Speakers

 

 

 

Since end 1998 I'm busy constructing my own 4-way active vented loudspeakers. I wanted to build no compromise speakers build with the best materials available and premium drivers. It's a never ending story since I still want to improve them further.
 
 Some improvements that were considered :

  • Try out Butterworth 4th order filters in place of the current 4th order Linkwitz-Riley filters. They have steeper cut-off slopes so the overlap between the drivers will be minimal. The former analogue filters were in the mean time replaced by Behringer DCX2496 digital crossovers. Experimenting with different filter characteristics is easy with those digital filters. After a lot of experimenting I decided to use 48 dB/octave Butterworth filters throughout the 4 channels. I used some parametric filters also to iron out some unwanted peaks in the drivers responses, like a 2.7 dB hump at about 700 Hz in the ScanSpeak 8545 midwoofers (Q=3.2), which seemed to cause a slight nasal tone in voices. I wondered a long time where this slight nasality came from ...
     

  • Replace the Amplimo power modules with amplifiers based on the Blameless Amplifier design : the Amplimo modules work fine but have their drawbacks (see below).
     

  • I adapted the enclosure so that the SS7000 tweeters, which have a diameter of 104 mm compared to the original 130 mm of the SS9900 Revelators,  fit properly. I did this by adding a felt ring to cover the gap. I experimented with different materials for this ring, but the felt gave by far the best frequency response : the ripples in it were minimal with the felt.
     

  • Limit the lower frequency corner frequency to some 50 Hz and leave the sub-bass range to the two new subwoofer cubes that are nearing completion. These use the Adire Tempest 15 inch drivers with their high volume displacement.
     



The unfinished Adire Tempest equipped subwoofer cubes with 150 l internal volume and 36 mm panel thickness
 

 

The finished Adire Tempest equipped subwoofer.


One primer coating and 6 layers of polished black satin lacquer. Two of them are now doing subwoofer duty in my living room to fill in the bottom end, especially for DVD movies. The frequency range is limited to 50 Hz, filtered with 24dB/octave via the Behringer DCX2496 digital crossover.

 

Since august 2003 I replaced the analogue Linkwitz-Riley filters with Berhringer DCX-2496 digital crossovers which improved the sound considerably. See the appropriate link on the audio projects page for some details.
 

Fig 1 : the finished loudspeakers with the drivers mounted and amplifier unit + Behringer DCX2496-filter unit on top.

Here you see the latest version with the DCX2496 on top and the ScanSpeak Revelator SS9900 replaced by the new ScanSpeak SS7000 ringradiator tweeter, which has an extended frequency range compared to the 9900 in the high treble range.

The overall size is 130 cm high, 43 cm deep and 31 cm wide (about 110 liter volume for the bass drivers). Material used is 1 inch MDF, heavily braced and the largest panels are reinforced with an extra 1 inch layer in the centre of the plane. Both mid-speakers are in a separate internal enclosure. The low-mid speaker has a rear volume of about 10 liters. The port is tuned to 29 Hz giving a 3 dB lower frequency of about 34 Hz. I added separate compartments for the high-mid and tweeter since I found out the screws of those units loosened under the pressure of bass and mid-bass. I think the speakers now sound cleaner because of the reduced influence from the lower frequency drivers on the high frequency ones.

The latest modification I performed was adding quite some glass fibre damping material in the bass box. This seemed to improve the sound considerably. It gives a much cleaner impression to my surprise.

          

Pictures of the boxes without and with the extra damping material :

 

As drivers we have :

- 2 x VIFA 25-CF-300 10 inch bass speakers :  Tymphany, VIFA, Peerless
- Scan-Speak 18W/8545K 7 inch mid-bass speaker :  Scan Speak
- Dynaudio D-52 AF high-mid dome speaker : Dynaudio website  (drivers not available for DIY anymore)
- Scan-Speak D2905/9900 Revelator dome tweeter :
  Scan Speak,
now replaced by the ScanSpeak SS7000 ring radiator tweeter.
 

Picture of the ScanSpeak SS7000 ring radiator tweeter mounted with a felt adapter ring 

Crossover frequencies are :

- bass - low-mid : 160 Hz
- low-mid - high-mid : 800 Hz
- high-mid - high : 4000 Hz

Filters were compensated 4th order Linkwitz-Riley (see the appropriate link). Compensation was required because Linkwitz-Riley works only perfectly in pure 2-way systems. In 3 or 4 way systems the extra filters required introduce some extra phase shift on the crossover frequency that has to be compensated for. You're invited to experiment with the crossover spreadsheet to see the effect. Summed response of the 4 outputs was within 0.06 dB of a flat curve (ever seen such "ideal" filters ?).

About the opamps used in these filters : I used TL072 FET opamps which work quite well, but if I were to redo these filters I would design them based on NE5532 transistor opamps. This would lower the distortion even further (about 0.002% compared to the 0.03 % obtained with the TL's). This would increase the current consumption some, requires good decoupling of the power supply near the IC's (100 nF across the V- and V+ pins) and lower impedance passive filter components : increase the capacitors so that maximum resistance values of about 2k are required (keeping the product of resistances values multiplied by the capacitor values constant).

The filter schematic

The interconnection of the different filters has changed now (see spreadsheet). First the signal is split low-high at 800 Hz, then both outputs are followed by the crossovers at 160 Hz and 4 kHz. This gives the best response possible (see the filter link). Before this each output was filtered via a separate filter bank. An added benefit of the new approach is that the outputs filtered at 800 Hz have a roll off slope of 48 dB/octave now, compared to the 24 dB/octave before.

The PSU balanced input
The filter
PCB (146 mm x 133 mm)
The
parts list

Download the designs of the PCB's in the format of PCB-developer of Andreas Waldherr (Waldherr website ) : PCB.ZIP

Power amplifiers are based on Amplimo ( Amplimo website ) modules of 180 W/4 ohm for the bass region, 120 W/8 ohm for low-mid and 100 W/8 ohm for both high-mid and high.

I did some in-room testing of the individual drivers and the complete system so as to fine tune the active filters. The red curve is the measurement at 1m, the blue at 2m and the green at 3m (more or less the listening distance). The speakers sound a bit forward to me, probably because the hump between 300 Hz and 2 kHz. With the DCX2496 digital filters, this midband hump is of course corrected and the forwardness is completely gone now.

Some results with the old analogue filters are here :

Clicking on the picture brings a more detailed view of it.

High frequency roll-off above 10 kHz is supposed to be due to the sound card. The measurement microphone is German made MBC-550 delivered with an individual calibration chart showing a flatness better than +/- 0.5 dB from 20 Hz to 20 kHz. It's rather expensive for private use at about 300 $/ however, not including the balanced pre-amplifier needed (I build mine myself as you could guess).
 

The near-field measurements of the different drivers looked as follows with the Linkwitz-Riley analogue filters :

Click on the graph to shows the high resolution version.


Smoothing was 1/8th octave. HF-roll-off above 10 kHz is again due to the sound card used. I did measurements in a anechoic room with a calibrated measurement system. My driver vendor Speakerland, has also such a room available for it's customers which I used.

The differences in peak levels take into account the different sensitivities of the drivers and the different gains of the power amplifiers per channel.
 


 

Measurements in my living room at 1 m with the DCX2496 tuned for best response. All crossovers are set for 48 dB/octave Butterworth response and both on the input and the individual outputs parametric equalisers were set.

Smoothing was 1/4th octave. The ripples below 200 Hz are standing wave artefacts from the living room. The tweeter's response below 4 kHz is in reality as steep as the other slopes, but there was some overloading in the measurement system caused by the fact that signal levels are proportional to the bandwidth which is by far the highest for the tweeter (16 kHz for the tweeter versus only 3.2 kHz for the mid, 640 Hz for the midbass and only 160 Hz for the woofers).

Go to the
filters page for information about the analogue crossovers and dcx2496 filters for the digital crossovers.

A lot of other measurements were done, including impedance measurements, but since they are less important in an active system and far more predictable, they are not shown here.

Measurements are done with SpeakerWorkshop for which I made a measurement  jig and electret microphone :



 Fig 2 : the measurement JIG used with SpeakerWorkshop
 



Fig 3 : the DIY electret measurement microphone with preamplifier
 

Here you find the Block Diagram of the active loudspeaker with the analogue filters.


View the filter characteristics :  4-way Linkwitz-Riley filters

 

The same spreadsheet as the 4-way Linkwitz-Riley filters above, but here you can choose corner frequencies and Q's and then the response is calculated :  4x4th-l-r-auto.xls


Here you find an exercise with phase compensated subtractive filters :  allpass-butterworth.xls


The Linkwitz-transform spreadsheet of John Murphy with frequency and phase response calculations added by me that let you see how the frequency range of a closed speaker is extended by the transformation : linkxfrm.xls


Another spreadsheet that shows a combination of 2 second order high pass filters usable for frequency response compensation of closed box speakers is 2xhpf-2nd-order.xls

 

Here you have all the Excel files in one archive : luchen.zip

Current status :

The enclosures are finished with light oak veneer with varnish finish. Weight : about 85 kg with the drivers. Filters are made and tested. They are flat from about 0 to 60 kHz (without the 20 Hz - 30 kHz band limiting filters). The measurement results can be found in this Excel spreadsheet : Filter Measurements . Frequency responses are exactly as calculated. I have build my own amplifier enclosures that fit on top of the speakers to minimize the length of the cables to the drivers. It houses the 4 amplifier modules and 3 large toroidal transformers. I can tell you they are quite heavy (over 20 kg each).

Some pictures of the home build amplifiers :

I added a LED-VU meter on the front based on a Velleman kit. I had to make a logarithmic pre-amplifier for it to extend the range from +3/-16 dB to 0/-60 dB. The design of this pre-amp took some time to get it right. Details can be made available if someone shows interest. Some detail information : as said the amplifiers themselves are Amplimo modules of 180 W/4 ohm for the bass section, 120 W/8 ohm for the low-mids and 100 W/8 ohm for both high-mid and tweeter. The power supplies are also genuine Amplimo ones, but for the bass and low-mid I decided to use heavy-duty Philips electrolytic capacitors of 33,000 uF/63V each. The high-mid amplifier has 4 x 8,000 uF/63V capacitors. Amplimo only provides 8,000  uF/63V which I considered too small for the low frequency amplifiers. The power switch on is done via a circuit with a power resistor and relays to limit switch on currents to the 3 large toroidal transformers. The series power resistor of 47 ohm 20 W is switched out of the circuit after about halve a second, after the switch on current has dropped to a more reasonable level (I don't want the lights in my living room to flicker when I switch on both amplifiers). I had to modify the amplifiers' protection circuit to make sure that the output relays' switch before any transients are generated that could damage the drivers. I think Amplimo could improve this, as well as the rather large DC-offset at the outputs (up to about 30 mV). The amplifier modules are AC-coupled at the inputs which is also unfortunate : this choice for AC-coupling should be left to the user since it introduces an extra phase error on the low side of the frequency range which I would prefer to avoid.

The photo below shows the filter unit on top, the amplifier in the middle and the top of the speaker on the bottom. Cables between the amplifier and the speakers are heavy duty 4 mm2 copper strand of only 20 cm length (the same cables are used internally in the speakers). Low-level input to the filter units is balanced to suppress any possible hum and RF pickup. The result is that the speakers are dead quiet : no perceivable hum and noise at listening distance (I was worried for that up front so I took any possible measures to prevent this kind of problems). By the way : after experimenting a bit I have put 47 nF WIMA capacitors on the AC-inputs of all the bridge rectifiers which eliminates completely the HF-ringing caused by the switching of the diodes. This switching excites the resonant circuit formed by the transformers leakage inductance and the stray capacitances. The ringing frequency was about 500 kHz originally ! The 100 Hz repetition frequency was clearly audible through the drivers before I added the 47 nF capacitors on the rectifier bridges.


 

My living room had virtually no soft materials inside, so it resonated as hell ! I now adapted the room with a 10 cm thick glass fibre absorbing wall behind the speakers to get better acoustic properties :

 

Here you see the finished loudspeakers, with the active DSP filters and the amplifiers in front of the absorbing wall :

 

This absorbing panel is covered with sound transparent decorative cloth to improve the looks. This doesn't seem to impair the absorbing properties. Quite unexpected, this treatment improved the 60 Hz resonance caused by the ceiling height considerably.

Finally a view of the left side of the open enclosure with the speakers mounted temporarily:

If you have any questions regarding this project write me at  luc.henderieckx@telenet.be