DAC Spectrum Measurements

On this page we show some measurements that we've performed on different points inside our DAC. The goal of these measurements is to verify proper operation of the PLL, which should improve the stability of the clock signal used to perform DA-conversion.

At the bottom of this page two pictures are included showing the distortion spectrum on the DAC audio outputs.

The first two figures below show measurements on the 11.3MHz master clock output of the Crystal CS8412 input receiver.

Figure 1: MCLK CS8412, no signal
Figure 2: MCLK CS8412, 1kHz signal
The left picture shows the master clock when no audio signal is present on the SPDIF (CD player is in pause mode). The right picture shows the same signal whena 1kHz tone is played. As you can see, the noise level increases, and the 1kHz tone is modulated into the carrier. This clearly shows that you don' want to use this clock signal to drive the DA-convertor directly.

The same measurement as in Figure 2 is shown in Figure 3, but with a wider bandwidth. It shows that besides the 1kHz tone, also the 44kHz sample frequency is present as well.

Figure 3: MCLK CS8412, 1kHz signal
Figure 4 shows the results of Figure 2 again. Figure 5 shows the master clock signal as generated by the VCXO in our PLL, both while playing a1kHz tone. These figures clearly show that the VCXO clock firstly has better noise performance, and secondly is not affected by the audio signal. (Be aware of the different vertical scale!)
Figure 4 (= 2): MCLK CS8412, 1kHz signal
Figure 5: VCXO output, 1kHz signal
The PLL internally compares these two signals in its 'phase comparator'. The DAC circuit provides this comparator signal for audible verification, called 'PLL-sound'. This comparator signal is low-pass filtered to control the VCXO. Figures 6 to 9 below shows this 'PLL-sound' signal.

First Figure 6 shows the PLL-sound which results as difference from the signals 4 and 5. Figure 7 shows PLL-sound when the player is in pause mode. In this case the CD-player is a Denon CD player.

Figure 6: PPL-sound in play mode
Figure 7: PLL-sound in pause mode
The figures clearly show that the spectral contents of the noise signal is larger for the play mode, as you would have expected from the previous figures. The same is true fo the 1kHz modulation of the clock signal, clearly seen in the difference signal in Figure 6, demodulated back into the audio domain. Auditioning the PLL-sound also reveals music signals, which also implies that the music signal is correlated in the receiver clock.

Another interesting observation is the difference between CD transports. Using a TEAC CD-transport, modified by Guido Tent, in exactly the same set-up, shows the following figures.

Figure 8: PLL-sound in play mode, modified TEAC transport
Figure 9: PLL-sound in pause mode, modified TEAC transport
These Figures 8 and 9 show clearly improved spectral noise levels, and reduced contamination by music. This leads to the conclusion that the Crystal receiver is not the only source of trouble, apparently the quality of the SPDIF input signal affects its clock output.

For other jitter measurements related to CD-transports, see our CD-transport page.

Finally, figures 10 and 11 show the spectral output of the audio signal of the DAC. Figure 10 shows the filtered output signal of the PCM-63 DAC chips (called passive). Figure 11 shows the output signal after amplification to normal line-levels (using an 6N1P output tube).

Figure 10: Audio signal passive
Figure 11: Audio signal after amplification
The passive output shows a second harmonic distortion of -92dB. Figure 11 shows the increased second harmonic distortion (-51 dB) caused by the tube amplification stage without feedback. The nice property of the tube amplifier is the absence of any higher-order harmonics.

Copyright © 2001, Marc Heijligers and the DAC group - All rights reserved.