Headphones Amplifiers: 0-ohm output impedance and misleading specifications

Thanks for posting that.

Actually, the first article doesn't say that measurements are tweaked to look better than they are, but rather that manufacturers often don't provide meaningful measurements under realistic operating conditions.  What you want to see is excellent performance driving headphones-level impedances (and hopefully not just pure resistive loads) at high listening volumes.  Almost everything can do well when not loaded with a low impedance:  have low THD and flat FR and so on.
 

One interesting point here:

Coming from a different background where noise is noise and matching the signal is very well understood, I sometimes forget that human hearing can be good at detecting patterns below a noise floor.  I can't comment on 30 dB, but it's obviously true that if you hear someone speaking when there's lots and lots of static in the background, you can still often understand what's being said.

It would be interesting to see how the THD differs with damping factor using planars or dynamic headphones with relatively flat impedance curves. Both the V6 and HD650 have fairly large variances in impedance.

Thanks for posting these. I've got a ton of respect for Benchmark, partly because of how much respect they have for real, solid measurements. 

While the harmonics may raise by 30db, they're still 70db down. The best headphones may have a THD close to the -60db range :-/

I agree with Benchmark's efforts to support good design practices and prod the industry: an HP out should be able to support everything from a 16ohm IEM to an HD800 to an HE-6. Yet it's obvious we're getting into perfectionist territory here.

It would have been really nice to see some impulse and waterfall plots there. 

While I'm sure distortion goes down, I sometimes feel that headphones can be overdamped.

Quote:

Originally Posted by anetode 

... an HP out should be able to support everything from a 16ohm IEM to an HD800 to an HE-6. Yet it's obvious we're getting into perfectionist territory here.

not likely - dynamic headphones have 3 orders of magnitude differences in sensitivity, combined with >30x difference in load impedance over the many models, styles

it is not practical to make an amp "noiseless" with high sensitivity 16 Ohm iem and have enough gain for orthodynamics or "studio monitor" low sensitivity 600 Ohm headphones

I would say a couple of amps should be designed for the extremes, with gain switching to overlap some in the middle of the dynamic headphone drive requirement range - electrostatics will require a 3rd wholly different amp

headphone amp manufacturers need to supply more info than the standard home audio power amp because of the greater range of headphone loads, sensitivity

noise really needs to be quantified in addition to gain, non-distorting I,V capability, any output Z built into the amp

I'm kind of surprised by those measurements. I honestly did not expect such a shift in the THD+noise. The reason this really surprises me is because they have done studies about damping factor in typical speakers and found that these super ridiculous damping factors have massively diminishing returns. I find it hard to believe that headphones would change so much. I also find it hard to believe that an MDR-V6 can produce such a low THD+noise, even on a top headphone amp/DAC.

Quote:

Originally Posted by Armaegis 

It would have been really nice to see some impulse and waterfall plots there. 

 

While I'm sure distortion goes down, I sometimes feel that headphones can be overdamped.

Thank you for reading my whitepaper. I wanted to include some impulse response plots, but we had difficulty separating the impulse response of the microphone from that of the headphone. All of the measurements in the papers are taken at the input to the headphones. As pointed out in the paper, these are not acoustic measurements. Acoustic measurements of the headphone outputs are much more difficult to perform with adequate repeatability. If we are able to make some meaningful and repeatable measurements of the acoustic output, we will release the results in another paper.

Regarding overdamping:

Overdamping can only occur when mechanical damping (such as viscous fluid) is employed. Mechanical damping can slow the transient response of the driver (on both leading and trailing edges). Electrical damping provides more drive current to the driver (due to the low source impedance). This increased drive current quickly accelerates the driver to follow transients, and then quickly stops the driver after the transient. A high electrical damping factor will actually improve a driver that has too much mechanical damping.

Here is an analogy:

Consider the axle on your car. The axle is analogous to the diaphragm in your headphones. The tires ride over bumps in the road (transients), and shock absorbers provide viscous mechanical damping. High tire pressure is analogous to a low headphone source impedance (high damping factor). If your tires are soft, transients are partially absorbed by the tires, and are not transmitted to the axles. The mass of the axles (headphone drivers), and the low-frequency damping provided by the shock absorbers (viscous fluid in drivers), impede the motion of the axle (diaphragms). Soft tires increase the absorption of transients. This may be desirable for a smooth riding car, but it not desirable when controlling the motion of audio transducers. If the tire pressure is very high, every transient is transmitted to the axle. Likewise, if source impedance is low (high damping factor), the diaphragms follow every transient and are quickly damped after the transient.

John Siau

Benchmark Media Systems, Inc.

Hmm, very interesting analogy there. I had always assumed that electrical damping followed in a similar suit to mechanical damping and thus had regions of over/under/critically damped (I guess that's what I get for being a ME rather than EE), but this seems to be not the case? Are there any instances in audio where a lower damping factor may be desirable?

edit: nevermind, John Siau already covered it:  measurements were at the output of the source, not acoustic ones.

Quote:

Originally Posted by Armaegis 

Hmm, very interesting analogy there. I had always assumed that electrical damping followed in a similar suit to mechanical damping and thus had regions of over/under/critically damped (I guess that's what I get for being a ME rather than EE), but this seems to be not the case? Are there any instances in audio where a lower damping factor may be desirable?

The damping factor (for driving loudspeakers, headphones, etc.) means something different than the damping ratio for oscillatory systems.  In a traditional mass-spring-damper (or electrical RLC) system with some quantity governed by a second-order differential equation, the damping ratio is part of the coefficient to the first derivative term and corresponds to over/under/critically damped.

The electrical input to the system provides the input force, and so increasing it will not cause overdamping.  I think.

Correct

Quote:

Originally Posted by jcx 

dynamic headphones have 3 orders of magnitude differences in sensitivity

 

it is not practical to make an amp "noiseless" with high sensitivity 16 Ohm iem and have enough gain for orthodynamics or "studio monitor" low sensitivity 600 Ohm headphones

Noise was the most significant difference between the three amplifiers we tested.  The Benchmark had an "unfair" advantage over the other two amplifiers because the Benchmark has 3 user selectable gain ranges, and analog gain control.  The Benchmark  signal to noise ratio only fell to 96 dB while the others fell to 73 dB (when driving 60-Ohm MDR-V6 headphones at normal listening levels).  See figure 5 here:

http://benchmarkmedia.com/discuss/feedback/newsletter/2011/12/1/do-specifications-lie-do-our-ears-lie-where-does-truth-lie-examination

These tests demonstrate the importance of multiple gain ranges, and analog gain control.  The vast differences in headphone sensitivity can only be accommodated with multiple gain ranges.  Within each range, analog gain control can maintain a nearly constant SNR over a limited +/- 10 dB gain adjustment.  This implies that the gain ranges should be separated by about 10 dB.  This is why we chose 0 dB, -10dB, and -20 dB gain ranges.  These 3 ranges combined with +/- 10 dB analog adjustment provide a range 40 dB without a significant loss of SNR.  Most headphones can be accommodated within this 40 db range (two orders of magnitude).  A fourth range could be added for high-sensitivity 16-Ohm headphones.  Without this 4th range the 16-Ohm headphones will be limited to about a 90 dB SNR.