Spectral SDR-2000 Professional D/A processor Measurements

Sidebar 3: Measurements

The SDR-2000 Pro's maximum output level was lower than that of most processors, measuring just 2.06V from the balanced outputs and 1.03V from the single-ended outputs. These low levels are a result of the 6dB attenuation in the SDR-2000 Pro's analog stage to conform to the HDCD requirement. When playing HDCD-encoded discs, the maximum levels will be 6dB higher: 4.12V from the balanced outputs, 2.06V from the single-ended outputs.

The output impedance measured 200 ohms from the unbalanced jacks, 400 ohms from the balanced outputs at any audio frequency. DC levels were below the 500µV limit of my voltmeter. As expected, the SDR-2000 Pro didn't lock to a 32kHz input signal, but decoded both 44.1kHz and 48kHz datastreams with no trouble. Many processors with trick reclocking schemes have trouble locking to different sources; the SDR-2000 Pro locked immediately to any source—even the highly jittered AES/EBU output from the Audio Precision System One.

The frequency response (fig.1) showed a surprising 0.3dB rise at 20kHz. The SDR-2000's response with a pre-emphasized signal (lower pair of traces in fig.1) was identical. I have no explanation for this (footnote 1).


Fig.1 Spectral SDR-2000, frequency response (top) and de-emphasis error (bottom) (right channel dashed, 0.5dB/vertical div.).

Channel separation (fig.2) was excellent, measuring better than 100dB at any audio frequency and 115dB at 1kHz. Although this is lower channel separation than the 125dB indicated in the owner's manual (albeit with no frequency specified), the SDR-2000's interchannel crosstalk performance was still excellent.


Fig.2 Spectral SDR-2000, crosstalk R–L (L–R dashed, 10dB/vertical div.).

Looking at a third-octave spectral analysis of the SDR-2000 Pro's output when decoding a 1kHz, –90dB dithered sinewave (fig.3), we can see a low noise level and good DAC behavior, but some 60Hz power-supply noise in the audio signal—particularly the right channel (dotted trace). I attempted to minimize the noise through floating the various grounds and different grounding connections between the Audio Precision System One and the SDR-2000 Pro, but this was the lowest noise I could get.


Fig.3 Spectral SDR-2000, spectrum of dithered 1kHz tone at –90.31dBFS, with noise and spuriae (1/3-octave analysis, right channel dashed).

Fig.4 is a wider-band spectral analysis of the SDR-2000 Pro's output when driven by an input signal of all zeros. We can again see the power-supply noise, but the most interesting feature of this plot is the rapid rise in noise energy above 20kHz. This noise is presumably the optional high-frequency dither generated by the PMD100 HDCD decoder/filter chip (footnote 2) [While this ultrasonic content is low in level, it might be a problem in non-Spectral systems that are marginally stable.—Ed.] The rolloff in the noise-floor energy starting at about 60kHz is due to the analog low-pass filter, which precedes Spectral's 1.2MHz-bandwidth output circuitry.


Fig.4 Spectral SDR-2000, spectrum of digital silence (1/3-octave analysis, right channel dashed).

As would be expected from the UltraAnalog DACs in the SDR-2000 Pro, the linearity plot (fig.5) was superb. The SDR-2000 Pro has virtually perfect linearity to below –105dB, with noise dominating the measurement below that level.


Fig.5 Spectral SDR-2000, departure from linearity (right channel dashed, 2dB/vertical div.).

Driving the SDR-2000 Pro with a 1kHz, –90dB undithered sinewave to look at the converter's low-level performance produced the plot of fig.6. If you think this looks like a dithered sinewave, you're right—the dither is introduced by the PMD100 decoder/filter chip. This makes it impossible to look at the undithered step transitions. Note how the dither appears to have a specific frequency around 25kHz rather than a random character—just what you'd expect after seeing in fig.4 how the dither energy is concentrated just above the audioband (footnote 3).


Fig.6 Spectral SDR-2000, waveform of undithered 1kHz sinewave at –90.31dBFS.

Fig.7 is the SDR-2000 Pro's noise-modulation plot. This is one of the best-looking results I've seen, with very tight trace groupings. This indicates that the SDR-2000 Pro's noise floor doesn't shift in level or change its spectral balance as a function of input level.


Fig.7 Spectral SDR-2000, noise modulation, –60 to –100dBFS (10dB/vertical div.).

Driving the SDR-2000 Pro with a full-scale mix of 19kHz and 20kHz and performing an FFT on the output created the plot of fig.8. This is the best performance I've measured in a digital processor. The 1kHz difference component is almost 100dB down, and no other intermodulation products are visible. Note that the level was increased by 6dB after the measurement was taken, to compensate for the 6dB of attenuation provided by the digital filter on non-HDCD data (like the test signal).


Fig.8 Spectral SDR-2000, HF intermodulation spectrum, DC–22kHz, 19+20kHz at 0dBFS (linear frequency scale, 20dB/vertical div.).

Finally, I was initially unable to measure the SDR-2000 Pro's jitter performance: the clock signal is buried inside a four-layer circuit board, and is inaccessible to a probe. Spectral provided me with a special test fixture that accessed the clock signal, but because of a ground interaction between the SDR-2000 Pro and the Meitner LIM Detector, I got readings that appeared to be invalid.—Robert Harley

Footnote 1: It might well make the Spectral sound "airy" in direct comparison with other processors, particularly as these tend to be a few tenths of a dB down at 20kHz.—Ed.

Footnote 2: I believe that, in their implementation of the PMD100 for the Mark Levinson No.30.5 and No.36 processors, Madrigal's design engineers decided not to use the chip's optional dither.—Ed.

Footnote 3: The higher components of the dither noise will be rolled off by the Audio Precision's anti-aliasing filter.—Ed.

Spectral, Inc.
442 Oakmead Parkway
Sunnyvale, CA 94086
(408) 738-8521

dial's picture

They still exist. Their products are well built, although the older ones use a lot of parts now unavailable. So take care in case of repair. But I don't know any owner of a Spectral unit with a single problem.
And they're beautiful. Like old Goldmund gear when it was french or during their first swiss years.

Herb Reichert's picture

converter as vivid and exciting - the class of its field

(at the time I was selling converters that competed directly with it)


rascs's picture

I worked at a dealer that sold these too. Spectacular, beautifully made top of class component. At one point we were back-ordered nearly a year despite Spectral's regular production. Would love to see the reprint of the Spectral/ Avalon/ MIT system that was positively reviewed. After hearing Spectral at the store and at CES it's hard to want anything else.
Thanks for reposting this.