Naim NA CDS CD player Measurements
The CDS put out 2.12V (left channel) and 2.08V (right) when decoding a full-scale, 1kHz sinewave. This is just slightly higher (half a dB) than the standard 2V output level. Frequency response was flat, though the treble showed some passband ripple from the digital filter. The CDS had no amplitude error when decoding pre-emphasized test signals. Frequency response and de-emphasis error are shown together in fig.1. The right channel is 0.16dB lower in output level than the left.
Fig.1 Naim NA CDS, frequency response at 0dBFS (top), and with de-emphasis (bottom). (Right channel dashed, 0.5dB/vertical div.)
Crosstalk between channels, shown in fig.2, was superbly low, measuring 110dB at 1kHz. The 6dB/octave upward slope of the curves, indicating lower channel separation in the treble, is indicative of capacitive coupling between channels. Note how closely the left-right and right-left plots overlap.
Fig.2 Naim NA CDS, channel separation (10dB/vertical div.).
Playing a test track of "digital silence," in which the encoded audio data are all zeros, is revealing of a digital system's noise as well as any converter-induced artifacts in the audioband. Fig.3 is a 1/3-octave spectral analysis of the CDS's output when decoding digital silence. This is among the best performance I've measured from any digital product. Notice the very low noise level and the almost total absence of power-supplyrelated noise. There are tiny blips at 60Hz and 120Hz, but they are lower than 125dB in levelsuperb performance.
Fig.3 Naim NA CDS, spectrum of digital black with noise and spuriae (1/3-octave analysis, right channel dashed).
Performing the same type of spectral analysis when the CDS decoded a 90.31dB, 1kHz dithered sinewave, but this time just looking at the audio band, produced the plot of fig.4. (Because of the use of record dither, any distortion products produced will be due to the player's decoding circuitry.) Again, the low noise floor is exemplary. Note that the scale has been expanded downward (to 130dB) to show the plot. When I ran the test with the typical scale of 60dB to 120dB, the spectrum was below the lowest horizontal division. When comparing the CDS's plot to those of other digital processor plots, keep in mind the different scaling. Note, however, that the 90.31dB test signal doesn't reach the 90dB horizontal division, indicating negative linearity error at this level.
Fig.4 Naim NA CDS, spectrum of dithered 1kHz tone at -90.31dBFS, with noise and spuriae (right channel dashed).
While I was measuring the CDS's linearity error, JA walked by and saw the intermediate plot of the "fade to noise" test used to calculate departure from linearity. Ideally, the line should be perfectly straight. As the curve was slowly plotted (over 30 seconds), JA remarked, "It looks like it uses the Philips TDA1541 DAC." After a few more seconds, he added, "And the Crown version." I thought he knew in advance that the CDS did indeed use the TDA1541 Crown DAC and was making a joke. He wasn't. He could tell by the slight but telltale negative sway in the curve that it was the 1541. Moreover, the relatively small amount of sway indicated that it was the premium Crown version! (footnote 1).
Getting back to the CDS's linearity error (fig.5), we can see the negative error typical of the TDA1541. The right channel is shown in the upper trace. This is fairly good performance, but not exceptional; some of the new DACs (the Burr-Brown PCM63P, for example) have much lower linearity error.
Fig.5 Naim NA CDS, left- and right-channel departure from linearity (2dB/vertical div.).
The CDS's reproduction of a 1kHz, undithered sinewave at 90.31dB is shown in fig.6. The overall shape is symmetrical and the quantization steps appear to be uniform.
Fig.6 Naim NA CDS, waveform of undithered 1kHz sinewave at -90.31dBFS.
Using the noise modulation test described in last February's Stereophile with the test signals on our Test CD 2, I measured how the CDS's noise floor shifts as a function of input level. The tighter the groupings, the better. The CDS has excellent performance above 3kHz (fig.7), but the curves depart significantly below that frequency, especially below 1kHz. When the traces cross, it means the noise floor's spectral balance is shifted by the input level. This is a relatively new test; we haven't yet established a correlation between performance on this test and any sonic characteristics.
Fig.7 Naim NA CDS, noise modulation (5dB/vertical div.).
The CDS performed well on the 19kHz and 20kHz intermodulation test (fig.8). This plot is made by playing track 30 from Stereophile's Test CD 2, which has an equal mix of 19kHz and 20kHz tones (the combined waveform reaching full scale), and performing an FFT analysis on the output. The 1kHz difference component is well suppressed, and the sidebands around the test tones are very low in level. The two spikes at 24.1kHz (sampling frequency of 44.1kHz minus 20kHz) and 25.1kHz (44.1kHz minus 19kHz) are more apparent on this plot than other digital products I have measured. This is characteristic, however, of the relatively slow Philips 4x-oversampling digital filter rolloff.
Fig.8 Naim NA CDS, HF intermodulation spectrum, DC-30kHz, 19+20kHz at 0dBFS, 100k ohm load (linear frequency scale, 20dB/vertical div.).
The CDS's reproduction of a 1kHz squarewave at full-scale (fig.9) is typical of the linear-phase 4x-oversampling digital filter used in the CDS. Output impedance was a low 99 ohms across the band, matching Naim's 100 ohms spec. Output polarity, checked by playing a test disc with a positive-going impulse and looking at the output with a 'scope, was non-inverting. Note that the CDS doesn't provide absolute polarity switching.
Fig.9 Naim NA CDS, waveform of undithered 1kHz squarewave at 0dBFS.
Playing the Pierre Verany test disc's intentional data-dropout tracks revealed the CDS to have adequate error-correction ability. The CDS played through track 33 but glitched on track 34, typical performance for a CD player or transport. Finally, I measured 200µV of DC at the left-channel output, no DC at the right-channel output.
Overall, the CDS measured fairly well on the bench. Its low-level amplitude linearity was not as good as that available from some newer DACs, but the player's very low noise floor and absence of power-supplyrelated noise was impressive.Robert Harley
Footnote 1: I once worked with a guy who could look at a CD and tell what mastering machine it was made onfrom any CD manufacturing plant in the world. Every mastering machine (except the one we built) has imperfections in the lead-screw that moves the turntable (or optics). These imperfections leave telltale variations in track pitch at the same disc radiusthe machine's "fingerprint."