Tube Power Amp Reviews

I measured one of the the Margules U280-SC Blacks (S/N B-50003950517) using my top-of-the-line Audio Precision SYS2722 system (see the January 2008 "As We See It"). With two modes of output stage—Triode and Ultralinear—three choices of output-transformer tap, and the option of having its two channels strapped for mono operation, the Margules U280-SC is actually 14 different amplifiers in one enclosure. Although Jason Victor Serinus exclusively used the amplifiers as a pair of monoblocks, I performed a complete set of measurements for all 14 mono and stereo modes of operation, to comprehensively characterize its behavior. (The gain control for each channel was set to its maximum for all testing.)

After the amplifier was turned on, it took a few minutes for the autobiasing to function, after which the blue LED for each output tube illuminated. I let the amplifier drive 1Wpc into 8 ohms for an hour before performing any measurements, but one thing should be noted: this amplifier gets unacceptably hot with continuous tones. After that hour, the top panel's temperature adjacent to the Triode/Ultralinear switches was 220°F/106.4°C—you could literally boil water on it. To avoid burning my fingers during the testing, I operated these switches with a pencil. See the "Manufacturer's Comment" on why this shouldn't be an issue with music.

The Margules amplifier preserved absolute polarity in all 14 operating modes, and its input impedance was 47k ohms at 20Hz and 20kHz, dropping inconsequentially to 40k ohms at 1kHz. The gain into 8 ohms depended on the output tap and operating mode, and ranged from 26.1dB, 2 ohm tap and stereo Triode mode, to 36.6dB, 8 ohm tap and mono Ultralinear mode. The output impedance also depended on tap and mode. It was highest from the 8 ohm tap in stereo Ultralinear mode, at 15 ohms at 20Hz, 10.5 ohms at 1kHz, and 9.4 ohms at 20kHz. These impedances halved each time the nominal value of the output-transformer tap was halved, and were significantly lower in Triode mode. For example, from the 8 ohm tap in stereo Triode mode, the impedance was 7 ohms at 20Hz, 5.5 ohms at 1kHz, and 5.4 ohms at 20kHz. As expected, the lowest impedance in stereo mode was obtained from the 2 ohm tap: in Triode mode, I measured 1.4 ohms at 20Hz, and 1.15 ohms at 1kHz and 20kHz. The output impedance of the U280-SC in strapped-mono mode was significantly lower than in stereo operation. In Ultralinear mono mode, the impedance from the 8 ohm tap was 5.5 ohms at 20Hz, dropping to 3.8 ohms at higher frequencies. In mono Triode mode, the corresponding 8-ohm-tap impedance was 2.9 ohms at 20Hz, 2.25 ohms at 1kHz, and 2.6 ohms at 20kHz.

With its generally high output impedance, the Margules amplifier's frequency response will be significantly modulated by how the partnering loudspeaker's impedance changes with frequency. The gray trace in fig.1, for example, shows the U280-SC's response from its 8 ohm tap in stereo Ultralinear mode with our standard simulated loudspeaker. The response varies by a very audible ±3.1dB! Even with the lower impedance in mono mode (fig.2, gray trace), the variations are ±2.3dB, which leaves me puzzled as to why JVS preferred the sound of his Wilson Alexias from the amplifier's 8 ohm tap. (The http://tinyurl.com/y85n3use">Wilson's impedance varies from 2 ohms in the upper bass to more than 9 ohms in the top octave. By contrast, the impedance of the Harbeth M40.1, the speaker that JVS found to be more compatible with the Margules amplifier, is significantly higher, varying between 7.5 ohms and 23 ohms.) In mono Triode mode (fig.3) the response varied by ±1.4dB, and in figs. 1–3 you can see that the increasing output impedance at very low and high frequencies starts to roll off the response into resistive loads. In stereo (fig.1), you can make out the presence of an ultrasonic transformer resonance, though this is well down in level. This resonance correlates with a small amount of overshoot and a single damped cycle of ringing with a 10kHz squarewave (fig.4).

Channel separation in stereo mode was good rather than great, at 75dB at 1kHz, increasing to 84dB at 200Hz but decreasing to 52dB at the top of the audioband. Measured at the 8 ohm tap with the input shorted to ground, the unweighted, wideband signal/noise ratio ref. 2.83V was 63.5dB in Ultralinear mode and 66dB in Triode mode. (Both ratios are the average of the two channels.) There was a slight improvement when the measurement bandwidth was restricted to the audioband, and with an A-weighting filter the ratios further improved, respectively to 75 and 78.5dB. Spectral analysis (fig.5) revealed that the Margules's noise floor was dominated by power-supply–related spuriae at 60Hz and its even- and odd-order harmonics.

As a stereo amplifier, the U280SC is specified to deliver 50Wpc (17dBW) in Ultralinear mode, 25Wpc (14dBW) in Triode mode. In mono, the respective specified power outputs increase to 100W (20dBW) and 55W (17.4dBW). I examined how the percentage of THD+noise changed with power in all 14 operating conditions. Fig.6 shows the behavior in stereo Triode mode with the 8 ohm tap matched to an 8 ohm load. The THD+N is <0.2% at low powers, but rises to reach 1%, our usual definition of clipping, at 34W, which is 1.3dB higher than spec. However, in Ultralinear mode with again the 8 ohm tap matched to the load, just 17W was available at 1% THD+N (fig.7), and the amplifier delivered its specified 50Wpc only at 4% THD+N.

The picture was a little better in Ultralinear mode from the 4 ohm tap (fig.8), where the Margules delivered 46Wpc into 4 ohms (13.6dBW). Again, the specified 50Wpc was reached at 2% THD+N. Operated in mono in Triode mode (fig.9), the U280-SC clipped at 37.5W (15.75dBW) from the 8 ohm tap into 8 ohms at 1% THD+N, with a lower level of THD+N at low powers than in stereo mode. It reached the specified 55W at 21% THD+N. In Ultralinear mono mode, the amplifier clipped at 60W into 8 ohms from the 8 ohm tap and at 44W into 4 ohms from the 8 ohm tap. However, fig.10 reveals that the U280-SC does deliver the specified 100W into 4 ohms at 8.5% THD+N.

In stereo mode, the Margules U280-SC offered suitably low distortion only when the nominal output transformer tap was equal to or lower than the load (fig.11). However, even at the low level used to create this graph there is a dramatic rise in distortion in the bass, presumably due to the output transformer core starting to saturate. The distortion decreased dramatically when the two channels were strapped for mono operation, particularly in Triode mode (fig.12). In JVS's preferred Ultralinear mode, the THD+N was around twice as high as in Triode mode, while the curves followed the same shapes as in fig.12 (fig.13). However, with the Wilson Alexia's impedance of 3 ohms or lower between 60Hz and 200Hz, the amplifier will be clipping with low frequencies higher in level than a few watts.

Midband distortion in Triode mode was predominantly low-order in nature (fig.14), though in Ultralinear mode higher-order harmonics crept in (fig.15). At low frequencies the distortion in both Ultralinear and Triode modes was high, with the spectrum having the second harmonic the highest in level and the higher-order harmonics regularly decreasing in level (fig.16). Intermodulation distortion was relatively low in level (fig.17), though the presence of power-supply–related spuriae in this graph does look a little alarming.

I was not impressed by the Margules U280-SC's measured performance when it was used as a conventional stereo amplifier. Putting to one side the fact that its chassis got too hot with continuous tones, in my opinion its distortion and noise are too high, unless its nominal output-transformer tap is lower than the load impedance. Operated as a mono amplifier, as JVS auditioned it, the Margules offers lower distortion coupled with greater power, but I'm puzzled why he preferred it in the higher-distortion Ultralinear mode and used its 8 ohm output tap—I would have thought the lower-impedance taps in Triode mode would prove a better match with his Wilson Alexia speakers. I suspect that matching such an amplifier with a compatible loudspeaker will be more complex a task than is usually the case.—John Atkinson