Just to clarify that I wasn't stating this to be power drive
-----------------------------------------------------------------------------------------
I can't really see why one would strive for power drive anyway - after all there will still be the problem of compression due to voice coil heating.
Current drive, on the other hand, is inherently free from this effect.
annex666 said:
I can't really see why one would strive for power drive anyway - after all there will still be the problem of compression due to voice coil heating.
Current drive, on the other hand, is inherently free from this effect.
to try something new and completely different?
on the other end, while power drive would only reduce (but not completely eliminate) voice coil heating and its nasty effects, it has the advantage of providing some (I would say just enough...
) speaker damping to make it more practically useable than current drive. All IMVHO, of course.
How would power drive reduce power compression?
After all, driving a dynamic driver with a constant power will still heat the voice coil, hence changing its resistive component and lowering the current though it - which, in turn, will lower the sensitivity.
Current drive does not suffer from this as, while there is still voice coil heating, it does not lower the current through the voice coil and hence sensitivity is not lowered. We do, of course, need to redefine "sensitivity" in this case (as the drive signal is current, not voltage) - as the ratio of audio power output (not electrical power) to input current (not voltage).
After all, driving a dynamic driver with a constant power will still heat the voice coil, hence changing its resistive component and lowering the current though it - which, in turn, will lower the sensitivity.
Current drive does not suffer from this as, while there is still voice coil heating, it does not lower the current through the voice coil and hence sensitivity is not lowered. We do, of course, need to redefine "sensitivity" in this case (as the drive signal is current, not voltage) - as the ratio of audio power output (not electrical power) to input current (not voltage).
Hi UnixMan,
I mention "claim" for Pwrtron amplifier but it is not. I am very aware it is, as you wrote, a very hard task to make a true "power-drive".
I wonder what would be the interest as such an amplifier.
Note that many loudspeakers manufacturers express frequency responses for 1 Watt, which is a pure fallacy, confusing many people: it would require to maintain constant a power of 1 W in the voice coil. Frequency responses are measured with constant voltage source, usually 2.83 V which is 1 W only for an 8 Ohm resistive load.
Speaking of watts for amplifiers and loudspeakers is good to calculate dissipated heath and not much else.
I mention "claim" for Pwrtron amplifier but it is not. I am very aware it is, as you wrote, a very hard task to make a true "power-drive".
I wonder what would be the interest as such an amplifier.
Note that many loudspeakers manufacturers express frequency responses for 1 Watt, which is a pure fallacy, confusing many people: it would require to maintain constant a power of 1 W in the voice coil. Frequency responses are measured with constant voltage source, usually 2.83 V which is 1 W only for an 8 Ohm resistive load.
Speaking of watts for amplifiers and loudspeakers is good to calculate dissipated heath and not much else.
For a given audio power the required electrical power will be the same in all circumstances, regardless of the drive mechanism (for want of better terminology) - it is dictated by the physical properties of the driver and nothing else.
Hence there will be the same heat dissipated in the voice coil using current, voltage or power drive (for a given SPL) - the difference with current drive is that the audible output level is as expected across the entire output power range (not compressed at higher output levels).
annex666 said:
that would be true for an ideal transducer. Or at least a "decent" one, with small losses and/or a (almost) purely resistive load.
Unfortunately speakers are all but decent transducers. They have an efficiency close to 0 (that is almost all of the incoming power are losses) and -what's more important- they are highly reactive.
With such a load, the heat losses in the v.c. depends a lot on phase shift and the amount of peak (reactive) currents.
Thus the "active" power dissipated in the voice coil is NOT equal to the audio power (I mean output SPL) and drive impedance does matter.
Am I wrong?
You are wrong in so far as you misread my post
The following statement:
...states, not that the audio power and electrical power are the same (this is clearly not true as I have been talking about heating losses in the voice coil), but rather to state that the electrical power wasted for a given output SPL will be identical, regardless of the drive method.
This is always true; it does not matter in the slightest that the load impedance is reactive - it is the same impedance regardless of the way it is driven!
I hope this is more clear
annex666 said:
not quite, IMHO. I would say there is the same effect as for the losses on AC power lines... the higher the phase shift, the higher the resistive losses...
The impedance of the whole output circuit (including amplifier output) does change. The higher the driving impedance (if it is resistive and/or produce phase shift complementary to that of the load) the smaller the phase shift -> the lesser the losses on the v.c. resistance.
The minimum dissipation would be on a purely resistive load (the v.c. resistance itself only). Adding reactive elements increases the currents and thus the wasted power.
I think you are confused.
You are contemplating the output impedance of the amplifier and the driver as a combined load - ignore for a second the output impedance, instead think of the voice coil by itself.
Fact: For a given driving voltage (across the voice coil) a relative current will flow (with phase shift and magnitude based on the voice coil impedance), likewise for any driving current a voltage will be produced (with phase shift and magnitude based on the voice coil impedance).
Fact: The power developed in the voice coil is based on the current passing through it and the voltage across it.
From the above two points it is clear that the power dissipated by the voice coil is based purely on the impedance of the voice coil and the driving signal.
Fact: For a given steady-state SPL (i.e. where the voice coil has reached a stable operating temperature) the driver will have a fixed efficiency - by this I mean the ratio of audio power to electrical power.
The above facts support my proposition and are fairly fundamental.
1: The maximum thermal losses in the voice coil will actually occur when there is zero phase shift between the current through it and the voltage across it (think about it - there is peak cross-correlation when two similar waveforms are congruent). Thermal compression is based on heating - which requires real power not imaginary power.
2: If you increase the output impedance then you will draw less current, which will inherently provide lower audible output power - and I did state "For a given audio power".
Anyway, I'm not going to argue further on this point - I don't think this discussion is a valuable enough use of my time; I believe the facts I have stated to be fairly clear.
Maybe if I am wrong I will forfeit my First Class Masters in electronics
...but then again, maybe not
You are contemplating the output impedance of the amplifier and the driver as a combined load - ignore for a second the output impedance, instead think of the voice coil by itself.
Fact: For a given driving voltage (across the voice coil) a relative current will flow (with phase shift and magnitude based on the voice coil impedance), likewise for any driving current a voltage will be produced (with phase shift and magnitude based on the voice coil impedance).
Fact: The power developed in the voice coil is based on the current passing through it and the voltage across it.
From the above two points it is clear that the power dissipated by the voice coil is based purely on the impedance of the voice coil and the driving signal.
Fact: For a given steady-state SPL (i.e. where the voice coil has reached a stable operating temperature) the driver will have a fixed efficiency - by this I mean the ratio of audio power to electrical power.
The above facts support my proposition and are fairly fundamental.
1: The maximum thermal losses in the voice coil will actually occur when there is zero phase shift between the current through it and the voltage across it (think about it - there is peak cross-correlation when two similar waveforms are congruent). Thermal compression is based on heating - which requires real power not imaginary power.
2: If you increase the output impedance then you will draw less current, which will inherently provide lower audible output power - and I did state "For a given audio power".
Anyway, I'm not going to argue further on this point - I don't think this discussion is a valuable enough use of my time; I believe the facts I have stated to be fairly clear.
Maybe if I am wrong I will forfeit my First Class Masters in electronics
...but then again, maybe not
Iv'e been following this thread with some interest, and have a few questions about current drive for speakers. I get the concept, but seeing as pretty much all drivers out there are designed for voltage drive, implementing current drive seems to present some obstacles, no?
The major problem of course is a speaker's impedance varying with frequency, thus unless passive impedance correction is applied, frequency response will be a mess.
However, once you have a bunch of passive networks in parallel with the driver, you lose true constant current drive, since the passives will suck current away from the speaker, or what am I missing here?
By the time you do all this passive correction, as far as the speaker is concerned, your'e right back at voltage drive in practice, so where is the benefit?
Applying EQ to cancel out the frequency response variations, before the amp will restore true current drive into the driver itself, so that seems preferable to me.
I'm pretty interested in exploring this, as my main speakers are over 20 years old, and need to be replaced, so now seems to be a good time to look outside the box - no pun intended
Lukas
The major problem of course is a speaker's impedance varying with frequency, thus unless passive impedance correction is applied, frequency response will be a mess.
However, once you have a bunch of passive networks in parallel with the driver, you lose true constant current drive, since the passives will suck current away from the speaker, or what am I missing here?
By the time you do all this passive correction, as far as the speaker is concerned, your'e right back at voltage drive in practice, so where is the benefit?
Applying EQ to cancel out the frequency response variations, before the amp will restore true current drive into the driver itself, so that seems preferable to me.
I'm pretty interested in exploring this, as my main speakers are over 20 years old, and need to be replaced, so now seems to be a good time to look outside the box - no pun intended
Lukas
Conventional ("dynamic") drivers have an output proportional to the current through the voice coil - they are current-controlled devices. Despite this, they may, of course, be voltage-controlled by the application of Ohm's law - unfortunately this means that the relationship between the control signal and the output is now frequency-dependent and complex.
Unfortunately it has become convention that dynamic drivers are generally voltage-controlled; this may have been a consequence of the state of amplifier design. Designers therefore attempt to "optimise" the voltage-SPL curve rather than the current-SPL curve which may cause problems when designing a current-driven speaker.
Firstly, I'd like to clarify that when I use the term driver I am referring to a single transducer that converts electrical energy to acoustic energy; I use the term speaker when referring to one or more drivers, mounted in a cabinet, connected to a crossover (possibly including compensation circuitry).
I am assuming when you use the term "speaker", above, you are referring to what I call a driver.
Your statement is only true when the driver is driven from a voltage source - when current-driven, the driver impedance should not directly affect the frequency response as the current through the voice coil is controlled by the amplifier and is not affected by the varying impedance of the driver itself. The output voltage of the amplifier will, of course, vary across the frequency range (dependent on the driver impedance), however the driver response will not.
Your thinking here is slightly confused, I think. It is possible to use passive compensation networks and still maintain current drive, in exactly the same way that it is possible to use passive compensation networks and maintain voltage drive.
When performing (passive) voltage compensation you aim to linearise the output by reducing the voltage applied to the voice coil wherever there is a peak in the output response - this is done by inserting components that have a higher impedance at that point, such that a higher voltage is developed across them instead of the driver itself. I would term this "series compensation".
This is directly analogous to current-compensation, however instead of using a series component to reduce the voltage at the peak frequency, you would use a parallel device that would reduce the current through the voice coil (i.e. lowering its impedance at the given peak frequency). I would term this "shunt compensation".
The matter is complicated slightly by the fact that shunt compensation can be used in a matter similar to series compensation by transforming the current to a voltage using a resistive pass component, - I would term this "series-shunt" or "mixed-mode" compensation but you can ignore this for now.
As noted above, you don't need to do this to maintain current drive, however this method is preferable for other reasons:
* Crossover and compensation can be combined easily.
* Active crossovers all bi-amping.
* DSP can be used to obtain responses that would be otherwise impractical using passive components. Also room compensation can be easily applied.
I look forward to reading your findings
annex666 said:
Yes, but that's only a measure of the (instantaneous) power transfered to/from the load. Given that the load is reactive, that has nothing to do with the amount of power (energy) actually "consumed" by the load, that is converted to heat and/or to/from mechanical work.
The power dissipated (converted to heat) in the voice coil always depends ONLY on its DC resistance, NOT on the complex impedance appearing across its terminals!
That is, the power dissipated on the voice coil is aways given by (i^2)Rdc, no matters what the actual voltage developed across the voice coil is!
(voltage across the v.c. depends on impedance, not just resistance. On the other end, all of the current will aways have to "go trough" its DC resistance producing (i^2)Rdc power losses. To see it in another way, the Rdc is in series with all the other elements composing the complex voice coil equivalent circuit. But the the only element which produces heat in the voice coil is its Rdc).
Thus, turning our point of view from instantaneous powers to the amount of heat produced, this depends only on the integral of the v.c. current over time (and of course on the v.c. Rdc).
Now let' have a somewhat wider look at what happens with either current or voltage drive.
With current drive we apply a "controlled" current (which is always proportional to the input signal) and allow the voltage to follow accordingly.
On the other end, with voltage drive we apply a "controlled" voltage (which is always proportional to the input signal) and allow the current to follow accordingly.
Now these seems to be two perfectly equivalent ways of supplying power to a load. And in fact they would be so... would the load be a resistor. But in our case it is not.
In fact, being it and e.m. motor, the current is proportional to the force applied to the cone while the voltage is proportional to the actual v.c. speed.
That is, with current drive we apply a controlled external force (proportional to the input signal) to a mechanical system (v.c., cone, etc) and allow its speed to "naturally" follow accordingly.
On the other end, with voltage drive we try to achieve a "direct" control of its speed. In a sense this basically implies (is) a form of electro-mechanical NFB!
With voltage drive we try to force the cone speed to instantaneously follow the input signal voltage variations... which of course is physically impossible (given that the moving mass is non-zero, it would require infinite power). What happens in practice is that the v.c. is subject to higher accelerations. Which implies stronger forces (-> higher currents in the v.c) and larger amounts of energy exchanged back and forth between the electrical and mechanical "domains". Which in turn means... more heat losses in the voice coil Rdc.
The higher is the effective DF the higher are the currents and thus also the losses (and their side-effects).
Current drive not only avoids heat-related compression altogether by completely "ignoring" the voltage appearing across the v.c., but it actually produces less heat, too.
Of course, a "low DF" or "power drive" is somewhere in between the two extremes.
(and, usually... in medio stat virtus.
)
Note that I said "based on" - i.e. not "the product of". Complex power can be calculated from the rms of the voltage and current and knowledge of the phase difference between them. From this, the real, reactive and apparent power can all be calculated.
My assumption was that the ratio of thermal power to audible power output would be constant and since the reactive power transfers no net energy to the load it would be possible to calculate thermal energy based on the above calculation of real power.
Agreed - when (if) I get time, I plan to build a pair of speakers that are current-driven, with full-range drivers in a dipole arrangement.
OK did a quick informal comaprison test of voltage vs current drive frequency responses with a 5 1/4" mid I have lying on my desk for another project.......
It's raining out, so I did not feel like schlepping the PC, LMS, amp and the rest of the parapharnalia to the backyard.
The measurements were crudely done with the driver in my lap and the mic handheld about a 1/4" away from the dustcap, but should be good enough for quickie comparison purposes. Voltage across the speaker terminals was set the same at 500 Hz for both situations. 100 ohm series resistor, simply because one was at hand on my bench.
As expected the constant current SPL graph shows a peak at resonance and a rise at the top end due to inductance.
Unfortunately I had major noise issues with my LMS setup today, so coudl not do comparitive distortion measurements. Have to sort that out somehow, as less distortion is what current drive is supposedly all about........
Lukas
It's raining out, so I did not feel like schlepping the PC, LMS, amp and the rest of the parapharnalia to the backyard.
The measurements were crudely done with the driver in my lap and the mic handheld about a 1/4" away from the dustcap, but should be good enough for quickie comparison purposes. Voltage across the speaker terminals was set the same at 500 Hz for both situations. 100 ohm series resistor, simply because one was at hand on my bench.
As expected the constant current SPL graph shows a peak at resonance and a rise at the top end due to inductance.
Unfortunately I had major noise issues with my LMS setup today, so coudl not do comparitive distortion measurements. Have to sort that out somehow, as less distortion is what current drive is supposedly all about........
Lukas
Impedance curve:
http://home.comcast.net/~lukaslouw/CurrentDrive/Impedance.pdf
SPL curves:
http://home.comcast.net/~lukaslouw/CurrentDrive/SPL.pdf
http://home.comcast.net/~lukaslouw/CurrentDrive/Impedance.pdf
SPL curves:
http://home.comcast.net/~lukaslouw/CurrentDrive/SPL.pdf
Hi LukasLouw,
Blue SPL curve shows Qms related energy storage - in driver time - not input (constant current) waveform time !
Dynamic waveform transduction amplitude linearity is ruined with current drive !
Red SPL curve shows Qes related control - less steady sine output but better dynamic control.
The blue curve has constant current and thus to get the cone resonating at this (70Hz) amplitude takes TIME ! Energy from the first couple of half cycles becomes stored by the driver/system and is returned as increased amplitude after those first one or more half cycles.
In other words some energy which should have been moving air for the first one/two half cycles of current drive is not being radiated because the driver is storing it. The dynamically changing amplitude of the radiated LF output does NOT match the original audio waveform at Fs; it is weakly defined and returned with resonant emphasis !!!
Some designers go for a tuned or critical Q SPL response between the blue and red curves, but any amplitude generated which is above the red characteristic is due to the driver/system, and is thus more parasitic than the minimum error due to the driver alone, because the increased amplitude (resonance) takes time to develop wrt the original input waveform, whether this is due to the nature of the electrical (current) input or the driver (high Qes) itself.
Cheers ........ Graham.
Blue SPL curve shows Qms related energy storage - in driver time - not input (constant current) waveform time !
Dynamic waveform transduction amplitude linearity is ruined with current drive !
Red SPL curve shows Qes related control - less steady sine output but better dynamic control.
The blue curve has constant current and thus to get the cone resonating at this (70Hz) amplitude takes TIME ! Energy from the first couple of half cycles becomes stored by the driver/system and is returned as increased amplitude after those first one or more half cycles.
In other words some energy which should have been moving air for the first one/two half cycles of current drive is not being radiated because the driver is storing it. The dynamically changing amplitude of the radiated LF output does NOT match the original audio waveform at Fs; it is weakly defined and returned with resonant emphasis !!!
Some designers go for a tuned or critical Q SPL response between the blue and red curves, but any amplitude generated which is above the red characteristic is due to the driver/system, and is thus more parasitic than the minimum error due to the driver alone, because the increased amplitude (resonance) takes time to develop wrt the original input waveform, whether this is due to the nature of the electrical (current) input or the driver (high Qes) itself.
Cheers ........ Graham.
Hi Graham,
are you sure? I mean, have you tried to actually record the (first cycles of) acoustic waveform produced by a speaker in the different conditions?
I would say that it's just "delay" (phase shift, actually). But mine is just an educated guess. I have not done any actual measurement, so of course I may be wrong.
are you sure? I mean, have you tried to actually record the (first cycles of) acoustic waveform produced by a speaker in the different conditions?
I would say that it's just "delay" (phase shift, actually). But mine is just an educated guess. I have not done any actual measurement, so of course I may be wrong.
BTW:
ANY cone movement takes time, no matter what kind of driving technique is used... though indeed driving impedance (together with voice coil Rdc) does alter the maximum current->force->acceleration to which the v.c. is subject, and thus the "settling time".
I'd bet that a "critical damping" of the cone is the best thing to do, but that's another story...
ANY cone movement takes time, no matter what kind of driving technique is used... though indeed driving impedance (together with voice coil Rdc) does alter the maximum current->force->acceleration to which the v.c. is subject, and thus the "settling time".
I'd bet that a "critical damping" of the cone is the best thing to do, but that's another story...
Graham has just described the effect of any filter circuit (be it mechanical or electrical) on the time domain appearance of the waveform. If the filter is minimum phase and the process is linear time invariant (the latter may be less true with actual drivers than the former), then he inverse filter can restore the "perfect" waveform. Limit is only the unavoidable highpass behaviour of practially all drivers. All this has nothing to do with drive impedance, exept that drive impedance obviously alters the filter characteristic.
BTW I never understood the term "critical damping" with the highpass behaviour. There is no critical damping with anything else than a lowpass, the term is simply not applicable for any other filter function. Still the Q=0.5 alignments are sort of optimal in a variety of aspects, but (time domain) "damping" isn't one of them, IMHO.
- Klaus
BTW I never understood the term "critical damping" with the highpass behaviour. There is no critical damping with anything else than a lowpass, the term is simply not applicable for any other filter function. Still the Q=0.5 alignments are sort of optimal in a variety of aspects, but (time domain) "damping" isn't one of them, IMHO.
- Klaus