Another potential tweak was considered briefly this month, but abandoned on finding that there was no need. It was to possibly change the bass/mid crossover frequency. The LR4 crossing is at 102Hz and it’s been working well there since around 2010. This frequency was selected on the basis of the published crossover points of 100Hz and 3kHz – the latter most definitely being incorrect. The mid/tweeter must cross at 2kHz. There was still some confusion about the correct bass/mid crossover point. Anechoic chamber measurements of the Oriels done professionally in Melbourne back in 1992 showed very clearly an 80Hz crossing. That was for the early version of the speakers with a different midrange driver however. It could be that 102Hz is slightly out of band for the fourth order bandpass design of the subwoofers. The anechoic measurements certainly showed no problems with the midrange extension down to 80Hz, but that was very low and they were known for midrange driver failures. Hardly surprising with the shallow second order filters of the passive crossovers! My active filters are much steeper, so there was room for exploration. The problem is that low frequencies measured a “normal” way with REW are dominated by room modes which totally swamp the raw driver output. Gating down works for higher frequencies at close microphone range, but LF info is lost. More to come…
Late August 2020 update
Keele style microphone measurements of both the midrange driver and one port opening were done using REW with very wide gating. The super near-field technique works to allow low frequency measurements despite the wide gating because the speaker output almost completely swamps the room reflections at the microphone. The microphone tip was at the port exit on the one hand, and within 10mm of the midrange dust cap on the other. The responses show perfect phase alignment at 102Hz with (or without) the active crossover in the signal path. Here is the SPL overlay:
The wobbles in the response below crossover are cause by the speaker/enclosure/port alignment and a subwoofer graphic equaliser ahead of the active crossover in circuit (set to the sofa – not this super near-field mic position of course). The shape of each trace mimics those of the electrical signal measured at the active crossover’s RCA output sockets. →
Phase (done using a loopback cable in a second channel for identical timing datums):
The phase traces are wrapped and most probably 360 degrees apart at 102Hz.
It’s “too perfect”. At 80Hz there is 10° misalignment, but around the current 102Hz crossing point there isn’t even a single degree of deviation! Being “too good to be true”, I set up an experiment to confirm REW’s low frequency phase data (which comes with some warnings from the software developer related to the broad nature of LF impulse response peaks, but the loopback “anchor” ought to address those) using “Old Faithful” – my BWD Aussie made analogue oscilloscope and a tone generator:
In the photo, the microphone is at a lower port opening (top port was covered by hand during tests but made no difference) and feeds a Focusrite 2i2 USB audio interface. The analogue tone generator set at 102Hz is sitting on the ‘scope. The laptop simply provides USB power to the 2i2 and the “Direct Monitor” switch on the 2i2 passes the mic input directly to Channel 2 of the ‘scope. Channel 1 of the ‘scope receives the 102Hz generated tone to activate the triggering and provide a reference trace – to which the port response on the one hand, and the midrange response on the other – can be referenced. The generated tone is split off to the input of the active crossover. No go! The ‘scope picked up a whole lot of high frequency “digital hash” from the 2i2 (probably 48kHz) and could not produce clean traces. I’d need some kind of low pass filter to clean it all up.
Here are the two comparisons and they confirm that there is perfect phase alignment between mid and port at the crossover frequency. The speaker wires from the woofer amplifier are connected “backwards” to the woofer terminals as usual because the bandpass design of the woofer enclosures does weird things to the phase – like near-inversion:
Photos show a T-piece at the output of the generator. This splits the signal to the scope and the active crossover input as noted above.
In both photos the reference/crossover input trace is to the left with the acoustic trace lagging to the right by precisely the same amount each time. Since the output signals of an LR4 crossover are always in phase with one another (but 360° apart), the shift in the traces can be attributed solely to the acoustic filters of the speaker drivers in their enclosures. Both overlays are identical, so the speakers – as complex as they might be – are phase-aligned perfectly and nothing needs attention at all. 102Hz is already the ideal crossover point. Two more photos taken with an 80Hz signal confirmed a shift consistent with REW’s 10° deviation at that frequency.
So the LR4 accounts for 360°, the inverted woofer wires account for 180°, so whatever that gap is (say around 90°) is attributed to the drivers and the enclosures themselves. Fortunately they just happen to be the same. 🙂
- Aside: An easier way to have done this might have been with a pair of identical dynamic microphones connected directly to the inputs of the 2-channel oscilloscope. I have two such mics, but at the time didn’t have the necessary adapters from ¼” TS jack to BNC. It would simply have shown aligned sine waves in this case.
February 2019 Update
The page on the three-stage phase shifter details an interesting tweak to the tweeters.
October 2018 Update
Mild EQ using JRiver DSP Studio’s Parametric Equaliser has provided a near-perfectly flat on-axis frequency response (within ±0.68dB) from 500Hz (the lower limit of the gating applied) to 20kHz per the following quasi-anechoic REW measurement through JRiver’s WDM:
Moreover – that is an acoustic measurement without smoothing taken at 1.5m from the baffle. For polar measurements, read on.
March 2018 Update
All of the C-J power amps lurking on this page have been relegated to the junkpile. Replaced by inexpensive but competent DIY solid state power amps (A and B) and together with the crossover tweak of Jan 2018 below, the Oriels are now simply sensational!
January 2018 update!
In late 2017 I learned that the owner’s manual for the Oriels (as well as nearly all the promotional literature and magazine hype) incorrectly specified the crossover point between midrange and tweeter as 3kHz. My original LR4 active crossover point of 2.88kHz was as close to 3kHz as I could achieve using standard resistors and capacitors, but it was wrong! 😳
A tweeter/mid crossover point of 2.88kHz, whilst having provided excellent on-axis performance for some years, was too high for optimised directivity. Moreover:
- This video featuring Dr Floyd Toole is a great refresher course for parts of his book. He’s about the only audio guru who makes proper sense to me. Freddes, Zinkfritz and Weaner (sorry – either I can’t spell or can’t remember their actual names 😈 ) for example each come across as either insane, or with more than a mere taint of apparent commercial and/or irrational bias. Inspired by Toole, I took some near-field horizontal polar measurements with narrow 3ms gating to produce quasi-anechoic amplitude traces from around 400Hz up. With the active crossover set at 2.88kHz there was significant beaming of the Eton Hexacone midrange driver seen by progressively worsening dips at about 2.2kHz as the microphone was moved beyond 30 degrees off-axis. This is bad because it’s well below 10kHz and therefore the first audible lateral reflections from side walls can’t track the timbre of the direct sound.
- Also relevant to the above is an article published about the Oriels way back in the August 1992 issue of Electronics Australia magazine – a copy of which I finally managed to acquire in 2017. Despite the article’s various inaccuracies, the anechoic measurement plots don’t lie and indicate very clearly a 2kHz mid/tweeter crossover point. → I soon realised that the 3kHz crossover point as specified in the later owner’s manual and sales literature was simply false. Those bastards! 😆 Jamo changed the midrange driver during production, but it was a near-identical Eton Hexacone nonetheless. There can be no rationale whatsoever behind an apparent increase in crossover frequency when crossing with an identical tweeter on an identical baffle. The Bass/midrange crossover point specified as 100Hz for later production Oriels might also be erroneous as 80Hz is evident in the published anechoic traces.
- Off-axis speaker performance is IMO not highly relevant in large rooms nor rooms with absorbent side wall “treatments”. I don’t play the silly audiophile room treatment game anyway. In reference to small rooms Toole says “rooms are good for music, even loudspeaker reproduction of music” and “what we hear substantially is dictated by what happens off-axis”. In my case the speakers are just 1.2m off quite reflective sidewalls. The early reflections from these surfaces render the off-axis performance of critical importance. Indeed I now open the curtains to reveal the glass for the 60° off-axis incident angle and it sounds crazy good that way. Some people apply expensive absorptive or diffusive panels at such locations. If that truely sounds better to them, then that only suggests to me that the off-axis performance of their speakers is poor.
- I borrowed a MiniDSP unit over New Year’s 2018 (thanks Allex) to determine that an LR4 at 2.09kHz reduced the off-axis (45° – 60°) 2.1kHz suck-out by 6-10dB without adversely affecting the on-axis measurements at all. The 28mm Dynaudio Esotar T-330D dome tweeters can reach down below 2kHz with apparent ease and their off-axis performance is superb at these lower frequencies. Indeed the acoustic measurements done after the crossover point reduction showed precisely -6dB at 2kHz from both the midrange driver and tweeter measured separately and that’s what’s expected from LR4 when used well within the natural bandwidth of each driver.
- The MiniDSP unit (model “2×8”) whilst useful as a test tool, was very noisy and generated an inordinate level of distortion and noise in the measurements. There was significant audible white (and other spurious) noise during the sweeps. It can’t be used for hi-fi! That certain people advocate the set-up and permanent use of such low-fidelity equipment for their dodgey open baffle flims-panels just shouts of where their standards lie. Anyway who cares? Compare the two midrange line level measurements to the sides here. To the left ← is my ESP-based analogue active crossover and to the right → is the MiniDSP set to mimic the analogue curve (i.e. a low shelf is included for baffle-step compensation). MiniDSP’s noise floor is 13 times higher, and THD is more than double at 1kHz (typical but that’s where the cursor is). Both tested using identical equipment and without changing any settings on the soundcard (Focusrite 2i2).
- With the analogue active crossover resoldered with 5k6 resistors for a 2.09kHz crossover point, the Oriels simply disappeared into the room. The tweeters are now responsible for a much more significant share of the bandwidth in a region where hearing sensitivity is highest. It’s been somewhat of a revelation.
- I did not waste any time testing LR2 or other low order crossover regimes which might only have increased IMD.
- The polar measurements indicated a tweeter resonance at 4.8kHz. A -2dB PEQ filter (set in the playback software) with a Q of 5 eliminates this.
- Also evident from the polar traces was that a general improvement in flatness of the tweeter response can be achieved with a broad -2dB PEQ filter of Q=1, Fc= 10kHz – not that anyone would hear it.
- The sound is astonishing and the on-axis response is within ±2.4dB including the 4.8kHz tweeter resonance!
Since the crossover point was to be moved a fair way down, it was necessary to increase the tweeter protection capacitance. As such, this second capacitor was added across the binding posts – to the other side of which the existing capacitor was already connected internally. This avoided the need to open the speakers again. It’s just an MKP soldered to parts of an original tweeter attenuation resistor bridge intended for the same spot. →
The on axis, 45° and 60° traces (1m radius mid-way vertically between driver centres, 1/24 octave smoothing 2.8ms IR window):
(cursor at 4.8kHz to highlight the minor tweater resonance)
- If I were to redo those today (see October 2018 Update), the traces would be “text book perfect”. I might just do that. 🙂
That may not “look” quite as great as similar measurements → made recently of my upstairs DIY system, but the Oriels now sound heaps “better” with most music. It’s probably that the rooms are very different. The upstairs system sounds quite sterile by comparison!
↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓ Old page: ↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓
Purple text additions/amendments to the old text below reflect the above findings.
This supersedes the bi-amplification project.
After listening to many sound systems in hi-fi stores and at friend’s places over the years, I still haven’t heard anything to match or come remotely close to the amazing bass extension of the bi-amplified Oriels. However, a few really nice speakers with push-pull valve amps in front of them certainly had midrange and treble performance to rival them quite convincingly.
These are late 90’s (around 1998) production Oriels with different Eton Hexacone midrange drivers than the originals. These are identifiable by six mounting bolts. The early drivers (1997 and earlier) had four bolts and a slightly smaller dustcap. →
Below is a photo of the back of a 6 bolt midrange driver. It has an extra magnet attached – achieving what? Well it’s probably a bucking magnet to tweak the driver performance – perhaps increasing efficiency slightly and improving LF extension for the low crossover point to the woofers. I don’t know, but it’s certainly not there as a “shield” to reduce flux emission to nearby CRTs since neither the tweeter nor bass drivers have them or any other kind of magnetic shielding.
The fabulous 1.6Kg Dynaudio Esotar T-330 D 28mm soft dome tweeters with (an alleged) 1kW uncompressed pulse response were standard throughout production. These are considered by many to be one of the best tweeters ever made.
So there is certainly nothing wrong with the drivers. There must have been something wrong with the crossovers. According to the owner’s manual, the standard crossover point between the midrange and tweeter was at 3kHz. Simulations of the original filter found that it rolled off the midrange input by 6dB/octave from around 630Hz up. It also showed a giant electrical peak at 200Hz. The tweeters had very complex filters – each with a component count of fourteen! Three inductor coils, eight capacitors and three wirewound resistors in a circuit designed (I guess) to compensate for a slight dip in the tweeter’s frequency response at around 6kHz. Simulations of the original tweeter filter showed that it rolled off part of the frequency range at 6dB/octave and the remainder below around 4kHz at 30dB/octave.
Anyway it was all over the place. Great drivers in an over-the-top cabinet structure but with elaborate passive strangulation networks designed to cater to the market by enabling simple connection to a single amplifier with colouration to boot.
- July 2018 Addendum: The 1992 article in Electronics Australia shows a most horrendous impedance curve presented by the Oriel’s to an amplifier – impedance swinging from around 2Ω! at 80Hz to around 30Ω at 2kHz.
- Also a closer investigation of the original passive tweeter filter schematic from the factory workshop manual finds that as well as a basic first order filter, an L-pad and a conjugate impedance matching section, there’s a “lattice phase equaliser” smack in the middle of it all:
- The lattice phase equaliser (L4/C7 & L5/C8) is an all-pass filter (AKA phase shifter) intended to compensate for physical off-set of the drivers on the baffle by shifting electrical phase to the tweeter. However such filters are not suited to low order crossovers like these! It was suggested to me that the crossover produces a null in the output to the tweeter at a frequency where the impedance of the inductors equals that of the capacitors and indeed a null is evident in the Electronics Australia measurements. I do not believe that the equaliser is solely responsible however, since it affects only the phase. Nonetheless there is an uwanted notch just above 1kHz with rising response below that. C6-C6A reduce the output there, but wow – who needs that? Right across the midrange (apart from at this notch), tweeter output is down from the midrange by a mere 10dB. As a result of the low order crossover and extended LF response of the Dynaudio tweeter, the unwanted low-end phase manipulation wreaks havoc when summed with the midrange’s output. None of this is mentioned in the EA article of 1992, but you can actually see it in the anechoic chamber measurements when comparing the individual driver measurements with the full range sweeps:
- In the above, the midrange output looks fine when measured separately, but look what happens in the summed response below. Clearly the shallow filters are allowing unwanted phase interactions across right across the midrange – in particular those 5dB ripples at around 400Hz. It’s quite abominable:
- Remember – that’s anechoic. Room interactions cannot be blamed!
Something had to be done and the obvious answer was to remove all the garbage and actively tri-amplify the system – adding another amplifier for the tweeters. I also considered replacing the midrange drivers with new ones and toyed with the idea of Scan-Speak Illuminators (for a minute or two 😉 ) – dismissing them for having incompatible mounting hole spacings. I’m not prepared to drill the cabinets and although a steel adaptor plate could be fashioned I decided to maintain the original appearance of the Oriels by retaining the standard drivers and see how things went.
Modifying the Speakers:
Over they go again:
After briefly considering further modifications of the passive crossover boards (lifting everything in parallel with the drivers and installing jumpers across anything in series), I decided it would be better to remove them altogether and connect the internal speaker wires directly to the binding posts. It’s completely reversible, should I ever decide to sell them which I won’t!
Small problem: Stupid Japanese “Isoda” cables running to the midrange and tweeters having delicate aluminium conductors twisted around copper ones. Regular solder won’t work. The factory service manual warned of these. Why would they use such dumb wires? Surely not so save a few dollars! Maybe they thought it affected the sound in some way. 🙄
Replacing the wires with regular copper ones was not a viable option as they are sealed through several internal partitions. I didn’t want to go there, so I got some special zinc/tin alloy aluminium-compatible solder to secure eye terminals to them for attachment to the binding posts. Dual wall heat shrink with internal hot-melt glue seals and fortifies the connections and thin regular heat shrink extensions form strain reliefs:
Direct wiring of the tweeter would be risky if an ordinary solid state power amplifier (without a protection circuit) were to be used. It’s OK for a valve amplifier though because DC cannot pass their output transformers. But for some protection against accidental low frequencies (like “somebody” inadvertently connects something to the wrong amplifier or crossover output), I installed a 25μF Solen metalized polypropylene (so-called “fast”) capacitor between the tweeter negative binding post and one of the “NC” (Not Connected) terminals and connected the internal tweeter return to that. The external part of the NC binding post got the heat-shrink treatment so a speaker cable cannot be connected to it by mistake.
This capacitor is tuned with the tweeter’s impedance curve to around 530 Hz which is several octaves below the active crossover frequency. At 530 Hz the phase twist of 45° is completely inaudible due to attenuation by the active crossover’s sharp roll-off from 2.88kHz. It could shift the phase response of the tweeter very slightly at the active crossover frequency where the wavelength is about 12cm, but this is good because the acoustic centre of the tweeter is less deeply recessed into the cabinet than that of the midrange (i.e. being closer to the listener a slight delay to the tweeter might be desirable). A block of soft foam was installed over the “hatch” (and that capacitor) before sealing the speakers again.
The speaker cables from the valve power amps were then connected positive-to-negative because the C-J preamp ahead of the crossover is phase-inverting. The bass speaker cable was connected conventionally (positive-to-positive) because the ports acoustically screw up the driver output phase at the bass-to-mid crossover frequency (the original passive filters inverted whereas the active crossover does not). The system is therefore what C-J describe in their pre-amp literature as “phase-correct” for better transient response. I don’t believe that for a minute because many crossovers invert phase at several places either side of their cross-over points anyway. The midrange here for example is competely inverted at the middle of its pass-band and my concern about that is non-existent. Be that as it may, the crossovers are phase-coherent and that’s all that matters.
A C-J MV-55 valve amplifier with two EL34s per channel drives the tweeters. I modified it from the standard ultra-linear configuration to 22 Watt per channel triode configuration per C-J factory instructions. Cleaner sound with more than enough power for the tweeters:
I tried out a solid state C-J MF2500 (about 420W/4Ω) on the bass drivers. It’s a mediocre amp good enough for bass, but its EI transformer vibration was too loud and annoying. It also has no soft-starter which it needs badly! It doesn’t have the sheer grunt and high damping factor of the 400W Class-D amp so I used that. It’s connected with 9 AWG cables to the woofer terminals:
Edit (Dec. 2012): I put the MF2500 power amp back into the system after finding a way to silence its noisy transformer! How do you silence a noisy C-J power amp without even opening it? Easy – just stick another transformer with an opposing radiated magnetic field on top of the chassis! 😀 This 250/218V autotransformer positioned accurately over the EI transformer of the C-J almost completely silenced it. Both have a magnetically reluctant/tranparent aluminium chassis:
Each midrange channel is powered by a C-J Premier Eight-XS monoblock (eight EL34s per channel push-pull triode output stage – 130W).
Despite what the marketing hype claimed, these are in my opinion very poorly engineered and excessively complex amplifiers. When working properly they sound superb but I couldn’t (and certainly wouldn’t) attempt to build anything like them. Unfortunately their overall layout provides horrendous hot spots to assure failure of their fancy pants marginally specified components (which do nothing better than standard parts anyway). The MV-55 is a much simpler and more reliable amplifier which sounds just as good within its power limits. Anyway, I learned how to fix them myself with some guidance from others and gave them a general overhaul and upgrade. But that’s another story (now here).
The 2.3kVA soft-starter/step down transformer gently starts the power amps – each of which is left with its power switch in the ON position.
The Active Crossover:
I looked at purchasing an Australian commercial digital active crossover/equalisation processor which uses algorithms to digitally divide and manipulate a PCM digital input signal, but I’m not overly keen on evolving (i.e. obsolete yesterday) digital technologies anyway, so I decided to take the already jam-packed active two-way analogue crossover chassis from the bi-amp project and cram another ESP P09 board (tuned at around 3kHz) into it for three-way use.
The new PCB (on top) is populated with very high quality components because it provides the critical tweeter signal and adds an additional “obstacle” to the remainder of the audio signal which it divides off. The red tuning capacitors are carefully matched Wima 10nF polypropylenes and the large white output coupling caps are ERD 1μF polypropylenes (yep – still an audiofool at the time of building). The opamps are National Semiconductor LME49720s (yep- should have used NE5532 I know) in gold-plated machined pin sockets. Even the metal film tuning resistors are of special grade. 😉 Actually all of this amounts to zilch, but costs were trivial and I thought “why not?”
The board is tuned at 2.88kHz which is slightly lower than the target figure due to RC pairing with obtainable components, but that’s fine and still within the tweeter’s lower range. It might even be benificial to limit the upper response of the midrange a bit.
The back panel of the chassis was removed to add two new output sockets.
The old 12V 4-pole muting relay had to go to make way for the extra sockets. Three 2-pole relays with 5V coils connected in series now fit between the sockets.
The Common pole of each relay is grounded whereas the NC poles tie to the individual outputs. The NO poles were snipped off:
The input sockets connect to the new PCB first. The high-pass outputs connect directly to the new “treble” output sockets. The low-pass is tapped off from underneath the new PCB to provide input signals to the old PCB beneath:
The “high-pass” outputs of the lower board now provide the band-pass signals to the midrange output sockets. The low-pass outputs of the old board go to the bass output sockets via the stereo subwoofer equalisers and rumble filter as before.
Prior to installing the new PCB (which blocks physical access), the input buffer opamp of the bottom PCB was temporarily left in place to enable calibration of the midrange gain trimmers.
I used the SPL meter just to get a starting point and ended up relying on my ears with swept tones and favourite music tracks to set the levels so that things sounded balanced. As with the original passive networks, the tweeter required reverse connection for a much flatter transition with the midrange.
There was a suggestion that some tweeters produce excessive frequency fluctuations when driven directly by valve amplifiers (and that the only solution was to use a solid state amplifier), but fortunately the swept response sounded flat (and impulse responses done later confirmed this). An expected Dynaudio Esotar dip at around 6kHz was not apparent at all. I couldn’t hear anything above 11kHz though, so tried this transistor amplifier and the result was the same – old ears. 🙄
A minor ground loop buzz in the midrange driver remained after connecting the preamp to the D/A converter. I thought I might get away with all but one amp having a ground loop breaker, but no – so I squeezed one into the MV-55 (tweeter amp):
… and the buzz diminished to almost nothing.
Then someone who ought to know what he was talking about suggested that the inexpensive NE5532 bipolar input opamps were about the best for bass performance, so I swapped all fourteen of the OPA2134s in the subwoofer equalisers and the rumble filter for them. It looked the same on the scope with sine waves and I’m probably kidding myself, but the bass might be a bit tighter when listening to music. Edit: NO – placebo!
First off I should say that I do like giant dynamics, brass and bloom in the sound of a hi-fi system. C-J valve pre- and power amps might be “to blame” a bit there. I do not like clinically accurate systems. This system does not produce that kind of dreary undynamic sound at all, but a colourful yet very full and tight room-filling and explosive sound, terrific front-to-back depth with amazing attack and bass that shudders deeper than you could possibly hear with no obviously harmonics from the woofers.
The direct-wired midrange without the passive garbage in front now has a proper frequency response and is so revealing that a very slight system hum brought about by the old under-shielded interconnects reared its ugly head, so I replaced them all with the good coax cables:
The system has now reached such a high level of performance that any further modification will be completely unnecessary.
I listened again to some of my favourite CDs including Liszt Piano Concertos 1 and 2 with Richter at the piano and the LSO under Kondrashin (Philips CD recorded by Mercury). The sound has almost reached a “close your eyes to be there” quality which absorbs you right into the recording. My only criticism is that the system is now so revealing that tape hiss on the old recordings that I love so much has become quite prominent… will have to tune that out mentally. 😉
There is a particular harpsichord CD that I love which now has the amazing instrument right in the room! Recorded in 1962 by Mercury. Absolutely brilliant presence (sorry for that). As well as the incredible music on the disc, there is deep, deep rumble (from something) and you can even hear people talking (at some distance – maybe in a nearby corridor or somewhere) at the end of some of the tracks during the prolonged sustain of the final chords.
Absolutely thrilled to have put together the best sounding hi-fi system that I have ever heard including some costing twenty times as much. It’s a shame that this silent web page can’t possibly do it justice.
This is still the most extended and most relaxing sound system that I have ever heard including my own active 3-way system built later on with double the bass power, a Linkwitz Transform and more than double the bass driver surface area! The Oriel’s mid and treble in particular are very smooth by comparison. The other system is superior in bass punch and is probably more revealing overall, but this can be at the expense of relaxing listening.
Well. I may have to eat eat my words. 😳
After learning a lot by doing impulse response measurements on the upstairs system and making certain changes, the two systems are getting closer. After reading Floyd Toole’s text book, it was time to measure this Oriels properly. Here is a graph of the left channel taken from the seat some 5m away:
Click for high res.
This is not a spatial averaging. It’s just how the system/room measures from the listening position after calibrating the whole thing crudely with an SPL meter and by ear. It shows that I like my bass, but it also shows a gradual upward slope from around 150Hz to around 1000Hz. The tweeter and midrange transition reasonably well (with non-inverted tweeter connection there is a giant suck-out). However it seemed that baffle step compensation might help flatten out the typical midrange response. So I added a 4.85dB circuit which has this curve (red = amplitude):
The circuit is installed at the inputs of the active crossover and looks like this:
R7 and R8 estimate the preamp’s output impedance up front, whereas R3 is the input impedance of the first P09 crossover board.
Anyway here are R5, R6, and C1 which replace coaxial cables between the input RCAs and the P09 inputs:
One of the terminals is not connected so is just used as an anchor point. Also the output levels needed adjusting to compensate for the upper range attenuation. Pretty basic stuff when you’ve done it before. And it sounds very, very nice! I’ll take some more measureements and post them here later. This is so much simpler and better than inductors and high current parts in a passive crossover.
Here’s the line level midrange output of the active crossover showing the downward slope produced by the compensator filter:
And here’s how the acoustic near-field (about 500mm) measurement looks with the mic at listening height and randomly off-axis horizontally:
Unlike the original passively crossed Oriels, there is no upward trend or anything untoward across the entire band. That was done in a room and no smoothing is applied! The measurement to the right → is not mine. It was done in 1992 by Electronics Australia Magazine and shows how poorly the original Oriels measured even in an anechoic chamber. The 4-5dB incline from around 100Hz to 1kHz is typical of cheap speakers of a similar baffle width when measured at a similar microphone distance and is related to the shape of the enclosure, but is inexcusable in a speaker retailing at around AUD20K. Jamo either was not aware of this massive problem or turned a blind eye to it. None of the audiophile reviewers even noticed it! Two resistors and a cheap capcitor completely eliminate it in my active system. The slope is not evident in the earlier graph on this page from that same publication due to the very close (5cm) mic distance used.
BTW if anyone ever says to me again “your frequency response is out across the mids”, it’ll only bring into question what their “reference” might be – and if it’s basically their own ESL or other panel-type atocities, note it down in my little black book as an interesting yet false opinion. 😉 BTW that graph has no smoothing applied. Simultaneous mid + tweeter polar plots to follow (est. Jan. 2018).
A few phase tweaking ideas
Having succeeded with a little phase shifting upstairs, it was time to look at twiddling the crossover phase responses of the Oriels to improve their behaviour. There wasn’t anything particularly horrendous, but I did decide to move the speakers further out from the walls to the exact ¼-¼ positions in the room for better imaging and also to move the listening seat forward from the back wall a bit. There was some silly talk on the ‘net by some famous “golden ratio” nutcase cable marketer that would have you place the speakers practically in the middle of the room, but that was rightly ignored. I also went completely against a couple of so-called “lectures” on room acoustics that I had attended just prior. Actually they were disguised sales pitches by salesmen pretending to be experts, but admitted to having never read Floyd Toole! Anyway in each of these the main speakers were placed in “smarty pants” positions smack-up against the front wall and it resulted in a complete lack of decent imaging and hopeless audible amplitude reponses on both occasions.
Well, I did read Toole and wanted to do it properly for a regular room that is not a near field monitoring studio! Problem was that at the new positions everything was reasonable except for at the crossover region between midrange and bass and there was a 400Hz suck-out (maybe a “floor” bounce effect) as well.
Some say that the proper way to achieve driver alignment is with digital delay using DSP, but I’m not interested in downsampling to some lesser processing format for the sole purpose of time alignment! Also my own measurements show that back where you sit (5 metres away from the speakers) the room swamps any high-precision alignment done to near-field measurements. Moreover I know that the brain ignores early reflections which the microphone sums into the graphs so most of the graphical anomalies can be considered “indicators” at best.
Step 1 was to tweak the bass response for the new speaker and seating positions using the stereo ESP P84 subwoofer equalisers. Here are the before and after traces of the left speaker alone (right was done separately and looks pretty much the same):
Red is before. Blue is after.
There’s a broad dip around the 102Hz bass/mid crossover region that EQ cannot address. You can see the feeble attempt to do so with the 100Hz subwoofer EQ adjustment which was later dialled out. This is a general misalignment of the 4th order bandpass subwoofers with the midrange driver – somewhat exacerbated by the P99 runble filter which poiltely screws the phase of the subs by about 50° at 100Hz! Some “contribution” is made by the baffle step circuit and the bass equaliser at line level too, but something was needed to address this. More below at Step 3.
Step 2 was to look more closely at he mid/tweeter crossover response. It would be easy enough to add phase shifters having done it once before (see the DSP or Not page). This time measurements found that nothing was needed. The Jamo designers did an excellent job in selecting drivers in which the acoustic centres (with inverted tweeter connection) were perfectly aligned around the crossover region. Of course they did not envisage a steep Linkwitz-Riley 4th order active, but nonetheless check it out:
In that near-field measurement with 1/24 octave smoothing applied red is the correct “inverted” tweeter connection and blue is the incorrect speaker wire reversal. The apparent taper over 8kHz is the cheap uncalibrated Behringer microphone. The cross-hairs are at the 2.88kHz crossover point. Nothing needs attention. Notice the midrange response – now much flatter due to the baffle step compensation circuit. The woofers were not powered for these measurements.
Back to the seat position measurements (previous graph) there are two remaining anomalies. One is a small dip at around 400Hz. It’s smack in the lower midrange and is probably a floor bounce suck-out, so I didn’t bother with it since I couldn’t hear it.
Which leaves Step 3 – to address that broad dip across the bass/mid crossover region. Flipping the bass speaker wires made it worse and I figured it might be a bit of a tall order to fix since the midrange driver is stretched very low already and the subs go wonky above 100Hz. As an aside, the Hypex plate amps in the upstairs system have a variable phase dial and that provided a 6dB improvement around their crossover region. So it seemed I might achieve some improvement here as well by adding an all-pass filter/phase shifter circuit. So I measured the very near field phase responses of the bass and mids (via the active crossover) separately. Here are the phase responses with the cross-hairs where they cross at 102Hz:
There’s only about 60° to play with so it wasn’t promising. Also the phases cross (are equal) just slightly above the crossover point at around 110Hz. The magnitude traces of the individual drivers (not shown) were each too low down at 102Hz. They could probably never sum to flat no matter what phase (or even digital timing) alignment were to be attempted.
Anyway after much deliberation and confusion I actually went to the extreme of modifying an ESP P99 PCB into a subwoofer 60° phase shifter… don’t believe me? Here it is ready to stack under the existing (unmolested) P99 rumble filter:
… with under-PCB hack job and all:
Then I changed my mind after having a little brainstorm (a little one is all I’m capable of 😉 ). I played a 102Hz tone and moved around the room to find where it peaked with a view to taking advantage of it! Low and behold – about 50cm forward from where I had been measuring at the sitting position there was a lovely 102Hz peak (SPL antinode) – easily located by ear. So I moved the lounge and microphone forward to that position and started over. So what if the drivers have inadequate overlap? Move into the crossover frequency room mode and recalibrate! You won’t find that one in the texts! Here are before and after responses with the 8-band subwoofer equaliser recalibrated to the new position:
Notice minus nothing at 20Hz in-room! Minus 5dB at 15Hz! And the in-room resonse would go even lower if not for the P99 rumble filter! 😯
Blue is the before trace (it repeats one shown previously). Red is after – no 102Hz dip but some new ripple and compromises above there, but somewhat better with no phase shift network added.
Well having listened to dozens of ultra high-end sound systems in people’s homes, stores as well as at a certain audio club, this system with all of its tweaks is by far the smoothest sounding hi-fi system that I have ever experienced.
Well my cheap little camera distorts the sound, but here is some fun anyway:
Here is a spectrogram taken from the lounge with both channels driven:
You can just make out the 220Hz concentration (outside the range of the graphic EQ bands) and some narrow very low frequency inter-band flares.