If the movement of the cone is being compared to voltage input, it occurs to me that the reactance of the impedance curve either side of resonance (which of course is a reflection of the mechanical resonance via the generator action of the motor) will cause phase shift between input voltage and voice coil current, and that because its the current that is directly proportional to force, there will be phase shift between input voltage and applied force.
Does this reactive impedance fully account for the phase between applied voltage waveform and cone acceleration ?
More interestingly, what happens in the case of a pure current drive system, now the drive waveform (which is current) will always be in phase with motor force and hence acceleration regardless of any mechanical reactance in the speakers impedance. The current (therefore force) is "forced" to follow the signal waveform.
What does this do to the phase relationship between input signal, acceleration, velocity and displacement ?
Of course now you no longer have any electrical damping, so you will have a completely different transfer function with a high Q peak in the response at resonance equal to Qms.
What if we now apply active equalisation for this high Q peak before the current amplifier in the form of a parametric EQ notch of the correct width and depth to restore the same amplitude response we had with electrical damping.
Despite still using a pure current drive does this now return us to exactly the same phase relationship we had before, if we compare the input signal before the parametric EQ to the cone acceleration, velocity, and phase ? (In other words, we aren't getting something for nothing)
Current drive doesn't really change anything except that the voltage appearing across the VC will vary as V = I*Z where Z is the same drive impedance instead of the current through the VC being I = V/Z as in the voltage drive case. . Making current drive doesn't mean there is no back EMF and electrical damping The physics doesn't change. The back emf is generated in a current drive system. The generator effect doesn't go away. All that happen is that the current supplied by the amplifier is increased to over come the back current so that the net forward current remains constant. As you point out, if you equalize the response of the driver to be the same with a current sources as with a voltage source you will find the current through and the voltage across the driver are identical in both cases. The driver characteristics don't change because it is driver differently.
Start by noting that the force is BL*I, then for the voltage source we write BL*I = BL*(Vs/Re -(BL*U)/Re), were BL*U is the back EMF.
When we consider current drive we typical never make the distinction between where the current comes from. We just impose constant currant and say, "see, no electrical damping". Wrong! What we need to recognize is that the constant current come from two factors,
BL*I = BL *(Is -(BL*U)/Re ), or I = Is - BL*U/Re, where Is is the current from the source. Then we can recognize that Is = Vs/Re and we can see that constant current is just a manipulation of Vs so that I remains constant. Electrical damping is still present, it has just been countered by the increase in the applied voltage.
Start by noting that the force is BL*I, then for the voltage source we write BL*I = BL*(Vs/Re -(BL*U)/Re), were BL*U is the back EMF.
When we consider current drive we typical never make the distinction between where the current comes from. We just impose constant currant and say, "see, no electrical damping". Wrong! What we need to recognize is that the constant current come from two factors,
BL*I = BL *(Is -(BL*U)/Re ), or I = Is - BL*U/Re, where Is is the current from the source. Then we can recognize that Is = Vs/Re and we can see that constant current is just a manipulation of Vs so that I remains constant. Electrical damping is still present, it has just been countered by the increase in the applied voltage.
100% correct. This points out that what's called an "audio power amplifier" is almost always a voltage amplifier with sufficient voltage gain and a final stage that can deliver enough current from a power supply to drive a loudspeaker. This can be seen on any amplifier with VU meters calibrated in watts into 8 ohms. Disconnect the speakers and the meters will continue to swing wildly with the input even though not a single watt is being delivered to a load. For a true current amplifier OTOH the meters should peg at the power supply rail voltage trying to drive current through an open circut with even the slightest input.
A true power amplifier would adjust voltage and current instantaneously to power the real axis (resistive) impedance of the load at each frequency. The input would be power delivered to a resistive load. I don't know of a single one that exists or what it would sound like if it did. Probably not much like any amplifier you could buy.
I recall reading back in the 70s when "phase coherent" or "time aligned" speaker systems as they were also called was the fad du jour, reviewers discussed "group delay" the time between application of voltage and response of the driver. These were usually a few milliseconds. If so the phase relationship between driving voltage and resulting physical acceleration, velocity, and displacement is probably meaningless because it would be in the thousands of degrees at mid frequencies where the crossover between the tweeter and midrange or mid woofer occurs. Nice Bodie plots on the earlier posting but still no measured data for these variables related to driver response.
Oh my. 🙁
Yes, the phase curve on the acceleration plot should be flipped.
Another thing that just came to mind is the fact that the "Flat" displacement curve below resonance would only be the case if the woofer was critically damped.(ie Q=0.7071).
Just noticed essentially the same plot(and errors) appear on page 88 of "Loudspeaker & Headphone Handbook"
Loudspeaker and headphone handbook - John Borwick - Google Books
I don't have the energy to debunk this. Anybody else?
He sure is confident in his assertions.🙁
David S.
I wrote a driver siulation code a few years back. If I can ding it on my hard driver I post some results.
For the most part, once you negate any air path delays, the phase shift will always tend to zero in the pass band. It also always leads (rises) when rolling off at the low end and lags (falls) when rolling off at the high end. This is why the region around zero degrees shifts in each case, because each case has a different region of flat response. (Above resonance, below resonance and at resonance.)
Amplitude (excursion) should be flat for the closed box case for any Q, as long as you are well away from resonance.
John Borwicks book is otherwise excellent, as I recall. Didn't it have different authors for every chapter?
David
Speakers from the 70's had "a few" milliseconds worth of group delay did they ? Somehow I don't think so 🙄
I'm not aware of any passive driver which could induce a delay of several milliseconds between application of input signal and acoustic output.
If we conservatively call "a few" ms 2ms, that is a delay equivalent to the apparent acoustic centre of the driver being 688mm, or about 2 feet behind the driver.
Although absolute acoustic centre can't be measured as accurately as relative acoustic centres, it can still be determined down to +/- 20mm or so using impulse based measurements with a known microphone distance, and all normal cone drivers have an acoustic centre very near to the voice coil cone junction over nearly all of their range.
A single driver with a flat response in its passband has group delay that is almost entirely from the air path distance to the listener.
The other possible sources of frequency dependant group delay (other than non-flatness of frequency response) is driver acoustic offset and the crossover. Two crossed over drivers with acoustic centres misaligned by 2 feet would be required to get a broadband error in group delay on the order of 2ms.
The crossover will also induce a group delay peak at the crossover frequency (regardless of drivers being time aligned) but this is in the order of 0.1 to 0.5ms with most typical crossover slopes and types.
"A few" milliseconds of group delay would be well past the audible thresholds of group delay at mid and high frequencies.
(A few ms of group delay at low bass frequencies on the other hand is common and generally considered inaudible)
There is such a thing as unwrapping the linear phase portion of the phase by estimating path delay time you know. 😛 Looking at a raw phase curve without removing the transit time from speaker to microphone is meaningless.
Yes it would be thousands of degrees before the phase is unwrapped, but most of that is linear phase air path delay.
Last edited:
The real issue is absolute delay and absolute phase shift. This is because the reference was made to the comparison between input voltage and driver phase response. Normally phase response is useful when considering it in relative terms. This is what you see in a Bode plot, the phase response compared to the amplitude response.
Bode plot - Wikipedia, the free encyclopedia
This is valuable for many reasons, for example for a negative feedback system to be stable the gain must be less than unity when the phase angle is 180 degrees. But it doesn't tell us anything about absolute delay between system input and output in an electromechanical energy conversion.
Group delay only tells us about the variation in time between different frequency components relative to each other.
Group delay and phase delay - Wikipedia, the free encyclopedia
However there is also an absolute delay;
"All frequency components of a signal are delayed when passed through a device such as an amplifier, a loudspeaker, or propagating through space or a medium, such as air."
As a frequncy of 1khz has a period of 1 ms or .001 seconds, each millisecond of delay is 360 degrees. Therefore a delay of two or three milliseconds will be hundreds of degrees. Unless the driver overcomes its inertia and responds almost instantaneously the relationship between the applied voltage and the motion of the driver in terms of absolute phase angle is absurd. At 5 khz where many tweeters cross over 0.2 ms or .0002 seconds is 360 degrees. When designing loudspeaker systems it is the group delay, not the absolute delay which is of concern.
It is lamentable that such basic notions have to be a source of contention here and even more lamentable that those in this industry not only don't seem to understand these basic concepts but have to have them explained after ridiculing those who don't know less than they do but far more.
Speaker Dave wrote;
“I don't have the energy to debunk this. Anybody else?”
I have not been able to check in often enough to say I have followed the whole thing but I can offer the view Heyser took.
The drivers acoustic phase is the acoustic pressure response to the input signal after all the fixed time delays are accounted for.
From a driver point of view, there is a time delay, several within the driver. Normally the largest delay is the low pass corner produced by the Rdc /Le, it also takes a finite time for the force to be transmitted up the cone body (dependant on the speed of sound in the material).
Where exactly the mean radiator position is physically can also be less than clear so what I say is a proper measurement of the real thing trumps a computer model or an assumption any day.
A band pass response of any kind, if minimum phase, has a corresponding phase rotation associated with it. Simple Filters produce delay (they can’t be predictive after all) in a band pass , one finds that infinitely high in frequency we can say there is no time delay but the output is infinitely rolled off.
When on the flat portion, the band is delayed relative to the very hf and as you get anywhere near the low corner, the phase change is much greater / sooner. On the high pass slope on the lf side, one is delayed much more than the midband region.
If you model this in a computer, it uses a real zero time reference and so one see’s the phase response roll down to the mid band and roll down again at the low pass hf corner.
How can you positively identify where the speaker is in time /distance at any frequency?
Heyser was I think the first person to ask this or at least he found a really clever way to do it. Unfortunatly, his process was a little like the beta / vhs issue, beta was better but narrowly licensed, vhs inferior but widly licensed.
Anyway the TDS process will unambiguously measure the acoustic phase of a loudspeaker. To do that, here is clever trick number one. Richard Heyser was a ham radio guy as well as an audiophile that worked at JPL.
He fed a linearly swept sine wave tone into the speaker and into one input of a modulator /analogue multiplier, the microphone signal fed into the other input.
The output of a multiplier is the sum and difference tones, the sum is discarded and the difference tone is fed in to an FFT. With a linear sweep, a time delay puts the mic signal X Hz behind the sweep tone and so knowing the sweep rate, one rescales the frequency scale on the FFT to time and now one knows how far away the source is because it’s X Hz behind.
The display it the ETC that one can see in arta, it is the Energy vs time response, the REAL part of this is the systems impulse response.
When one has established the first arrival peak, then the system knows how much time to remove when it measures acoustic phase.
To exclude close reflections and out of band noise, clever radio trick 2 is applied.
In a radio there are IF or intermediate frequency transformers which in reality were bandpass filters. In an am radio for example, the broad band antenna signal is fed to one side of a modulator and a high frequency oscillator signal fed into the other. The huge roar that comes in is narrowed down to one station by feeding the multiplier output into the IF filter which passes about 10Khz wide at 455khz (usually). Thus, your tuning knob controls an oscillator running 455KHz different than the frequency your listening to (by then rectifying the amplitude modulation at 455k into an audio signal.
In Heysers clever design, the sounds that arrived late, were at a lower frequency and so excluded from the IF filters. By detecting phase electrically and having a concrete time reference, one has a pretty reliable phase measurement right?
Yes and a bit no. The correct phase appearance depends on knowing where time=0 but at the actual time=0, there is no signal coming from the speaker, it’s too far down. AS the frequency falls, the sound is detected and one can declare time zero. In reality the error between where theoretical Time 0 is and where one starts to get a signal is equal to a short distance (part of a wl at 20K). Heyser was also clear on GD, it is relative to the shortest time, delay can can’t be predictive. Personally I don’t think GD is a very useful measure for a broad band signal use as time is a fixed scale where the range of times of the natural period are 1000:1. On the other hand, it is the phase shift that causes the steps and wiggles in GD and that is always in the right dimension relative to frequency..
Best,
Tom
“I don't have the energy to debunk this. Anybody else?”
I have not been able to check in often enough to say I have followed the whole thing but I can offer the view Heyser took.
The drivers acoustic phase is the acoustic pressure response to the input signal after all the fixed time delays are accounted for.
From a driver point of view, there is a time delay, several within the driver. Normally the largest delay is the low pass corner produced by the Rdc /Le, it also takes a finite time for the force to be transmitted up the cone body (dependant on the speed of sound in the material).
Where exactly the mean radiator position is physically can also be less than clear so what I say is a proper measurement of the real thing trumps a computer model or an assumption any day.
A band pass response of any kind, if minimum phase, has a corresponding phase rotation associated with it. Simple Filters produce delay (they can’t be predictive after all) in a band pass , one finds that infinitely high in frequency we can say there is no time delay but the output is infinitely rolled off.
When on the flat portion, the band is delayed relative to the very hf and as you get anywhere near the low corner, the phase change is much greater / sooner. On the high pass slope on the lf side, one is delayed much more than the midband region.
If you model this in a computer, it uses a real zero time reference and so one see’s the phase response roll down to the mid band and roll down again at the low pass hf corner.
How can you positively identify where the speaker is in time /distance at any frequency?
Heyser was I think the first person to ask this or at least he found a really clever way to do it. Unfortunatly, his process was a little like the beta / vhs issue, beta was better but narrowly licensed, vhs inferior but widly licensed.
Anyway the TDS process will unambiguously measure the acoustic phase of a loudspeaker. To do that, here is clever trick number one. Richard Heyser was a ham radio guy as well as an audiophile that worked at JPL.
He fed a linearly swept sine wave tone into the speaker and into one input of a modulator /analogue multiplier, the microphone signal fed into the other input.
The output of a multiplier is the sum and difference tones, the sum is discarded and the difference tone is fed in to an FFT. With a linear sweep, a time delay puts the mic signal X Hz behind the sweep tone and so knowing the sweep rate, one rescales the frequency scale on the FFT to time and now one knows how far away the source is because it’s X Hz behind.
The display it the ETC that one can see in arta, it is the Energy vs time response, the REAL part of this is the systems impulse response.
When one has established the first arrival peak, then the system knows how much time to remove when it measures acoustic phase.
To exclude close reflections and out of band noise, clever radio trick 2 is applied.
In a radio there are IF or intermediate frequency transformers which in reality were bandpass filters. In an am radio for example, the broad band antenna signal is fed to one side of a modulator and a high frequency oscillator signal fed into the other. The huge roar that comes in is narrowed down to one station by feeding the multiplier output into the IF filter which passes about 10Khz wide at 455khz (usually). Thus, your tuning knob controls an oscillator running 455KHz different than the frequency your listening to (by then rectifying the amplitude modulation at 455k into an audio signal.
In Heysers clever design, the sounds that arrived late, were at a lower frequency and so excluded from the IF filters. By detecting phase electrically and having a concrete time reference, one has a pretty reliable phase measurement right?
Yes and a bit no. The correct phase appearance depends on knowing where time=0 but at the actual time=0, there is no signal coming from the speaker, it’s too far down. AS the frequency falls, the sound is detected and one can declare time zero. In reality the error between where theoretical Time 0 is and where one starts to get a signal is equal to a short distance (part of a wl at 20K). Heyser was also clear on GD, it is relative to the shortest time, delay can can’t be predictive. Personally I don’t think GD is a very useful measure for a broad band signal use as time is a fixed scale where the range of times of the natural period are 1000:1. On the other hand, it is the phase shift that causes the steps and wiggles in GD and that is always in the right dimension relative to frequency..
Best,
Tom
Last edited:
Tom, thank you for reviewing the basic principles of superheterodyne radio reception which was universally adopted for AM, FM, and analog television among others. People may have forgotten that this was not the only method developed historically in our digital internet era.
Time = 0 can be established by inserting a single parasitic impulse marker superimposed on a sine wave, say at its crest, the impulse used as a trigger for a storage scope or just recorded on a data logger. The mechanical response can be determined accurately by measuring the cone movement with a laser interferometer. I'd bet that you'd get different delays depending on what part of the cone you are measuring, lower near the center, greater towards the periphery as the internal pressure wave in the cone travels outward. The time difference between the applied marker at the voice coil and the observed marker can be translated into degrees of absolute phase shift for the sine wave frequency applied.
Time = 0 can be established by inserting a single parasitic impulse marker superimposed on a sine wave, say at its crest, the impulse used as a trigger for a storage scope or just recorded on a data logger. The mechanical response can be determined accurately by measuring the cone movement with a laser interferometer. I'd bet that you'd get different delays depending on what part of the cone you are measuring, lower near the center, greater towards the periphery as the internal pressure wave in the cone travels outward. The time difference between the applied marker at the voice coil and the observed marker can be translated into degrees of absolute phase shift for the sine wave frequency applied.
The only significance of GD in a loudspeaker is whether or not it is constant. and if it is not constant (which is generally the case), whether the variation with frequency is audible or not.
Agreed...well away from resonance.
The smooth/quick transition to flat shown in the plots just below resonance implied critical damping to me.
Yes, each chapter had different authors. Chapter 3 by Peter Baxandall on electrostatic loudspeakers was particularly well written I thought...one of the few technical sources on the subject. Although Chapter 3 was supposed to be virtually unchanged for the 3rd revision, it surprised me to see at least 5 printing errors that had crept into the formulas that were not present in the 2nd revision.
Again correct but unfortunately the irrelevant point of applied voltage versus direction of acceleration was brought up. A spurious issue that was made far too much of.
Is there a point to your message ? It seems to be a long ramble that doesn't really have a destination, or much connection to your original claim of "a few" milliseconds of group delay in speakers.
Fact, no reasonable speaker has "a few" (2+) milliseconds of group delay at anything other than bass frequencies, and you've offered nothing to back up your claim that this is the case.
Fact, nearly all the "absolute delay" between input signal and acoustic output is the propagation time delay from the speaker to the microphone/listener.
Any electromechanical delay in the driver itself can be represented by the apparent acoustic centre of the driver being slightly further back, and is of little consequence. In fact since most drivers are not flat and have some physical depth, and the wave front is produced by the summation of many cone points at differing depths, the concept of an "absolute" acoustic centre from which additional "delay" in the driver can determined is nebulous at best.
The reality is that the driver has an "apparent" acoustic centre somewhere near but not exactly at the voice coil cone juncture which can be considered the delay free "time zero" reference for that driver relative to an input signal.
Fact, "achieving" multi-milliseconds of group delay in a multi-way speaker is only ever going to happen from having huge driver acoustic centre mismatches on the order of several feet, as no reasonable crossover will have several milliseconds of group delay at midrange or treble frequencies.
Any group delay is going to smear your parasitic impulse. If you just trigger on the leading edge of the measured impulse you're going to be establishing the time zero for the highest frequencies reproduced by the speaker - with group delay present lower frequencies may be delayed (or ahead) of this. Not an accurate way to do it.
In the piston range the centre and the outside of the cone will be moving together in time - no delay between the two. In the breakup region an impulse will travel from the centre to the outside at the propagation speed of the cone material, which is finite, so the outside of the cone will be delayed, however it is also closer to the listener.
For sound its the summed acoustic response at the listening position that matters, measuring a few spot radius locations on the cone with a laser won't tell you much if the cone is in its breakup region.
In a driver designed to operate well through the breakup region (some full range drivers for example) the directly forward vector of the propagation velocity of the bending waves from the centre to outside of the cone is chosen to match the speed of sound in air, thus the cone launches a coherent wavefront even at frequencies where it is large.
More gratitous insults.
"The reality is that the driver has an "apparent" acoustic centre somewhere near but not exactly at the voice coil cone juncture which can be considered the delay free "time zero" reference for that driver relative to an input signal."
How do you reconcile that with the original assertion that the applied voltage and acceleration are in opposite directions? If this statement is correc they are always in the same direction.
Do you have any proof for any of your assertions? That's what you ask of others, where's yours?
No data was offered proving the original assertion either.
It is not my problem if you cannot follow the arguments in my postings. If you take constant issue with them then I strongly suggest you don't read them and skip over to something else.
No, just frustration at your attempts to lead us around in circles.
And this has what to do with propagation delay from a speaker to a microphone ? If I reverse the phase of a driver it changes the phase relationship but doesn't change the absolute delay, nor the apparent acoustic centre.
It's you that made the initial claim that speakers had several milliseconds of group delay so its up to you to prove it, not for me to disprove it.
It's difficult if not impossible to prove a negative - I could provide measurements of one or more speakers that have less than 2ms of group delay but it still wouldn't prove that there weren't some somewhere that exceeded that amount. You would claim that I cherry picked ones that backed my position.
On the other hand if you provided a single measurement or reference that backed up your claim the point would be proven. So the burden of proof falls on you to prove a positive, not me to refute it by trying to prove a negative.
I suspect most of the people reading this thread take constant issue with your wild claims, but they are getting tired of the effort of constantly refuting them. I don't give up quite so easily.
You can't measure something like this with a microphone. You must be joking. You haven't even talked about the propagation delay between the time the sound hits the sensing element and the time corresponding voltage appears at the output. Only an optical device is fast enough to make this kind of measurement.
The initial claim remains unsubstantiated by any measurement, that the applied voltage to a loudspeaker and its acceleration are always in opposite directions. Do you have any data to support this assertion? So far I haven't heard even one iota of evidence, just a lot of noise and insults for challenging it.
The initial claim remains unsubstantiated by any measurement, that the applied voltage to a loudspeaker and its acceleration are always in opposite directions. Do you have any data to support this assertion? So far I haven't heard even one iota of evidence, just a lot of noise and insults for challenging it.
The impulse response of a speaker is the integral of the entire radiating area of the cone, unless your laser is measuring every part of the cone area simultaneously the result won't be valid, especially when the effective radiating area varies with frequency in the breakup region.
A laser will also detect when a particular part of the cone starts to move, but it is not allowing for how far that part of the cone is from the listener. The centre of the cone may start moving first at high frequencies, but it is also further from the listener.
Depending on the angle of the cone and the cone material the impulse from the centre could arrive first, or the impulse from the cone edge could arrive first. (Ideally you'd want them both to arrive at the same time, but not many drivers do that)
If the purpose of your laser measurement is to measure the time delay between an electrical impulse being applied and the first movement of the centre of the cone, my response is "so what" ? How much of the delay is electromechanical and how much is path delay to the listener is immaterial.
All it does is very slightly alter the apparent acoustic centre depth. Absolute acoustic centre depth is not relevant to speaker design, only relative depth of different drivers, (which is much easier and more accurately measured) and there are those that would say even that doesn't matter.
It wasn't me that made this assertion in the first place, I'm not sure why you're pestering me about it as if I did ? I didn't respond to this point in any of your posts, and I don't have an opinion on the matter one way or the other.
- Status
- Not open for further replies.