Theta Digital Generation VIII D/A converter Measurements
Continuing a Theta tradition, the Generation VIII's maximum output level was very high, at 9V single-ended and 18V balanced. Its volume control, which operates in accurate 1dB steps, will therefore tend to be used around its -20dB setting in systems with typical gain architectures. Both sets of outputs preserved absolute polarity and featured very low source impedances: 11.5 ohms for the unbalanced RCA jacks, 23 ohms for the balanced XLRs.
Fed CD data, the Gen.VIII's frequency response was absolutely flat (fig.1, top pair of traces), and its de-emphasis error was negligible (fig.1, bottom traces). The processor had no problem locking on to high-sample-rate PCM data, though its response showed a touch of passband ripple before dropping off rapidly above 45kHz (fig.2). Channel separation (not shown) was astonishingly good below 2kHz, at better than 120dB in the L-R direction, 130dB R-L, and worsening only slightly at the top of the audioband.
Fig.1 Theta Generation VIII, CD frequency response at -12dBFS into 100k ohms, with de-emphasis (bottom) and without (top). (Right channel dashed, 0.5dB/vertical div.)
Fig.2 Theta Generation VIII, 96kHz frequency response at -12dBFS into 100k ohms (right channel dashed, 0.5dB/vertical div.).
The top pair of traces in fig.3 shows a 1/3-octave spectral analysis of the Theta's analog output while it decoded dithered data representing a 1kHz tone at -90dBFS. The noise floor's smoothness is entirely due to the 16th-bit dither noise in the signal. Increasing the word length to 24 bits dropped the noise floor by 15dB in the midrange and treble, suggesting that the Theta's DACs have almost 19-bit true performance, which is excellent. The lowest two traces in fig.3 show a similar spectral analysis for dithered 24-bit data representing a 1kHz tone at -120dBFS. These peak only slightly above the -120dB level and can be clearly be distinguished from the background noise. Though the Theta gets a helping hand in these measurements from its high output level, it does offer truly extraordinary dynamic range. Repeating the spectral analysis with a digital "black" signal (not shown) revealed the Gen.VIII's noise floor to rise above the audioband, reaching -75dB at 200kHz, due to the noise-shaping used to wring such high audioband performance from the DAC chips.
Fig.3 Theta Generation VIII, 1/3-octave spectrum of dithered 1kHz tone at -90dBFS, with noise and spuriae, 16-bit CD data (top) and external 24-bit data (middle), with 24-bit dithered 1kHz tone at -120dBFS (bottom). (Right channel dashed.)
Fig.4 shows the Gen.VIII's linearity error measured with a dithered 16-bit signal, which remains below ±2dB to below -110dBFS. In fact, by looking at fig.3, it should become apparent that this error is entirely due to the dither in the incoming data. The Theta's true linearity error is minimal to below -120dBFS. As a result, its reproduction of an undithered 16-bit, 1kHz tone at exactly -90.31dBFS is perfect (fig.5), with the Gibbs Phenomenon "ringing" at the onset of each of the three discrete DC voltage levels clearly defined. Increasing the undithered bit depth to 24 bits gives a pretty good sinewave (fig.6).
Fig.4 Theta Generation VIII, left-channel departure from linearity, 16-bit CD data (2dB/vertical div.).
Fig.5 Theta Generation VIII, waveform of undithered 1kHz sinewave at -90.31dBFS, 16-bit data.
Fig.6 Theta Generation VIII, waveform of undithered 1kHz sinewave at -90.31dBFS, 24-bit data.
The Gen.VIII's output stage is extremely linear (fig.7), with the harmonic spuriae associated with a full-scale 1kHz tone lying at -94dB or below (0.002%). The second harmonic is the highest in level; at low frequencies, the third and second swap places. Intermodulation distortion is also very low (fig.8), with the 1kHz difference tone resulting from a full-scale mix of 19 and 20kHz tones lying at -96dB (0.0017%). None of these distortion components will have any audible effects.
Fig.7 Theta Generation VIII, unbalanced output, spectrum of 1kHz sinewave, DC-1kHz, at 0dBFS into 8k ohms (linear frequency scale).
Fig.8 Theta Generation VIII, unbalanced HF intermodulation spectrum, DC-25kHz, 19+20kHz at 0dBFS into 8k ohms (linear frequency scale).
The Generation VIII has two choices for rejecting word-clock jitter on its data inputs, Reclock and Jitter Jail, the latter said to achieve the highest performance. Even so, assessing the Gen.VIII's jitter rejection with the Miller Audio Research Analyzer, I got a mere 230 picoseconds peak-peak of jitter driving the TosLink input from a WAV file on my PC with Reclock engaged. Switching in Jitter Jail dropped this already-low figure to 180ps. Fig.9 shows the spectrum of this jitter. The central spike is the 11.025kHz tone, which is used as the carrier for the LSB-level low-frequency squarewave that exercises the DAC's bit transitions. The sidebands spaced at 229Hz and its odd harmonics to either side of the spike (indicated with red numeric markers) are due to word-clock jitter. They are very low in level. The highest-level sidebands, at ±15.6Hz (purple "1" markers), contribute 59ps to the total.
Fig.9 Theta Generation VIII, unbalanced output, high-resolution jitter spectrum of analog output signal, WAV source on PC via RME Digi96/8 Pro and TosLink connection, Jitter Jail connected (11.025kHz at -6dBFS sampled at 44.1kHz with LSB toggled at 229Hz). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz. (Grayed-out trace is data from PS Lambda via S/PDIF connection.)
This graph indicates more random noise spikes than I usually see. More interesting, when I drove the Theta with an electrical S/PDIF connection from a PS Lambda transport, although the jitter level dropped even further, to 168ps, the central peak broadened (grayed-out trace in fig.9), which suggests the presence of very-low-frequency random jitter components with this source.
Finally, as the Gen.VIII has analog inputs, I tested it as a conventional line preamplifier. Its balanced input impedance at 1kHz was a respectable 20k ohms, but only 7k ohms unbalanced, which will give lightweight low frequencies with some tubed source components. The maximum gain was 5.8dB for both inputs with the volume control set to "86"; the unity-gain setting was "80." Neither input inverted absolute polarity.
The frequency response was the same for both unbalanced and balanced inputs, and had a very wide bandwidth: -1dB at 200kHz (fig.10, top pair of traces). This was with the volume control at its maximum; set to -20dB, which is more likely where it will be used, the Theta's bandwidth was a little wider, at -0.75dB at 200kHz (fig.10, bottom traces). Channel separation (not shown) was excellent, at around 100dB across the audioband, though this is not quite up to the standard of the digital inputs.
Fig.10 Theta Generation VIII, balanced analog input/output frequency response at 1V into 100k ohms, with volume control at +6dB (top) and -20dB (bottom). (Right channel dashed, 0.5dB/vertical div.)
The Gen.VIII's analog inputs were basically overload-proof—it took more than 12V to drive the balanced input into clipping, 6V for the single-ended input. Fig.11 shows how the measured THD+noise percentage changed with output voltage into both 100k ohms and 600 ohms: the distortion level actually lies at the limit of the Audio Precision System One, revealed by the sawtooth effect in the traces as the AP changes its gain. Even into the punishing 600 ohm load, the Gen.VIII will put out 19V RMS at clipping in balanced mode; the corresponding figure for the unbalanced output was 10.5V.
Fig.11 Theta Generation VIII, balanced analog input/output, THD+N (%) vs output voltage into 100k ohms (right) and 600 ohms (left).
As inexpensive digital products get better and better, it takes superb engineering skill to justify the price of a high-end digital processor. The Generation VIII's extraordinary measured performance demonstrates that Theta Digital's design team has that skill.—John Atkinson