HeadRoom Total BitHead headphone amplifier Measurements

Sidebar 2: Measurements

I loaded the Total BitHead with four fresh AAA batteries and looked first at its performance via its analog input. The maximum voltage gain into 100k ohms was 11.4dB, and the DC offset varied between 5 and 7mV. This offset will not be high enough to damage headphones, but its presence led to a scraping noise when I rotated the BitHead's volume control. The input impedance at 1kHz was a moderate 47.5k ohms, dropping slightly and inconsequentially to 36k ohms at 20kHz. The unit preserved absolute polarity; ie, it was non-inverting.

As befits a headphone amp, the BitHead's output impedance (including 1m of interconnect) was a very low 0.3 ohm across most of the band, rising to a still low 2 ohms at 20Hz. The high-frequency response with the HeadRoom imaging processor switched out of circuit was flat to the 200kHz limit of my test system, though the low-frequency response was down 1dB at 18Hz. With the processor engaged, the BitHead's response differed according to whether: the signal was dual-mono (fig.1, top pair of traces below 700Hz; a mild bass boost and presence cut); both channels were driven but out of phase (fig.1, bottom traces below 700Hz; a presence boost but a mildly suppressed lower midrange and bass); or one channel only was driven (fig.1, middle traces; the same as when the processor was out of circuit). As a result of the wide bandwidth, the Total BitHead's reproduction of a 10kHz squarewave (fig.2) revealed very short risetimes and an excellent square shape.

Fig.1 HeadRoom Total BitHead, analog frequency response at 1V into 100k ohms (from top to bottom at 100Hz): both channels simultaneously and in phase; both channels individually; both channels simultaneously and out of phase (1dB/vertical div.).

Fig.2 HeadRoom Total BitHead, 10kHz squarewave into 100k ohms.

Fig.3 shows how the percentage of distortion and noise in the BitHead's output changes with output voltage into loads ranging from 100k ohms (bottom trace, way higher than the unit will see in practice) to 50 ohms (top trace, typical of, say, Grado headphones). The downward slope of the traces on the left-hand side of this graph indicates that true distortion is below the background noise level, which is low. But when the BitHead clips, it clips hard. Even so, it still gives out just above 1V into 50 ohms at 1% THD, which should be sufficient to drive headphones to high, if not deafening levels.

Fig.3 HeadRoom Total BitHead, distortion (%) vs 1kHz output voltage into (from bottom to top): 100k ohms, 10k ohms, 1k ohm, 100 ohms, 50 ohms.

Fig.4 shows that, into high impedances, the character of the distortion is almost pure second harmonic, which is subjectively benign. However, into a load more representative of real headphones, some high-order harmonics make an appearance (fig.5), these associated with the waveform's zero-crossing points and possibly due to the need for the BitHead to use a low-power-consumption output stage. (The op-amps used in the Total are the Burr-Brown OPA4743; the less well-specified BitHead uses the National Semiconductor LM6134. Both are low-power devices that can swing volts up to the rail values.) The highest-level distortion components are still the second (fig.6), but higher-order harmonics also make an appearance. However, intermodulation distortion was low (fig.7). (Ignore the presence of low-level AC-supply components in these last two graphs; they are related to my test setup, not the BitHead.)

Fig.4 HeadRoom Total BitHead, 1kHz waveform at 1V into 100k ohms (top), 0.0165% THD+N; distortion and noise waveform with fundamental notched out (bottom, not to scale).

Fig.5 HeadRoom Total BitHead, 1kHz waveform at 1V into 150 ohms (top), 0.027% THD+N; distortion and noise waveform with fundamental notched out (bottom, not to scale).

Fig.6 HeadRoom Total BitHead, spectrum of 50Hz sinewave, DC-1kHz, at 1V into 8k ohms (linear frequency scale).

Fig.7 HeadRoom Total BitHead, HF intermodulation spectrum, DC-24kHz, 19+20kHz at 1V peak into 8k ohms (linear frequency scale).

Turning to the Total BitHead's performance with digital input signals, it uses a Burr-Brown PCM2902, a single chip that includes a USB 1.1 data receiver, an oversampling digital reconstruction filter, and a two-channel 16-bit D/A converter. This chip can handle sample rates up to 48kHz, meaning that the BitHead will not decode high-resolution PCM data. (In any case, the USB 1.1 interface has too slow a data rate for such signals.) My measurements were taken using either iTunes or Bias Peak 4.0 playing AIF files on an Apple PowerBook with the computer's volume control set to its maximum in System Preferences, as recommended by HeadRoom.

The BitHead inverted signal polarity via its USB input, and a 0dBFS tone resulted in an output of 1.025V (the red clipping LED illuminated just below that level). Fig.8 shows the BitHead's frequency response when fed 44.1kHz-sampled data. The bass is more rolled-off than it was via the analog input, with the -1dB point at 31Hz compared with 18Hz. A 0.5dB channel imbalance can also be seen in this graph, as can a slight but possibly just audible rise in the top two audio octaves, over which is laid some passband ripple from the PCM2902's digital filter. Channel separation (not shown) was better than 90dB in the bass, decreasing through the usual capacitive coupling to a still adequate 56dB at 20kHz.

Fig.8 HeadRoom Total BitHead, D/A frequency response at -12dBFS into 100k ohms (right channel dashed, 0.5dB/vertical div.).

The PCM2902 is specified as having a 93dB dynamic range, slightly less than 16-bit PCM is capable of. Fig.9 shows the spectrum of the BitHead's output while it decoded 16-bit data representing a dithered 1kHz tone at -90dBFS. (Feeding the BitHead with 24-bit data from Peak 4.0 gave identical traces, confirming that the PCM2902 truncates incoming data to 16 bits.) As expected from the specification, the noise floor in this graph is higher than you get from state-of-the-art 16-bit DACs. More important, some negative error in absolute level is apparent that is associated with some second-harmonic content, though this is worse in the left channel than in the right.

Fig.9 HeadRoom Total BitHead, 1/3-octave spectrum of dithered 1kHz tone at -90dBFS, with noise and spuriae, 16-bit USB data (right channel dashed).

Plotting the linearity error against signal level (fig.10) confirms that the error becomes increasingly negative as the signal level drops below -80dBFS, because the DAC is diverting energy away from the fundamental into the second harmonic, an octave higher. Looking at the waveform of an undithered 1kHz tone at -90.31dBFS (fig.11), the obscured shape of what should be a well-defined wave with three discrete voltage levels indicates a highish noise floor and poor low-level linearity.

Fig.10 HeadRoom Total BitHead, left-channel departure from linearity, 16-bit USB data (2dB/vertical div.).

Fig.11 HeadRoom Total BitHead, waveform of undithered 1kHz sinewave at -90.31dBFS, 16-bit USB data.

Finally, examining the Total BitHead's rejection of word-clock jitter when fed USB audio data was somewhat problematic, as the measured amount of jitter was surprisingly dependent to some extent on the setting of the volume control. With the Miller Jitter Analyzer, the worst case was with the volume control full up, with a measured jitter level of 1.12 nanoseconds peak-peak. The best case was with the volume thumbwheel set to midway, resulting in a measured level of 897 picoseconds. However, the spectrum of the BitHead's output revealed the presence of some significant low-frequency jitter-related sidebands (fig.12, purple numerical markers), as well as a pair of data-related sidebands (red "7" markers). The noise floor in this graph is about 12dB higher than I see with the highest-performance 16-bit D/A converters, implying a true resolution of around 14 bits.

Fig.12 HeadRoom Total BitHead, high-resolution jitter spectrum of analog output signal (11.025kHz at -6dBFS sampled at 44.1kHz with LSB toggled at 229Hz). Source: Apple iTunes on Apple PowerBook playing 16-bit AIF file via USB input. Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz.

Through its analog inputs, the Total BitHead offered excellent measured performance. The introduction of higher-order distortion harmonics as the load impedance dropped suggests that high-impedance headphones will probably work better than lower-impedance models. Through its digital USB input, the BitHead offered only modest measured performance, but on the desktop, using data-reduced sources from a PC, that probably won't be an issue.—John Atkinson

2020 Gilkerson Drive
Bozeman, MT 59715
(800) 828-8184