Announcement

Collapse

INTRODUCTION - PLEASE READ FIRST TO UNDERSTAND THIS FORUM!

"This Harbeth User Group (HUG) is the Manufacturer's own managed forum dedicated to natural sound from microphone to ear, achievable by recognising and controlling the numerous confounding variables that exist along the audio chain. The Harbeth designer's objective is to make loudspeakers that contribute little of themselves to the music passing through them.

Identifying system components for their sonic neutrality should logically proceed from the interpretation and analysis of their technical, objective performance. Deviations from a flat frequency response at any point along the signal chain from microphone to ear is likely to give an audible sonic personality to the system at your ear; this includes the significant contribution of the listening room itself. To accurately reproduce the recorded sound as Harbeth speakers are designed to do, you would be best advised to select system components (sources, electronics, cables and so on) that do not color the sound before it reaches the speakers.

For example, the design of and interaction between the hifi amplifier and its speaker load can and will alter the sound balance of what you hear. This may or may not be what you wish to achieve, but any deviation from a flat response is a step away from a truly neutral system. HUG has extensively discussed amplifiers and the methods for seeking the most objectively neutral among a plethora of product choices.

HUG specialises in making complex technical matters simple to understand, getting at the repeatable facts in a post-truth environment where objectivity is increasingly ridiculed. With our heritage of natural sound and pragmatic design, HUG is not the best place to discuss non-Harbeth audio components selected, knowingly or not, to introduce a significantly personalised system sound. For that you should do your own research and above all, make the effort to visit an Authorised Dealer and listen to your music at your loudness on your loudspeakers through the various offerings there. There is really no on-line substitute for time invested in a dealer's showroom because 'tuning' your system to taste is such a highly personal matter. Our overall objective here is to empower readers to make the factually best procurement decisions in the interests of lifelike music at home.

Please consider carefully how much you should rely upon and be influenced by the subjective opinions of strangers. Their hearing acuity and taste will be different to yours, as will be their motives and budget, their listening distance, loudness and room treatment, not necessarily leading to appropriate equipment selection and listening satisfaction for you. Always keep in mind that without basic test equipment, subjective opinions will reign unchallenged. With test equipment, universal facts and truths are exposed.

If some of the science behind faithfully reproducing the sound intended by the composer, score, conductor and musicians over Harbeth speakers is your thing, this forum has been helping with that since 2006. If you just want to share your opinions and photos with others then the unrelated Harbeth Speakers Facebook page http://bit.ly/2FEgoAy may be for you. Either way, welcome to the world of Harbeth!"


Feb. 2018
See more
See less

6dB, 12dB, 18dB or 24dB crossover filters - surely the gentle 6dB (per octave) is the best?

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • 6dB, 12dB, 18dB or 24dB crossover filters - surely the gentle 6dB (per octave) is the best?

    This is a classic tale of misinformation, obfuscation and downright deceit!

    In a perfect world - the audio world of tomorrow - there would be no need for loudspeakers to have 'crossovers' because a single, miraculous drive unit would be able to cover the entire audio range from 20Hz to 20,000Hz. Until that wonderful day arrives, high quality reproduction demands at least two drive 'units'; one covering the fundamental notes in the bass/midrange and another smaller, lighter and more delicate unit (the tweeter) covering primarily the harmonics of music. We can do without the tweeter (AM radio is like listening without a tweeter) but the fidelity of reproduction is diminished, especially on classical music.

    So, like it or not, to design a hi-fi loudspeaker system that can fill a room with sound using today's technology, we have to divide the sonic spectrum between drive units .... and hope that when the two or more sound beams arrive at the human ear, that they can be combined in our brain to sound like one, full-range, original sound. Clearly this is asking a lot of our brain, and of the speaker designer. It is nothing more or less than an illusion, and one which a dog, for example, may not be capable of appreciating.

    The design of a hi-fi loudspeaker crossover has many aspects, and must be considered a holistic solution, just like the design of an aircraft. You cannot consider the design of the wings without considering their capability to generate lift and simultaneously and inescapably, their introduction of drag. You must also consider - simultaneously - their weight and their rigidity and at the same time, their flexibility in bad weather and their resistance to snapping off. Cost is another factor as is ease of servicing. So to tightly focus on one aspect of the design to the exclusion of all others is poor engineering, and most unlikely to achieve the overall best result. In engineering, a genuinely open mind plus rich experience about what has and has not worked in the past has taken us safely from the 707 to the double decker A380 in merely fifty years.

    The loudspeaker crossover's basic functions is to send bass/mid frequencies to the bass/mid driver and high frequencies to the tweeter. When expressed conversationally like that, it sounds rather mundane, even trivial. Is it? Superficially it's a very simple concept and can be implemented in a passive speaker with electrical components which would have been recognisable (although somewhat larger) well over 100 years ago. We use wound wire coils (which resist high frequencies) to feed the electrical signal to the woofer/midrange) and we use capacitors which block low frequencies to feed the electrical signal to the tweeter. So the absolutely simplest two-way crossover filter could be implemented with one component, a capacitor, blocking woofs getting to the tweet unit if it could be tolerated that some tweets were reproduced by the woofer/midrange. A more sensible start point would be two components as a bare minimum - a coil filtering off some of the tweets being fed to the woofer and that same capacitor as above blocking some of the LF from reaching the tweeter. So there you have it! A two-driver loudspeaker and its two-component, two-way crossover: the so called '6dB' crossover. Audio nirvana? Er... no. Audio madness? Probably.

    You can see an illustration of the basic one coil (green, marked L) and one capacitor (yellow, marked C) speaker here. The amplifier provides the voltage (left side of drawing) and you can just about make-out the coil and capacitor in the speaker cabinet. There is a schematic of the simple coil-capacitor network here.

    What we call the loudspeaker 'crossover' has, just like the aircraft wing, several other simultaneous functions aside from dividing the sonic spectrum between bass/mid and tweeter. You can get an overview from this article I prepared here. To list those functions:
    1. Division of sound between the bass, mid and HF drivers
    2. Adjustment of the relative loudness recognising that the tiny, lightweight tweeter is almost certainly louder than the heavy-coned bass/midrange driver
    3. Adjustment of overall response shape from the lowest to the highest frequencies - flat? tilt up? tilt down? hole in the middle? boost in the middle....?
    4. Adjustment of phase relationship between the drive units for flattest possible integration below, through and above crossover point
    5. Setting the vertical on-axis reference plane - level with tweeter? half way between mid and tweeter? above tweeter ....?
    6. Adjustment (or not) of 'baffle step' to correct for rising efficiency due to sound being trapped on front face of speaker cabinet with increasing frequency
    7. Impedance compensation (or not) to adjust electrical load seen by amplifier
    8. Fuse or protection of HF drive units
    9. Notch filtering to suppress peaks in drive units' sound output
    10. Adjusting the exact contribution from the bass/midrange and tweeter in the crossover region to get the very best sonics from the drivers, recognising woofers make 'dirty' sounding tweeters and tweeters can 'bark'

    Note! Whilst 1) can be crudely implemented with one or two components (read above) as a minimum, all other functions need more components! No designer should proudly claim that he has set component minimisation as his design goal (implied when the '6dB' crossover type is boasted), which means that he has turned his back on implementing points 2-10!

    To be continued.
    Alan A. Shaw
    Designer, owner
    Harbeth Audio UK

  • #2
    Effect of 6 dB/octave filters on stereo imaging

    The 6 dB/octave filter does have a "good" feature in that it allows for the perfect summation of the filtered highpass and lowpass components of the audio signal. Apart from the expected summing to unity, it can achieve zero phase shift in the summed response over the full frequency range. However, by virtue of the slow rolloff rate (6 dB/octave), the crossover region of interaction is very wide.

    It is apparent that individual loudspeaker drivers have their own response shapes, which are convolved with the filter response functions in any given loudspeaker design. For example, the woofer in a two-way system is likely to have a natural high-freqeuncy response that probably starts rolling off at around 12 dB/octave (or more) somewhere around 45 kHz. The tweeter's natural low-frequency response will start rolling off at around 1 kHz or so, also at around 12 dB/octave. The natural responses of these drivers will interact with the 6 dB/octave filters, so that the filtered low-pass and high-pass responses will no longer be purely 6 dB/octave.

    I've noted that at least one loudspeaker manufacturer has a model that appears to be based on low-cost 6 dB/octave filters, and a "high-performance" edition of this model is available that uses higher-cost higher-order filters (12 dB/octave I believe). This crossover is described by the manufacturer as delivering noticeably less overlap between midrange and tweeter, which is as expected.

    Assume that a woofer and tweeter combination with quite flat responses in each of their respective passband responses are chosen for a two-way system. For the sake of this example, let the woofer be 200 mm in diameter, and the tweeter 25 mm in diameter. In one prototype system we apply low-pass and high-pass 6 dB/octave filters, and get a reasonably flat summed response. In another prototype system we apply 12 dB/octave or even higher-order filters, and also get a reasonably flat summed response.

    All other things being equal (at least theoretically), in general, what can we expect from the stereo imaging of the speakers using 6 dB/octave filters versus those using the higher-order filters?

    Comment


    • #3
      Dandy theory ....

      Originally posted by witwald View Post
      The 6 dB/octave filter does have a "good" feature in that it allows for the perfect summation of the filtered highpass and lowpass components of the audio signal. Apart from the expected summing to unity, it can achieve zero phase shift in the summed response over the full frequency range. However, by virtue of the slow rolloff rate (6 dB/octave), the crossover region of interaction is very wide....
      This may be true and relevant when designing electronic filters in the lab, but is not relevant (or even true) when applying this theory to the design of the the crossover in a multi-drive unit loudspeaker. You have ignored physical offset of the drive units; the fact that the tweeter is physically closer to the listener than the woofer. That means that there is a time difference between the same note radiating from the tweeter and woofer, and in the crossover region where they are simultaneously pumping out sound, this phase difference will, if not corrected for, produce cancellation and/or addition of sound as the drive units fight each other: you'll hear that.

      We've discussed the story of speaker crossovers here and here.

      The secret of great crossover design is to implement a holistic solution that takes full account of the physical/acoustic/electrical realities of drive units screwed to a loudspeaker baffle. Not one textbook covering crossover design approaches the subject from a sufficiently objective and real-world perspective to make it possible to put the book down and actually devise a useful crossover. 6, 7, 8, 9, 10.... 14.5, 16... 21.... 27dB or whatever filter order (or any intermediate values) are truly irrelevant as design targets when it comes to applying speaker solutions in the real world. That explains why, 20 years after the first crossover simulators arrived, most hifi speakers sound far too unnatural and hard: the theory is only the beginning of a long and complex design journey.

      I have little patience with theoretical crossover solutions: they are quite useless in the real world and are likely to produce a disappointingly unmusical, unnatural listening experience.
      Alan A. Shaw
      Designer, owner
      Harbeth Audio UK

      Comment


      • #4
        Does filter order affect stereo imaging?

        Originally posted by A.S. View Post
        The secret of great crossover design is to implement a holistic solution that takes full account of the physical/acoustic/electrical realities of drive units screwed to a loudspeaker baffle. Not one textbook covering crossover design approaches the subject from a sufficiently objective and real-world perspective to make it possible to put the book down and actually devise a useful crossover. 6, 7, 8, 9, 10.... 14.5, 16... 21.... 27dB or whatever filter order (or any intermediate values) are truly irrelevant as design targets when it comes to applying speaker solutions in the real world. That explains why, 20 years after the first crossover simulators arrived, most hifi speakers sound far too unnatural and hard: the theory is only the beginning of a long and complex design journey.
        The holistic solution to crossover design is the only relevant one, of course. Your examples clearly demonstrate this, as did the papers published by Harwood in Wireless World all those years ago.

        Let's assume that enough filter components are used in a crossover design to perform suitable driver response equalisation in a multiple driver loudspeaker system. Is there any inherent benefit for stereo imaging when using higher-order filters versus a first-order filter design?

        Consider a popular system configuration based on a 150200 mm woofer and a 25 mm tweeter, with a typical crossover frequency located in the 24 kHz region. Could a loudspeaker system designer fully expect that a first-order filter network, with its attendant broad overlap of the woofer and tweeter acoustic responses, would have poorer stereo imaging characteristics than if a higher-order crossover filter had been used?

        Comment


        • #5
          Two issues

          Originally posted by witwald View Post
          ...Consider a popular system configuration based on a 150–200 mm woofer and a 25 mm tweeter, with a typical crossover frequency located in the 2–4 kHz region. Could a loudspeaker system designer fully expect that a first-order filter network, with its attendant broad overlap of the woofer and tweeter acoustic responses, would have poorer stereo imaging characteristics than if a higher-order crossover filter had been used?
          That's a very valid question. But are we overcomplicating the question by wrapping together two issues which are not, logically or actually, causally related?

          You mention ....
          • broad overlap of woofer/tweeter sonic output when the filters are 'first order', that is, only slowly fade out/fade up the contribution of the woofer/tweeter
          • stereo imaging

          How are these two issues related if at all? We could probably measure overlap with a microphone and a test signal, but of 'stereo imaging', what does that actually mean and how would we measure it? Is there any possibility that stereo imaging is (only) a personal, internal psychoacoustic experience and will necessarily vary from listener to listener as does colour sensitivity.
          Alan A. Shaw
          Designer, owner
          Harbeth Audio UK

          Comment


          • #6
            Modern trends in many loudspeakers - unnatural sound

            Originally posted by A.S. View Post
            20 years after the first crossover simulators arrived, most hifi speakers sound far too unnatural and hard...
            I can only but agree with the above observation. I wish I didn't have to, though.

            Having auditioned a wide number of speakers in the past few months, I have been disappointed with the general sound quality of most of them. Unnatural and hard seems to be a fair description of what I've heard.

            Classical acoustic guitar (gut strings) should not have a somewhat metallic edge to its reproduction. I've found vocals to be thin sounding, reproduced in a way that somehow lacks natural body. Backing vocals aren't well separated from the lead singer's voice, or if they are, then both sets of voices don't sound very natural at all. Metal-coned woofers, with their shiny and sometimes colourful finishes, seem to be a popular stylistic choice, while the overall quality of sound reproduction is far from satisfying. Cabinets using colourful plastic mouldings are common. Speakers within any given model range can sound significantly different from one another. And then there's even more variety in sound quality between the different model ranges offered by some manufacturers.

            It appears that very few manufacturers have a cohesive product range, where each speaker embodies fundamental aspects of a "family" sound quality, that is based on an engineering quest for natural sounding reproduction.

            Yes, most "hi-fi" loudspeakers these days do seem to sound far too unnatural and hard. It appears that the careful results-driven design approach pioneered at the BBC, as reported by Harwood in his 1968 and 1976 Wireless World articles, has bowed to the demands of marketing for multiple product ranges and price points.

            Comment

            Working...
            X