Tube Power Amp Reviews

Like most tube amplifiers, the 300SE LX20 had some detectable microphony in that it I noted a very slight bloom in the upper midrange. The use of good vibration damping and resisting supporting sound tables is most rewarding. However, in one listening session I was checking some aspect of speaker performance and operating the Pause control on the CD player in the middle of a noisy, mid-dominant track. I was surprised to hear a low-level post-pause echo that lasted a second or two. If this was microphony, then not using loudspeakers would deal with the issue. But when I wired up some headphones via an adapter box, I was equally surprised to find that the "echo" remained—in other words, it was also set off by the electron currents in the LX20 tube. The degree of electron-induced ringing was similar for four tubes tried, and the frequency was also remarkably similar, pointing to a consistently fabricated structure within the tube.

I had also noted a tiny glitch in the steady-state frequency response, and tried some MLSSA measurements. shows the result. Ringing is evident at 600Hz in fig.1 and has not decayed significantly in the first 10ms of decay. Some other secondary modes are also present above 2kHz before the clear, fast decay spectrum can be seen.

Fig.1 Cary CAD-300SE LX20, cumulative spectral decay (10dB/vertical div., 10ms time window)

I therefore carefully looked at the steady-state frequency response (fig.2)—the vertical scale in this graph is just 1dB overall. The lower trace shows the standard output tube, the upper the luxury T100 titanium tube. With the former, the resonance blip was much smaller but is split in two. This result correlates with the still more neutral midband I heard with the T100 tube. However, there did not appear to be significant or repeatable differences in output power, load matching, or distortion between the two output-tube options.

Fig.2 Cary CAD-300SE LX20, frequency response at 1W into 8 ohms (4 ohm tap) (from top to bottom): LX20 and T100 output tubes (0.1dB/vertical div.).

On a resistive load, the amplifier's frequency response measured 26Hz-16kHz, with +0.1/–0.5dB limits. For the -3dB half-power points, the bandwidth reached from 8Hz to 36kHz—pretty good for a design without negative feedback! Channel balance was within 0.2dB with matched output tubes. (The gain varied slightly with bias level.)

Input impedance was an easy 150k ohms in parallel with around 100 picofarads, an easy load for any source. The amplifier needed a fair input voltage despite its modest output power: 1.8V for clipping, 450mV for 1W IHF.

Given the output power, the signal/noise ratios were more than satisfactory: for a 1W reference, 60dB unweighted, 80.3dB A-weighted. Against full level I got 71dB unweighted, 98dB excluding hum, and 93.5dB A-weighted.

Rated at 20W into 8 ohms (13dBW), the 300SE LX20 was a little less powerful than the integrated SEI version, which I looked at a few months ago for Hi-Fi News & Record Review in the UK. That model scraped 23Wpc midband for 3% distortion. Running about 1.3dB lower than the specification, probably due to a slightly lower voltage on power line, the 300SE LX20 gave a load-matched average of 18.5W midband (4 ohm tap to 4 ohm load, 8 ohm tap to 8 ohms), though 4 ohm loading on the 8 ohm tap severely lowered the output to 5W at 1kHz.

There wasn't much current available—around 1.9A peak—but since the current clipping was just as aurally benign as the voltage clipping, this was no hardship.

The 300SE LX20 is unconditionally stable, laughing in the face of reactive loads—an 8 ohm plus 2µF load barely touched the squarewave response, as seen in fig.3. There was negligible overshoot and no trace of ringing, while the flat top indicates a fundamentally flat pass-band. A complex load will not by itself significantly change the sound.

Fig.3 Cary CAD-300SE LX20, small-signal 1kHz squarewave into 8 ohms in parallel with 2µF.

The 300SE LX20's damping factor was predictably poor: about 3 for the 8 ohm tap and 2.2 for the 4 ohm tap, referenced to an 8 ohm load. Such a source impedance will alter a speaker's effective frequency response (fig.4), while this and the consequent changes in bass balance and damping might also indicate a revision to your preferred speaker placement to help compensate. In the case of the WITT II (fig.5), the differences amounted to a small balance shift in the bass and a slight upper-bass emphasis. Moving the speaker 3–4" forward proved sufficient.

Fig.4 Cary CAD-300SE LX20, frequency response at (from top to bottom at 3kHz): 1W into 8 ohms (4 ohm tap), and 2.828V into Wilson WITT II loudspeaker (2dB/vertical div.).

Fig.5 Wilson WITT II loudspeaker, electrical impedance (solid) and phase (dashed) (2 ohms/vertical div.).

In the listening tests I tended to favor the 4 ohm tap, even with the 8 ohm Wilson WITT 2. There seemed to be little loss in maximum output: with the lighter 8 ohm loading on the 4 ohm tap, the benefit of about 30% improvement in damping factor could be heard as a more neutral output from the speaker, and superior bass control. The measured loss in level was barely 1dB, bearing out the listening-test result.

As expected, full power was not available at 20Hz. Fortunately for the 300SE LX20, significant audio power below 30Hz is rare in naturally balanced music recordings. For a psychoacoustically acceptable distortion level at 20Hz, this Cary could produce about 10W but near full power was available by 40Hz, in context a satisfactory result. There was no problem with power delivery at high frequencies, and nominal full power was possible at 20kHz. Thanks to the choke-fed power supplies, there was negligible power-supply intermodulation.

Experiments with bias showed that it was possible to fine-tune the tube characteristics for lowest distortion—for example, at 1W—and, for a given load, optimize the linearity. Higher-than-recommended bias levels offered no real advantage: Peak power did not increase, while low-level distortion actually increased as the ideal load-matching point was overshot; in addition, tube life will be reduced.

I suspect that some keen-eared listeners, aided by sensible cross-checking with a bias meter, will be able to fine-tune these settings to their own satisfactions. For my UK supply voltage, I found that bias values between 105 and 110mA were fine; even at 95mA, the quality loss wasn't very great. Cary recommends 110mA.

The high-frequency intermodulation performance was barely satisfactory measuring -40.2dB at 10W output, or just over 1% of difference tone. At normal listening levels of 1W, a comfortable figure of -60dB (0.1%) at 1kHz was obtained (fig.6). I also looked at the linearity difference between the T100 and LX20 output tubes at 1W, and found little to choose between them. The low-order harmonic distortion spectrum seen in fig.7 was typical for both tubes, with minor balancing of bias currents and about 0.2% (-56dB) of distortion. Note the satisfactorily low supply-ripple components below the 1kHz fundamental.

Fig.6 Cary CAD-300SE LX20, HF intermodulation spectrum, DC-22kHz, 19+20kHz at 1W into 8 ohms (linear frequency scale).

Fig.7 Cary CAD-300SE LX20, spectrum of 200Hz sinewave, DC-2kHz, at 1W into 8 ohms (linear frequency scale).

But have no illusions—for me at least, this low-order harmonic distortion was mildly audible, to and beyond full power. In increasing measure it sounded like a purring in the bass, and a soft fuzz in the mid and treble. Uncannily, while peak overdrive levels sounded fairly innocuous, mid- and low-level detail remained almost intact when the linear power dam was breached. With the Cary operated well below peak, distortion levels were academic; the sound was so naturally sweet, vocal lines so articulate, that any technical criticisms became pointless.—Martin Colloms