External vs. integrated audio
About those few problems with on-board DACs and headphone amplifiers.
Oct. 19, 2020
There are a lot of hype surrounding high-end DACs and headphone amplifiers. Some audiophiles attribute magical qualities to high-end electronics, we can read reviews about how an external DAC changes the listening experience, sometimes dramatically...
However, there are very few problems with on-board DACs & amplifiers in notebooks, tablets and even cell phones today. The number of cases when switching to an external DAC can improve the listening experience is rather small and most of the time investing in an external DAC or amplifier is questionable. The problems that may occur with integrated audio are easy to identify and fall into three categories:
Noise is a real problem in analog electronics and not in digital electronics, however, some low-end products may have audible noise level. The headphone out on cell phones have SNR (signal-to-noise ratio) of 85-95 dBA and these are pretty good values. The noise floor of 95 dBA SNR is very unlikely to be heard even with a sensitive over-ear headphone, and the noise floor of 85 dBA is just barely audible in a quiet room. The other contributor to the high noise level is the 'SNR loss' when a high sensitivity headphone is used with a high gain headphone amplifier (see below).
The volume is limited by the maximum voltage of the headphone amplifier. Cell phones, some portable audio players and sound cards have low maximum voltage that is not enough to drive high impedance headphones to loud levels.
The output impedance is critical with certain types of open dynamic headphones
Open dynamic headphones with low impedance (<100 Ohm) from Focal, Sennheiser and Grado require amplifiers with low output impedance (~10 Ohms). I suppose that this is the main (and only) reason why sometimes audiophiles hear night and day difference between DACs or headphone amplifiers. (A comprehensive article about headphone amplifiers, impedance and response error can be found here.)
Non-linear distortion, frequency response errors and jitter (jitter: distortion caused by the fluctuations of the clock signal in DACs and ADCs) are intentionally omitted from the list. DACs and amplifiers have flat frequency response and extremely low distortion even with complex loads (THD<0.1%) in the audio band. Jitter was a serious problem in budget portable CD players two decades ago, but in modern audio players and sound cards jitter is below the hearing threshold.
Another source of distortion in a digital playback chain is the sample rate conversion (SRC). There is nothing wrong with sample rate conversion if it is done correctly, but fast SRC algorithms (linear, cubic, spline) generate audible non-harmonic distortion. Due to the huge number of sampling frequencies (44.1k, 48k, 88.2k, 96k, 192k) the sample rate conversion is inevitable. A simple sine sweep measurement can be used for checking the quality of the sample rate conversion or a quick audibility test: if the frequency of the test signal (pure tone, sine sweep) is higher than the top of the human hearing range, then only the distortion products are audible.
The level of the distortion products of jitter and sample rate conversion is proportional to signal frequency. It means that bass tones, vocals and instrument tones that lack high frequency content (such as piano) are the least affected by them.
The SNR loss
The sensitivity of headphones and the gain of the amplifiers have to be matched. A low sensitivity headphone will sound quiet with an amplifier that capable of delivering only 1 Volt (RMS). On the other hand a high sensitivity headphone will be too loud with a high-gain, high output amplifier designed to drive low sensitivity headphones. In this case, the volume (gain) has to be reduced and the signal-to-noise ratio of the amplifier will be reduced also.
Here is an example. An amplifier has max. voltage of 10 Volt and the SNR is 110 dBA. If we connect a high sensitivity headphone that becomes ear drum killer loud above 1 Volt, we have to reduce the volume by 20 decibels (for voltage 20 dB reduction is a division by 10). However the SNR is reduced by 20 decibels also so we have 90 dBA instead of the 110 dBA and the noise floor becomes slightly audible. A standard sound card has better performance.