This is an old page – now superseded. It details the active separation of the woofers and their EQ. There is an oversight in the phase alignment of the bass to midrange sections as corrected on the tri-amp page.
An upgrade of the upstairs stereo system to bi-amplification with an ESP P09 active cross-over was so successful that I hadn’t listened to music on the downstairs system in months. And the downstairs system had the potential to be even better so this project just had to happen.
The downstairs system is a small home cinema setup, but the main front left and right speakers can be used for stereo music CDs without the multi-channel surround processor in-line. For this, there is a Sony SCD-1 SACD player used as a CD transport, a (C-J) Premier Nine valve output stage DAC, a C-J Premier Ten valve preamp, a pair of C-J Premier Eight-XS valve mono-blocks and a pair of Jamo Oriel 3-way tower speakers standing about 1.78 m tall. (UNDER CONTINUOUS OFF-SITE VIDEO SURVEILLANCE, IR DETECTED, DATA DOTTED, ENGRAVED AND GUARDED BY A FIERCE KILLER DOG and Space Enforcer Claudius!) 😯
The bass sections of the Oriels each consist of four chambers and two drivers. I.e. two pairs of 4th order band-pass sub-woofers in each of which an 8 inch Peerless driver (real Danish ones) separates a sealed chamber from a ported chamber. The Eton Hexacone midrange crosses over to the bass at 100 Hz and to the Dynaudio Esotar tweeters at around 3 KHz. The midrange is mounted on a neoprene-clad narrow curved front baffle. The substantial range of musical frequencies emanate from the midrange drivers for excellent imaging and depth of sound stage. The bass response is fantastic – fast and dynamic and easily extending below 20 Hz which is deeper and cleaner (for music reproduction at ordinary listening levels) than any subwoofer that I have heard. It’s like having four small sub-woofers acoustically coupling with each other and complementing two-way main speakers but with their output directly adjacent to the main drivers for near seamless transition to the midrange and with stereo separation of the lower frequencies when called upon. Some might scoff at that last bit and suggest that bass frequencies are “non-directional” or impossible to localise and settle for a mono subwoofer. Apart from missing unsubtle things like the discrete left and right cannon shot on the Tchaikovsky’s Nutcracker Soundtrack (Telarc CD) for example, they would miss out on a whole lot of spatial information and 6 to 9 dB of free emphasis brought about by acoustic coupling when stereo sources contain essentially monaural bass content (in this case 6 dB below around 33 Hz due to doubled input power plus acoustic coupling between left and right speakers spaced at a 2.3 metres plus 3 dB right across their pass band due to acoustic coupling alone of the close vertically spaced ports).
In their standard form however (relying entirely on their internal passive crossovers) the midrange and high frequency sound only came to life when valve power amps were added to the system. Previously several different solid state amplifiers were tried with disappointing results.
Anyway here is a picture of one of the Oriels and a cross section:
Conventional band pass sub-woofers have a ported chamber that is smaller than the sealed chamber, but the Oriels are different with tight sealed chambers behind each driver. I guess this was done to limit cone excursion at very low frequencies with some compromise perhaps in the system “Q”. Pressure mode sound propagation in the listening room seems just as good as with sealed subwoofers, but the bottom end extension is much better. It’s heaps tighter than most bass reflex designs which seem to lack impact and do unload when called upon to do anything really serious. The Oriels do not require LF boosting like regular sealed enclosures do, so power requirements are very much lower. I guess there is a small compromise in group delay response, but this I wouldn’t notice that with the type of music that I like.
So I came up with a combination of ESP project PCBs to actively cross-over the system for bi-amplification and assembled a new chassis which includes a 15-0-15 toroidal transformer, a full wave rectified power supply (P05b), a P09 24dB/octave Linkwitz-Riley crossover set at 102 Hz and a 36dB/octave rumble filter (P99) set at about 15 Hz to further protect the four bass drivers from over-excursion and more importantly, to reduce amplifier power wastage.
Unlike the room upstairs however, there was a “boom” at the seating position at around 25-30Hz which needed taming, so I included a stereo pair of P84 8-band “constant Q” subwoofer equalisers (20-100Hz) after the low-pass outputs of the crossover board and before the rumble filter.
Here are some photos of the finished chassis:
The chassis-grounded aluminium internal partition may look silly, but it works very well to isolate the audio circuitry from the 240 Volt AC wiring to the transformer. This can be demonstrated by waving an oscilloscope probe around the partition.
I kept the mid-to-tweeter passively crossing but removed/bypassed the low-pass section of the passive networks to the bass drivers and the high-pass sections to the midrange as I did upstairs.
The crossover inputs receive the stereo signal from the CJ pre-amp. The high-pass outputs go directly to the upper outputs for the C-J mono blocks. The outputs of the rumble filter go to their respective output sockets for connection to a more powerful solid state stereo power amp for the bass drivers. A 4-pole muting relay (powered by an unregulated output of the power supply) shorts the four outputs on powering down.
I decided to stack the equaliser boards (right channel above the left) and used unusual horizontal trim pots extending off the sides of the PCBs with their through-holes aligned to allow a vertical square shaft to pass through and turn two pots simultaneously as though they were dual-gang pots but without the need to tie one board to the other with 17 short wire links! The equalisers are a set-and-forget arrangement requiring removal of the lid to gain access for re-setting. And yes – the pots do turn backwards which is a little confusing at first. 🙂
There was a small problem of chassis height. I wanted everything to fit in a slim-line chassis and the one I liked had only 46mm internal height. I figured it could be done with 17mm above each equaliser PCB. The only real problem was the on-board power supply capacitors and power supply regulators of each board and the usual vertical header pin sockets that would have the wires extending upward, but I carefully mounted the regulators over and under the PCBs with their legs bent in odd directions and was able to source squat Japanese filter caps that fitted straight in vertically. I also sourced some neat screw terminal headers allowing the connecting wires to extend sideways.
Here is a side-on photo of the stacked equaliser boards (bit tight eh?). The diodes at the left in front of the filter caps were removed after the photo was taken.
I was going to raise the crossover frequency slightly (to 112 Hz) so that standard 20 – 100 Hz bands of the equalisers could be used without the top band coinciding too closely with the crossover point, but on the PCB designer’s recommendation I chose the closest possible crossover frequency (102Hz) to the original passive design. I took his advice that being of band-pass design the sub-woofers might not be very tolerant of out-of-band signals and that with the last pot of each board centred, the equalisers would be almost transparent at the top of the woofers’ pass band. Also, if the natural 4th order upper roll-off of the sub-woofers happened to overlap with the active low pass slope, the resultant 48dB/octave slope would cause an unwanted dip. Best hope that the natural roll off is somewhat higher (luckily it was). And it is probably better to retain the lower range of the midrange drivers for better sound quality compared to that emanating via the ports anyway.
There are plenty of boards and opportunities for unwanted ground loops and mess, so a four-pole terminal block came in very handy for some of the power supply wiring and circuit grounding.
The equalisers’ on-board power supplies initially received low voltage AC, but due to slight noise in the low-pass outputs (caused either by the low capacitance value of the on-board filter caps or a duplication in ground references) I rewired them to receive unregulated DC +ve, -ve and GND tapped directly from the underside of the main power supply PCB just before it’s regulators. The bridge diodes were removed from each equaliser but their regulators were retained. This obviated any additional burden on the main power supply regulators which already ran warm. An upgrade of the standard 1W carbon resistors to 3W metal film was all that was needed. There is now no measurable hum or noise in any output.
The internal signal cabling is thin Teflon coaxial. I really like this coaxial cabling. It is easier than using shielded twisted pair and the PTFE does not melt while soldering. Apart from polypropylene (which melts) it’s the best man-made dielectric too.
I used high grade components throughout without going overboard (say by using esoteric Analog Devices opamps, WBT RCA sockets etc). The crossover uses LME49720 super low distortion + noise opamps. The low frequency boards use cheaper OPA2134PAs as they are more than adequate for bass. The RCA sockets are Neutrik and the electrolytic caps are Japanese Panasonic where available. The filter caps on the rumble filter are Vishay Dale polyester and the crossover filter caps are Panasonic polyester. The regular MKT polyester I found to be too inaccurate and imprecise requiring the purchase of about 5 times as many as needed to get the ones you wanted to match at the specified value. That wasn’t a problem with the equaliser as only a few of each MKT were needed. The high pass outputs of the crossover have Vishay polypropylene coupling capacitors mounted underneath the PCB.
We tipped the Oriels over to access the passive crossovers:
Here are the binding posts:
Looks like a submarine hatch and you practically have to dive to get there! And they took some cracking after 12 years of being jammed (thank goodness Richard Basehart didn’t step out!). 🙂
And the passive crossover:
The board is intended for passive bi- or tri-amping, so there are three discrete filters which were quite easy to modify. The posts marked “NC” for “Not Connected” are intended for attaching attenuating resistor links for the tweeter in a bright room.
Loaded with horrible (but necessary given their high value) electrolytic capacitors amongst some nice French polypropylenes as well as a couple of nasty iron-cored inductor (distortion) coils:
The under-side photo is flipped vertically so it aligns logically with the top photo.
Two yellow lines indicate new link positions. Using the above, I drew up the bass and midrange filter circuits. I nice person later emailed me a factory service manual which confirmed my brilliant circuit tracing skills. 🙂 All but one of the electrolytics were no longer needed and were either bypassed with links, or lifted. Also, one of the iron-cored inductors was no longer needed. Great! 🙂
The tweeter sections were untouched apart from replacing the NP electrolytic capacitor with a new one of higher voltage rating, so I didn’t bother drawing up the circuit for it. The high-pass functionality of the midrange section was disabled as the active crossover does this, but the mid range low pass and Zobel networks were retained of course. The bass drivers were direct-wired to the input posts (bypassing the low-pass with wire links as the active crossover does this too).
Calibration was half a day’s work with a tone generator, oscilloscope, SPL meter and ears. One surprise was that when both amplifiers were turned on together after individually calibrating the SPL with the crossover output trimmers, there was a suck-out at the crossover frequency instead of the expected 6dB gain. Strange 😕 . All speaker connections seemed correct – but then I remembered that the original passive filters which crossed the bass to midrange were second order Butterworths each contributing 90° of phase shift for a summed inversion at the crossover point, whereas the 4th order Linkwitz-Riley active crossover is phase-correct (actually a 360° shift) at the crossover point. As mentioned above, the only output from the bass drivers is via the ports. Therefore the port tuning of the band pass sub-woofers must acoustically invert the driver phase at the crossover frequency. Reversal of the speaker cables to the woofer terminals was an easy fix. The scope was then used to ensure that the levels of both L and R outputs of the crossover were identical. The 100Hz band of the equalisers were first carefully “centred” by matching scope traces with the high pass at the cross-over frequency and then all the other bands were adjusted using a tone generator and listening from the best seat. The calibration is correct only at that seat. Move just a little bit forward and its all wrong again due to room nodes.
There was a minor buzz in the midrange drivers due to a ground loop, so I added a safety ground loop breaker to each of the power amps – which worked.
The performance of the Oriels is much improved. The midrange has freed up a bit and things seem to be better placed. I can’t say that the mid and upper frequencies are greatly improved, but the bass is simply astonishing. Audibly flat to 20 Hz with only slight tweaking of the equalisers. It’s fast too and without boom at any frequency. My hearing is only good down to around 20 Hz, but the room continues to shudder silently way down to 15 Hz where the rumble filter cuts input to the power amp sharply (36dB/octave). Evidence of the response is in what can be “felt” and in sympathetic resonances that develop in various things around the room at different frequencies.
While doing a sweep at about 90dB (about 2 or 3 Watts input) I was able to capture evidence of a 14.3 Hz room response by picking up a vibration around an internal window frame with my digital camera’s microphone. The SPL meter was still registering 90 dB, but it’s specified lower frequency response is only 20 Hz so the reading can only be considered “indicative”.
Here it is for the doubters (and I know who you are):
The only audible sound in the room was this rattle – i.e. no second or higher harmonic distortions from the speakers which would have been plainly audible had they been present at any significant level.
A highly rewarding project.
The midrange and treble were still not quite right. See here for an update. 🙂