Hello all joenetters, old and new. As you may know (or not, or just sigh at) I find the subject of speaker distortion and ways to reduce it VERY FASCINATING (more so than building and listening to tube amps, even). A while back, I believe I shared with you some translations and/or adaptations of some old articles written by our (Scandinavian) local hero of speaker design/understanding, Ragnar Lian. He's inspired me before, and he does it again these days with some very intriguing and inspirational postings on the Norwegian newsgroup no.audio, in response to someone's curiousity regarding current drive of speakers. If any of you people paid attention to all my past raves on these subjects, you will see that Ragnar sums up many of the things I've been trying to summarize myself. He does it way better, though. But some "new" things may be of universal interest, as when Ragnar explains "alnico sound", in the text below (which I hastily translated from Norwegian this afternoon, please pardon the lack of lingual refinement and listen to this aging man instead).
I can probably not blame any of you for not knowing who Ragnar Lian is or what he's contributed to the field of audio and speakers in partcular. But he can look back at half a lifetime's involvement in the speaker research and industry, having worked with Tandberg, Radionette, Seas, Bang & Olufsen, and co-founder of Scan-speak, affiliation with Dynaco and numerous other audio companies etc. etc. He now works designing industrial magnetics for a Norwegian firm called Contec Design (CT-scanners and such). It's a shame he's not a member here, really.
Personally I think the below text (directly translated from three successive postings to no.audio) is packed full with stuff that should send people like us scrambling to the drawing board/shop/think tank, whatever...or at least rob us of *some* peace of mind and induce some humbleness with respect to the goals we all pursue in some way or other. If he repeats something, it's for a reason, so read it and THINK!
Here we go:
Oh, my - granddad Lian feels touched. Words about current-drive sound like music to my ears, and it's about time! So...one problem will naturally appear, the damping factor! But what IS that? It's a wool blanket we hang in front of the speaker to obscure the mechanics. A loudspeaker is a piece of mechanics! Those who can't picture mass and springs physically may look at the speaker's equivalent diagram, where BxL forms a feedback loop. It does the exact same thing as in amplifiers by producing pretty measurement data for an unconscious pig-design. It does nothing wrong, but it obscures a problem.
Same thing with the damping factor - it doesn't do anything wrong, but it provides an acceptable frequency response curve for a poorly designed bass speaker. An oscillating circuit consisting of a weight and a spring stores energy, and if this energy is not absorbed mechanically or is minimized in the acoustic design, it will be sent back to the amplifier at the most bizarre points in time. Wasn't it the good old NS-1000 that sent 40 amp spikes back to the amp, paralyzing it? I have a dim recollection of a demo by Otala some time in the late 70's, where one of the most respected amps was loaded by an 8 ohm resistor and an NS-1000, respectively. The amp's output was coupled to a tape recorder. At the demo, this tape was played back, with the recorded difference of the two different loads. The difference was like night and day!!!!!
The moral is: A speaker should be loaded in a controlled manner, that is, resonances and other tendencies of energy storage should be treated where they appear, not hidden by the damping factor. When a speaker is properly loaded, the damping factor naturally doesn't do any harm, but doesn't improve anything either! If one is short of ideas, examine patents from 1925 up to about 1950, back then the problems had to be solved without the aid of damping factor, which meant that there was no other way than fixing the problems where they originated. Lots of more or less intelligent patents.
Date: Thu, 22 Oct 1998 11:21:48 +0200
I have received some response to my posting on this subject a week ago, and some questions of "how to do it? " have appeared. There is no simple answer, it has more to do with a mindset and "programming". Some times it may be necessary to take a step back, not to stay there, but to get a better overview.
In the sixties, some idiot took the loudspeaker out of the radio. Pretty smart, with regard to placement and furnishing, but with time it led to a disaster. The disciplines of amp design and speaker design were separated, with a minimum of communication between the two professional groups. A voltage interface was defined, in which amp designers made amps loaded by 8 ohm resistors and speaker designers made speakers based on a source impedance of 0 ohms from the amplifier.
The amp designer, on his side, made a great effort to achieve marginal figures of harmonic distortion into 8 ohms (0.0000001%), whereas the speaker designer was so busy playing with Thiele/Small parameters and theoretically flat frequency response that he never had the time to philosophize on distortion, perhaps occasionally some harmonic, but certainly not intermodulation.
A sad, but overlooked fact is that the electrodynamic speaker is controlled by current through a voice coil, not by a voltage across it. As voice coils are more or less filled by solid iron, this causes a pronounced cubic current distortion, usually on the order of 0.5-5% !!!!!!
This demonstrates two things; 1: The nice distortion characteristics of the amplifiers seem wasted, as the driving current is distorted by iron in the magnetic system, but, 2: It can still be heard that the amplifier has been improved far below the level masked by speaker distortion. This shows that distortion measurements by themselves don't say everything about sound quality, rather that it is the TYPE of distortion mechanism we hear. This is encouraging!
But now, if we had an amplifier that provided undistorted current rather than undistorted voltage, we would have achieved a drastic improvement of the total distortion.
The paradox is that the loudspeaker is controlled by current, and the amplifying elements, whether tubes or transistors, are by nature current generators. There's ALWAYS anode CURRENT, collector CURRENT and drain CURRENT which is being controlled. Using various components and trick circuits this is converted to a varying voltage which is supplied to the speaker, where the driving current is strongly distorted by an iron filled voice coil.
Pages can be written about distortion mechanisms in loudspeakers, this I will skip this time, instead I have a small digression about transistors. Has anyone pondered why it's called emitter, base and collector? See, when the transistor was developed it proved to be a current controlled device, actually it was a diode consisting of collector and base, and with the emitter a current was injected, emitted into the PN layer between collector and base, and a larger current variation ocurred between collector/base than the injected emitter/base current. The first germanium transistors were PNP, an amplifier was drawn with the base upwards from ground like a T, the collector was connected to - (minus) (here, + was connected to ground), and a - (negative) signal current was injected into the emitter, the arrow pointing inwards.
Note that I am not saying that this is the only right hookup, but some weakly programmed engineers should allow themselves to think the thought, perhaps even process it. Some really fine transistors with linear current amplification existed around 1970, wasn't it the D44 and D45 series from General Electric, perhaps they have gone into hibernation in some corner at HARRIS ?
You can't make current controlled speakers and current amps just like that, and make them communicate. For 40 years we have been strictly indoctrinated as to what was correct, and the whole design philosophy both in speaker design and amp design is based on "the way" ! This will take time and effort to change.
Regarding speaker design: we need some communication between disciplines. Tor Forsman's demo at the AES meeting showed that not only the classical radiation impedance is important, but that this can be supplemented by aerodynamics. Isn't there an airplane designer with a passion for loudspeakers somewhere? It's nothing mystical, just a different discipline.
For the impatient - the introductory question - yes, it's possible to get started. You take a super high current op-amp such as LM12, and make a current feedback loop around it. This will not be a hi-fi amp, but is a good start for observing effects.
Regarding current controlled speakers, there's a bunch of creative patents from the 1925-1950 period, when practically all amps were high impedance and thereby current generators.
And a little warning, particularly in loudspeakers there are many distortion mechanisms at work simultaneously, which in part of the frequency range are in opposite phase. Therefore one may experience that when one distortion mechanism is removed or reduced, a higher total distortion will be measured while the speaker sounds cleaner.
Is it possible to stir anyone's curiosity?
Subject: V?rluft om h?sten/current drive
Date: 26 Oct 1998 22:04:20 GMT
I'm thankful for the response which has been mixed and diverse, and which partially calls for a little more information on the matter.
Some have also mentioned servo controlled speakers. I'll treat this first, as it's a fairly concrete matter. I would not say that they don't work, I struggled with them a great deal in the mid-70's. It worked nicely at bass frequencies. If you tried to push the cone inward, it struggled to oppose this motion, using all the power available from the amplifier. But it had its cost: Nice deep bass, but rotten midbass. This was because of phase shift in the servo loop, popularly speaking: the problems were shifted upwards to a higher frequency. The conclusion back then was that it was better to use money and resources to improve the driver units themselves. Philips did market some motional feedback speakers, but they sounded unmistakably like Philips!
Now, back to the real issue, current driving speaker units. I don't have any simple recipes, for the simple reason that this has been a non-issue the past 40 years. Current drive is not the sole solution either, but it's practically necessary to go this way in order to understand how a speaker works, then go on to design drivers with improved properties. Looking at today's speaker units, one is amazed that they don't sound a lot worse, they're packed full of mistakes.
The reason why I bring this forth on the net instead of designing and getting rich myself is that I'm beginning to get old, and have a list that it would take half a lifetime to follow up. And when I say "tease somebody's curiosity" it's not because I have exciting completed stories to tell, but because I have lists of interesting problems that should be solved by those crazy enough to do so. Developing "canned music" requires, apart from a certain amount of theoretical knowledge, musicality, freedom from prejudice and a dose of insanity!
The perfect speaker is absurd, and will only be created in advertisements, but a huge potential for improvement exists compared to what we've seen the past 25 years. It was tremendously exciting when Benson, Thiele and Small presented their calculation models for loudspeaker enclosures, and they helped understanding bass reflex enclosures in particular. But at the same time something like a blockage was introduced, as everything circled around these transfer functions, and the computational models for the speaker were adapted to these functions. In itself nothing wrong, but it caused everything not described by these models to be regarded as nonexistent !!!!!!
But now back to my list of problems and paradoxes:
1: Current drive vs. voltage drive:
The force generated by the voice coil is equal to B x I x L, where B is the magnetic field, I = voice coil current and L = effective length of wire in the magnetic field. We struggle to make an amplifier with extremely low distortion, and then the force from the voice coil, caused by the current through it, is dramatically distorted (0.5-5%) because the coil is partially filled with iron, which has eddy currents and hysteresis losses which in turn cause distortion.
Furthermore, the coil has a self-induction that causes a slower buildup of current and thereby a time delay or loss at higher frequencies. Furthermore this self-inductance is dependent on the amount of iron within the coil, and will vary when the coil moves. This causes intermodulation, as a lower frequency that produces a fair amplitude will modulate a higher frequency. (As far as I recall, the woofer of the AR3A modulated a 600Hz tone by 50% (!!!!!!) when it was supplied with a 20 Hz tone at nominal power.)
Also, the voice coil gets warm, and the resistance increases by 0.4% per degree celsius so that the current, and thereby the output drops correspondingly. Typically this causes a compression on the order of 4-6 dB. If the speaker is driven by a pure current rather than a pure voltage, these effects disappear or are dramatically reduced.
(Symmetric drive, a copper cylinder long enough to short the voice coil in all positions, was my answer to these problems associated with voltage drive, and was used by Scanspeak from 1973 on, and by Dynaudio since 1976. The system was not perfect, but gave an improvement by a factor of 3-10.)
If one thinks that introducing current drive fixes everything, one is stuck in the glue. Since 1960 we have been accustomed to "pig design" a loudspeaker's acoustic rear load, the damping factor has concealed our lousy craftsmanship. Here I thought of the old patents from the midwar years. Not to rehash them, but the idea was to make rear loads which reduced or eliminated the speaker's resonance, that is, that the mechanical system was damped in itself.
I would like to highlight a Carlisle patent from about 1930, describing all conceivable variations of enclosures with coupling ports/slits to free air or to one or several separate chambers - a wellspring of ideas and a valuable aid each time one ends up in an argument with someone or other who's reinvented something.
2: Modulation of the magnetic field.
This is a neglected area. In the 50's they philosophized a great deal over why the voice coil was displaced in the magnetic field under conditions of strong drive, among other things Cunningham defined the "solenoid effect". If one takes the trouble to measure the force from the voice coil inwards and outwards respectively, at say an applied coil current of five amps, one will realize that there is a large difference. The coil pushes harder outwards than it pulls inwards. Somewhat dependent on the design, this difference will be on the order of 5-50% (!), causing both offset and compression.
I have, just to tease, done calculations on three different magnet systems; my old Scanspeak system with 18 mm coil, 100 windings and a 120 mm ferrite magnet, but without SD (Symmetric Drive). Then an imagined system with a 40mmx40mm alnico magnet, and one system with a 40x10 neodymium magnet. These three systems have roughly the same magnetic field, statically speaking. The ferrite system has an asymmetry of 25%, the alnico system has 4% and the neodymium system at about 2% ! Mysterious? Not at all! If we picture the voice coil, 18 mm long, 100 turns, at 5 amps, this produces 500 amp-turns, or a field strength of 28000 A/m, this is a magnet "let loose" around the air gap. It is directed with or in opposition to the main field, depending on the direction of the current, and acts directly upon the field in the air gap, and also shifts the operating point of the magnet. The ferrite magnet's operating point is at about 2500 gauss, and has an incline such that it can easily be shifted 2-300 gauss, that's quite a few per cent! Both the alnico magnet and the neodymium magnet have operating points up around 10-11000 gauss, and an incline which makes them somewhat "stiffer", meaning that they can't be moved much. Both neodymium and alnico have the property of having a permeability close to 1, in other words the same as air, and they do not affect the inductivity of the voice coil! Then we can go back to the early 70's, when JBL spoke of "The Alnico Sound" - that was no lie!
3: The loudspeaker as accumulator of energy:
All reactive components, both electric and mechanic, store energy. This energy is converted to electric or acoustic energy at a point in time over which one has no control. The ideal speaker has neither mass nor springs, self-induction or capacitance. Therefore the ideal speaker doesn't exist, but we don't have to remove it as far as possible from the ideal. Low mass, compliant suspensions and low voice coil inductance.
All of this about speakers, and I have barely touched the voice coil. No solutions, just a lot of exciting work, for those crazy enough!
I don't have a lot of experience with amp design, but it should basically be easier to make a current amplifier than a voltage amplifier, if we disregard the past 40 years of programming. Even though telephone technology is primitive with regard to distortion and frequency response, I believe there is a great deal to learn from it in terms of methods. It's really impressive, before satellites one could phone all across the world over a pair of twisted wires thanks to current-based technology. That never would have worked with voltage!
- -- end of forwarded message --
_/\_ Thomas Dunker \ The Horn Speaker Home Page: / \ P.O.Box 2811 \ http://invalid.ed.ntnu.no/~dunker/horns.html | | 7002 Trondheim \ | | NORWAY \--\ "Those with head above water \____/ firstname.lastname@example.org \ see only the tip of the iceberg" |||| phone (+47)7391xxxx \ (Gene Dalby)