My logic might be totally faulty with the following so please correct me if I'm wrong.

Many people talk about transient response with headphones and some use this as a selling point for headphones like electrostatics. But is this really an issue? Is there a point where a headphone driver's transient response is fast enough and have many headphones above a certain price point ($100?) passed this minimum point?

Let me explain my point using another audio analogy. For amplifiers slew rate is important. Yet, based on the voltage swing required through what maximum frequency, we can say that X slew rate is fast enough. Still, audiophile manufacturers will throw slew rate numbers around to sell or justify their products.

Can the same idea of a maximum threshold (beyond that it doesn't practically matter) said of the transient response of a headphone driver?

A lot of people seem to talk about some headphones being "fast" and some being "slow" with regards to transient response. They common theme is that "fast" headphone reveal details in the music whereas "slow" headphones cannot reveal the details or do so with a slow decay.

Yet, it seems to me that revealing details and decay (or the illusion of decay) has more to do with the particular frequency response of the individual headphone and less to do with the actual transient response. Headphones that are very bassy with a rolled off high end sound slow, meaning the details are not as clear. Headphones that are more treble/upper-midrange focused seem to reveal details clearly and quickly.

But many of the headphones (from dynamic to electrostatic) I've heard reveal all details, they just reveal them differently. With some headphones these details are more obvious than others.

Are particular frequency responses being interpreted as transient response? Could I make the case (to myself) than many headphones (dynamic or electrostatic) have the transient response to reveal all details in the music, it is just revealed differently based the FR of each headphone?

So, that being the case, could I conclude that many modern dynamic drivers are fast enough in terms of transient response?

Good transient response should indeed directly correspond to (flat) frequency response. A headphone that has a sufficient performance at high frequencies (meaning flat, not boosted or attenuated!) implies it has a good enough transient response.

Does anybody know if modern dynamic headphone drivers perform well at high audible frequencies (assume 20k - 22k)? There are a number of graphs out there, but almost all headphones contain various dips and peaks above 10 khz. However, this is not necessarily a defect as headphones are designed to interact with the head and a measured flat EQ is NOT desirable.

So it's difficult to tell from these graphs if these high frequencies dips and peaks are from poor performance or intended design.

I don't believe transient response and frequency response need to have a direct relationship. Transient response, as the name suggests, is a measure of a device's ability to respond to transients. Ultimately a headphone's (or speaker's) diaphagm has to move air in order for us to be hear the sound. To use a crude illustration, a heavy diaphragm will not be able to respond as quickly to an electrical stimulus (transient) as a light diaphragm due to inertia so it will not cause the air to move as quickly. Thus it will have inferior transient response or seem "slow". As to what is an adequate response and whether most modern cans achieve it, I can't say.

I don't know what transient response sounds like, or how it's measured or quantified... So, when it comes to specs, "advertising claims" or comments by reviewers, I ignore it! When a reviewer mentions it, I assume it has something to do with the high frequency detail. But for me, these impressions are just too vague to be of any use.

I was just 'looking around" and I see that that Sennheiser claims "Excellent Transient Response" for the HD-600. What does that mean? When I look at the "Technical Data", I don't see a number for that, no graph, nothing!


As always, the most important thing is to listen to the overall sound.

I believe that you can "characterize" a system (including its frequency response) by using a single impulse (which is a transient). This is not the standard way frequency response is measured on audio equipment, and I've never heard of anybody doing it this way.

It is directly related. The steepest transients you can have is a sequence of -1,1(,-1,1,-1,...) samples. At the same time that's the highest frequency at full amplitude that your playback system can output. If two speakers have the same flat frequency response up to this frequency (and maybe above) it is impossible that one of them has worse transient response than the other.



This model speaker will have an attenuation of higher frequencies in its frequency response plot.

Two opposite extreme examples of bad transient response would be

(1) You apply a transient to the headphones and nothing happens.

(2) You apply a transient to the headphones and something happens and keeps happening long after the transient has passed.

Transient response is tested by applying transients and then measuring what happens. Three common transients that are used for testing are:

(1) Impulse.

(2) Square wave.

(3) Tone burst.



Agreed.

Headphones are just relatively small speakers in 2 relatively small rooms that are each shared with an ear. Because the speakers are small, it is relatively to make them have good response, particularly at high frequencies. Headphone design is as much about the room as the speakers, a concept that more would do well to consider strongly.



HD600s are great-sounding headphones. I don't think that a great-sounding headphone would have bad transient response.



Particularly true since the actual way that headphones sound is somewhat depdendent on how your outer ear is made. Everbody is different, there.



It is true that the canonical transient is an impulse, and impulse response is related to frequency response by means of the Fourier transform.



The problem with testing using impulses is that the amount of energy that you can pack into an impulse of a height that is small enough to be processed with no clipping or other nonlienar distoriton is limited. Because the energy is limited, there is a problem with noise in the measured response. One solution would be to use multiple impulses, but the impulses must be widely separated if low frequency response measurements are desired. Now, the time to make the test becomes an issue.

It has been known for some time that you can synthesize the impulse response of something by means of lots of calculations if you know both the exact input signal and output signals. Before high performance computers, this was not feasible. It now is.

Interestingly enough, the applied signal can be anything that has lots of energy at any frequency at which you desire a good measurement - even music. However, measurements are fastest if you use carefully-tailored test signals, and a means for generating optimal test signals named MLS is known.

I didn't know either, until I listened on Stax electrostatic headphones at work. A colleague and I sat down with a high-quality Beyerdynamic (300 Euros) and a Stax Lambda (more than 2000 Euros with the amp). We played music and switched between the two. The surprising thing was, the biggest sound difference wasn't during music or castanets, but when listening to speech. The Lambda sounded much "crispier" than the dynamic phones, much more like the person was standing in front of you. Given the different technology, I'm pretty sure the Stax sounded better because they more accurately reproduce the pitch transients in voiced speech.

Chris

Optimal frequency response + optimal phase response = optimal transient response.

Issues surrounding transient response largely turn into a larger discussion about phase, which is a somewhat lively topic.

Could you summarize the effect of phase response? I've got some vague intuitive idea of what you're referring to, but one of your lucid explanations would really clarify this for me.

For what it may be worth, the Headroom website provides graphs of square wave responses on their "build a graph" page.

The Sennheiser 600s show a recognizable square wave at 500 hz, with the 650's doing much better and the 800's better still. Just what the audible consequences of this may be i am not knowlegable to comment on, but the measurements are interesting.

OK, I think I may be misunderstanding the use of the word "transient". If you record a single bass drum hit the level will go from silence to very loud very quickly. Isn't that a transient? How does hf behaviour influence a speaker's ability to reproduce it? Or to put it another way I've always interpreted transient response as a time domain parameter. Is that incorrect?

Simply not true. Under ordinary circumstances, Newton was right, F=MA or A = F/M. IOW you can accelerate either a light or heavy diaphragm as much as you want to, just provide enough force. What this means is that all other things being equal, the speaker with the heavy diaphragm will be less efficient, but not necessarily have any "slower" response when it comes to following a transient in an audio signal.

IOW you have have two tweeters, one with a light diaphragm and one with a heavy diaphragm. They can both respond to rapid transients and high frequencies with equally flat, smooth response.

For example, people say "an electrostic diaphragm is so light, it must respond very fast". Unfortunately, the force that drives it may not be nearly as strong as a good old fashioned wire coil suspended in a strong magnetic field. This is even more clearly true for magnetic planar speakers. Yes, the diaphragm is light per square inch, but there are a lot of square inches. It is far more difficult to produce a strong magnetic motor that is distributed over a large area than one that is concentrated in one small place.

People who compare electrostatic versus dynamic earphones geneally have no idea whether or not the frequency response is well-matched. Generally, they have no idea about level matching, either. So the comparison has gone south and we haven't even got around to talking about listener bias!

Yes, but your amplifier doesn't know whether it's dealing with a light diaphragm or a heavy one. It doesn't/can't adjust its output accordingly. So, wouldn't a heavy diaphragm simply be slower to respond to the same stimulus than a lighter one? Also I'm still not clear why everybody is focussing on high frequencies. Is it not possible to have lf transients?

My point is that any light versus heavy diaphragm issue is automatically dealt with when you increase the force applied to that diaphragm.

In the real world there may be not even a need to turn up the volume to increase the force, because the system components that apply force the heavy diaphragm can easily be more efficient.

Definitions:

Electrostatic motor - two surfaces with a varying voltage between them, which creates a varying electrostatic force that acts on the diaphragm.

Planar Magnetic motor - a conductive diaphragm or a conductive area of a diaphragm that is acted on by an magnetic field, so that varying the current flowing through the conductor applies a varying force to the diaphragm.

Electromagnetic motor - a traditional loudsepeaker voice coil placed inside a strong magnetic field that is attached to the diaphragm and applies a varying force to it.

When we compare speakers with planar magnetic or electrostatic motors to speakers with electromagnetic motors, it is a general rule that the planar magnetic and electrostatic motors are weaker and far less efficient.

At this time, Planar Magnetic motors can be more efficient than electrostatic motors, but both pale in comparison with electromagnetic motors. The most efficient drivers with electromagnetic speaker motors have efficiencies on the order of 35% or more, compared to 1% or less for the others.

I think we may at at cross purposes as that's my point too - I'm not talking about electromagnetic Vs electrostatic. Given
2 diaphragms of the same type, 1 heavier than the other, the heavier one will need more force to be applied in order to move it.

What I think you're not understanding here is that there is a 1:1 relationship between the time domain and the frequency domain (gain AND phase).

Transients are all about good HF response.

I believe good square wave response does imply better transient response but I'm not sure what the correlation is between the square wave graph and our ability to hear the differences. Headroom (and just about everybody on Head-Fi) raves about the transient response of the Senn HD800, but there are a number of headphones that have square wave graphs that look good and also better than the HD600. Many IEMs (take the Sleek Audio 6) have a square wave response that makes the HD800 square wave graph look quite bad.

What does this mean in a practical sense? I have no idea. I wonder what the square wave graph for an electrostatic headphone looks like.

That being said, I have long since stopped paying serious attention the graphs on Headroom. It suspect something is going on but I don't care to make accusations on a public forum.

First, note a fundamental property of linear time-invariant systems (which include all systems of linear distortion, like headphones neglecting THD): the impulse response and the complex transfer function are interchangeable, between the time and frequency domains. And the transfer function can be decomposed into a magnitude function (the usual "frequency response" measured in dB) and a phase function (measured in degrees). The impulse response is essentially what a transient response plot is all about - the transient response is simply the convolution of the input signal with the impulse response. (Of course this equivalence only exists when the plots have sufficient accuracy in the x- and y- axes.)

Thus, if the transient response somehow represents something lacking in the magnitude response, that something must also be in the phase response. And in fact, the phase response is easier to analyze in a lot of cases. The "ideal" phase response is zero degrees at all frequencies, and non-ideal phase response can be analyzed in rather clear and understandable ways. It's a lot harder to compare non-ideal transient responses than to compare phase responses.

The "common wisdom" about phase response, in the non-audiophile sense, is that the ear is more or less insensitive to it. The psychoacoustics involved generally suggest a predominantly magnitude-sensitive cochlea (although there are some interesting bits involving a sensitivity to locations of zero crossings that would be way beyond my competence to explain). Nevertheless, the nonlinearities in the ear - and specifically even-order harmonic distortion - imply that different frequencies can intermodulate to the same frequency. If one of those frequencies changes phase, constructive/destructive interference can occur and the power at that frequency can (audibly) change. And there are gobs of tests indicating a significant amount of phase sensitivity in the ear, although perhaps not as much as is claimed by some audiophiles. And of course, the phase response plot is far more useful for evaluating this sort of audibility than the transient response. It's also really hard at times to tell if a test is being compromised by even order nonlinearities in the headphones/speakers My old forum posts on absolute polarity here on HA is a pretty clear example of all of this, but it's also coming up on the Stereophile forums right now, where a professor is using all of this (plus blind test results) as a rather clever justification for high res.

If I'm not mistaken, the authoritative citation on the audibility of phase response is still: Lipshitz, Stanly P., Pocock, Mark, and Vanderkooy, John, "On the Audibility of Midrange Phase Distortion in Audio Systems,' J. Audio Eng. Soc., Vol. 30, No, 9, Sept. 1982, pp 580-595.

I don't know enough about the specifics of headphone and speaker measurements to say for certain how big of an issue this is - WmAx or jj would be better people to ask, or perhaps even JA! - but I think it's safe to say that it is an issue with some transducers and not with others.

Also note that a transducer can have insanely good magnitude response, and insanely good phase response, and still have sh*tty THD problems, which as usual are audible to varying degrees. None of this - frequency response, phase response or transient response - will tell you anything about THD, because it does not exist in an LTI system model. Etymotics phones (although I still love mine) are virtually the poster child for this sort of issue.

OK, clearly I hadn't understood the term "transient response" correctly. Can you clarify for me then what term is used in audio engineering to describe a sytem's or component's ability to respond to sudden changes in level. IE to accurately mirror the dynamics in music?

Thanks



Edit for typo

That requires both good frequency and phase response as well as freedom from nonlinear distortion.

THD and IM are ways to measure nonlinear distortion.

I know we're off topic here but I really would like to understand this so please bear with me.

Suppose you have a recording of a symphony orchestra. Suppose the 2nd movement opens with the full orchestra playing at maximum volume (ffff or whatever). To reproduce this your audio system will need to go from the silence - or stricly speaking, the noise floor - of the inter-movement gap to the 100dB+ SPL of the full orchestra across pretty much the full frequency spectrum in a very short space of time. For the sake of this example let's assume that we have enough amplifier power to produce the required SPL without clipping. Is there a measure used in the audio world that indicates a system's or component's ability to reflect this sudden peak accurately? If so, what is it called?

Thanks

Transient response! And it is directly related to phase and frequency response.

Thanks. So, to go back to my earlier question why are people saying things like :
"Transients are all about good HF response"

"A headphone that has a sufficient performance at high frequencies (meaning flat, not boosted or attenuated!) implies it has a good enough transient response."

Why don't low frequencies matter? Is it just that phase errors are less noticeable at lower frequencies?

How much low frequencies matter depends on which transient you choose to test with.

There's a major problem - there is no unique, generally-agreed-upon definition for transient response.



Actually, lower frequencies are the only place where phase errors can matter at all. Above about 1 KHz, the ear is relatively insensitive to phase errors that are equally applied to both channels.

A heavy diaphragm also needs a stiff suspension so that the movement is controlled (damped). A dynamic system with little damping will tend to oscillate (peak overshoots and long settlement time).
This is what the waterfall spectrum (a 3D plot of frequency vs. attenuation vs. time) is for but I haven't seen anything like that measured for headphones, only saw it for Visaton loudspeaker drivers.

The relevant measures of the fidelity of audio reproducers are the same for *any* audio signal:

(1) Amplitude or power, which answers the question: Is it loud enough?
(2) Frequency response, which along with (3) answers the question: Is the timbre or spectral balance right?
(3) Phase response
(4) Freedom from nonlinear distortion, which answers the question: are any there any internally generated spurious signals that are responses to the input signal?
(5) Dynamic range which answers the question: are any there any internally generated spurious responses which are *not* responses to the input signal?

The accuraty of response to test signals like square waves and other impulses are defined for a given frequency response if the reproducer has a property known as minimum phase, which is the common relationship between amplitude and phase response. If a reproducer is minimum phase then flat frequency response and ideal response to impulses is a given. Most loudspeaker drivers are essentially minimum phase but most multiway loudspeakers systems are not.

That's wrong. The stiffness of the suspenstion is not what damps a speaker diaphragm.

The stiffness of the suspension and the diaphragm's mass determines the lower resonant frequency.

Diaphragms have two types of damping. Mechanical damping comes from losses in the suspension (which is different than stiffness), and electrical damping comes from the interaction between the voice coil, the magnet, and the source.



Which is a good thing, within bounds.



The waterfall plot of a speaker has only a small area at lower frequencies that relates to the stiffness of the suspension.

Thank you very much, Axon. Fascinating as always.