VTL Reference D/A processor Measurements

Sidebar 3: Measurements

The VTL's output level when decoding a 0dB, 1kHz signal was 3.26V (left channel) and 3.22V (right channel). This is 4.2dB higher than the standard 2V output for a full-scale signal. Frequency response was flat (fig.1), down just 0.2dB at 20kHz. Note also that the right channel is about 0.1dB lower in level than the left. Fig.2 shows the VTL's de-emphasis error. This plot is actually the decoder's frequency response when playing an emphasized test signal, but since the VTL's frequency response was flat, this graph represents only errors in the VTL's de-emphasis curve. The sharp corners in the trace are a result of the measurement being taken at five discrete frequencies (125Hz, 1kHz, 4kHz, 10kHz, 16kHz) rather than with a swept sinewave as is frequency response. At any rate, this is the worst de-emphasis error I have measured in a CD player or digital processor. Typically, digital decoders with some de-emphasis error exhibit a few tenths of a dB deviation. The VTL's error was +0.5dB at 3kHz and worsened to +1.5dB at 16kHz. This will make pre-emphasized discs sound bright or fizzy. The fact that both channels have the identical error indicates that component tolerances are probably not responsible.

Fig.1 VTL Reference, 44.1kHz sampling, D/A frequency response at –12dBFS (right channel dashed, 0.5dB/vertical div.).

Fig.2 VTL Reference, 44.1kHz sampling, de-emphasis error (right channel dashed, 0.5dB/vertical div.).

Looking at the spectral content of the VTL when decoding a dithered –90.31dB, 1kHz sinewave (fig.3), we can see a fairly high level of 120Hz full-wave rectified power-supply–related noise, as well as harmonics of the power-supply noise at 240Hz. The amount of high-frequency spuriae is generally low, with no harmonics of the 1kHz signal apparent. Notice also the exceptional linearity, indicated by the 1kHz tone's amplitude falling exactly at the –90dB horizontal division.

Fig.3 VTL Reference, spectrum of dithered 1kHz tone at –90dBFS with noise and spuriae (1/3-octave analysis).

The departure from linearity plot (fig.4) reveals the extraordinary linearity of the UltraAnalog DACs. The Fade to Noise is virtually a straight line, while linearity error at –100dB is remarkably low.

Fig.4 VTL Reference, left-channel linearity error, 16-bit data.

Channel separation (fig.5) was moderately good, with the L–R slightly different from the R–L. Note the different curve shapes. This measurement is made by decoding discrete frequencies in one channel and measuring the crosstalk in the other channel. The best separation was R–L at 125Hz (95dB), the worst was R–L at 16kHz (62dB). This is generally adequate, but is far from the Stax DAC-X1t's separation of better than 130dB through most of the band, decreasing in one channel to 110dB at 16kHz.

Fig.5 VTL Reference, channel separation ref. 0dBFS, R–L top, L–R bottom (10dB/vertical div.).

No interchannel phase error was measured, typical of dual D/A converter designs. Output impedance was 324 ohms at 1kHz, 341 ohms at 17Hz, but 3456 ohms at 20kHz. The 1kHz, 0dB squarewave (fig.6) is typical of the NPC digital filter in that the Gibb's phenomenon ringing is clipped. Looking at a –90.31dB dithered 1kHz sinewave was fairly difficult due to the somewhat high level of 120Hz noise. (There was also some very-low-frequency noise present on the unit's output.) However, we can see from fig.7, which shows an undithered 1kHz waveform at the same level, the remarkable precision of the UltraAnalog DAC. The steps at +1, 0, and –1 are clearly visible, indicating the DAC's precision.

Fig.6 VTL Reference, 1lHz squarewave at 0dBFS.

Fig.7 VTL Reference D/A, undithered 1kHz sinewave at –90.31dBFS (16-bit data).

Finally, fig.8 shows the VTL's output spectrum while decoding a signal representing a 1:1 mixture of 19kHz and 20kHz tones with the composite waveform peaking at 0dB. While the aliased image of the 20kHz component at 24.1kHz (44.1kHz–20kHz) is buried in the noise—which is excellent—the 1kHz intermodulation product at 1kHz is higher in level than usual at –62.1dB, shown by the cursor position. This is a severe test of a CD player's or D/A processor's HF linearity, however, though this behavior might contribute to my feeling that VTL homogenized instrumental outlines when compared with the Stax.—Robert Harley

Fig.8 VTL Reference, HF intermodulation spectrum, 300Hz–30kHz, 19+20kHz at 0dBFS into 100k ohms (log frequency scale).

Vacuum Tube Logic of America
4774 Murrietta, Suites 9 & 10
Chino, CA 91710
(909) 627-5944