Some rambeling thoughs on Speaker design issues (part 1)
The following is abstracted and extracted from several pieces I have written over time and which reference fundamental design issues relating to speakers. I post this for information and discussion as many threads seem to focus on specific items in this but usually fail to attempt to percieve the larger picture.
Requirements of Transducers for Loudspeaker Systems
An often heard or read statement is:
"It is more or less established there that a real fullrange speaker goes from 20 Hz to 20 kHz +/- 1.5 dB"
In the same breath it also implied or intimated that this is the ONLY quality criteria for a good speaker.
I find this very a very interesting illustration of what preconceived notions can do to the way we view a subject.
First of all, the 20Hz-20kHz "audible" range is something concocted on a green table, by essentially drawing a line on a piece of paper. There was some reasoning and research behind this for sure, but that was also a lot that was ignored.
I would like to see some discussion of what matters in reproducing music and musical instruments, instead of waving specifications and numbers written on paper around and saying "It must this otherwise it's not Fullrange or Hi-fi or whatever....".
Now I am well known for having substantial "prejudices" about Audio Technology. I often quite dogmatically declare certain technologies to be superior. Much of the background for these dogmatic declarations comes from a study of music itself, working professionally in recording and reproducing music and in recent years from carrying out a number ďall else being equal" blind tests in a relaxed, unpressured but very controlled environment.
Moreover, much of I find to matter from a rational view of the subject, based on physioaccoustic research and psychoacoustic research as well as electronics has been born out in listening tests, indicating usually a quite substantial correlation of theory and realityÖ
But not, I have not yet formulated the GUT (Great Unified Theory), not even for music reproduction, nor do I intent to. I'm an Engineer in the end, all I want are parameters to which I can design equipment so it makes a better job out of reproducing music. And that is where we come back to transducers, especially in our case Speakers.
And one of the things that matter in absolute terms least, but which are being used as primary criteria in establishing the quality of Speakers is Frequency Response. Now before we have a close look at what REALLY matters in speakers, let's first look at a few connected and highly intriguing statements (both are from memory and correct in content, but not verbatim):
1) Spectral analysis on a Harpsichord shows that a played Harpsichord will produce significant energy above around 8Hz reaching up to beyond 72kHz. (someone actually researching the behaviour of musical instruments).
2) The Frequency range needing to be reproduced for a music lover (as opposed to an Audiophile) is 100Hz to 9kHz. (a well known critic of classical music, - he later revised his view as to mean 50Hz - 12kHz in the light of hi-fi Equipment allowing this range easily to be reproduced).
3) The 3m in room smoothed response of the Speaker fitted into a commendably tight 10db Window between 100Hz and 8kHz. (D.B. Keele in a review of a well known audiophile Speaker in "AUDIO". The accompanying graph from a Crown TEF showed appx. +15db and -20db extremes for the unsmoothed deviation from 0db in the same 100Hz to 8kHz band. The same speakers measured almost perfectly flat anaechonically. The Speaker also was judged quite favourably subjectively in this and many other reviews).
So from the above it appears that we have two substantially competing or even contradicting views.
Many people will agree that the 100Hz - 10kHz range is the range within which most of the musical "information" is encoded. Yet we will also find that in the real world and at our listening position a perfectly flat response Speaker will be subject to many narrow band rises and drops, overall making about 35 db between the lowest and highest point and that is only in the musically "relevant" 100Hz - 10kHz range.
At the same time, many Instruments (If we take into account "parasitic" sounds - like the mechanisms on the Harpsichord giving rise to extremely low tones at quite high intensity) will produce sounds over an incredibly wide range of frequencies.
So, do we need speakers capable of 8Hz - 72kHz (+/-1db), or is 100Hz - 10kHz (+/-10db) more than good enough in the real world? And is the Frequency response actually particularly important as measure of quality, particularly in the context of current 2-Channel Audio?
Allow me to take as an example one specific Instrument in order to illustrate the kind of demands made by a 2-Channel (say ORTF Array miked) recording of the instrument chosen. For this example we shall use the "Kettle Drum" or the "Timpani". This a very large Drum, tuned to a fairly low resonance, depending on a number of issues often in sub 50Hz Range.
When struck by the "mallet" the impact will move the Skin quite far off it's resting position, instantly generating a strong pressure wave of very low frequency, which is actually only a "spike" into one direction, having a very steep rise. In addition, the interaction between mallet and skin will produce a kind of "white noise" wideband sound reaching up to above 2kHz. Both the pressure wave and the noise originate from the same point in space, the point of the Skin where the mallet struck.
We will then (after the initial Impulse, when the mallet has been literally bounced back by the restoring force in the Skin and has left the skin) see a starting oscillation at the tuned resonance frequency of Skin and Drum "Kettle". This Oscillation was stared by imparting a Impulse to the Skin and will reach quickly reach a maxima and fade out slowly unless damped out. The origin of the resulting sine wave (and it's harmonics) will be much less localised than the first, as the whole skin now participates as well as other parts of the Instrument.
If we where to place a flat response transducer at the position of the Drum and where to take a Frequency response sweep with the two Mic's of the ORTF array placed as usual we would find a severely non-flat frequency response for all of this.
Yet, energy between a few Hz and a few kHz is being produced, even with the relatively low tuning of the drum. This is because a non-linear resonating system (like a string, a Drum Skin and so on) produces not only Harmonics of the Fundamental, meaning tones with twice trice and so on Frequency but also Subharmonics of the Fundamental, meaning tones with halve, a third and so on Frequencies.
Moreover, a particular phenomena in Human hearing (it is based on a mixture of the physioacoustic behaviour, or so to speak the mechanics, of the ear and some psychoacoustic components, or so to speak the signal processing in the brain) also has some impact. If we hear the correct set of Harmonics from a tone with the fundamental (and even subharmonics) fully suppressed we will still hear the actual tone (or note) as our ear and brain system "reconstructs" the actual note from it's harmonics.
Similarly, the absence of one specific harmonic again has little impact on the perception of the tone of a note.
At the same time, the perception of the placement or position of the Instrument is actually being determined by initial wideband impulse and pressure wave, which hence if not reproduced time coherent will result in a diffuse perception of the Instrument. Interestingly it seems that also our perception of dynamics is involved here.
BTW, the perception of the note itself is of course based on the actual fundamental Frequency, while the perceived "tone" of the instrument is based purely on the harmonics (and to a degree subharmonics). Lastly, it is interesting to note that all natural events follow a roughly gaussian distribution, with usually quite identical shapes of the distribution curve to both sides of the peak. This also holds for instruments and their harmonics and subharmonics, which are also forming (on a linear frequency scale) the usual common "bell" shape known to anyone ever having studied statistics.
All this is quite complex a subject and one that it would have enough material for a few dozen dissertations in it, so don't expect to even make a full outline of what all this implies. Let me just come to some of my conclusions here.... (in part 2)
Requirements of Transducers for Loudspeaker Systems (part 2)
A Transducer intended to reproduce music (including human voice and the like) should have a fundamentally balanced response over a fairly wide frequency range (as wide as possible) but it MUST UNDER ALL CIRCUMSTANCES be balanced over the 100Hz - 10kHz range, using 1/3 Octave averaging. Narrow-band deviation from a flat response are fairly harmless, especially notches. It is desirable to have the transducer to offer some Energy outside the 100Hz - 10kHz band, ideally on par with the 100Hz - 10kHz range.
By only reproducing the 100Hz - 10kHz range it is still possible to capture all relevant fundamentals and harmonics, including the harmonics needed to "restore" the relevant very low Notes below 100Hz and the Subharmonics to "restore" the noise type happenings above 10kHz.
Indeed, all the music, all notes and phrases will be able to be followed and pretty much all nuances will be captured.
The slight impact is on the tonality of certain instruments, especially at the high end with such as Harpsichord, Percussion and the like, as well as the physical sensation form the fundamentals and pressure waves of the lowest notes. While these items contribute to a closer "illusion" in many cases, it is entirely possible to enjoy music without them.
It appears therefore also quite important that our Transducer shall not alter the distribution of Harmonics, so any distortion should be low, of the type giving a simple Gaussian spectrum and ideally of very low order. Iíll keep the coverage of Distortion at that, past mentioning that most speakers distort a lot and that in conventional Cone/Dome drivers there are mechanisms that tend to link certain types of distortion to sensitivity in an inverse way (eg higher sensitivity for a given bandwidth = less distortion).
Furthermore, if any semblance of correct dynamics, timing (I know, I did not cover this earlier) and positioning are desired, it appears that the larger part of the 100Hz - 10kHz range should be reproduced in a manner and fashion that is time coherent. This usually implies the use of a single transducer (even multiple transducers driven in unison seem to lead to some blurring due to their no longer being in unison at higher frequencies.
The above points strongly towards the absolute need of using high(ish) Sensitivity, wideband Transducers if our goal is the musically correct reproduction of the Music.
The reasons for this are as mentioned the time coherence which even coaxials of the Tannoy or Altec mould cannot manage (though the old RCA LC-1A 15" Coaxial did). Also the presentation of two point sources (matching the 2 point "pickup" from our Microphones) and the low Distortion are relevant.
As we have seen, Frequency response as such is not a primary issue at all. IF the "sensation" of "being there" is desired also on a physical Level (impact) it appears the use of well designed Subwoofers (or the use of wideband drivers with a capability to go low enough) seems desirable.
In addition, for those people who still have a decent HF hearing (myself being a case point) it might be desirable to have drivers with more extended HF or indeed Supertweeters. I prefer to use and can hear the difference made by a Supertweeter crossed in at a nominal 22kHz [-6db] covering up to about 40kHz.
Considering the extensive body of knowledge that exists on room acoustics, the appearance of the "average room" and so on it comes again and again as a surprise to me how ill considered and conceived most so-called "High Fidelity" loudspeakers are, especially if one compares the situation to the sound reinforcement sector and serious studio monitors (meaning NOT the Yamaha NS-10 or BBC LS 3/5 or similar "Scherzo's").
Speakers interact with the room in various ways, both due their own behavior and due to the different regions of the room discussed above. The key to much of perceived tonality is the behavior of room and speaker though the reverberant range, because in this range we find the fundamentals and formants of most instruments and the human voice as well as all the harmonics. Above I commented that in reverberant range the speaker behaves a lot like a light source.
For a long time two specific types of speakers have been sold to the unsuspecting public that have by design severe behavioral problems in this range. One type is the Full Range Dipole (Electrostatic or Planar) and the full range omnipolar speaker. Neither type is found in serious Studio or sound reinforcement applications for exactly this reason.
Dipoles and Omnipolar Speakers create way more problems than they solve, by radiating energy over a very WIDE frequency range and also a very wide area. Think simply of a naked lightbulb in a room. This is your omnipolar speaker and to a lesser degree your dipole. This makes the frequency response in room within the reverberant range maximally dependant on the room symmetry, absorption etc. and thus will require major efforts on room treatment to correct for the problems caused. Of course, a notable minority of listeners actually likes the presentation of omni's and full range dipoles and they by all means are welcome to their preference.
Another major problem source are modern "conventional" HiFi Speakers. They misuse a semi omni directional HF unit (the dome tweeter was designed by Stu Hegeman for applications radiating upwards and giving semi omni directional hemispherical response) in combination with a cone unit that at higher frequencies coming up to the crossover frequency is actually quite directional and as a result the directionality changes strongly with frequency. The use of non coincident sound sources also adds to the problem as now in at least one plane (lateral or vertical depending upon orientation) the radiation is turned into several "beams" going off on sides of the main (forward) one.
All in one unholy mess and one that has done much to cause the rather unnatural, upper midrange emphasised sound of modern HiFi systems, by placing a depression in the upper fundamentals and a strong emphasis onto the lower harmonics for the response off axis, assuming a flat on axis response.
As an aside, Dipole Speakers using dynamic cone drivers tend to remain dipole only to around 300-500Hz, with the rear output attenuated at higher frequencies by the lowpass filter formed by basket and trapped air plus the acoustic "shadow" of the magnet system. So, unless extra rear midrange and tweeters are used to re-establish the dipolar pattern at higher frequencies cone driver dipoles tend to minimise the rear output through much of the reverberant range, one more advantage to recommend them.
Above I said that in the modal range the room becomes resonant. Pressure Maxima and Minima are scattered around the room in a complex pattern. All normal enclosed dynamic speakers operate in the modal region effectively as omni directional radiator (regardless of being a domestic LF Horn, Reflex, sealed), using a DIPOLE to cover the modal range will actually result in a rather well behaved Bass reproduction mostly free of extreme resonanceís in room modes. The way the dipole interacts with room modes will result in a much more even LF response by exciting room modes less.
A dipole will maximally energies room modes if placed in the areas that with conventional speakers turn into "black holes" and the least if placed where normal speakers energies room modes most. I am sure you have heard remarks that "speakers sound best where they do the maximum visual damage" and "speakers sound worst where my wife wants them...". Well, dipoles, for low frequencies at least are the opposite. They sound best where your wife wants them and worst where they do the greatest visual damage to room decor.... :D
What follows from this is of course that 90% of ALL HiFi speaker designers got it all *** forward. They make hybrid Electrostatic or ribbon Dipole speakers that are dipoles where they should directional (midrange/treble) and that omni directional where they SHOULD be dipolar (bass) or they make speakers that have a totally uncontrolled and widely varying directivity with frequency. Normal Box speakers really can only take refuge to special room setups or equalisation, they DO NOT WORK without them, if High Fidelity is desired.
It seems in the last 70 or so years of sound reproduction only three or four companies ever "GOT IT" (or at least got it to some degree), namely Hartley, Celestion with their SL-600 & dipole Subwoofer combo, Gradient in Finland and the now defunked Audio Artistry headed up by Sigfried Linkwitz and the Linkwitz DIY Designs. Surprisingly, the pre-stereo area music center and/or large Valve radio tended to be equally dipolar at low frequencies and even many early stereo music centers where equally dipolar in nature at LF. The old Hartley Concert Master Speaker (22" Dipole Woofer and 10" sealed box fullrange speaker) is surprisingly close to an ideal domestic speaker and the usual common 2-Way floor standing of stand mounting 2-Way "High Fidelity" speaker is surprisingly far from being such.
Well, as we know, that was when designers trusted their ears more than their measurements (as accurate measurements where difficult and measurements showed poor correlation with what was heard), before people started to believe that because it could be measured it was meaningful and that a good performance in the few and one-dimensional areas that where measured actually somehow equated to audible performance. But that as they say is another story, back to room acoustics.
Below the modal range our best choice (should we wish to extend our response that low) is a simple monopole subwoofer, either sealed, passive radiator or vented with a suitably low cutoff. Much about such is available all over the net, no need to add.
Requirements of Transducers for Loudspeaker Systems (part 3)
Anyway, back to some numbers so we can start designing something.
These numbers focus on the MINIMAL ability to reproduce acoustical Jazz and classical Orchestra, as well as Vocal Works like choral music and opera. Modern styles like rock and pop which are usually fully electronically created and have no "acoustical" Yardstick against which to compare are not being considered, yet they are usually also well served by Transducers covering or exceeding the noted specifications. Some of the requirements are not based on material discussed here, I'll come back to this later.
The Transducer(s) should be able (IMHO) to reproduce 96db SPL at the listening position with less than 1% THD of the simple type in the 100Hz - 10kHz range (this includes also the amplification devices) and with less than 0.3% THD in the 300Hz to 3khz range. (For stereo this applies to a pair of transducers in typical positioning, being driven by equal amounts of power).
The Transducer(s) should be free from significant Power Compression for 96db SPL at the listening position. (For stereo this applies to a pair of transducers in typical positioning, being driven by equal amounts of power).
The Transducer(s) should have a even energy balance in the 100Hz - 10kHz range.
The Transducer will desirably have notable amounts of energy in the whole range covered accurately by the recording, meaning 30Hz to around 50kHz on LP, 10Hz to about 12kHz on CD.
(Note: above 12kHz anything coming from the Output of the DAC/CDP playing a CD was not part of the original recording, but is spurious sound generated by the digital Filter or simple Distortion if no oversampling is used, and donít go quoting Nyquist on me until you at least understand what he really wrote and what the limiations of his theorem are).
The transducer should become more directional with rising frequency to reduce the room influence by promoting direct sound over the reverberant field and should have a well defined directivity to at least the lower reach of the Modal range of the listening room.
The transducer must be capable to preserve the fundamental Waveform of a triangular wave and a squarewave at least for the 300Hz - 3kHz range, a much wider range being desirable.
At any extend, it should be clear that as far as I'm concerned a truly flat 20Hz - 20kHz frequency response is nothing that would be per se required for the reproduction of music. Indeed due to room influences having such a transducer would actually result in severely lifted Bass response in most rooms and hence an ill defined and boomy response.
As often a narrow HF dispersion is also being criticised, I would like to add that wide, uncontrolled dispersion of the upper midrange and treble destroys imaging. It actually promotes the rooms reverberant field over the direct sound and at the same time it promotes a overly bright sound as the usual "housecurve" for higher frequencies is not promoted, instead the energy balance in the reverberant field remains flat or worse, severely uneven.
Lastly, it should also be clear now that using multi-miking and multitracking during recording will severely compromise all aspects of the musical performance recorded even WITHOUT massive additional "doctoring".
I hope all these slightly unconnected ramblings and notes help to clarify some of the reasons why I feel that wideband Drivers MUST be used and why I feel that the nowadays very common 3 or 4-Way, flat response, low efficiency Cone/Dome speakers are mostly incapable of producing "high fidelity", if fidelity to the musical event is our goal.
I guess I should state another philosophical qualification for the missive posted above - High Fidelity of the reproduction only really makes sense and is applicable if the recordings are nade in a manner and fashion that allows a largely unaltered reproduction to create a convincing illusion. Many recordings are not made such.
Very interresting read. Thank you!
Very very good read.
But remember, I need a subwoofer flat to 8 Hz to play the lowest organ note...!
Otherwise, another option is the "chimney transmission line", if you live in a suitable house that has Chimneys from old fireplaces but central heating. The usual 2-Story house will have a chimney around 7 - 8m tall, in other words a nice 12Hz quarterwave transmission line if blocked up on all fireplace entries but one. In that entry place your 15" 'free air' woofer with low FS (if neccesary weight the cone) and voila, with a little EQ an 8Hz capable "Supra-Woofer" is born.
As usual, merely a little thinking out of the Box will get a lot fo results.
just went to a concert monday, twas "hawkwind", some aging space cowboys.
pa was kinda small but right loud.
all the music I heard came through these speakers. estimated about 5Kw output medium efficiency, say 97 db broad range. certainly more headroom was available in the higher frequency ranges.
sound quality was pretty rough.but all instruements and voices were well defined. micing person permitting
after the gig had a look at the amps on stage. some bog standard 12" and 25" cabs , plain amps, sure mikes, accoustic drums(amplified during gig) and a plethorer of keyboards, computers and effects units.
has this anything to do with what you say,
only maybe if I'd made a bootleg with a MD
but if its straight from the mixer ?
I like electronic and accoustic music, I like my speakers
and certainly a worthwhile tweak after design is flat response 20-20
+-1.5 if you can smoothed in room.
Good to hear Hawkwind is still in action. Did they play 'Silver Machine'?:)
Re: Some rambeling thoughs on Speaker design issues (part 1)
|All times are GMT. The time now is 09:39 AM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2015 DragonByte Technologies Ltd.
Copyright ©1999-2015 diyAudio