NuForce CDP-8 CD player Measurements

Sidebar 3: Measurements

To measure the NuForce CDP-8 CD player, I used Stereophile's loaner Audio Precision SYS2722 system. (See and "As We See It" in the January 2008 issue.) For some tests, I also used my Audio Precision System One Dual Domain and the Miller Jitter Analyzer.

The NuForce CDP-8's maximum output level at 1kHz was 2.58V, which is 2.2dB higher than the standard 2V. This player will sound louder in any comparison with another product, something that will need to be compensated for. The output impedance was to specification, at 100 ohms across the audioband, but unusually, the NuForce's output was in inverted polarity. Error correction was good rather than great, the output producing no audible glitches with the Pierre Verany Test CD until the gaps in the data spiral reached 1.25mm in length. With 1.5mm gaps, the player muted its output and its front-panel display read "00:E4."

While the CDP-8's frequency response (fig.1, top pair of traces) was as flat as is usual with a high-quality CD player, the response with a preemphasized CD featured a lack of mid-treble energy (fig.1, bottom traces). Although there are relatively few preemphasized CDs around, these will sound rather lifeless on the NuForce. Channel separation (not shown) was superb, however, at >120dB in both directions below 8kHz, which is better than the specification.


Fig.1 NuForce CDP-8, frequency response at –12dBFS into 100k ohms with regular (top) and preemphasized data (bottom, right channel dashed, 0.5dB/vertical div.).

For consistency with the digital tests I have performed since 1989, I first examine a CD player's resolution by playing a dithered 1kHz tone at –90dBFS while sweeping the center frequency of a 1/3-octave bandpass filter from 20kHz to 20Hz. The result of that test with the NuForce CDP-8 is shown in fig.2: The traces peak at exactly –90dBFS, suggesting excellent DAC linearity, and other than a very small spectral bump at 60Hz in the left channel, presumably due to power-supply interference, all that can be seen is the effect of the dither noise used to encode this signal on the CD. Repeating the test with an FFT analysis gave the traces shown in fig.3: Again, the traces peak at –90dBFS, but the superior resolving power of this technique uncovers the presence of low-level sidebands either side of the 1kHz tone, at 876 and 1124Hz. These frequencies are equal to 1kHz, ±124Hz, so the sidebands are not connected with the full-wave-rectified power-supply frequency of 120Hz. I'm not sure from where they arise, therefore.


Fig.2 NuForce CDP-8, 1/3-octave spectrum with noise and spuriae of dithered 1kHz tone at –90dBFS with CD data (right channel dashed).


Fig.3 NuForce CDP-8, FFT-derived spectrum with noise and spuriae of dithered 1kHz tone at –90dBFS with CD data (left channel blue, right red).

Plotting the amplitude error as I reduced the level of a dithered 500Hz tone from –60 to –120dBFS resulted in a graph that showed only the effect of the recorded dither noise (fig.4). As predicted by figs.2 and 3, the CDP-8's linearity error and self-noise are both very low down to the 16-bit resolution limit of the CD format. As a result, the NuForce's reproduction of an undithered tone at exactly –90.31dBFS was essentially perfect (fig.5), with a symmetrical waveform and the three DC voltage levels that describe this signal well differentiated.


Fig.4 NuForce CDP-8, left-channel linearity error, dBR vs dBFS, 16-bit data.


Fig.5 NuForce CDP-8, waveform of undithered 1kHz sinewave at –90.31dBFS, 16-bit data (left channel blue, right red).

The CDP-8 offered very low levels of harmonic distortion. Even into the demanding 600 ohm load, the highest-level harmonic was the second, at –100dB (0.001%), which will be inconsequential (fig.6). However, there are now many more sidebands visible around the 1kHz fundamental than in fig.3, and the overall noise floor looks very granular. Comparing the two graphs suggests that the CDP-8's level of background noise and spectrum are related to the signal, which is referred to as "noise modulation." This behavior can also be seen in fig.7, which shows the spectrum of the NuForce's output while it reproduced an equal mix of high-level 19 and 20kHz tones. Ultrasonic images of the two tones can be seen at 25.1 and 24.1kHz, suggesting that the CDP-8's low-pass reconstruction filter is a littleleaky—something generally felt to correlate with better sound quality—and that its conventional intermodulation distortion is very low. But again, the overall noise floor looks granular.


Fig.6 NuForce CDP-8, spectrum of 1kHz sinewave at 0dBFS into 600 ohms (left channel blue, right red; linear frequency scale).


Fig.7 NuForce CDP-8, HF intermodulation spectrum, 19+20kHz at 0dBFS peak into 100k ohms, 16-bit data (left channel blue, right red; linear frequency scale).

Fig.8 shows that the background noise indeed depends on the signal being played. The cyan and magenta traces are the spectrum of the NuForce's output while it played a track containing data representing an undithered 1LSB DC offset. (I use this track to examine a player's noise floor with no signal playing rather than one containing "digital black" because many CD players will turn off their output when presented with digital black.) The noise is generally very low, other than a regular series of low-level spikes at the power-supply–related frequencies of 60, 180, 300, 420Hz, etc. But when I played a full-scale tone at 1kHz (blue and red traces), not only did the noise floor rise overall, presumably due to the recorded dither at the LSB level; also, the 1kHz tone developed a wide "skirt," and the 124Hz-spaced sidebands seen in earlier graphs made an appearance, along with others at different frequencies. It is very difficult to predict the effect of this noise modulation on sound quality, but I would have expected the NuForce to sound rather muddy and uninvolving.


Fig.8 NuForce CDP-8, low-frequency spectrum of 1kHz sinewave at 0dBFS into 600 ohms (left channel blue, right red) and of 1LSB DC offset (left cyan, right, magenta; both 16-bit data, linear frequency scale).

From where does this noise modulation arise? Looking at the spectrum of the CDP-8's output while it played a track with the diagnostic J-Test data (fig.9), a veritable picket fence of sidebands can be seen. While, commendably, none of these lies at the data-related frequencies of 229.5Hz and its harmonics (or, at least, they don't rise above the residual level of the harmonics in the test tone), numerous sidebands can be seen at 11.025kHz, ±124Hz and the harmonics of ±124Hz. This graph was taken with the Audio Precision SYS2722; the Miller Jitter Analyzer gave the same spectrum and calculated the jitter level to be 3.13 nanoseconds peak–peak. This is one of the highest levels of jitter I have measured, and about 20 times higher than in the best products I have measured.


Fig.9 NuForce CDP-8, high-resolution jitter spectrum of analog output signal, 11.025kHz at –6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz, 16-bit CD data. Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz (left channel blue, right red).

It is impossible to say where this jitter comes from. I wondered if it was related to the behavior of the switch-mode wall wart, but repeating the measurements with a conventional linear supply didn't make any difference. I must admit that I am at a loss as to why WP liked the sound of this player as much as he did.—John Atkinson

TakisJK's picture

... (Normally, a CD player constantly varies its speed, from 200rpm at the innermost data spiral to 500rpm at the outermost data spiral, in order to provide the DAC with a steady datastream.)...

I think this is wrong. The truth is exactly the opposite.

The first track on a cd is in the inner spiral and the last one in the outermost, so the cd spins at higher rpm on the 1st track and reduces speed as the time elapses and the laser moves away from the center.
The angular velocity is going down so that the linear velocity remains constant (at any point on the disc the laser reads, any given moment)

John Atkinson's picture
Yes, you are right TakisJK. I will amend the text accordingly. - JA
dcolak's picture

@John Atkinson: WP liked it so much because it was 2.2dB louder.

It's an old trick.