Evolution of the Crossover
The Ariel started out with a truly minimal crossover; a 2 uF capacitor in series with the tweeter, and a 16 ohm resistor in parallel with the tweeter (to damp variations in tweeter impedance). In theory it should have worked just fine. It measured very well, as I expected, since the drivers are exceptionally flat and easy to drive. It just didn't sound very good. It sounded like lots of other audiophile speakers out there; overetched, unnatural, with a sort of crude and annoying "look-at-me, don't I have great treble" quality. I played around with various values for the C and R, but the annoyance-factor stayed the same. So much for the minimal crossover approach. What's fashionable isn't always best.
After a while, I realized the source of the problem ... even the most deluxe dome tweeter doesn't like too much excursion, and that's exactly what a minimal crossover forces the tweeter to do. All direct-radiators (tweeters, mids, woofers, etc.) have diaphragm excursion that increases with decreasing frequency at a rate of 12dB/octave. (Drivers are constant-acceleration devices.) To prevent the excursion from actually increasing below crossover, you must use at least a 2nd-order (12dB/Octave) crossover ... even this rate simply keeps the excursion constant below the crossover point.
I didn't need too much additional persuasion to increase the rolloff rate on the tweeter; the one-cap crossover also required the tweeter to be connected in reverse-phase in order to match the midbass units (which have their voice coils about 1.5" behind the tweeter.) I'm not too fond of drivers wired in reverse phase ... I prefer to adjust the rolloff rates and other factors so I can keep all drivers wired in the same polarity. That way the group-delay errors are only at the crossover frequency, not spread across the entire treble region.
So I reverted to my familiar pattern of designing the crossover to carefully optimize the performance of each driver. The tweeter received a low-Q 2nd-order network (C1 and L1), a damping resistor in parallel with the tweeter (R2), and the standard voice-coil inductance compensation (R3 and C3). Much better! That's what the tweeter wanted all along.
It still needed work, though. The tweeter was fine, but I now could hear the generic MTM/D'Appolito forwardness from the midrange drivers. That aggressive and unmusical "home theatre" sound that is so common these days. This was something I was a little worried about when I started the Ariel project. Every commercial MTM system I'd heard to date had a forward and rather unpleasant midrange that was also quite directional ... the snarl didn't really go away until you walked into another room. I was now hearing it from the Ariels, and I wasn't too sure I could do anything about it. Time to find out.
Even though the Vifa midbass units measured beautifully flat with a textbook 2nd-order Bessel rolloff, something wasn't quite right. The mids were too aggressive, smooth MLSSA curves notwithstanding. When you set down the microphone and actually listened to music and pink-noise, it was quite obvious. So I took the path of least resistance, thought a little while, and picked a multi-tapped inductor (L5) from North Creek Music Systems so I could play around with the woofer rolloff.
I shut off the tweeters and just listened to the midbass units, adjusting the tapped inductor as I walked around the room. Much better, but not home free yet. I picked out several different capacitors and increased the rate to 2nd-order, starting with a low-Q Bessel slope. The original midrange coloration disappeared, replaced by a new one about an octave or two higher, near the corner of the crossover slope. Hmm, what's going on here? Adjusting the filter Q by changing the C/L ratio didn't fix the problem, so I took the direct approach and added an R in series with the C6 capacitor. Ah. That sounded good. The MTM coloration was gone for the first time. How interesting.
Now it was time to have a little fun. I wired a L-pad to 15 feet of cable, connected it in series with the C, and walked around the room while I turned the pot. There was a critical value at about 1/3 of the rotation; below that, MTM coloration, above that, all gone. Turn the pot all the way up, and the mids reverted to the L-only sound. I repeated this several times, marked the value on the pot, and measured the resistance. It turned out that 3 ohms was the critical threshold, so I wired two 6.8 ohm resistors in parallel and left it at that.
(Looking at the circuit a little more closely, it is evident that R6 and C6 in the woofer crossover approach a pole-zero cancellation for the woofer inductance, allowing the main 1.0mH inductor to reshape the response without interference from the woofer's rise in inductance.)
Now it was time to match up the tweeter crossover, but this was easy, since by now the midbass units had acquired a very gentle 4th-order Gaussian slope ... the way out of the troublesome D'Appolito coloration. I wouldn't have suspected at the beginning of the Ariel project that doing very subtle phase adjustments on the woofer crossover would remove the forwardness in the midrange.
The Ariel now sounded sweet, relaxed, and natural. The 2-meter on-axis measurement (shown above with no response smoothing) followed the intended 2dB slope from 100Hz to 10kHz with a very mild recession around the 3.8kHz crossover region. This is the classic "BBC dip", and much preferable to a "forward" emphasis in the upper midrange. Since ear is approaching its greatest sensitivity in the 2-5kHz region, even very small peaks create an unpleasant and unnatural sibilance. By contrast, a small dip in this region results in a slightly more distant perspective, and a more relaxed sound.
The "BBC dip" is in the most sensitive region of hearing (refer to a Fletcher-Munson curve) and very small adjustments in the value of the tweeter capacitor can "trim" the perspective of the soundstage. If the sound is too "polite" and distant for your personal taste, increase C1 to 5.6 uF; if it is too forward and "in your face", drop back to 5.0 uF. Whatever you do, please don't use generic metallized-film polypropylenes like Solen for the C1 tweeter cap.
I strongly recommend Auricap, Hovland Musicap, or North Creek film-and-foil capacitors, with an additional 0.1 or 0.22uF Teflon bypass for the last measure of transparency. Yes, these parts are expensive, but the tweeter cap is the single most important component in your high-fidelity system - more important than interconnects or speaker wires. Keep your priorities straight!
(For those of you who have instrumentation, you can check your work by intentionally reversing the phase of the tweeter. If the crossover is working properly and you measure at a 2 metre distance, you'll see a 20 to 30dB null at 3.8kHz. This means the phase-match between the normally connected midbass and tweeter is better than 6 degrees, which means the polar pattern is very stable throughout the crossover region. Don't forget to reconnect the tweeter in "normal" phase!)
All that's left is the impedance correction (conjugate) network at the input of the crossover. This isn't audible per se, but some power amps don't like reactive speakers. Without the impedance-correcting network, the impedance rises to about 30 ohms around 2kHz (due to the wide frequency separation between the low and high-pass filters). With the network in place, the impedance rise is limited to 8 ohms. The impedance curve stays between 3.5 ohms and 8 ohms from 10Hz to 100kHz, a simple load for nearly any amplifier. The average value in the 200Hz to 800Hz range is 4.5 ohms, so if you have a tube amp, connect the Ariel to the 4 ohm tap.
The whole process of crossover tuning took about 3 months, checking and re-checking with MLSSA and LMS each time I tried a slightly different crossover. 15 versions later, I was done, and Karna and I were very glad to settle back and simply enjoy our new Ariels.
The parts quality of the crossover components has a significant effect on the subjective balance of the speaker, and can easily outweigh re-balancing the crossover or even changing a driver. My experience with the Ariel crossover is that changing an important capacitor or resistor for one of equal value but a different construction requires at least some re-balancing to get the speaker "zeroed-in" again. For this reason, it is essential to use the specified parts (or better ones) for all sections of the crossover.
The usual audiophile preoccupation with costly interconnects is almost completely wasted when you consider that many $10,000 audiophile speakers use $2 metallized-polypropylene caps - if $150 had been spent on top-quality tweeter caps, you could save many thousands of dollars elsewhere in the system, and still end up with much better sound. But since consumers don't see the crossover, and don't realize the critical importance of crossover parts quality, they waste thousands of dollars on much less effective system upgrades. (When you buy the tweeter caps, ask for a tested and matched pair ... 1% matching would be nice if possible.)
Even non-polar film capacitors are slightly directional. Ask the vendor which orientation sounds best. Side A will go towards the input, or the signal side of the circuit. Side B will go towards the driver, or the ground side of the circuit. The only reason I can imagine for this phenomenon is that the caps are slightly self-shielding, and little bit of benefit is derived from placing the outermost foil towards the low-potential side of the circuit.
Other crossover parts are important as well, although not as sensitive as caps in the direct signal path. I recommend Jensen flat-copper inductors, North Creek Music System 10 and 12 gauge air-core inductors, and Ohmite, Mills, or Caddock 1% 10 watt non-inductive wirewound resistors.
Warning! Avoid conventional sand-cast power resistors, Mylar or electrolytic capacitors, and ferrite-core inductors. If you use low-quality crossover parts, the sonics of the Ariel and ME2 will be degraded all the way down to mid-fi status. Don't even think about it. On a related note, there are parts vendors who are making their own decisions about Ariel crossover parts without ever having built one; if they try to steer you away from the recommendations on the Web page, run, don't walk, to another vendor. If you're looking for reliable second opinions, the enthusiasts listed on the Index page are a great place to start. These dedicated builders have tried every part under the Sun, and you can take their recommendations seriously.
If you want to have fun experimenting with truly exotic parts, such as Teflon or silver-foil oil caps, start with the 5.0 uF to 5.6 uF series cap (C1) in the tweeter circuit, then go after the 10 uF shunt cap (C6) in the woofer circuit. The rule of thumb is that series parts that carry the entire signal are far more audible than components that shunt portions of the audio signal to ground. This rule-of-thumb applies to electronics and volume controls as well, so for components that carry the entire signal, use the very best parts you can get.
The best way to mount the crossover is to give it a box of its own. This isolates the capacitors from vibration (which they don't like) and lets you tweak in comfort while you listen. Just keep the crossover at least 12 inches away from any large iron or steel objects, which means don't put it next to your power amp or TV set.
Keep the tweeter and midbass filters electrically isolated with independent star grounds ... also, separate the inductors by at least 6 to 8 inches and place them at 90 degree angles to each other.
Bi-wiring (independent sets of speaker wires for the tweeter and midbass) is strongly recommended, particularly if the amplifier has a low damping factor (like many tube amps). My initial experiments indicate that single wires sound less dynamic and a bit "flatter" than bi-wire pairs, even if the single wire is quite exotic and costly. This is another area where experimenting can make a real difference to the overall sound of the system.
A Brief Digression about Crossover Design: There is misinformation promoted in some Internet news groups that all you need is the currently fashionable expensive driver and an off-the-shelf active crossover to equal $1,000 to $10,000 commercial designs. This, to be blunt, is just plain wrong. The most exotic and advanced drivers are notoriously hard to integrate into a speaker system, and professional designers know this. The best people in this field take anywhere from 6 to 18 months to design a system that is ready for sale, and that's with years of design experience and daily access to advanced test equipment.
If you found some unusual drivers on eBay, good for you, but now it's up to you to find a professional designer and pay them for their time (I am not volunteering). To give you a idea of what this might cost you, I generally charge US$20/hour for design services, and the Ariel took me 6 months to design (pretty fast for me), so by a quick reckoning that's $20,000 of my time, at least 80% of which was spent refining the crossover. Matching the crossover to the specific drivers in the Ariel took 25 years of experience, $16,000 of hands-on time, and $5,000 of test equipment. That was with drivers that were very easy to work with.
Other drivers, particularly Kevlar, carbon-fiber, Aerogel, or metal, take more time, since they have much more ragged response curves and require significant equalization and phase/polar pattern correction. Every deviation from flatness, especially in the crossover region and above, requires additional design time; sometimes after several months of difficult work, the particular combination of drivers has to be abandoned, and entire speaker system re-thought from scratch. This happens to the best professional designers, not just beginners; every pro, in the more honest moments, will tell you stories about the "the one that got away." Some drivers work together, and others don't.
The individual speaker drivers are not the star of the show; it's the harmonious combination of drivers, crossover, and cabinet that make it all come together. No matter how talented a soloist is, if they can't play with other musicians, they're no good at all. It's the same with speaker drivers. In a very real sense, the crossover is the conductor, responsible for harmonizing a complex dynamic system. That's why buying a generic crossover off the shelf is unlikely to give you the best results.
Since every system sounds different, there are two portions of the crossover which are intended as "user adjustments." The first is the discrete resistor L-Pad (R1 and R2), which raises or lowers the tweeter level, and the second is trimming the value of tweeter capacitor between 5.0 uF and 5.6 uF, which controls the 3kHz presence range. The Ariel crossover is quite sensitive to small variations in component value, so keep your adjustments small and do a lot of listening before making the next change. This is why an easily accessible external crossover is a good idea. When you make changes to the rest of your system, the Ariel can be trimmed to correct for any changes in tonal balance.
If you want the most extended treble, remove R1 and select 12 to 15 ohms for R2. If you want a slightly soft treble, select 1 ohm for R1 and 8 ohms for R2. After listening to many different amplifiers over a 2-year period, I chose the middle path of 0.5 ohms for R1 and 10 ohms for R2. The tweeter level is very much up to you, and I expect different people to select different levels depending on their system.
Trimming the value of tweeter capacitor between 5.0 uF and 5.6 uF alters the impression of "forward" vs "distant" sound. (Larger values bring the sound forward.) Transistor amps tend to have a a slightly distant midrange, so 5.6 uF is probably a good starting point. Tube amp owners will probably find 5.0 uF to be ideal for their system. The goal is for voices to sound natural and "in the room," but without edginess, shouting, or artificiality.
The values above 5.0 uF are especially convenient for combining a 5 uF cap with a smaller, and higher-quality, bypass cap. For example, good combination might be a 5 uF polypropylene film-and-foil paralleled with a 0.22 uF Teflon cap. Avoid combining a cheap metallized-film cap with a more costly metal-foil or oil cap; you can get the worst of both worlds when you do this. The Ariel and ME2 are extremely sensitive to both the value and the parts quality of the tweeter cap; this point cannot be emphasized enough.
As mentioned elsewhere on this Web page, the Ariels are designed for tube amps, not transistor amps; in the seven years I've had this site on the Internet, the rare e-mail grumbling about the sonic balance of the Ariel always came from a transistor-amp owner. If your existing speakers need 200 watts to make them sound good, yes, I can see why people buy transistor amps. But if your speakers have more reasonable power requirements, tube amps pretty much win hands down, especially if you include nicely restored vintage amps from Dynaco, Eico, H. H. Scott, or Fisher. A well-restored 1950's amplifier can quite easily hold its own against a $5,000 to $10,000 audiophile tube amp; try it and see for yourself.
Balancing the Speaker System
Since speakers are electroacoustic transducers and antennas at the same time, they have complex behaviour that cannot be reduced to simple graphs or curves. Even the most sophisticated instrumentation only hints at what's really going on, so careful listening is still an essential part of designing a speaker system. I use a combination of listening to music and pink-noise, and measuring the speaker with several different MLSSA test modes, such as impulse response, group delay vs. frequency, frequency response at different angles, and the cumulative decay spectra.
Careful listening to pink-noise is a very sensitive way to discover resonant colorations (more so than music, and much more repeatable), but it can fail to detect notches in the spectrum unless they are very sharp and deep. So you have to be careful when you tune a system so you don't inadvertently create broad depressions. In addition, pink-noise testing tells you nothing about dynamic qualities, so you can end up with speakers that are smooth and inoffensive but don't sparkle and sing on real music. Still, pink-noise testing lets you quickly detect and remove resonant colorations; just alter the crossover and add or remove cabinet damping until the speaker really begins to sound like an actual waterfall.
MLSSA, FFT, LMS, or 1/3 octave measurements provide an essential cross-check to make sure that you're not subtly skewing the spectrum as your ear gradually adapts to the sound of pink-noise. (Adaptation is a serious problem with pink-noise testing. Listening and tweaking sessions should be kept under 10 minutes so your mental reference point doesn't begin to shift.)
All tests have their blind spots, so cross-checking is very important, especially when you listen to music. You may have to choose between a sense of verve and directness and a tuning that is relaxed and neutral; this is your call. The drivers in the Ariel are exceptionally flat, which makes the system tuning easier. If the drivers had large peaks like metal-dome tweeters, Kevlar, or carbon-fiber drivers, the tuning process becomes far more difficult, and requires a lot of experience in knowing what to equalize and what to leave alone.
In practice, subjective tuning results in a 3-way round of testing, using pink-noise, measurements, and music listening. This is the sequence I use:
- Aim for the intended acoustic target slope (4th-order Gaussian or whatever) by using computer optimization with LEAP or XOPT, or use old-fashioned cut-and-try with many repeated measurements made at 2 meters distance at 0 degrees, 15 degrees, 30 degrees, and so on.
(WARNING: Do not make system measurements at a 1 meter distance. For all multiway speakers, not just the Ariels, the crossover radiation pattern doesn't fully "gel" until you get at least 2 meters away. If you fine-tune the system at a 1-meter distance it will almost certainly be wrong at 2 meters. Fortunately, the measurements at 2, 3, and 4 meters are very similar, so the 2-meter measurment is valid for greater distances.)
- Fine-tune the new crossover in half-dB steps by ear and repeated MLSSA, LMS, or IMP measurements. Keep doing this until you are satisfied with the overall technical and subjective performance.
- Record the measurements, crossover description and topology, and the version number.
- Have music-listening sessions using several amplifiers and a wide variety of sources. Instead of listening critically, ask yourself if you're enjoying what you hear. Do you feel the music? Does it move you? Focus on the emotional qualities, not just the usual audiophile sonics. In other words, what does it do well?
Go back to Step 2 until you feel genuinely satisfied that the whole system has reached its full potential. For example, crossover on this Web page is actually the 15th go-round using this multi-step procedure above. So don't expect perfection the first time around, and don't expect music-listening or measurement sessions to do it all. It takes both.
Speaker Location and Image Quality
The best imaging and spatial qualities result with 50 to 55 degree spacing between the left and right speakers, the tweeters placed towards the inside, and both speakers aimed at a point about 1 to 2 feet in front of the listener. In effect, this forms an equilateral triangle that ends about a foot short of your nose. When the speakers are correctly aimed, you will see the large-radius edge appear slightly closer to you than the small-radius edge, with an inch or two of large-radius sidewall showing.
Visualize a sphere with a 1 metre radius extending in all directions outward from the tweeter; keep this imaginary sphere free of obstructions of any kind. If your room is big and uncluttered enough to provide a free radius of 1.5 metres, so much the better. The less junk there is in the immediate vicinity of the speaker, the better it will sound. (No, various room-damping tricks won't help if the free-space criteria can't be met. You can't fool Mother Nature!)
If you can set them up like this, and have a pretty good system, you'll hear a smooth and even halo of sound extending about 2-3 feet outside the speaker pair and about 5-10 feet in height. If you're lucky enough to have top-quality directly-heated triode amps, these figures double, providing spatial qualities approaching the best and most natural multichannel systems.
By contrast, if the speakers are too far apart, or not toed-in enough, the center-fill will be weak and "phasey," and the intermediate center-left and center-right localizations will tend to jump towards the speaker. If the speakers are too close together, or toed-in excessively, the stereo image will compress, lose its natural quality of airiness, and off-axis images will disappear.
When the toe-in and spacing are correct, you will hear a perfectly even distribution of sound and a well-proportioned impression of space on nearly all stereo recordings. You should be able to move left-to-right over several feet and not have the image shimmer, wander, or collapse back into the speakers. If it does, you'll experience listening fatigue just from muscular tension trying to keep your head in the "sweet spot."
Unfortunately, most systems I see in homes, dealerships, and the CES have the speakers aimed with zero toe-in and spaced too close together. This gives the worst of both worlds: narrow, unstable images that collapse off-axis, intermittent depth perception, and a very limited impression of the spatial texture of the original recording.
Since many people, including audiophiles and reviewers, have never heard stereo as it's intended to be heard, try a simple experiment with a boom-box that has detachable speakers. Go ahead, put the thing on a kitchen table, space the speakers 60 degrees apart with yourself at the tip of an equilateral triangle, and aim the speakers at a point about a foot in front of you. For once, don't listen to the awful sound quality; just pay attention to the stereo impression.
The point of this simple experiment is to demonstrate that good, in fact, excellent stereo can be delivered by the crudest and most basic systems. There's no reason you can't have this same experience with much larger systems in your living room.
A Brief Note about Efficiency
There seems to be a misunderstanding out there about transmission lines and efficiency. Keep in mind that driver efficiency is almost purely a function of cone mass, radiating area, and magnetic strength in the voice coil gap, and very little else. Believe it or not, for a given driver, the box design has no effect on the driver efficiency. (Read the Neville Theile and Richard Small articles in the 1972/73 AES Journal if you don't believe me.)
If the crossover is really greedy, it can attenuate a lot of upper-band energy, but very few designers try to remove energy in the 100-500 Hz region, where any correctly operating driver is naturally flat.
So how did TL's get a reputation for being inefficient? Well, if you mass-load the driver by putting damping material too close to the cone (3 inches or less), that will indeed remove a couple of dB ... as well as altering the frequency response. Also, the heyday of TL's was in the mid-Seventies, when driver efficiencies were at an all-time historic low of 81 to 84 dB/meter.
Actually, TL's are well-suited for very efficient drivers that would otherwise have high cutoff frequencies if you used them in a conventional vented or closed-box Thiel-Small alignment. For example, the Vifa P13-WH-00 drivers used in this system typically provide an F3 of 83 Hz in a conventional vented QB3 alignment, while the Ariel transmission line system is still strong at 55 Hz. As you might imagine, there's a pretty big difference between 55Hz and 83Hz of bass response!
In addition, the transmission line rolls off more gently than a 4th-order vented system ... indeed, if the TL has the appropriate rolloff frequency, it can actually match the low frequency room lift and not seem to roll off at all. (To do this, though, the Ariels would have to roll off at 25 to 30 Hz. That's another project altogether.)
The Vifa midbass drivers in the Ariel are operated in parallel and produce an efficiency around 92 dB/meter (calculating directly from the Theile/Small parameters). The Scan-Speak D2905/9500 tweeter produces a measured efficiency around 90 dB/meter. These numbers are consistent with the target slope of an overall 2 dB tilt from 100 Hz to 10 kHz.
By measuring efficiency in the power center of the musical spectrum, which is a band between 200-800 Hz, the Ariels provide an efficiency of 92 dB/meter ... more efficient than most audiophile-class loudspeakers, and much more efficient than planar speakers and mini-monitors. In practice, a good 8 watt SE triode will fill the room, and a 20-30 watt amplifier will have all the power you could ask for. Remember that amps that clip smoothly (absence of feedback and good design helps here) can easily sound 2 to 3 times more powerful than conventional amps. Tube watts really are bigger watts.
Amplifiers and Sound Quality
The perennial question ... which amplifier? For most readers, it's simple; just get a classic old EL84/6BQ5 push-pull amplifier, and there you go. There are plenty of old Fifties-vintage Eico's, Dyna's, and Scott 20-watt amplifiers still kicking around, and restoration can be done any competent guitar-amp tech. The newer Single-Ended 2A3 and 300B amplifiers are a mixed bag; some are designed by newcomers who think that tube amps should sound like old AC/DC table radios ... warm and muffled. Anyone who's heard a properly restored Fifties amp will be quickly disabused of this notion ... they sound fast in comparison to many contemporary amplifiers.
When a highly-reviewed "audiophile" amplifier is outperformed by a Dyna Stereo 70, it isn't even really hi-fi, much less "audiophile." So for all practical purposes, a Stereo 70 is entry-level hi-fi. At $119 in 1958 dollars, it was entry-level back in 1958, and remains so today. That you can spend upwards of $3000 on a highly-reviewed audiophile amplifier, and get a less musical and realistic sound than a Stereo-70, is a unspoken comment on "progress" in the hi-fi industry. In a way, the good old Dyna Stereo 70 makes our lives easier; it makes a convenient go-nogo benchmark for assessing amplifiers, and a good fallback for those times when we get ear-burn from the current crop of electronics.
All right, now that we've established the minimum standard, where do we go from here? Good question! There are hundreds of tube amps out there ... triode, pentode, single-ended, push-pull, ultra-linear, pentode, etc etc. Some sound terrific, and some are so-so. Keep in mind that 90% of the commercial designs are variants of the Williamson, Dynaco, or Acro circuits of the late 1950's, with the most of the sonic differences coming down to parts selection. (If you look at amplifiers with KT88, 6550, EL34, or EL84 tubes, the percentages rise to about 95 to 98%) By contrast, amplifiers that use direct-heated triodes like the 2A3, 300B, SV572, 211, or 845 have a startling variety of circuit designs, and not surprisingly, sound pretty different from each other.
It's a cliche, but you have to listen for yourself, and remember that power amps sound strikingly different on speakers with different efficiency ratings. The amp that sounds best on an audiophile minimonitor almost certainly will sound different, and worse, on an efficient speaker ... and vice versa.
There are some very isolated folks who just can't get their hands on tube gear, or are afraid to make the leap to what to them is an alien technology. Unfortunately, good transistor amps are extremely rare; they measure nicely all right, but not very many are lifelike in the sense of delivering that "you-are-there" quality to music ... indeed, I would guess that many audiophiles have never heard that effect, even once, due to the degraded condition of modern electronics. Twenty years ago, transistor amps were brittle-sounding, grainy, and very two-dimensional sounding. Nowadays, transistor-amp designers have tweaked the "audiophile" breed of transistor amps to emulate the sound of mediocre tube amps ... sweet and rolled-off sounding, but with little realism or sense of depth. The grain and edge are gone, but so is much of the sense of life and sparkle in the music.
I know of exactly one transistor amplifier that avoids the trap of grain-n-grit on one hand and audiophile flatness on the other ... in short, it sounds just like a really good triode amp, but with more power! Unfortunately, it's not well known, so don't expect to see it at most dealerships. This unit is: the R.E. Designs LNPA-150. This 75-watt transistor power amplifier has the immediacy, the clarity, and the musical rightness of direct-heated triode amps ... but is a completely different technology. How is it done? Don't ask me ... transistor electronics are not my specialty. But it is possible if you move away from universe of conventional high-end audio.
Moving to the more specific question of how tubes and transistors sound with the Ariels, I've noticed that the midrange is more prominent and immediate-sounding with tube amps (clearer, more prominent vocals), while transistor amps seem to have a distant and recessed midrange, with a stronger emphasis on deep bass and extreme treble. Many transistor amps (especially audiophile models) just don't work with the Ariels, giving a dull and undynamic sound. Switch to a tube amp, and the entire system opens up and really sings.
It's not a matter of damping factor; adding a 1-ohm resistor to a transistor amp just makes it sound worse, moving even further away from the clarity and naturalness of the vacuum-tube amps. My only guess for what's going on is the transistor amplifiers are not at their best at the low power levels required by the Ariel, which much of the time is in the milliwatt region. Although theory indicates that Class A transistor amps would sound best with an efficient speaker, in practice I have found the reverse to be true, with the SE transistor amps being the worst type of amplifier, and bearing no sonic resemblance to SE triode amps at all. Looking a liitle deeper, though, and it turns out that SE transistor amps frequently use generous amounts of feedback to stabilize the DC parameters of the circuit, something that is not required by transformer-coupled triode circuits.
In short, avoid SE transistor or OTL tube amps with the Ariel. SE combined with feedback seems to magnify the defects of transistor sound, and OTL's (which frequently have high levels of feedback) can't deliver the current that a 4-ohm speaker requires. Not that there's anything wrong with OTL tube amps; they sound great with (some) 8 and (most) 16-ohm speakers. (Note to the reader: no, the Ariel cannot be converted to 16-ohm operation. Designing a good 16-ohm speaker is not as easy as it looks.)
Perhaps more surprisingly, the highly regarded audiophile favorites don't always come off very well with the Ariels. In particular, the Ariels are allergic to Krell, Cary, Pass, and Audio Research products; I'm sure they sound great with different speakers, but I can say from experience that these amps just don't work with the Ariels. To sum up, 3 watts isn't enough, 8 to 60 watts is just fine, and 100 watts or more is too much. The big tube amps with smoldering banks of EL34's or 6550's, or big transistor amps with heat sinks on all sides, are not a good match for medium to high efficiency speakers like the Ariel.
Well, that certainly sounds gloomy, doesn't it? Which amps sound good on the Ariels, if the ones that most folks know about aren't a good match? Well, I can say from experience the Audio Note Ongaku is a stunner, but then, at $85,000 or whatever it costs these days, it ought to be! The Ariels also sound wonderful with various different SE triode amps, such as the Reichert Silver 300B, the parallel-feed confections from Electronic Tonalities, and various others I can't remember right now. I was knocked out by the WAVAC 833 at the 1998 Winter CES, and I must say it is the most hard-core exotic design I have heard or seen, with rare and costly NOS WE473A inputs, NOS Genelex KT88's as driver tubes, and a spooky-looking 833A transmitter tube pumping out a solid 100 watts of single-ended power. This is one amp where the $35,000 price tag doesn't seem out of line, considering the extreme rarity of the parts, made-to-order Tango transformers, and the NC-milled 60mm-thick aluminum chassis. Oh yeah, it sounded good too, good enough to make anyone forget about audiophile amps forever.
Having had (more than one) glimpse of heaven, Karna and I were frustrated by the mundane sound of the amps that came trooping through for review by Positive Feedback magazine. (I don't do that anymore, so don't send review samples to me, send them to the magazine!) Don't go looking through all the back issues of PF trying to find the dogs; it's standard policy at PF not to print bad or even "ho-hum" reviews, and just send the product back to where it came from without comment. The products Karna and I did like were far, far out of our price range, even at deep discounts ... I'm in no position to spend more than the price of a new car on an amplifier!
So ... off we went on the adventure of designing our own electronics. Read about the Amity, Raven, and Aurora. If you want additional information, visit the tail end of the Aloha Audio page, and you'll see links to people all over the world.
Unlike the Ariels, I can tell you about our experiences, but I can't hold your hand, or offer any additional information, when it comes to vacuum-tube electronics. This stuff is dangerous! Not only that, things like good grounding technique aren't something you can learn over the Internet. Do yourself a favor, find an experienced tube-amp builder, guitar-amp technician, or ham-radio old-timer, and have them train you in the art of vacuum-tube electronics. Good engineering practice is not something you learn over the Internet, any more than you can learn to shift a manual-transmission car, or fly a plane. Some things require hands-on instruction from an experienced professional. Building a vacuum-tube amp is one of them. Good luck!
Ariel Tweaks and Bug-Fixes
Easy Tweak: Fill the upper chamber of the Ariel (or the side chambers of the ME2) with fine, dry, "silver sand" or lead shot. If the base of the Ariel is large and heavy enough, the added weight will not make it top-heavy. (A hazard for those with small children and active pets.) Some people have built large bases for the Ariel out of marble, slate, or Corian; that should be both heavy and very inert.
It's desirable to have these heavy masses at the top and bottom of the cabinet; that way, mechanical vibrations within the cabinet structure are absorbed at both ends with a minimum of standing-wave reflection. (I haven't done this, but others have, and have reported good results.)
Extreme Tweak: One reader in Holland mentioned removing the rear plastic assembly from the tweeter and using the built-in tweeter sub-enclosure in the Ariel & ME2 to replace the tweeter's own integral baffle.
Those of you who have more mechanical aptitude than I can succeed in removing the rear plastic assembly; all I succeeding in doing was knocking sharp little chips off the ceramic magnet. If you go ahead and carefully remove the rear chamber, make absolutely sure the air-seal where the wires go into the rest of the enclosure is air-tight. Any air-leak at all between the woofers and the tweeter dome will result in very high distortion and eventual failure of the tweeter.
If you succeed in modifying the tweeter, resigning yourself to the loss of warranty, and making absolutely sure there are no air-leak into the tweeter chamber, yes, there will be a performance increase for the D2905/9000 or D2905/9300. By making the rear chamber of the tweeter about 20 times larger in volume, the free air resonance of the tweeter goes down (good), the minor bump around 1.5kHz is removed (good), and midrange distortion probably goes down as well (assuming the air seal for the tweeter chamber is very, very good).
If you want to get even more tweaky, you can drill three very small holes between the tweeter chamber and the outside of the cabinet; this acts as a pressure-relief, or resistive-vent, for the tweeter, and would lower distortion even more. I would suggest drilling three 2mm or 3mm holes (in a vertical line) on the Left side of the cabinet, about 4.5" back from the plane of the front baffle. (This is for the Right cabinet; for the Left cabinet, drill the holes on the right side.) No matter what you do, the tweeter chamber should be lightly filled with long-fiber wool or crimped-Fortrel filling material, the same as the straight section of the transmission line. You want to get even more fancy, use the same Deflex or felt damping pads used elsewhere to damp the rear panel directly behind the tweeter.
This tweak is a major hassle, and I haven't done it myself. For those of you who are adventurous, though, it would almost certainly improve the D2905/9000 or D2905/9300 significantly, and make the L4, R4, and C4 notch filter unnecessary.
Bug Fix #1: On close reading of North Creek's crossover assembly manual, there appears to be an error on connecting the 0.3mH inductor used in the tweeter circuit. In the assembly manual, the "hot" end of the inductor is incorrectly shown connected between R1 and R2. The correct connection is between C1 and R1, as shown in the Web schematic. As you folks probably already know, the Web page is the reference master, and supersedes everything in the PF articles, the "Soul of Sound" book, or the North Creek manuals. I expect this error will be resolved in the next press run of the manual.
Bug Fix #2: Out of the hundreds of people who have written me about their experiences with the Ariel or ME2, myabe five have mentioned problems with dull, boomy, or recessed sound. If this happens to you, check the following items:
- The Ariels do not work in rooms with suspended ceilings ... they become very boomy and heavy-sounding. Of course, rooms with suspended ceilings have wretched acoustics anyway, which is why you never see them in performing spaces used by musicians. A very quick test of any room is the quality of speech ... when you first walk into a room and speak out loud, does your voice sound lively and natural, or is it dull and distant sounding? This tells you a lot about how a speaker will sound in that room. Rooms with low ceilings (8 feet or less) can also be difficult for transmission-line speakers, which energize the room boundaries.
For a "difficult" room, the ME2 is probably the best choice ... one of the charms of a minimonitor is they work in almost any acoustical space, a genuine advantage in an unpredictable hi-fi show environment. Yes, there is less bass, but what's there is less room-dependent
- If you must use a transistor amp, and are not happy with the sound you're getting with Ariels, try increasing the C1 tweeter cap to no more than 5.6 uF. This increases the "Q" of the tweeter crossover and adds a bit more overlap between the midbass and the tweeter, raising the level in the 3.5kHz region. This makes the sound more "forward" and aggressive - but it also diminishes the sensation of depth. I regret to say that many high-end transistor amps just don't sound good with the Ariels; if the transistor amp has banks of power transistors, electrolytics the size of beer cans, and complex floating bias schemes, it probably won't sound good on the Ariels. The Ariels are intended for simple, moderate-power amps, which in the transistor world is 60 watts or less.
- Did you remember to bi-wire the crossover going all the way back to the output terminals of the amplifier? This may seem like a small detail, but using a single speaker wire for both the tweeter and midbass sections results in a flat and undynamic sound. (This happened to me until I took my own advice and went back to the recommended bi-wire configuration!)
With bi-wire, of course, you are free to select different types of wire for the tweeter and midbass sections. I would particularly avoid stranded wire for the tweeter; microscopic corrosion can subtly increase distortion, impairing resolution of fine detail. Good alternatives to stranded wires are: flat (Nordost, Goertz), solid-core (Kimber), or Litz (Cardas, Nirvana). The "SuperCables Cook Book" by Allen Wright has terrific-sounding wires you can make yourself; this is advertised in Glass Audio or Sound Practices magazines.
- Are the crossover parts the recommended ones, or simply what was available from the vendor? The Auricap, Hovland Musicap, or North Creek film-and-foil capacitors are the best choice, along with Ohmite, Mills, or Caddock precision wire-wound resistors. I also recommend 10 or 12 gauge air-core inductors, even for the tweeter inductor. It may seem counter-intuitive, but as the tweeter inductor gets larger, the tweeter crossover becomes more effective at very low frequencies, greatly improving the clarity of the treble range.
Some folks are allergic to external crossovers; this is unfortunate, since capacitors are quite microphonic, acting both as electrostatic speakers and low-quality condensor microphones. You can't do much about the emission of sound (except consider oil-filled caps in metal cans), but you can isolate them from the heavy vibration of the enclosure by using an external (non-metallic) enclosure.
- Is the lining material on the interior cabinet walls the recommended 3/8" 85% to 100% wool felt? If other materials (like foam) are used, the sound will be adversely affected.
- For filling the straight part of the transmission line, never use fiberglass. Drop in small handfuls of cleaned and carded wool (best), or Fortrel or Dacron fiberfill. The filling should be very loose and fluffy; don't compress the filling material or try to pack it tightly.
- Did you "cut corners" and not use the recommended large radii on the cabinet corners? I know the large radii are a hassle, but they make a big difference!
If none of these fixes work, the most drastic solution is to adopt a "minimal" crossover for the Ariel, consisting of a single 2 to 3uF capacitor for the tweeter, and a single 10 to 16 ohm resistor across the tweeter terminals. No crossover at all for the Vifa midbass drivers. Some listeners prefer the more "raw" presentation of the minimal crossover; the choice is yours!
© 1996 to 2001 Lynn Olson.