Q. Why are there no distortion measurements?
There will be in due course but this is contingent on me building a replacement impedance converter for the microphone in the artificial ear. This is necessary because the principal barrier to achieving high-resolution headphone distortion measurements is self-noise in the measurement microphone, which typically restricts the dynamic range to about 60dB at 90dB SPL, and considerably less at lower frequencies due to 1/f noise in the impedance converter. The noise problem can be overcome by a combination of time and spectral averaging but to achieve the results I wish to this would involve impractically long measurement duration. If all goes to plan, the repacement impedance converter will lower the 1/f noise corner frequency sufficiently to reduce the amount of averaging that's required by a large factor. Then I will be able to make and present distortion measurements that are world-class, in which the inherent distortion of the measurement microphone sets the resolution limit. The issue of microphone noise is described in greater detail here
Q. Why are there fewer measurements for some headphones than others?
Primarily this is because some of the headphones included here I tested before I'd developed the complete suite of measurements I now use. A notable example is the Sennheiser HD 800 S
, which has the cleanest CSD waterfall I've ever measured from a headphone. (Note that the HD 800 S has now been fully retested here
.) In other cases where certain measurements are missing it's usually because they're not applicable to that particular headphone. For instance, headphones with a four-wire connecting lead cannot have electrical crosstalk caused by the shared earth impedance of a three-wire connection.
Q. What is a polarity inverting headphone and why does it matter?
By convention, a positive-going signal voltage should give rise to a positive-going change in sound pressure. In loudspeakers this can be complicated by certain crossover networks which require the drive units to be connected in anti-phase to ensure a flat frequency response. In most headphones, which use a single drive unit to cover the entire audible frequency range, this is not a factor and there is no excuse for the polarity convention to be flouted. The audible consequences of polarity inversion have been the subject of controversy for many decades since the 'absolute phase' issue came to prominence. Some claim that polarity inversion is clearly audible (assuming an input signal in which correct polarity has been scrupulously maintained), others that it is inaudible. What isn't disputable is that no passive headphone should be guilty of polarity inversion. It is carelessness by the manufacturer, plain and simple, although some ANC headphones are polarity inverting in passive mode only because of how they operate. The PSB M4U 8
is an example. It is, of course, possible that having been alerted to what is generally a wiring problem, a manufacturer will fix it. I am very happy to acknowledge that change in HTL, provided that the manufacturer supplies a new series-production sample for testing which resolves the problem. It will be a first if this happens – but the opportunity is there. Headphones which prove to be polarity inverting in HTL's tests are indicated by this symbol on their test results page:
Q. Why are some of the uncorrected frequency responses noisy?
They aren't. What looks like noise in the measurement is actually caused either by closely-spaced resonances, usually in planar magnetic headphones, or – with open-back and floating headphones – by room reflections passing back through or around the headphone capsule to the measurement microphone in the artificial ear. This unevenness could be suppressed by smoothing the response traces but that risks obliterating what, in the case of diaphragm resonances, is genuine and relevant information. Reflection effects are minimised in HTL's testing by performing measurements as far from reflective surfaces – particularly the walls, floor and ceiling of the room – as possible, and using an adaptive window analysis method described here
. The presence of resonances can be verified by checking the headphone's CSD waterfall, where they will appear as ridges, and the fitted impedance versus frequency graph, where they will appear as small peaks or a similarly 'noisy' trace.
Q. Why in the legacy tests is there often low frequency unevenness in the impedance versus frequency traces?
These older impedance tests were performed using pink-spectrum periodic noise as the test signal, and this unevenness was an unfortunate artefact. It was to eliminate this, and make the impedance data more consistent, that the change was made to the stepped-sine test method now used. Impedance testing takes longer to perform this way but the quality of the results is ample justification.