RME Digi96/8 PAD computer soundcard Measurements
Other than the addition of a two-channel A/D converter, RME's Digi96/8 PAD is identical to the Digi96/8 Pro I reviewed in the November 2000 and January 2001 issues. I refer readers to those reviews for measurements of its D/A section.
To test the card's ADC, I drove it from the unbalanced analog outputs of my Audio Precision System One DSP. The program used was CoolEdit Pro, which can operate at sample rates of up to 96kHz, the fastest the RME is capable is of. I also used RME's excellent DIGICheck program. The resultant WAV files stored on the host computer's hard drive were analyzed by using the RME card's AES/EBU output to feed either the Audio Precision's digital input or the digital input of a PrismSound DScope 2.
I looked at the card's A/D and D/A frequency response from its analog outputs. The result, with a 96kHz sample rate, is shown in fig.1. Flat in the audioband, the response has a slight rise apparent above the audioband, before sharply rolling off in the transition band above 40kHz. The analog input impedance was a fairly low 9.7k ohms, which will not be a problem, other than with tubed sources that might have a high output impedance at low frequencies. The sensitivity, however, was quite high, only 1.24V RMS being required to drive the input to digital full-scale. (Internal jumpers can be removed to reduce the sensitivity to 6.9V RMS and increase the input impedance to 50k ohms.)
Fig.1 RME Digi96/8 PAD, A/D/A frequency response at -12dBFS from headphone jack, 96kHz sampling (right channel dashed, 0.5dB/vertical div.).
Set to 24-bit precision—and DIGICheck both confirms that all 24 bits are active and indicates that the sample rate is within ±50ppm—the Digi96/8 PAD's A/D converter offers low linearity error down to below -100dBFS (fig.2). As a result, the waveform of a -90dBFS 1kHz sinewave, converted with 24-bit precision (fig.3), is excellent, if a little noisy. Spectral analysis of the WAV file representing this signal (fig.4) indicates that the only distortion harmonic present with this low-level signal is the third, with a smooth noise floor indicating that the RME's converter is properly dithered.
Fig.2 RME Digi96/8 PAD, left-channel departure from linearity, 24-bit data (2dB/vertical div.).
Fig.3 RME Digi96/8 PAD, digital-domain waveform of 1kHz sinewave at -90dBFS, 96kHz sampling, 24-bit data. (Note that the displayed waveform includes the reconstruction filter behavior. The actual sample values are displayed as black dots.)
Fig.4 RME Digi96/8 PAD, digital-domain spectrum of 1kHz sinewave, DC-24kHz, at -90dBFS, 48kHz sampling (linear frequency scale).
Performing a similar spectral analysis for a 1kHz tone at -1dBFS (fig.5) shows that the third harmonic (at a very low -105dB) is joined by a smidgen of second harmonic (at -118dB!), but the noise floor remains white and free from spuriae. (It should be noted that these results include the residual distortion in the Audio Precision's analog output.) However, when the analog input level was raised to just below digital clipping (fig.6), many more harmonics appear, these joined by enharmonic products as well. The THD+noise percentage, calculated by the DScope software, rises from -101.3dB for 1kHz at -1dBFS to -93.6dB at -0.05dBFS. As with most A/D converters, the top 1dB of the PAD's allowable dynamic range should be avoided, except in emergencies.
Fig.5 RME Digi96/8 PAD, digital-domain spectrum of 1kHz sinewave, DC-24kHz, at -1dBFS, 48kHz sampling (linear frequency scale).
Fig.6 RME Digi96/8 PAD, digital-domain spectrum of 1kHz sinewave, DC-24kHz, at -0.05dBFS, 48kHz sampling (linear frequency scale).
When I measured the Digi96/8 Pro's D/A stage, I felt it was okay for monitoring but that the card's digital output would be best fed to an external audiophile DAC for the best sound quality. The PAD's A/D stage, however, offers excellent performance; you'd have to spend much more to get an appreciable improvement.—John Atkinson