Topic: Types of Driver Ringing (Read 11264 times)

ultrabike · « **Reply #10 on:** June 20, 2013, 06:28:17 AM »

I'll give it a shot with what I know (others with more experience and knowledge might chime in - and correct possible misconceptions on my part). If a headphone was excited by a very "fast" signal in the sense of "very small" rise and decay time (an impulse) the headphone system (acoustic and mechanical) may not be able to track the signal instantaneously. In other words, the cans will not decay immediately after the input signal (nor react immediately to it).

What the headphone will do is to continue to move a little for a while, sometimes way after the signal had stopped. It could decay slowly and perhaps without oscillations, or perhaps relatively fast and with oscillations. It could also optimally decay fast, within the limitations of it's bandlimited nature, and almost not oscillate...

These oscillations will happen at certain frequencies, and a properly parameterized CSD plot may provide for a straight forward visualization of their magitude and duration ("ringing") as a function of frequency.

These issues sometimes correlate well with perceived issues defined as "tremble razors of death", "midrange shout", "upper midrange" crap, "double/triple octave penetrator", among other things.

Some thoughts (skip-able stuff)

In my mind overdamped (slow non-oscillating decay), underdamped (fast oscillating decay), and critically damped (optimally fast non-oscillating decay) terms are given to second order systems (very useful for PLL characterization). I can visualize a headphone system as an equivalent system made up of several cascaded second order subsystems over, under, and critically damped at different "undamped" natural frequencies. Also, given the audio frequency range and the length of the cable, the amp and headphone together can be considered a single network system, where the different amps affect the overall system damping. In my mind this is not the whole story: non-linear distortion, response being a function of space/acoustic impedance/etc variations, and many things I have not yet thought off... but it's part of it.

If the headphone system oscillates at a particular said natural frequency in the audio band, after an impulse has been applied to it, as previously mentioned, it will "ring". In the time domain, the ringing can be seen as small oscillations (sometimes hard to see) riding the impulse response. A properly windowed and parameterized spectrogram (CSD plot) will conveniently display these oscillations as a function of frequency.

Sometimes oscillations at particular frequencies indicate the presence of a strong reflection (rough comb filter) which may result in a notch, peak, weird reverb, and all of the above defined ear ravaging issues.

Anyway, this is my understanding...

Kunlun · « **Reply #11 on:** June 20, 2013, 12:24:29 PM »

Thanks Ultrabike! If anyone else would like to add anything or give their own understanding, I'd appreciate that, too.

Tyll Hertsens · « **Reply #12 on:** June 20, 2013, 03:52:52 PM »

Thanks for the post, Marv. Good stuff.

xnor · « **Reply #13 on:** June 21, 2013, 06:26:19 PM »

I might add that ringing is a necessity for minimum phase systems. Sure, headphones are not perfectly min. phase but for all intents and purposes they behave pretty similar if you ignore non-linear distortions.

The easiest way to see this is to send an impulse (in digital audio a single full-scale sample, every other sample is zero) into a "system" with non-flat frequency response, for example a min. phase equalizer.
The filters will react to the input signal and produce an output signal. The filters cannot produce any output other than zeros before they actually "see" the impulse. This is causality, an important property of min. phase systems.

It should be noted that the impulse contains all frequencies from DC to Nyquist (half the sampling rate). Any distortion in the frequency response (aka linear distortion) has to produce ringing after the impulse.

Let's take a look at an EQ example:

Here all we have is a single narrow filter at 100 Hz that boost by a few dB. The only difference between the upper/lower output signal is that the Q (inverse of the bandwidth) of the filter was different. High Q, or in other words narrower peaks/dips, will cause longer ringing.

Why? The extra energy at 100 Hz has to be "released", and that happens immediately after the filter sees the impulse on the input. The extra energy produces the post-ringing we see. Check out the period of the post-ringing. It's 0.01 seconds, or 1/0.01s = 100 Hz which is exactly where we boosted.

But why does the higher Q ring longer? A little thought experiment: a perfect sine tone is indefinitely long because an abrupt start/stop would produce frequency content other than that of the tone. The opposite, an impulse, is theoretically indefinitely short and contains all frequencies.
On the one hand, if we increase the Q factor of the filter to infinity all that will be boosted is exactly 100.000... Hz causing indefinite ringing.
On the other hand, as we decrease the Q factor to zero all frequencies will be affected and since the impulse contains all frequencies all such a filter would do is output the impulse with higher amplitude - a simple gain "filter". All the extra energy will be concentrated at the instant of the impulse. No post-ringing.

Another property is stability, but it should be obvious that headphone drivers don't keep on producing post-ringing e.g. after an impulse steadily increasing in amplitude until they finally blow up.

--

So I guess you (may I say we?) should distinguish between FR related ringing and other ringing problems.

OJneg · « **Reply #14 on:** June 21, 2013, 07:04:05 PM »

Quote from: ultrabike on June 20, 2013, 06:28:17 AM

Some thoughts (skip-able stuff)

In my mind overdamped (slow non-oscillating decay), underdamped (fast oscillating decay), and critically damped (optimally fast non-oscillating decay) terms are given to second order systems (very useful for PLL characterization). I can visualize a headphone system as an equivalent system made up of several cascaded second order subsystems over, under, and critically damped at different "undamped" natural frequencies. Also, given the audio frequency range and the length of the cable, the amp and headphone together can be considered a single network system, where the different amps affect the overall system damping. In my mind this is not the whole story: non-linear distortion, response being a function of space/acoustic impedance/etc variations, and many things I have not yet thought off... but it's part of it.

Great explanation!

I think the second order analogy works well, but I don't quite see how it relates to PLL.

Also, would it be fair to say that an underdamped system would manifest itself as a long lived, high Q resonance, while an overdamped system would be a broader resonance at it's respective Fc? Or does an overdamped system kind of just blend in with the rest of the decay when you're looking at CSDs?

Maybe Purrin can provide some insight on this one. In headphones, do resonances seem to be mostly a function of the L & C components in the mechanical parameters (that is, compliance and mass) or is it the acoustic parameters (that is, acoustic C and inertance)? I suppose they could also be a function of electrical parameters, but something tells me those would be less of a factor in headphones.

Quote from: ultrabike on June 20, 2013, 06:28:17 AM

In my mind this is not the whole story: non-linear distortion, response being a function of space/acoustic impedance/etc variations, and many things I have not yet thought off... but it's part of it.

One more question. So when we see ringing that's an octave apart like in the K601, what we're actually seeing is the non-linear effects of the resonance at that first F, correct? With any given resonance, is there a good way to predict whether we'll see ringing an octave above? Could high THD generally lead to these Octave Penetrators, or does it have to do with the nature of the resonance more so?

Thanks

AstralStorm · « **Reply #15 on:** June 21, 2013, 08:16:29 PM »

From my silly simple Matlab FEM simulation, yes, a -1/2 octave, 1/2 octave, octave and 2 octave resonance can be caused by nonlinearity of the driver and perhaps reflection. Specifically, quadratic or tan-like nonlinearity. It will look like such on CSDs. This is almost always driver resonance.

Overdampened system will have little resonance and will sound like an anechoic room. Unless it's a driver, in which case high frequency extension and response will suffer.

In most high end headphones, almost all of the resonance comes from the casing.
Compliance/mass matters mostly in tuning open designs and even then, not as much as you'd think. Changes, unless really severe, will usually only shift resonances around.

Marvey · « **Reply #16 on:** June 22, 2013, 12:39:38 AM »

Cut From SHOUTBOX (the best way of describing it from what I've seen so far):

Ringing is mechanically and technically defined on that thread. Subjective listening can best be described as an artifact that stands out as an audible anomaly. I'm not the best of identifying ringing from listening as a causal factor, but I often hear things that are later seemingly correlated to ringing in a certain area when the measurements come out. For example, the 334 w/ sibilance and a certain treble timbre I described as mechanical. Ringing might be a cause or a correlation in this case, can't say for sure. I would say a peak comes of as an accentuation/over emphasis, ringing more a fatigue or annoyance issue. Lack of refinement. Though not all lack of refinement is from ringing. My 2cents

CONSIDERATIONS:
Normally, I wouldn't bother with CSDs (with speaker drivers that is.) Why is this? Because I can simply rely on FR graphs. With speaker drivers, I can very easily with minimal effort, get good measurements in a relatively confined space by gating the impulse response before the first major reflection (the wavelengths of the upper-mid / treble are short enough to fit within the gated interval).

This gating cannot be done with a headphone because the dimensions within the enclosure are often too small (one reason I rely on the sponge coupler - at least for CSDs - and BTW the techniques I have employed for FR differ significantly from the pictures which Jude posted on HF.) Although it should still be noted that FR peaks still often indicate ringing.

The short version of the story is that with speaker measurements, a peak (or severe null) in the FR is always indicative of ringing. With headphones, this is not necessary so because of the internal reflections causing odd resonances and nulls in the measurements. I was interested in CSDs because I wanted to isolate these "treble nasties" I was hearing with many headphones and find those trouble spots with precision.

In general, this is what I have concluded for myself (one looks for ridges in the CSDs.) Consider this a 2013 update:

see next post.

Marvey · « **Reply #17 on:** June 22, 2013, 01:05:34 AM »

5k Ringing

Examples: Grado, ATH-AD2000

It doesn't bother me that much.
Ringing at 5k tends to moderately broad with moderate decay length.
Adds a bit of edge, bite, and excitement to the sound.
I generally don't hear it unless I make a direct comparison with a transducer which does not ring. After that I'm like "oh, I hear it now".
For all I know, I may have slight hearing damage in this area. A 4-5k null. This is typical of industrial hearing damage.

6k Ringing

Examples: Senn HD800

Generally tolerable, but unpleasant.
Similar to 5k ringing in that it adds some bite.
Hardness bordering on glare.

7k Ringing

Examples: Ultrasone Pro900, Beyer T70, AKG K550 (variable), Fostex TH900 (minor - note ridge is very short lived),

Not good. Bad.
Glare galore with some sibilance

Marvey · « **Reply #18 on:** June 22, 2013, 01:08:07 AM »

8k Ringing

Examples: Grado, Beyer T5P, Beyer T1 (variable), LCD2r1 (harder to detect b/c overall treble is shelved down.), Beyer DT880 (slight), Beyer T5P

Very bad. Nasty. Painful
Sibilance with some tizziness.

9k-12k Ringing

Examples: Senn HD25-1 II

Adds definition.
More tolerable, but still unpleasant.
Tizziness and sizzle.

Marvey · « **Reply #19 on:** June 22, 2013, 01:23:13 AM »

Some important considerations:

Your experience may differ.
Take into account other behaviors (ringing elsewhere, overall frequency response curve / tonal balance, etc.)
Some headphones tend to vary from unit to unit. In some cases, the variance may be significant.
Consider length and width of ridge. In generally, lengthy ridges tend to be worse (and also more narrow.)
Orthos seem to have very narrow band ringing / resonance that looks like a wall. This doesn't appear to be audible and I think it gives us hints at how tight the diaphragm is tensioned (or if it's tensioned at all.)

Author Topic: Types of Driver Ringing (Read 11264 times)

ultrabike

Re: Types of Driver Ringing

Kunlun

Re: Types of Driver Ringing

Tyll Hertsens

Re: Types of Driver Ringing

xnor

Re: Types of Driver Ringing

OJneg

Re: Types of Driver Ringing

AstralStorm

Re: Types of Driver Ringing

Marvey

Re: Types of Driver Ringing

Marvey

Re: Types of Driver Ringing

Marvey

Re: Types of Driver Ringing

Marvey

Re: Types of Driver Ringing

(adsbygoogle = window.adsbygoogle || []).push({});

Author Topic: Types of Driver Ringing (Read 11264 times)