Tube Preamp Reviews

Following Dick's auditioning, I carried out a few measurements to examine whether there were any correlations between them and the CAD-5500's sonic performance. Looking at the output waveform with a unidirectional pulse revealed the CAD-5500 to be non-inverting. Channel separation was somewhat disappointing, measuring around 50dB in both directions across most of the audio band with the input of the measured channel shorted to ground (fig.1). Though this is significantly worse than the intrinsic separation offered by even inexpensive CD players, it is still low enough that soundstage width will probably not be affected.

Fig.1 Cary CAD-5500, channel separation, L–R (top), R–L (bottom), (0dB = 1V input).

The input impedance is quite low for a tube design—I measured 9k ohms with the volume control full, which might lead to a lightweight bass with tube CD players featuring a high, AC-coupled output impedance—the original CAL Tempest comes to mind.

Output impedance was, as DO stated, high at a measured 8650 ohms (left channel) and 7830 ohms (right), which will undoubtedly give a dark tonal balance with long or excessively capacitive interconnects. Looking at the Cary's frequency response, however, reveals that even with a low-capacitance, 220pF interconnect, the unit does tend to roll the high frequencies off a little early. Fig.2 shows the measured response of the processor's CD input from 10Hz to 50kHz with a 100mV input level and the volume control set to 12 o'clock, at which position the CAD-5500 features unity gain. Both channels are an audible 2.5dB down at 20kHz.

Fig.2 Cary CAD-5500, CD input frequency response at 100mV (right channel dashed, volume control at 12 o'clock, cable capacitance = 220pF, analyzer input impedance = 100k ohms, 1dB/vertical div.).

Note also the 2dB rise in the bass centered on 20Hz, which will also be audible as added weight or authority to the sound. (Both CD and Aux inputs showed identical behavior.) A 1V RMS signal was slightly compressed with the volume control wide open (by about 0.5dB), with then only a 1dB bump in the low bass, both these factors presumably being due to the processor's soft-clipping action.

Although the balance control was centered for these measurements, there is a 1dB imbalance between channels noticeable in fig.2. With the volume control wide open and the balance control centered, the left channel gave 12.1dB gain at 1kHz, the right 13.1dB with a 100mV input signal.

I was unable to examine the Cary's rejection of radio-frequency energy, but I did look at the ultrasonic rolloff in more detail than shown in fig.2. Fig.3 shows the CD input response measured with a swept sinewave up to the 200kHz limit of the Audio Precision setup. I was expecting rather more out-of-band rejection than –20dB at 200kHz, and the slope appears surprisingly shallow. The input level for this measurement was 100mV and both green front-panel LEDs stayed dark. They did light for input signals above 2kHz, however, when the level was raised to 500mV, and for signals above 1kHz when the input level was 1V.

Fig.3 Cary CAD-5500, CD input frequency response at 100mV (right channel dashed, volume control at 12 o'clock, 5dB/vertical div., measured with swept sinewave).

Perhaps the reverse-phase, ultrasonic-noise–canceling circuity is more effective with noncontinuous signals. I looked at the spectrum of wideband pink noise, therefore. Fig.4 shows both the spectrum of the Audio Precision's pseudo-random pink-noise output signal up to 200kHz and the spectrum of this noise after being processed by the CAD-5500. It can be seen that the suppression of the ultrasonic content of the noise is effectively identical to that shown in fig.3.

Fig.4 Cary CAD-5500, output spectrum compared with pink-noise generator spectrum, both 1–200kHz (right channel dashed, volume control at maximum, 5dB/vertical div., measured with swept bandpass filter).

How about the ultrasonic noise typical of a CD player? Fig.5 shows the spectrum of a Magnavox CDB472 while reproducing a 1kHz tone at –60dB, both fed straight into the analyzer and via the Cary CAD-5500 with the volume control full. (0dB on this graph is equivalent to –60dB.) The unprocessed output features peaks centered on the sampling rate and its multiples: 44.1kHz, 88.2kHz, and 176.4kHz. It can be seen that, apart from the lowest ultrasonic peak, the others are reduced in level by 20dB or more. Though this is good, it doesn't correspond to the much better ultrasonic rejection shown in Cary's own figures (figs.A & B in the main text). The power-supply ripple noise at 120Hz can be seen to be raised by 10dB by the Cary, though this is to a still innocuous –96dB.

Fig.5 Magnavox CDB472 CD player, output spectrum of 1kHz tone at –60dBFS, compard with same signal after processing by CAD-5500, both 100Hz–200kHz (volume control at maximum, 0dB = –60dBFS, 10dB/vertical div., measured with swept bandpass filter).

There were no differences in this behavior with the volume control set above 12 o'clock, unity gain, but below this position the ultrasonic rolloff appeared to diminish. Fig.6, for example, shows the CD input's response with the volume control set to 7 o'clock, apparently indicating that the response actually starts to peak above 200kHz. An enigma, perhaps due to inaccurate level matching between the main signal and the reverse-phase ultrasonic signal, but also an indicator that the Cary's volume control should be kept above the 10 o'clock position for the best ultrasonic noise rejection.

Fig.6 Cary CAD-5500, CD input frequency response at 100mV (right channel dashed, volume control at 7 o'clock, cable capacitance = 220pF, analyzer input impedance = 100k ohms, 5dB/vertical div.).

My examination of the Cary's noise rejection was limited to a bandwidth of 200kHz, well below the typical RF noise frequencies. Although the rolloff of noise below 200kHz was less than I was expecting, it will still result in good RF noise rejection.

Finally, to examine the CAD-5500's "soft-clipping" feature, I examined the measured THD+noise at three different input levels—100mV, 500mV, and 1V—the latter being some 6dB below the typical CD-player maximum output level of 2V RMS, with the volume control wide open. As can be seen from fig.7, the level of distortion is significant even at 500mV input, while the 1V input gives more than 3% distortion over most of the band. Backing off the volume control brought the distortion down—fig.8 shows the distortion vs frequency for 1V and 500mV input levels with the volume control set to unity gain—suggesting that it is the output stage that limits when asked to give much more than 1V RMS.

Fig.7 Cary CAD-5500, THD+noise (%) vs frequency for 1V RMS (top), 500mV RMS (middle), and 100mV RMS (bottom) input signals (right channel dashed, volume control at maximum).

Fig.8 Cary CAD-5500, THD+noise (%) vs frequency for 1V RMS (top) and 100mV RMS (bottom) input signals (right channel dashed, volume control at 12 o'clock).

With the volume control set for less than unity gain, there appeared to be an increasing level of distortion and noise imposed on high-frequency signals. Fig.9 shows that imposed on a 1V input signal with the volume control set to 9 o'clock. It is hard to say whether the rapidly rising level of spuriae for signals with frequencies of more than 1kHz is due to distortion or added ultrasonic noise—the waveform appeared "furry." If the latter, then this measurement, too, suggests that the Cary processor's reverse-phase ultrasonic noise-canceling works best with the volume control set to give unity or higher gain.

Fig.9 Cary CAD-5500, THD+noise (%) vs frequency for 1V RMS input signal (right channel dashed, volume control at 9 o'clock).

Fig.10 shows the level of distortion imposed on a 1kHz tone plotted against output level with the volume control full, thus revealing the soft-clipping nature of the Cary's output stage. If 0.3% is regarded as a reasonable threshold above which it will be possible to hear the distortion, at least on pure tones, it can be seen that this level occurs with a 1.5V output. As this is about the level necessary to drive most US amplifiers into clipping, it is probable that the Cary's higher distortion above this level will be academic.

Fig.10 Cary CAD-5500, THD+noise (%) at 1kHz vs output voltage (left channel only).

Overall, despite Dick's enthusiastic response to the sound of the Cary Audio CAD-5500, I was a little disturbed by its measured performance. Given that it will reduce the level of RF energy in the signal it passes, things it does in the audio-band still cannot be ignored. The rolled-off top octave, the bump in the bass, the limited channel separation, and the distortion from the soft-clipping circuit are all things which either shall or might be audible. They will certainly confuse the issue of whether the Cary's good sound is due to just the reduction in RF noise, or to a combination of that plus the audio-band changes, or even, as pointed out by Stanley Lipshitz in his letter this month, to just the latter.

When a device like this effectively acts as an equalizer, it would be more fair to its purchaser, in my opinion, for its manufacturer to make its effects capable of being defeated, as with the CD tone-shaping circuit featured by, for example, the Adcom GCD-575 CD player. Such things as added distortion should be optional rather than mandatory, I feel.—John Atkinson