Bentoronto,
Thanks for the reference - but something seems to be wrong this side of the ocean/this side of the hemisphere/muscle between my ears. I tried to upload the M1 schematic, but the PC comes back to me with a "file damaged, cannot download" ruling.
Never mind. What you describe is close enough; yes, that could be more than eliminating cable effect. I think we have thus dealt with the topic as being of negligible importance in normal systems.
For the rest, I also mostly agree with other posts on the loudspeaker matter. What little I might add would again be a matter of priorities. Loudspeaker imperfections viz-a-viz room acoustics - that is most difficult to quantize. Correction of the cone movement (in fact voice coil movement - not quite the same thing) should be easy what with modern digital capability, but I have difficulty in ordering my priority list with regard to the relative merit of each component of that.
Thanks for the reference - but something seems to be wrong this side of the ocean/this side of the hemisphere/muscle between my ears. I tried to upload the M1 schematic, but the PC comes back to me with a "file damaged, cannot download" ruling.
Never mind. What you describe is close enough; yes, that could be more than eliminating cable effect. I think we have thus dealt with the topic as being of negligible importance in normal systems.
For the rest, I also mostly agree with other posts on the loudspeaker matter. What little I might add would again be a matter of priorities. Loudspeaker imperfections viz-a-viz room acoustics - that is most difficult to quantize. Correction of the cone movement (in fact voice coil movement - not quite the same thing) should be easy what with modern digital capability, but I have difficulty in ordering my priority list with regard to the relative merit of each component of that.
Priorities can be pretty idiosyncratic. But by many standards, attacking speaker distortion has to be high priority as low hanging fruit, tall poppy, bang for the buck, matched-component concept, etc.
The hard part is estimating what effort, cost, educational value, and joy will be for the experimenter.
Here's help. Allow for destroying a speaker or two in the process. Does that challenge you or does that depress you?
Folks with resonant boxes... sorry, nothing here for you. Horns with sealed enclosures inside (like the wonderful Klipschorn) prolly are OK. Other horns and "horns" are maybe's. ESL's - big question mark.
The hard part is estimating what effort, cost, educational value, and joy will be for the experimenter.
Here's help. Allow for destroying a speaker or two in the process. Does that challenge you or does that depress you?
Folks with resonant boxes... sorry, nothing here for you. Horns with sealed enclosures inside (like the wonderful Klipschorn) prolly are OK. Other horns and "horns" are maybe's. ESL's - big question mark.
To get over the cable capacitance issue we could use this technique on the feedback cable. EDIT: actually, better to use dual shielded, ground the outside shield and attach the inside shield to the buffer. So use shielded coax or something similar with buffered shield. Page 96:
http://www.linear.com/pc/downloadDocument.do?navId=H0,C1,C1154,C1009,C1028,P1219,D4138
Since these are audio frequencies we can trade speed for linearity in the buffer... I have played in the simulator with circuits to do this.
- keantoken
http://www.linear.com/pc/downloadDocument.do?navId=H0,C1,C1154,C1009,C1028,P1219,D4138
Since these are audio frequencies we can trade speed for linearity in the buffer... I have played in the simulator with circuits to do this.
- keantoken
Last edited:
.... but yet again, one gains by putting matters in perspective by looking at practical figures. Even with the worst loudspeaker cable at 8 ohm, one finds that cable capacitance only begins to act at many 100kHz up to over 1 MHz, depending (again, sticking to domestic situations).
(This apart from some evidence that certain amplifiers can loose their cool with moderate cable capacitance, amazing as that might sound. But that is not the topic here.)
(This apart from some evidence that certain amplifiers can loose their cool with moderate cable capacitance, amazing as that might sound. But that is not the topic here.)
I'm not talking about the speaker cable capacitance. Most amplifiers are designed to tolerate cable capacitance. It's the cable taking feedback to the amplifier I'm talking about. This wire's characteristics directly affect amplifier stability as well as noise, etc. But by using coax and shielding it, we introduce capacitance that might affect the feedback loop. However, if we drive the shield with a voltage parallel to the conductor, the capacitance can't react, and if the buffer has low output impedance, EMI will drain through that into ground.
Although we could do it for the speaker cables as well... This would allow us to make tighter, more linear amplifiers by offloading cable capacitance. I'm sure it's been done for some time in the test equipment industry and RF.
- keantoken
Although we could do it for the speaker cables as well... This would allow us to make tighter, more linear amplifiers by offloading cable capacitance. I'm sure it's been done for some time in the test equipment industry and RF.
- keantoken
It seems like using a second, identical voice coil, just to provide feedback, which sounds like a great idea at first, might be problematical, since its dynamics would be different, due to the lack of the speaker.
But if we could somehow have the position of the center of the speaker cone as a feedback signal, we could wrap that back to the amplifier's input. We could take the input signal and put it through a delay filter, to time-synchronize it with the fed back output signal, scale the fed back output signal to the same amplitude as the input, and subtract the input from it, to get an "error" signal, which we would then subtract from the actual input signal, to try to correct the output (not entirely sure if I got all of the +/- signs correct, there; Look at Bob Cordell's thread(s) on "error correcting" amplifier feedback.).
I designed and simulated many such "error correcting" amplifiers, a couple of years ago (using just the power amplifier's output, not the speaker's center deflection, as feedback). It was usually very interesting to see what was fed to the power amp's input when, for example, a square wave was the original input signal but the power amp was a little lazy and didn't want to slew quite fast enough to reproduce the required slew rate, and/or would "bend" the rising and falling edges (or ring like crazy with a pure-capacitive load). The error-correcting feedback system would present, to the power amplifier input, square waves that had spike-like additions to the edges, which resulted in a nearly-perfect output from the power amplifier.
One particular 33-Watt amplifier that I simulated (based on three paralleled OPA541E chipamps for ease of simulation) was capable of driving square waves into up to 15 uF with no overshoot (or undershoot). Into 2.2 uF, it could do rise times of less than 4 us with zero overshoot & undershoot, for 30Vp-p square waves, although, the rise time was limited by the intrinsic rise time capabilities of the OPA541 chipamps, and the input was first passed through a first-order low-pass RF filter with F(-3dB) of 678 kHz. It was gratifying to see the output having _exactly_ the same slew rate as the input, and even (almost) the exact-same curvature and shape where the rising edges transitioned into the flat tops of the square waves.
That power amplifier had a gain of 0.9879 and a frequency response from DC to F(-3dB) of 375 kHz, with F(-0.1dB) of 40 kHz. Using an input sine of 20 kHz that had a THD-20kHz of 0.000019%, the output THD-20 was 0.000107%. (But I hadn't tried to optimize the THD-20.)
To delay the input to be time-synchronized with the output feedback, I used a six-pole Sallen-Key 0 dB Low-Pass filter with the appropriate Group Delay (800-something nanoseconds, in that case), modified for zero DC offset, with a response type that was Linear Phase, 0.05-Degree Equi-Ripple Error, with an F(-3dB) of 563 kHz. I used free software from ti.com (FilterPro?), to design the filter, and used fast LT1363 (1000V/us) opamps to implement it in the LT-Spice simulator (free, from linear.com). At the time, I noted that I might want to use a higher-order filter to get more delay with less low-pass filtering. (I think that the free version of FilterPro was limited to sixth-order filters, at most, for the type of filter I was designing.
In case anyone is wondering, I did also include parallel parasitic capacitances across all resistors in the circuit, and series parasitic resistances in all capacitors, and even parasitic inductance in the power supply rails. (But I did not bother to simulate ground-returns' parasitics, for that one.)
Sorry to have blathered-on, for so long, about all of that.
- Tom
Last edited:
In my opinion, the future is in powered speakers, with no substantial lenght of wire to the amplifier and no risk related to speaker feedback miswiring, not to mention easy transportation and set up and how much space is gained by getting rid of amplifier racks. I'm working in that field. Nowadays we can make very compact and efficient amplifiers.
The future is also in one amplifier per driver and DSP crossovers built into the speakers.
There are many old habits and myths to throw away.
The future is also in one amplifier per driver and DSP crossovers built into the speakers.
There are many old habits and myths to throw away.
Last edited:
Sorry!
My fault (and it wasn't even old year's eve). Agree fully with what you posted.
No prob Johan, it doesn't look like the original topic is being discussed anymore, though I don't know what the point would be.
Eva's comment about about amp+speaker combos really does sound like the future, as long as the signal gets to all the speakers digitally and without phase issues. Sounds like a wireless job, or IR/laser for those wanting something old-school. Speakers could be placed anywhere without worrying about cables.
I think the largest hurdle for direct cone-movement feedback is getting a device that measures movement accurately. A laser would work well I think, if you could find a way to mount it inside the cabinet without it vibrating, and mount a flat reflector to the driver. And then what do you detect the signal with? My, oh my... Well, some speakers have a hole going in from the back straight to the dustcap, I'll bet that would be the best place for some sort of transducer.
- keantoken
Eva's comment about about amp+speaker combos really does sound like the future, as long as the signal gets to all the speakers digitally and without phase issues. Sounds like a wireless job, or IR/laser for those wanting something old-school. Speakers could be placed anywhere without worrying about cables.
I think the largest hurdle for direct cone-movement feedback is getting a device that measures movement accurately. A laser would work well I think, if you could find a way to mount it inside the cabinet without it vibrating, and mount a flat reflector to the driver. And then what do you detect the signal with? My, oh my... Well, some speakers have a hole going in from the back straight to the dustcap, I'll bet that would be the best place for some sort of transducer.
- keantoken
Mmmmm ...
Over the years (yep; very old experience me, but hopefully not limited to old school) I sometimes had the impression that folks were trying too hard to make one thing for all purposes. For me and perhaps a few others here, my 'music centre' is a stationary thing; moved as little as the furniture. And I never had any problems with cables - not audible now, or even when I still had the ability to descern 21 kHz (lab experiment). Purposefulness sometimes suffered unrecognised in the dazzle of the newest 'convenience-for-a-day' - er - convenience.
So I will stop at courses-for-horses.
I need to be convinced that there exists a relatively simple alternative to sensing the loudspeaker signal current, everything concerned (two voice coils, accelerometers, what-have-you). Goodness knows, there were enough short-lived efforts, though the audible effect was something else. The problem with a driver is as much cone behaviour over its whole surface as other matters, not detectable with whatever 'ideal' device one pictures to follow cone movement. I have read reports that made good sense, that even half-accurate methods of such compensation need to be committed to a particular loudspeaker (system). Not to be over-sceptical (our good friend John Curl had a saying in his avatar: 'Condemnation without examination is prejudice' - or such), but I will be naughty to conclude that the Crown "Delta-Omega" points to just that: A 'small resistor' - thus the old series-R way of sensing loudspeaker current (or make it 'series-Z').
If so, it can be done as they possibly do. Although I previously mentioned that it can be trouble-prone depending on the circuit, I have tried that with success on a 20 cm main driver plus 20 mm tweeter plus cross-over system - the reachable amplifier output impedance was -3 ohm, thus cancelling a meaningful portion of the loudspeaker internal resistance. But the audible advantage was minimal except in the bass region, to which I subsequently limited the effect. (By the way, the year was 1970. And Connoisseur had this in their valve amplifier in 1955 - full frequency.)
But I ramble. Apology to Crown if I am wrong. As Keantoken said, this is 'much ado about how much in practice'?
But perhaps I am not post-"post-present".
The "resistor sensing" method for loudspeakers is quite the same as a current feedback type amplifier as opposed to a voltage feedback type. There are also mixes between current and voltage feedback...
The principle force in a loudspeaker is an inductor. This inductor produces a magnetic field which causes the cone to move.
And so I've always wondered, why in the world we drive these speakers with voltage instead of current? Because it's not voltage that determines the strength of the magnetic field as much as it is current. This issue was "fixed" by making speakers a fixed resistance, so that a specific voltage translates into a specific current and thus magnetic force on the cone, but this barely even works because of the reactivity of the mechanical properties of the loudspeaker. This "resistance" changes with frequency. So what you have is that the magnetic force of the coil, which depends on current, not voltage, is only indirectly affected by the amplifier's output, and is at the mercy of cone movement.
If we use a current feedback amplifier topology, the magnetic force is now directly controlled by the amplifier. And not only that, cable resistance has no effect, and the resistance of the voice coil now has no affect either.
I asked someone why no one makes current feedback amplifiers to drive speakers and the reply was that speakers nowadays are designed specifically for voltage drive...
So either tradition is biting our ankles, or there's just some really good reason that I was never told.
However, electrostatics are a different animal. The principle mechanism of an electrostatic speaker is capacitance; the polar opposite of inductance. The mechanical force that makes electrostatics move is the static attraction between two isolated plates. This force depends on voltage. So the ideal topology for an electrostatic is voltage drive. (so perhaps the reason they sound so good is because they were actually 'meant' for voltage drive?)
Anyways, those are my thoughts. Current feedback might be the best method, but the relationship between coil magnetivity and cone movement is not necessarily linear; so again, only a direct type of cone movement detection, it seems, could get even close to the .0001%THD real sound wave distortion. And that's a big IF, because of cone surface properties.
I wonder how close they are to making nanotube fiber speaker cones? Maybe they'll be the next fad?
Anyways, those are my thoughts. Just to walk my talk, I should probably start experimenting with current feedback.
- keantoken
The principle force in a loudspeaker is an inductor. This inductor produces a magnetic field which causes the cone to move.
And so I've always wondered, why in the world we drive these speakers with voltage instead of current? Because it's not voltage that determines the strength of the magnetic field as much as it is current. This issue was "fixed" by making speakers a fixed resistance, so that a specific voltage translates into a specific current and thus magnetic force on the cone, but this barely even works because of the reactivity of the mechanical properties of the loudspeaker. This "resistance" changes with frequency. So what you have is that the magnetic force of the coil, which depends on current, not voltage, is only indirectly affected by the amplifier's output, and is at the mercy of cone movement.
If we use a current feedback amplifier topology, the magnetic force is now directly controlled by the amplifier. And not only that, cable resistance has no effect, and the resistance of the voice coil now has no affect either.
I asked someone why no one makes current feedback amplifiers to drive speakers and the reply was that speakers nowadays are designed specifically for voltage drive...
So either tradition is biting our ankles, or there's just some really good reason that I was never told.
However, electrostatics are a different animal. The principle mechanism of an electrostatic speaker is capacitance; the polar opposite of inductance. The mechanical force that makes electrostatics move is the static attraction between two isolated plates. This force depends on voltage. So the ideal topology for an electrostatic is voltage drive. (so perhaps the reason they sound so good is because they were actually 'meant' for voltage drive?)
Anyways, those are my thoughts. Current feedback might be the best method, but the relationship between coil magnetivity and cone movement is not necessarily linear; so again, only a direct type of cone movement detection, it seems, could get even close to the .0001%THD real sound wave distortion. And that's a big IF, because of cone surface properties.
I wonder how close they are to making nanotube fiber speaker cones? Maybe they'll be the next fad?
Anyways, those are my thoughts. Just to walk my talk, I should probably start experimenting with current feedback.
- keantoken
Here's a nice motional feedback idea based on measuring the capacitance between concentric cylinders:
Capacitive motional feedback for loudspeakers
And here's a couple of ideas that could help performance and ease of construction:
(The text below is copy-pasted from an old email - hence any weird formatting etc)
_______________________________________________
The illustration below shows the existing system as well as a couple of proposals. In each case the output voltage (V out) is proportional to the velocity of the moving (red) plate. Static plates are shown in green.
In both of the proposed schemes, the moving plate is earthed, rather than connected to a high voltage source. This can be done by simply connecting the moving plate to one end of the voice-coil (preferably the earthed end). There is then no need for extra "flying" wires, drilling holes in the voice-coil former or attaching additional wires to the spider.
Proposal A
* With the values shown, response will roll off below about 1.6 Hz, which could be ignored or compensated elsewhere in the circuit.
* Depending on the driver, it may be possible to use the voice-coil former itself as the moving element. This is likely to be better behaved mechanically than a glued-on cylinder due to it's rigidity, good balance, and direct coupling to the voice-coil.
* Possible disadvantages could be pickup of hum and noise from ambient fields as well as the effects of nearby earthed objects, which may affect the electrostatic field and cause distortion.
* A couple of diodes from the op-amp input to earth would probably be a good idea to protect against high-voltage surges at switch-on / switch off.
Problems with distortion
A problem with the original design as well as with Proposal A above is sensitivity to sideways movement of the red plate. This effect is reduced with cylindrical capacitors, as a sideways movement will increase capacitance on one side, while reducing capacitance on the other side. However these effects do not cancel out - a movement to either side will result in an overall increase in capacitance. Thus any sideways vibration at the frequency being played will produce 2'nd harmonic distortion at the output.
Proposal B
* This avoids the problems mentioned above because the capacitance between the moving plate and either / both of the static plates is irrelevant.
* What is being measured is the capacitance between the two static plates in the region above the moving plate.
* The moving plate acts only as a shield.
* With solid, accurately aligned, well damped static plates, I would expect this system to give good linearity and be immune to the effects of spurious sideways vibration or resonance of the moving plate.
* It would also be easy to implement decent shielding around the static plates to prevent pick-up of hum etc.
Capacitive motional feedback for loudspeakers
And here's a couple of ideas that could help performance and ease of construction:
(The text below is copy-pasted from an old email - hence any weird formatting etc)
_______________________________________________
The illustration below shows the existing system as well as a couple of proposals. In each case the output voltage (V out) is proportional to the velocity of the moving (red) plate. Static plates are shown in green.
In both of the proposed schemes, the moving plate is earthed, rather than connected to a high voltage source. This can be done by simply connecting the moving plate to one end of the voice-coil (preferably the earthed end). There is then no need for extra "flying" wires, drilling holes in the voice-coil former or attaching additional wires to the spider.
Proposal A
* With the values shown, response will roll off below about 1.6 Hz, which could be ignored or compensated elsewhere in the circuit.
* Depending on the driver, it may be possible to use the voice-coil former itself as the moving element. This is likely to be better behaved mechanically than a glued-on cylinder due to it's rigidity, good balance, and direct coupling to the voice-coil.
* Possible disadvantages could be pickup of hum and noise from ambient fields as well as the effects of nearby earthed objects, which may affect the electrostatic field and cause distortion.
* A couple of diodes from the op-amp input to earth would probably be a good idea to protect against high-voltage surges at switch-on / switch off.
Problems with distortion
A problem with the original design as well as with Proposal A above is sensitivity to sideways movement of the red plate. This effect is reduced with cylindrical capacitors, as a sideways movement will increase capacitance on one side, while reducing capacitance on the other side. However these effects do not cancel out - a movement to either side will result in an overall increase in capacitance. Thus any sideways vibration at the frequency being played will produce 2'nd harmonic distortion at the output.
Proposal B
* This avoids the problems mentioned above because the capacitance between the moving plate and either / both of the static plates is irrelevant.
* What is being measured is the capacitance between the two static plates in the region above the moving plate.
* The moving plate acts only as a shield.
* With solid, accurately aligned, well damped static plates, I would expect this system to give good linearity and be immune to the effects of spurious sideways vibration or resonance of the moving plate.
* It would also be easy to implement decent shielding around the static plates to prevent pick-up of hum etc.
Attachments
A bit of both, I think.
Current drive to a speaker gives much lower distortion, but also much less damping so the frequency response looks like the impedance plot.
Could be excellent for "active" speakers, though - with current-drive amps, built-in equalization and maybe motional feedback to damp the bass resonance.
Malcolm Hawksford did some research and wrote a couple of papers on the subject. See J12 and J14 here:
Malcolm Hawksford - Publications
Godfrey,
Thanks for that reference. I am a poor-white on dial-up connection, so will download later. I have an idea what they (the papers) are about.
Keantoken
Mmmmm .... so have I.
I slaved through varsity being taught that magnetism has to do with amp-turns. Later on it becomes obvious - until one meets with hi-fi. Just to recap.: Normally overall voltage NFB is used with a current loop inside the former - comes out quite nicely and gives rise to many interesting things. For simplicity the two goes to the same NFB input point mixed in a way that minimally influences either.
If not seen before: Somewhere in Denmark/Sweden an experiment was done with either voltage or current NFB, all else remaining the same. (It was stated that in the latter case "something else'' was done to curb loudspeaker basic resonance). Result: The current drive was preferred by the audience.
No, there is nothing that says loudspeakers should be voltage driven! That is what generates much of the strife in the cable/high DF debate - it is not necessary to begin with, barring the main L.F. resonance (which we will leave there for the moment; agreed that it will need attention). The amplifier overload characteristic will have to be looked at - enough available voltage to maintain the current characteristic at peak impedances. And thereby might hang a problem with the funny impedance characteristics shown by some manufacturers.
[Rant: I have a problem with so-called superior loudspeakers but which need oodles of amplifier current because some 8 ohm systems can go to 2 ohms. Thus the G... D... engineer must put up. In this day and year (2010 now!), with the digital CAD at one's disposal, there is no excuse for a driver designer who cannot keep to within civilised levels of impedance. Rant over.]
So yes - are we all merely followers of habit? No, KT, there is nothing here that you have not been told. (Or perhaps more carefully, that I have also not been told.) Opportunity for others to broaden our horizons.
I will copy Hawksford to my files in the morning; thanks again for those references.
Thanks for that reference. I am a poor-white on dial-up connection, so will download later. I have an idea what they (the papers) are about.
Keantoken
Mmmmm .... so have I.
I slaved through varsity being taught that magnetism has to do with amp-turns. Later on it becomes obvious - until one meets with hi-fi. Just to recap.: Normally overall voltage NFB is used with a current loop inside the former - comes out quite nicely and gives rise to many interesting things. For simplicity the two goes to the same NFB input point mixed in a way that minimally influences either.
If not seen before: Somewhere in Denmark/Sweden an experiment was done with either voltage or current NFB, all else remaining the same. (It was stated that in the latter case "something else'' was done to curb loudspeaker basic resonance). Result: The current drive was preferred by the audience.
No, there is nothing that says loudspeakers should be voltage driven! That is what generates much of the strife in the cable/high DF debate - it is not necessary to begin with, barring the main L.F. resonance (which we will leave there for the moment; agreed that it will need attention). The amplifier overload characteristic will have to be looked at - enough available voltage to maintain the current characteristic at peak impedances. And thereby might hang a problem with the funny impedance characteristics shown by some manufacturers.
[Rant: I have a problem with so-called superior loudspeakers but which need oodles of amplifier current because some 8 ohm systems can go to 2 ohms. Thus the G... D... engineer must put up. In this day and year (2010 now!), with the digital CAD at one's disposal, there is no excuse for a driver designer who cannot keep to within civilised levels of impedance. Rant over.]
So yes - are we all merely followers of habit? No, KT, there is nothing here that you have not been told. (Or perhaps more carefully, that I have also not been told.) Opportunity for others to broaden our horizons.
I will copy Hawksford to my files in the morning; thanks again for those references.
You'll be delighted to hear that MartinLogan's latest marvel dips to 0.7 ohms minimum.
One good reason for "high end" speakers to torture amplifiers is that inflicting physical pain on them is the best way to get them to sound different.
This is good for marketing:
If different amplifiers sound the same in your system, then your speakers obviously aren't "good" enough to "reveal" the differences ....
<ducks before the lynch-mob arrives>
This is good for marketing:
If different amplifiers sound the same in your system, then your speakers obviously aren't "good" enough to "reveal" the differences ....
<ducks before the lynch-mob arrives>
Last edited:
GODFREY!!!
SEE YOU AFTER SCHOOL IN THE BICYCLE SHED!
....but just before you give me a bloody nose; you must be a salesman's nightmare (not sure what the male of that is ....)
I will not spoil this here festive day by being further honest.
Is that because of solder spillage somewhere signal-ground, that was not detected during quality control? (oops - sorry!)
SEE YOU AFTER SCHOOL IN THE BICYCLE SHED!
....but just before you give me a bloody nose; you must be a salesman's nightmare (not sure what the male of that is ....)
I will not spoil this here festive day by being further honest.
Is that because of solder spillage somewhere signal-ground, that was not detected during quality control? (oops - sorry!)
Member
Joined 2009
Paid Member
My thinking is that it makes more sense to put the power amplifier near the speaker rather than try to fix a problem of long cables at high power levels. For example, my PMC floorstanders have the option to bolt a small monoblock from Byrston on the back of them. You still have to run long cables to the power amp from the source but this can be done at much more reasonable power levels.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.