While the LX521 did indeed render the listening room benign I was still aware that I was listening to a speaker system. Now with the LX521.4 the promise of the room and speakers "disappearing" has truly been delivered upon.

April 2017

As we should all know by now, Siegfried Linkwitz never rests. On the 18th of December 2016 he published a circuit diagram for an analog version of a C4-LX521.4V23-2x8V108.xml file loaded into a miniDSP. On 23 January 2017 I built a prototype of the ASP.4 and on the 26th of that month Siegfried requested it for testing. He made some minor changes to the circuit which he published on the 27th of March 2017.

These changes have been incorporated into the new circuit board and testing has concluded on the third prototype design. Update May 2017: Production boards are available now from Hairball Audio. Update January 2018: Assembled units are available from both Hairball Audio and Madisound.