If IEC 61938 had been universally adopted, then designers could engineer their headphones to be driven by a standardised 120Ω output, taking into account the coloration born out of interaction between the amplifier output and the headphone's impedance. However, the standard was never taken up by amplifier designers in this way, and I can find no evidence that headphones were ever designed specifically for use with 120Ω outputs — only that studio and hi-fi headphones used to favour high voice-coil impedances, and thus were less impacted by the relatively high output impedance of the amplifier's driving circuit.
At Sonarworks we've carried out output-impedance measurements on most of the smartphones currently available. This was done to prepare for implementing our DSP calibration solution on smart devices, so that frequency-response deviations can be accounted for. Measurements were obtained from 170 devices, and most fall in the 1 to 10 Ω range. Notably Apple and Samsung smart devices had very little trouble providing low output impedance. Sony smartphones didn't fare too well, though, with output impedances in the teens, while the worst results were mostly from more obscure phones made by Gigabyte, Alcatel and OnePlus.
So what about pro gear? Sadly, we weren't able to test many audio interfaces, and although some manufacturers state the output impedance for their headphone outputs, most shy away from doing so. RME's Fireface UCX had a 30Ω output impedance, but their newer Fireface UFX II has a much lower output impedance of 2Ω, which is better. Antelope Audio seem to stick to the old 120Ω standard, although they have implemented an almost zero-Ohm output-impedance amp on their new AM RI converter. Interestingly, this device is also able to switch to a negative output impedance! Generally, it seems as though manufacturers are taking notice and are trying to design their headphone outputs with very low output impedances. I'd urge them to more clearly state the specs of their headphone outputs, as more and more audio creatives are using headphones as their primary monitoring device.
Many integrated amplifiers in hi-fi systems, meanwhile, have headphone outputs taken right from the power amp's output stage, with some resistors tacked on. It's a cheap way to do it, but one shouldn't expect wonders from budget-class equipment. The proper way to do it would be to add a dedicated headphone stage with a simple op-amp circuit, which costs very little.
The challenge of limiting power into low-impedance loads can be met in other ways with smart design. An obvious solution would be to implement a gain switch, providing a low-gain mode for low-impedance cans, and a high-gain mode for voltage-hungry high-impedance loads. Keeping an amp at low gain comes with numerous performance advantages: a stronger negative-feedback loop controls the amp more precisely, for lower noise and lower distortion. Dedicated headphone amps sometimes have this feature, but audio interfaces seem to omit it. The RME ADI‑2 DAC and ADI‑2 Pro are welcome exceptions, as they can be set to automatically increase the gain when volume is turned up. The ADI‑2 DAC even has a separate low-power output stage just for IEMs, implemented with lower gain and thus lower noise — which is especially important for super-sensitive models.
It might seem as though it's a jungle out there, with extreme variations in headphone sensitivity and load impedance, and audio interfaces with mystery specs, some of which outright ignore the requirement to keep the output impedance low. However, mostly out of necessity, there is some convergence going on. For one, it's extremely rare to see headphone manufacturers come out with truly inefficient designs. Headphone impedance is also generally going down, with the recently released 150Ω Neumann NDH20 being a notable exception. That particular model is reasonably efficient, though, and can be driven from most laptop outputs. Most other recent releases in the studio headphone world have tended to present about an 80Ω impedance, and only established models like the Sennheiser HD650 and HD600 or Beyerdynamic DT770 and DT990 stick to higher impedances that require an audio interface with a decent headphone driver to achieve a good volume headroom.
As mentioned previously, multi-driver IEMs have especially wild impedance swings due to their passive crossovers. Generally, our engineers have found the performance of IEMs too inconsistent to recommend them for critical applications, mainly due to their variable fit in the user's ears. They are, however, irreplaceable as stage monitors for performing musicians. Custom-moulded IEMs should at least mitigate the fit issue, but the problem of interaction between the output impedance of a wireless receiver's headphone amp and the tricky load presented by the IEM still looms. Specifications for the analogue outputs in popular wireless IEM receivers rarely list any measurements beyond frequency response, power delivery and noise level. Hopefully, their designers have taken into account that almost any output impedance will make delivering the promised tonal response impossible. On top of that, an output feed resistor will turn precious amplifier power into heat, thus shortening battery life.
It's unlikely that we'll ever end up in a situation where all headphone amps have near-zero output impedance. This being the case, there's another option: the tonal shifts resulting from output and load-impedance interaction can also be fixed digitally. Our lab uses Schiit Magni 3 amps, which all have less than 0.3Ω output impedance, and we've always stated that our calibration products are meant for headphone drivers with zero or near-zero output impedance. Potentially, someone using our product with a headphone output that has higher impedance would not be able to get the promised flat frequency response. To counter this real-world problem, we have thought about implementing an algorithm to account for the resulting tonal shift. The maths behind it is rather simple: all that would be needed is load impedance measurements for all 310 IEM model profiles and a way for the listener to find out the exact output impedance of their headphone amplifier. As seen in the tonal shift graph for the Shure SE535 above, the frequency response deviation follows the electrical impedance curve, with the magnitude changes depending only on the output impedance figure, so a correction curve can be easily calculated. The hard part will be getting hold of all the headphones to measure them again...
Rūdolfs Putniņš is Marketing Project Manager at Sonarworks.
As we've seen, the main reason why headphone amps are built with non-zero output impedance is to limit power delivery into low-impedance loads. However, it's not the only reason. Adding a feed resistor to the output of a headphone amplifier can also provide useful protection. Older op-amps, and even some discrete circuits, don't like being exposed directly to the complex load presented by the headphone and its cable, so a feed resistor is mandatory to keep the circuit stable. Thankfully, modern op-amps have become sufficiently rugged to make the series output resistor unnecessary, but removing it can still expose some amp circuits to risk due to the fact that a headphone's TRS jack plug momentarily shorts the amplifier outputs when plugged or unplugged. For example, SPL's Phonitor is a brilliant headphone amp, but it can be easily damaged or rendered unusable by careless unplugging of headphones when the music is playing — so be careful if you're lucky enough to own one! Or maybe it's time to start using XLRs for connecting headphones?
There are also distribution amplifiers for headphones that have output impedance of more than 12Ω, to ensure that connecting multiple headphones in parallel doesn't overload the output stage — each additional headphone connected reduces the resultant load impedance.
As well as potentially affecting the frequency response of connected headphones, adding a resistor to raise the output impedance of a headphone amp brings with it another problem. In 2011 John Siau, Director of Engineering at Benchmark Media, published a white paper titled The '0-Ohm' Headphone Amplifier, with measurements that illustrate an increase in harmonic distortion when a feed resistor is used to increase an amp's output impedance from near-zero to 30Ω. Sony's MDR-V6 — a popular 60Ω headphone — is shown to suffer an increase of total harmonic distortion of up to 55dB when the 30Ω resistor is added to the amplifier's output. This suggests that using feed resistors in amplifier outputs might not only change the frequency response, but also add distortion.
Siau suggests that this distortion increase is due to reduced electrical damping of the driver, as replacing the Sony headphone with a simple 60Ω resistor showed no change in distortion from the amplifier. The same test was done with Sennheiser HD650 headphones, which also showed a marked increase in distortion, albeit at lesser magnitude and mostly confined to the sub–300Hz region.
High output impedance thus increases the distortion of the whole system comprising both the amplifier and the headphone itself — and unlike frequency-response anomalies, this cannot be addressed by clever DSP trickery. At least not yet!