[AutumnCrown]

I've read in a few places that high impedance headphones require more voltage for a given SPL. I've also read that proper damping is important for driver control. Are these related? Maybe if I understood what damping was on a physical/mechanistic level, rather than conceptual (ratio of output impedance to driver circuit impedance), it would help me understand. Is there any science behind the idea that a headphone may require a much higher voltage swing for optimal response than for simple SPL level?

 

[NA Blur]

Like many oscillators the driving forces vs the impedance forces are constantly at war. If one begins to dominate you get unwanted resonance or damping. More volume can add distortion it, but oscillators have thresholds for optical performance and to get there you have to reach or just surpass the threshold. Too little driving force and you do not get the full movement of the oscillation which in the case of an audio driver can result in the lack of air movement. This might sound thin or too laid back.

 

[castleofargh]

  1/I've read in a few places that high impedance headphones require more voltage for a given SPL. 2/I've also read that proper damping is important for driver control. Are these related?3/ Maybe if I understood what damping was on a physical/mechanistic level, rather than conceptual (ratio of output impedance to driver circuit impedance), it would help me understand. Is there any science behind the idea that a headphone may require a much higher voltage swing for optimal response than for simple SPL level?

1/ higher voltage=more SPL. everything else working fine, voltage is loudness with twice the voltage resulting in +6db. don't worry this is one of the most misunderstood thing in the audio community. many have the idea that somehow we can always send more voltage or more power through a headphone while at the same loudness.
 
3/  you can see this from wikipedia

imagine that what you see is the movement of the headphone's membrane being pushed by a sudden fixed voltage. ideally it would go to a fixed position and stop, the magnet pushes the coil away in proportion to the magnetic field created by the electrical signal. so all those stuff are relative to each other and a given voltage means a given position. instead depending on damping variables we can get those movements. they do get at the same position in the end, but the trip changes for reasons.
-the dark blue would be what we call underdamped, there is very little to stop the membrane from moving freely, so it overshoots the final position and goes back and forth several times before stopping at some point because of air and stuff.
-the red one is the opposite, overdamped. there are a lot of forces that oppose the free movement of the membrane, so it doesn't move as fast and can't oscillate as much. think of this as the air being replaced by oil.
-the last one, clear blue is some in-between case, where the movement is fairly free and fast, but still dampened enough to limit ringing. 
 
 
2/from a purely electrical point of view, what we want is for the coil to move according to the variations of the signal. that means moving with the signal, but it also means limiting the potential influence from other variables!!!! we want reasonable efficiency when possible, we want the electrical forces to dominate the mechanical ones to some extent because we still have a signal to "copy". we also want to limit potential changes in voltage for electrical reasons, like various changes in impedance, be it per frequency, or because you're using another amp and it doesn't have exactly the same impedance as the old one. in short we wish for stability just as much as fidelity. strong impedance damping, often called impedance bridging is what we use to reach that goal and almost al modern headphones are built expecing to be used following that principle. in such a scenario, the headphone has high impedance compared to the amp(ideally at least 10 times higher). what that means for the electrical signal is that the current has a real hard time flowing through the coil, and voltage stays very stable(and that's what we really want as voltage is loudness). even if you go from a 0.2ohm amp to a 3ohm amp, and even if the headphone's impedance instead of being flat, looks like this:

if you don't care to respect the 1/10 ratio, among other things, in this case you will apply an EQ to the sound. some might not call it that because they think EQ is an insult, but that's what it is in this case with a hd650. where there is the impedance bump, the voltage will go higher, boosting that specific frequency area. and the higher the amp's impedance, the stronger the bass bump. keeping the 1/10Th ratio ensures that this signature change will be kept under 1db.
 
in a few cases with specific gears, it can sound nice and be desirable to some people to have a high impedance amp. on speakers some will like adding a few ohm to the amp to change how the subs will sound. bad damping in such a case often feels like the low end loosen up a bit. there are a few cases like that where some people enjoy what is basically doing it wrong on purpose. with multidriver IEMs, the changes in signature can be all over the place, more bass, less bass, rolled off bass (if the amp as protective caps), weird stuff happening at the crossover frequencies between each driver... like anything, sometime we're lucky and we somehow enjoy the resulting sound, but so very often the result sucks objectively and subjectively. unless I clearly have measurements of both the amp and headphone letting me anticipate what wil happen, I don't play with fire and prefer to pick a good impedance ratio. that's how it should be for almost all gears.
 
 
I hope it's clear, but I 'm usually not

.

 

[jagwap]

  1/ higher voltage=more SPL. everything else working fine, voltage is loudness with twice the voltage resulting in +6db. don't worry this is one of the most misunderstood thing in the audio community. many have the idea that somehow we can always send more voltage or more power through a headphone while at the same loudness.
 
3/  you can see this from wikipedia

imagine that what you see is the movement of the headphone's membrane being pushed by a sudden fixed voltage. ideally it would go to a fixed position and stop, the magnet pushes the coil away in proportion to the magnetic field created by the electrical signal. so all those stuff are relative to each other and a given voltage means a given position. instead depending on damping variables we can get those movements. they do get at the same position in the end, but the trip changes for reasons.
-the dark blue would be what we call underdamped, there is very little to stop the membrane from moving freely, so it overshoots the final position and goes back and forth several times before stopping at some point because of air and stuff.
-the red one is the opposite, overdamped. there are a lot of forces that oppose the free movement of the membrane, so it doesn't move as fast and can't oscillate as much. think of this as the air being replaced by oil.
-the last one, clear blue is some in-between case, where the movement is fairly free and fast, but still dampened enough to limit ringing. 
 
I hope it's clear, but I 'm usually not

.

 
Your diagram shows another possible voltage related issue.  When a driver overshoots, if that overshoot represents a voltage larger than the amplifier can produce, as it returns the back E.M.F. pours back into the amplifier, but this time not while the output is necessarily in a linear output region.  I've seen this in subwoofers, and the output stage can get driven into the rail. This is why clamp diodes are needed on some amplifier outputs.  This can cause a version of clipping even if the signal has not saturated.

 

[castleofargh]

I'll trust you on that, I'm a total noob when it comes to amps. past one or 2 general designs and typical behaviors, I'm Jon Snow. ^_^
intuitively I would expect the simple loss from efficiency to limit the risk of "high" voltage generated back into the coil from the movement itself. but for a massive overshoot it seems reasonable to consider the possibility.
 I guess it's yet another reason to keep a good impedance ratio.

 

[barryt]

Let me offer a thought here.
 
Voltage is not control. Voltage is about power and it takes 10 times the about of power to double the amount of sound and is separate to this discussion  More voltage is more headroom.  When you run out of volts you "clip" the signal and produce a new sound, you add energy to the signal typically frequencies.  If this is too fast for the transducer to convert into sound then heat is produced instead of sound, this is why an over driven small amplifier can burn out a headphone whereas a big amplifier won't (given the same signal).
 
Control is the achieved with feedback, that is the amplifier sensing what is going on in the headphone's mechanics and hopefully altering the signal to the headphone to force it to comply.  This is usually most perceived in the low end, is called 'damping', and has an defined protocol about how to measure and quantify.  It is arguably the reason for the difference in sound of tube and transistor amplifiers, the output of the amplifier is changed by the feedback so there is a signal going to the transducer that is not a true copy of the amplifier's input.  I effect all hell breaks loose and new/different sounds are produced.
 
In my years of designing amplifies I have come to avoid feedback to achieve superior sonics, but more on that later.

 

[pinnahertz]

  Let me offer a thought here.
 
Voltage is not control. Voltage is about power...

Voltage is not about power. Voltage is about voltage. Voltage, Power, Current and Resistance are inseparably linked mathematically.
 
Voltage V = I × R = P / I = √(P × R)
Current I = V / R = P / V = √(P / R)
Resistance R = V / I = P / I2 = V2 / P
Power P = V × I = R × I2 = V2 / R
 
Things get more complex when you look at an AC circuit with a complex impedance, but everything is still linked mathematically.

...and it takes 10 times the about of power to double the amount of sound and is separate to this discussion.

It takes 10x power to double the percieved loudness of sound, it takes 4X to double sound pressure.  There is no such quantity as the "amount of sound". 

More voltage is more headroom.  When you run out of volts you "clip" the signal and produce a new sound, you add energy to the signal typically frequencies.

Since voltage and power are inseparable, you also clip when you run out of power. The "new sound" produced does not add energy, it adds harmonics (spectral content). Total energy is slightly less than an unclipped wave at the same level.

 If this is too fast for the transducer to convert into sound then heat is produced instead of sound, this is why an over driven small amplifier can burn out a headphone whereas a big amplifier won't (given the same signal).  

This is a widely held falsehood. Clipping amps do not burn out drivers any more than unclipped amps driven to the same level. Drivers are burned out because of heating or over-excursion, with heating being the primary cause of death in tweeters. The RMS value of a waveform is also known as the heating value. The clipped waveform has, in fact, slightly less RMS value than an unclipped waveform at the same level. The increase in high frequency spectrum above clipping is well below any power handling problem the driver is already encountering.

 
Control is the achieved with feedback, that is the amplifier sensing what is going on in the headphone's mechanics and hopefully altering the signal to the headphone to force it to comply.

A very odd analysis. I see no case where an amplifier senses anything mechanical in a driver directly. It may have to absorb some back EMF, but that depends on a lot of other factors. In most topologies that employ feedback, the feedback is there for amplifier stability and distortion control. If you take it away you don't have an amplifier anymore.

This is usually most perceived in the low end, is called 'damping', and has an defined protocol about how to measure and quantify.

Damping factor is the ratio of the amplifier output impedance to the load impedance plus wire resistance. And is mostly not a factor at all because when you add the wire resistance the damping factor becomes much higher than typically expected.  

It is arguably the reason for the difference in sound of tube and transistor amplifiers, the output of the amplifier is changed by the feedback so there is a signal going to the transducer that is not a true copy of the amplifier's input.  I effect all hell breaks loose and new/different sounds are produced.  

The higher output impedance of tube amps has a lot to do with the potential sound difference, but not because of damping.  The impedance of the driver and whatever network may be involved creates a voltage divider with the amplifier output impedance affecting the voltage vs frequency a the driver.  There are other minor factors as well.
 
To say an amplifier with feedback cannot produce an amplified true copy of it's input at it's output is ridiculous.  It is far more likely to be closer to the input than one with less feedback and higher distortion.   Anyone performing a distortion measurement on an amplifier would know this first hand. Yes, some amps are better than others, and there are bad designs in all topologies.  But assuming an amp is designed well, it's entire goal is to produce an amplified replica of its input.

 

[barryt]

Well it is not my intention to get into a contest here but would like to point out a couple of things you might wish to reconsider.
 
First, you are right, I should have said perceived sound rather than SPL, I tend to think in terms of perception, thank you.
 
Second, in damping factor the measurements are made at the amplifier output not the load, they are made at the point where the feedback signal is derived, thus the cable resistance is considered as part of the load and not the source.  In an amplifier with an output impedance of say, 0.01 ohms driving an 8 ohm speaker the Damping factor is 800.  Add a say 10 feet of #10 wire at roughly 0.001 ohm per foot (20 feet round trip) you add .02 ohms resistance making the load 8,02 ohms.  That makes the damping factor 802.
 
If, by the way, you made the measurement at the load rather than the feedback point the amplifier output would be 0.01 + 0.02 or a total of 0.03 ohms, the damping factor would be 266.  
 
But what is damping?  It is an electro-dynamic abstraction base almost solely on the feedback of the amplifier.
 
Feedback is the key.  If a tube amplifier had same the open loop and closed loop gain as a transistor amplifier, it would have the same Damping Factor.
 
We used to demonstrate this by comparing a tube amplifier with a transistor amplifier that had an added 1 ohm resistor in the series with the cable, try it, the 'transistor' sound will go away.
 
Third is feedback, feedback is a dynamic property takes the output of the amp and adds it, out of phase to the input.
 
Every transducer is two-way, a speaker is a microphone, ever use an intercom, you speak into the loudspeaker to be heard.  The problem is that the crossover and speaker/box, store energy and then at a later time (typically 2 milliseconds for a 1 foot deep speaker box) send back to the amplifier which adds it to the input and produces a new signal, now 2 mS later, that is feed to the speaker.  It is a new sound, we call it the 'transistor sound' because tube amplifiers, with less open loop gain, have less of it.
 
Thus a no-feedback amplifier would not have the problem.  How do you do this?  Simple, current-mode amplification instead of voltage mode amplification.  Operating transistors in the current-mode rather than the voltage mode produces very little distortion and requires no feedback, thus you get the best of both worlds.
 
On the clipping question, ask any speaker manufacturer, he will tell you that under powered speakers are the root-cause of failure for far more warranty claims than over powering.
 
As far as the distortion issue I would refer you to Bob Carver's demonstration in the 70's, we can learn from the past.
 
http://thecarversite.com/yetanotherforum/default.aspx?g=posts&t=4481

 

[jagwap]

  Well it is not my intention to get into a contest here but would like to point out a couple of things you might wish to reconsider.
 
First, you are right, I should have said perceived sound rather than SPL, I tend to think in terms of perception, thank you.
 
Second, in damping factor the measurements are made at the amplifier output not the load, they are made at the point where the feedback signal is derived, thus the cable resistance is considered as part of the load and not the source.  In an amplifier with an output impedance of say, 0.01 ohms driving an 8 ohm speaker the Damping factor is 800.  Add a say 10 feet of #10 wire at roughly 0.001 ohm per foot (20 feet round trip) you add .02 ohms resistance making the load 8,02 ohms.  That makes the damping factor 802.
 
If, by the way, you made the measurement at the load rather than the feedback point the amplifier output would be 0.01 + 0.02 or a total of 0.03 ohms, the damping factor would be 266.  
 
But what is damping?  It is an electro-dynamic abstraction base almost solely on the feedback of the amplifier.
 
Feedback is the key.  If a tube amplifier had same the open loop and closed loop gain as a transistor amplifier, it would have the same Damping Factor.
 
We used to demonstrate this by comparing a tube amplifier with a transistor amplifier that had an added 1 ohm resistor in the series with the cable, try it, the 'transistor' sound will go away.
 
Third is feedback, feedback is a dynamic property takes the output of the amp and adds it, out of phase to the input.
 
Every transducer is two-way, a speaker is a microphone, ever use an intercom, you speak into the loudspeaker to be heard.  The problem is that the crossover and speaker/box, store energy and then at a later time (typically 2 milliseconds for a 1 foot deep speaker box) send back to the amplifier which adds it to the input and produces a new signal, now 2 mS later, that is feed to the speaker.  It is a new sound, we call it the 'transistor sound' because tube amplifiers, with less open loop gain, have less of it.
 
Thus a no-feedback amplifier would not have the problem.  How do you do this?  Simple, current-mode amplification instead of voltage mode amplification.  Operating transistors in the current-mode rather than the voltage mode produces very little distortion and requires no feedback, thus you get the best of both worlds.
 
On the clipping question, ask any speaker manufacturer, he will tell you that under powered speakers are the root-cause of failure for far more warranty claims than over powering.
 
As far as the distortion issue I would refer you to Bob Carver's demonstration in the 70's, we can learn from the past.
 
http://thecarversite.com/yetanotherforum/default.aspx?g=posts&t=4481

Yes, Carver knows his stuff.
 
Adding a resistance outside the feedback loop does not make things "better" or "less like transistors" unless something else is wrong.  As the previous post said, one of the things it does is alter the frequency response and crossover frequencies a little so it may just be compensating for something you don't like the sound of.  But you are right it depends where you measure the damping factor as to the value.  The best place is often on the feedback spur.
 
I suspect you are referring to "no-global-feedback" if we want to discuss the same thing.  All gain stages, current or voltage have local feedback built in.  "No-global-feedback" can sound good.  I've worked with such designs.  But well sorted feedback designs can be just as good, usually better..  
 
There is never one magic thing: valves, damping factor, noise, feedback, THD, IMD, phase distortion, DNR, headroom, materials, capacitors.  All in balance to fit the budget, comprehensively measured and proved, and confirmed in listening test.
 
To the original question: A little more than enough voltage to take care of the signal is needed, so that the output stage stays linear, doesn't saturate, and the feedback doesn't saturate the input stage.

 

[barryt]

Accepted, yes, it global feedback we are talking about as individually both tubes and transistors have intrinsic amounts of local (to the component) feedback.it is a Gestalt, the result of all element working together.  We are after an illusion from which we derive pleasure.  Analysis of the signals involved can come from many processes as you note, with the various measurements you suggest, it is the old 4 blind men and an elephant conundrum.
 
In the end it is all in our minds, as designers it is our responsibility to give the greatest amount of illusion for the buck that we can.  It is all an illusion, hard to fit a real orchestra in hour living room.

 

[pinnahertz]

  Second, in damping factor the measurements are made at the amplifier output not the load, they are made at the point where the feedback signal is derived, thus the cable resistance is considered as part of the load and not the source.  In an amplifier with an output impedance of say, 0.01 ohms driving an 8 ohm speaker the Damping factor is 800.  Add a say 10 feet of #10 wire at roughly 0.001 ohm per foot (20 feet round trip) you add .02 ohms resistance making the load 8,02 ohms.  That makes the damping factor 802.

But damping factor is meaningless at the amplifier terminals. It only matters at the actual load, in fact, at the actual driver terminals, post wire, post crossover (if any) where there could be an audible impact. You can raise the amp terminal DF to the thousands, it won't change at the driver if there's enough wire and crossover in between. Your example of wire is interesting, but not typical. In the context of headphones, there was a recent thread that cited something like 10' of 30ga wire. That's enough to make amp DF a moot point completely.

If, by the way, you made the measurement at the load rather than the feedback point the amplifier output would be 0.01 + 0.02 or a total of 0.03 ohms, the damping factor would be 266.  

Right, and the change the wire made in DF was pretty radical. If you scale that example to something more typical like 14ga and 40', you'll see why DF is really completely overblown as a factor.

Feedback is the key.  If a tube amplifier had same the open loop and closed loop gain as a transistor amplifier, it would have the same Damping Factor.
 
We used to demonstrate this by comparing a tube amplifier with a transistor amplifier that had an added 1 ohm resistor in the series with the cable, try it, the 'transistor' sound will go away.

Your demo confirmed the sonic difference is related to output Z, and has nothing to do with feedback.

On the clipping question, ask any speaker manufacturer, he will tell you that under powered speakers are the root-cause of failure for far more warranty claims than over powering.

First, that's not true, but anyway...
 
Yes, this is a typical misconception. However, it makes no sense.
 
1. Drivers are rated for max power. It's often a power vs time figure.  There's no mention in typical driver specifications about damage from a particular waveform.
 
2. There's no way a manufacturer would know conclusively if the defective driver was driven by a clipped amplifier or if it was damaged by clean but excessive power. 
 
3. A clipping amp has less capability to deliver power to the driver than a non-clipping one. Just ask yourself where that power is coming from once the maximum limit has been reached.  
 
4. If a clipping amp could damage drivers more easily then it must be some mechanism that doesn't involve heating related to RMS value. The next thing people usually go to is high frequency spectrum caused by clipping. Unfortunately, even very hard clipping does not produced harmonics of sufficient power to dominate or exceed that of the primary unclipped signal. (I can prove this if it gets down to it).
 
The next thing people do is claim clipping is just like DC. Not true at all. Do the waveform analysis, there is NO DC caused by clipping. A clipped waveform has a very definite crest factor, DC has none. It's not DC. Any other ideas?

As far as the distortion issue I would refer you to Bob Carver's demonstration in the 70's, we can learn from the past.
 
http://thecarversite.com/yetanotherforum/default.aspx?g=posts&t=4481

That page deals with the audibility of distortion, and yes, I'm familiar with all of that having been involved fist-hand with a similar demonstration. If anything, that shows that even the necessity of presenting a amplified true copy of the input signal at the output is in question. However, in practice, distortion levels are orders of magnitude below what he tested, in any successful amplifier design. There are other parameters that contribute to or inhibit a "true copy", but they all also have thresholds of audibility. Some of the shockingly high. But most  if not all current amplifiers operate far below those thresholds.

 

[barryt]

I appreciate the fervor of your beliefs.  Output impedance is the result of feedback.  I don't think you understand the implications, subtle and gross of feedback but I expect you will with time come to the that understanding.  I am just now releasing a no feedback, Class A, current mode headphone amplifier that is warmly being received, perhaps you will design an amplifier of your own soon and get it into production.  
 
As far as the clipping issue, please enjoy your beliefs, years of building high power 4-way sound re-enforcement systems, taking them on the road, my patents and being Chief Engineer of manufacturing for audio companies has me seeing things a little differerent that you do.  I wish you well.

 

[pinnahertz]

  I appreciate the fervor of your beliefs.  Output impedance is the result of feedback.  I don't think you understand the implications, subtle and gross of feedback but I expect you will with time come to the that understanding.  I am just now releasing a no feedback, Class A, current mode headphone amplifier that is warmly being received, perhaps you will design an amplifier of your own soon and get it into production.  
 
As far as the clipping issue, please enjoy your beliefs, years of building high power 4-way sound re-enforcement systems, taking them on the road, my patents and being Chief Engineer of manufacturing for audio companies has me seeing things a little differerent that you do.  I wish you well.

I would have welcomed a better explanation of the clipping/damage issue from someone in your position.  I have put more than passing effort into researching this, but if your years of experience show otherwise, I'm up for learning. 
 
So, a few questions: 
 
If higher RMS power isn't what damages speakers (clipping amps produce less power than non-clipping ones at the same output levels), then what is it about a clipping amplifier that does?  
 
 
I'm attaching a set of curves that I've made comparing the power of a clipping amp to a non-clipping amp of the same gain.  The data was taken from actual measurements.  The blue and violet traces are made with test tones, the others with program material.  Each data point is an increase of 1dB past the threshold of clipping.   Please explain how a clipping amp burns out drivers based on increased power. 
 
I'm also attaching a set of spectrum graphs that show an analysis of the spectral content of a densely processed pop tune.  Each trace represents an additional 2dB above amplifier clipping, with the bottom trace showing no clipping just before the threshold.  Please explain how the additional high frequency information would damage a driver.
 
It seems that unless I manufacture an amp and get it into production my "beliefs", even though well researched, can't possibly be correct.  Hmmm.  I'm not quite ready to throw over 4 decades in audio engineering out the window, but if you enlighten me perhaps I might correct my misunderstanding.  I do learn things every day, and I'm wrong about something at least that often too.
 
 

 

[castleofargh]

 I've been told that DC offset can damage drivers, I've been told that square waves can damage drivers, so I came to assume that any sort of constant non null voltage isn't a driver's favorite signal.
but why? I've honestly never read anything with clean evidence to support the many rationals behind those ideas.
and at some point TBH, I'm thinking that I just have to use a proper amp for my load instead of wondering if it's dangerous for the gear to listen to clipped or highly distorted audio. because that was never the audio I wished to listen to anyway.

 

 

[pinnahertz]

Certain types of amps can have failures that cause nearly the full + or - supply rail to be applied to the load.  Most of those are older designs.  Newer and better amps should not fail into a mode that can damage anything including their output devices.  
 
We should remember that severe clipping does not result in square waves.  Even exceeding the clipping threshold by 10dB doesn't result in square waves.  It results in clipped audio, which has a different RMS value and crest factor than a real square wave.
 
We usually don't listen to clipped audio, but there are number of pro sound scenarios where clipped audio could and often does occur.  Modern pro amps have means of avoiding clipping with a type of peak limiter.