Active hi-fi Loudspeakers

Loudspeaker construction is probably one of the most beloved activities of electronic and woodworking hobbyists. And this is for a very good reason: While commercial musical equipment is usually of quite usable quality, commercial speakers unfortunately are not, unless you go to exotic and extremely expensive brands. The reason is simple: Good speakers, by law of physics, need to be large. There is no way around this basic fact. And large speakers require more material to build, and are much more expensive to transport and stock.

For the home builder this is a minor consideration. He will build them right where he will use them, and the cost of a little more wood is well justified by the better quality obtained.

I built my first speaker system when I was 15 years old. The cabinets enclosed a volume of about 50 liters, the woofers had a diameter of 30 cm, and the tweeters were 7 cm travelling-wave units. Instead of a real crossover circuit, simple capacitive coupling for the tweeters was used. I still have these speakers, they sound OK, but nothing out of this world...

When I had bought my first own home, time had come to build some really good speakers. I wanted to get a sound quality that leaves no wishes unfilled, without spending an arm and a leg. I looked through magazines and books, learned more about acoustics, and after some time came up with the design presented here: A triamplified, phase-linear, subtractive crossover, expanded class-A, resonance compensated active Hi-fi speaker system. Now, how's that for a name?  :-)

But you shouldn't believe in names. Maybe you should believe in specifications. These speakers have a frequency response of 18 to 22000 Hz, flat within better than 1 db. Of course, this goes in open space only; inside your living room there WILL be resonances...   The distortion is undetectable by ear, and the distortion of the electronic part, which I can put into my measuring instruments, is unmeasurable there too. The noise level is low enough to be unhearable at more than 30 cm, but I can hear the background hiss when placing my ear against the tweeter. No problem in practical use...
It doesn't make much sense to speak about power rating and efficiency, since these are active speakers. But the nominal driving level is 1 V RMS, and this will produce enough sound pressure to be loud for home use. Be warned, however, that these speakers are not usable for PA in large rooms, despite their size.

Oh, did I mention size? The cabinets are 125 cm high, 55 cm wide, 75 cm deep. They are a bit larger than what would fit in your bookshelf. But after lots of calculations, simulations and practical experiments, I came up with this size as the smallest one that would work acceptably down to 18 Hz. The woofer is a 38 cm unit with soft mounting, instead of a squeaker I used a wide-range 10 x 25 cm oval speaker, and the tweeter is a 2 cm dome unit with cast-metal diffuser. Treble dispersion is 165 degrees, making these speakers usable even in wide rooms.

So, let's go to the innards. Inside each speaker box there is an electronic circuit that includes a regulated power supply, three expanded class-A power amplifiers, a phase-linear subtractive crossover, and a Linkwitz resonance compensator. Here is the schematic diagram for the entire electronic unit. This is a GIF file of several thousand pixels per side, so be sure to save it, print it, and view it in full resolution. You may want to print it NOW, before reading the following discussion, so you can follow the explanation on the schematic.

Let's start with the power supply: The entire circuit is powered from a very simple regulated +-15V power supply, that is able to deliver about 2 A continuous duty, and at least 6 A on musical peaks. Two three-terminal regulators with power transistors wrapped around them do the regulating job. Here in chile we have a 220 V power grid, but of course feel free to use a transformer with a primary suited for your local voltage.

The audio signal enters at U1A, which is a simple voltage follower, delivering the signal to U1B. This operational amplifier is wired as a Linkwitz resonance compensation filter. It has a frequency response that is the exact opposite of the low-end woofer response, thus expanding the low bass coverage into frequencies that fall below the natural resonance, and would be strongly suppressed without the filter. Note that some components are specified at rather strange values. These are the exact values computed to compensate the specific behavior of my woofers in my cabinets. If you don't use the exact same speaker, and the exact same materials and dimensions for the cabinet, you will need to recalculate these values. I used the standard values that come closest to the ones stated here, but again, don't copy this part of the circuit blindly, because it will NOT work properly if your woofer or cabinet are different from mine.

I copied the Linkwitz circuit from an article in the november 1990 Elektor Electronics magazine. It was entitled "Active Mini Subwoofer, part 1", written bt T.Giffard. I suggest you try to obtain a copy of that article. It gives all the necessary equations to calculate proper component values for the Linkwitz circuit, for your specific box and driver.

Now let's go on to the cross-over: Again this is not my own design, but is taken from "Active Phase-linear Cross-over Network" , published by Elektor Electronics in September 1987. It mentions Stanley Lipshitz and John Vanderkooy as designers of this circuit. I just changed component values, but kept all of the basic design.

U2A and B form a fourth-order low pass filter with a cutoff frequency somewhat below 500 Hz. U3A and B are used as an all-pass filter (no frequency limiting action), that has exactly the same delay characteristics as the low pass filter, plus an 180 degree phase reversion. Thus, at pins 7 of U2 and U3 respectively, the signals are exactly phase-opposed, and U2 is delivering only the low range, while U3 is still delivering the entire range.

U3C and D make up another fourth-order low pass filter, but this one cuts off somewhat below 5 kHz. And there are two all-pass filters with the same delay as this low pass one: U2C and D, and U4A and B. So, at U20 pin 14 we have the low range, which is directly applied to the woofer level potentiometer. At U3 pin 14 we have the low and medium ranges, in the same phase. Applying these two signals to a differential amplifier (U4D), we keep the medium range and suppress the low range. This signal is applied to the squeaker level pot. Lastly, at U4 pin 7 we have the entire audio range, in opposite phase to the bass-plus-medium-ranges signal at U3 pin 14. Simply adding the two signals together produces an output of only the high range, which is applied to the tweeter level pot. As a bonus, the three output signals are in the same phase. Beautiful, isn't it?

You could ask why I used this circuit, instead of a simpler low-pass, band-pass, high-pass approach. After all, the exact phase relationship anyway gets destroyed by the three speakers. Well, the phase linearity of this circuit is just a by-product, not the main thing. The real reason for using this approach is this:

If you use three separate filters, it is nearly impossible to match them so well that the composite frequency response remains flat. Even if you do a very careful design, and use parallel combinations of components to get precise values, you still have to calibrate each individual circuit, simply because of inevitable component tolerances. If you use first order filtering, like in most passive crossover circuits, this problem is not too serious, but it becomes unmanageable for high order filtering. Now, why would you want high order filters at all?  Simply because this minimizes the problem of a single signal coming from two separated sources, for signals that fall in the range where TWO speakers respond to it.

So, using the subtractive crossover, we can use fourth-order filters, with the resulting steep separation of individual speaker responses, and keeping the overall response very flat. If a low-pass filter cuts off a little higher or lower, simply the other speaker will fill in the missing sound, or make room for the other's response.

If you build two speaker systems, with three amplifiers in each, that makes six of them. So, it pays to look for a design that provides good performance, while being extremely simple. The circuit used in this project is a standard application mentioned in many op amp data sheets. An op amp is used as a driver, passing its supply current through the base-emitter junctions of two power transistors. There is a local feedback loop and an overall feedback loop, and the gain of these two loops has to be carefully balanced in order to get stability. Resistors in parallel to the transistor base-emitter junctions shunt a part of the op amp's current, thus setting the idling current of the transistors at a safe value. In this circuit, the idling current was chosen at around 100 mA, which gives 3 W of static power dissipation in the transistors. This is low enough to be easy to handle from the thermal point of view, while at the same time it is high enough to guarantee that the transistors never switch off, thus giving true expanded class-A operation, and eliminating cross-over distortion.

These amplifiers can deliver approximately 12 W each into 8 Ohm speakers, and despite my efforts, I could never measure or hear any distortion coming from them. They are extremely clean, despite their almost ridiculous simplicity!

The only part of the schematic now remaining to comment is the circuit around the relay. This is just a timer, that delays the speaker connection by about one second, so they come in when the entire circuit has had time to settle down. When the power is switched off, the relay shuts the speakers off almost immediately, before the amplifiers' DC levels can destabilize.  So, the timer completely eliminates clicking or popping in the speakers when switching the active speakers on or off. It has no other function, so if you want to save the cost of the relay, you can do so, at the expense of some slight popping...

Now let's start the woodworking! As mentioned previously, these boxes are quite large, 125 x 75 x 55 cm, giving a total volume of over 500 l. I built them from 19 mm thick chipboard. I bought the wood already cut to size, which spared me from a lot of additional work.
All joints are glued with common woodworker's white glue, which is still the best thing available for wood. In addition, rows of screws were used to hold the boards together while the glue dries. Approximately 140 screws went into each box.

Maybe you are interested in a discussion about closed cabinets (infinite baffles) versus vented ones (bass reflex enclosures)? Entire books have been written about this, and fanatics of each method have been known to shoot each other because of their differing opinions. I don't want to shoot anyone, but still I'm against the trend in present-day industry. A closed cabinet has a smooth response. The frequency where the response starts falling off is higher than for a well designed bass-reflex box, but the roll off is much smoother. At very low frequencies the closed box responds much better, since the vent hole of the bass-reflex box behaves basically like an acoustic short-circuit. This allows to force the closed box to give stronger low bass, which is not possible using a vented box.
Another problem of vented boxes is that they exhibit a strong resonance, and a somewhat uneven frequency response even much higher in the spectrum than this resonance. This makes for the typical boomy sound of bass-reflex boxes.

But they have advantages too: A vented box has, by physical law, higher efficiency than a closed one. So, if you need the most possible sound pressure level, it's an advantage to use them.

In short: When sheer loudness is more important than sound quality, use vented boxes. When quality is more important, use closed boxes.

My speakers, of course, use closed boxes. I have never heard a bass reflex box that goes down to 18 Hz flat...!

Using a drill stand is a great help when it comes to drilling holes straight. Without the stand, whatever you do, you WILL crook them! This photo shows the drilling operation on one side panel. When you do this, be sure you place the board on some supports high enough to avoid drilling into the floor! I know some people who usually forget this, and their floors look like Emmental cheese. 

Here you can see a good number of details about speaker enclosure construction. First of all, you may already have noticed that the large panels are braced. This is very necessary for boxes of this size. The reason for bracing them is that otherwise the panels would resonate within the woofer's range. And if a panel resonates, it extracts a lot of energy from the sound field, re-radiates it in different directions, and the result is a terribly boomy, boxy sound! This must be avoided by all means!

Some speaker builders go so far as to avoid wood altogether. They build their speaker enclosures from concrete. That's quite inexpensive, and really goes a long way towards avoiding vibrations, but it's easy to end up with a speaker enclosure that weighs several tons. In my opinion, it's reasonable only if you are building your home, and you can include "speaker rooms" right in the design. If you can do that, go for it, it's always better than using wooden boxes!

Another approach is using double wooden walls, and fill the space with sand. This is more reasonable, but still a wood-sand-wood sandwich speaker box will be very heavy.

So, the last acceptable choice is what I selected here: Bracing the wooden walls enough so they are reasonably stiff, and the resonances are pushed into higher frequencies, and then cover the inside of the box with sound damping material.

By the way, chipboard is better than plywood or solid wood, because it has higher internal friction and thus less acute resonances. 

The box is taking shape! Note the space left in the upper left of this photo, on the upper back of the cabinet. This is where the electronics will be fitted later. That space is vented to the outside, to aid in the cooling of the components, and to allow easy reach, should ever anything fail (it never did, knock on wood!).
Much of the acoustically absorbing lining is already in place, except that on the upper side wall. First the separation between the woofer and squeaker airspace has to be installed. It runs diagonally from the lower edge of the squeaker to the corner of the electronics' container. Unfortunately I forgot to make a photo of that step! 

The box has been closed! The cables are in place, the speakers will be fitted from the outside, and the electronics from the back. There is no way to reopen the box, except for using an ax.
Small rubber feet have been installed too, in order to avoid scratching the floor or vibrating against it. While the speaker box is upside down, an old blue jean serves as a cushion.

It's incredible to see how many commercially made speakers don't have any feet, and happily jump around on a hard surface, producing a terrible accompaniment to your music! 

The box is being prepared for the aesthetic part of the work. Wood filler is used to close the screw holes, and fill in any other dents, holes, scratches, imprecise assembly, etc.

And here is the part of the work that transforms an ugly chipboard box into a beautiful piece of furniture: Gluing on the veneer! I love this work, specially when working with such a beautiful wood as this: It's Alerce, a chilean hardwood of intense reddish color, long grain, often with white speckles. Unfortunately Alerce trees have been overexploited to a tremendous level, specially from about 1850 to 1960 or so. These trees take many hundred years to reach maturity, and can live for several thousand years! Alerce trees are now under absolute protection, and only those trees that died naturally can be cut up. Alerce wood is extremely resistant to rotting, so even trees that died hundreds of years ago can still be used. The veneer used for these speakers came from such a tree. 

The veneer has been applied, and at this point I reasoned that I could well listen to music while the varnish dries. So, before starting the paint job, I did the final assembly: Mounting the speakers and the electronics.

Here you can see the individual speakers used: The tweeter is a Selenium dome unit with cast metal diffuser. I don't remember the model number. The oval one is a japanese wide range speaker of unknown brand, but very good sound quality, and the woofer is a Selenium 38W89. Don't think that these are ideal speakers! I choose the best I could get locally, and the selection isn't too good in this part of the world. Specially the woofer leaves a lot to be desired: Its free-air resonance is at about 25 Hz, and rises to about 37 Hz when enclosed even in this rather large enclosure. I would have preferred a much softer woofer, with correspondingly lower resonance, but I couldn't find one!  This was the reason why I had to use a Linkwitz circuit, despite the large size of woofer and cabinet, to extend the frequency response down to 18 Hz.

A small anecdote: Before coming across these oval speakers, I had bought a pair of Selenium M120 squeakers. They were rated at a frequency response of 300 to 10000 Hz, so they should fill the bill. Oh boy, was I wrong buying those critters! I still don't understand how anyone can make such a bad speaker! Put your head into a plastic chamber pot, and then listen to music. That's what it sounded like!  I measured the frequency response, and found it to start around 1200 Hz, quickly peaking, then dropping off by 16 dB, rising again to something higher than the first peak, and then disappearing into the noise at around 3 kHz...  And the phase response went through several complete circles in that range! These things were jokes, not speakers!

During the next several months I searched for usable squeakers. There are none in this country. Whatever midrange speaker I tried, was either totally unusable due to uneven frequency response, or usable only in the very high range. This put back my project by several months, until I finally decided to use supposedly low quality extended-range speakers for the midrange. Even the cheapest ones had better midrange than purpose-made midrange speakers, and those japanese oval ones are really fine, giving usable response from about 100 to 16000 Hz, and almost completely flat response in the range I need.

Here are the two electronic modules. As you can see, they are much simpler than what they look on the schematic. Note that all components, except only for the transformer and fuse, are mounted on the boards. For the photo, I placed the two electronic units on top of the speaker box, so you can see both the top and bottom side.
If you want to copy the boards, here are the printed circuit layout, the drill guide, and a simple guide for placing the components and the jumpers. Note that some jumpers go on the component side, the others on the solder side. There are a fair number of jumpers, but it's still easier to use such jumpers than to make a double sided board in the home shop!

The ten power semiconductors, all using TO220 cases, are neatly placed in a row and mounted to a piece of aluminum angle stock. This angle then mounts against a finned heat sink about 16 x 10 x 4 cm in size, and the entire assembly mounts in the open back of the box, the heat sink flush with the rear wall. The heat sink gets warm, but never hot. 

The job is completed! The two speakers are ready, sitting in my still nearly empty living room. I had almost no furniture in my newly bought apartment, so I COULD build big speakers!

I applied three layers of varnish to the speakers, and added a black screen. But I was too lazy to sand the veneer, so the surface is slightly rough. Someday I will sand it flat and apply another layer of varnish. I just need some time...

The table carries a test setup, comprising instruments to measure frequency response, distortion, and several other parameters. I used this to calibrate the level potentiometers and do some tests. But then I stored away those instruments, and let my ears measure the outcome. I spent several days listening to nonstop music. Slow a capella music, to judge mid range cleanliness. A documental recording of the world's largest church organ (in Passau, Germany), to judge low bass response. Well recorded romantic orchestra works to judge spatial resolution. And Heinrich Schuetz's Psalms of David, for 4 choirs, soloists, orchestra and organ, as a test for REAL grand music...   My ears were happy. So was I.


22 years after building these active speakers, I built a new version, keeping the boxes, woofers, and transformers, while replacing the squawkers, tweeters, and the electronic modules. You may want to read about Version #2 of my active HiFi speakers!

Back to the homo ludens electronicus page.