Tube Power Amp Reviews

In the world of tubed amplifiers, designers don't need to be told about the weaknesses of our understanding of psychoacoustics and the relationship between steady-state measurement standards and sound quality. This is reflected in Jadis's specifications, which are sparse and, frankly, uninspiring. The output is quoted at 100Wpc into a matched load for 0.6% distortion, specified at 1kHz only. The high input sensitivity is ideal for passive control use for sources down to 2V output, while the input impedance is loosely quoted as >100k ohms. The frequency response is given at 1W with the –3dB points placed at 15Hz and 100kHz, a rather restricted bandwidth when compared, for example, with an equivalent Audio Research design. Distortion at 30W is quoted at better than 0.6%, 100Hz–10kHz (usually a manufacturer will quote 20Hz–20kHz). No details are given for output impedance, intermodulation, dynamic range, or S/N ratio.

Jadis makes no attempt to conceal the relatively poor measured performance of its products, arguing, I think correctly, that the figures have remarkably little to do with sonic performance. Nevertheless, it is possible to make a fine tube amplifier measure well. Audio Research is the leading exponent of this art, many of their tubed amplifiers doing almost as well on the bench as solid-state models—and sounding good as well. However, it is the particular failings of the Jadis which in one sense allows certain perceptual factors to be weighed. Because it sounds so very good, and because it fails in certain areas, we can make some tentative observations about the subjective importance of these technical parameters. These will come to light as the lab analysis proceeds.

The JA 200 had measured poorly in the past and there was no reason to expect much more from the Defy-7. I therefore took care to get all my listening done before measurement to avoid the engineering side of my personality getting the better of my judgment.

Beginning at the beginning, the '7 is rated at 100Wpc into a matched load. Set to the optional 4–8 ohm tap, and on a UK 243V, 50Hz supply (a little higher than the amplifier might normally expect), it could not quite raise 100W with one channel driven into 8 ohms, 1kHz, measuring 96W (19.8dBW, or 0.2dB short). What's 0.2dB between friends? Well, the 19.8dBW value is a visual clip result at high distortion; if we adopt a more stringent 1% distortion limit for maximum output, the Defy-7 reached only 12dBW, 16W. The picture was worse still at the frequency extremes—for example, the maximum power at 20Hz for 1% THD is 9.1dBW, less than 10W, though if some sort of visual clipping limit is adopted, 17dBW, 50W was possible. The situation was no better at 20kHz, where 10dBW, 10W was the continuous power limit for 1% distortion.

These results were repeated into a 4 ohm load (remember that this is ostensibly the 4 (to 8) ohm tapping on the output transformer). With both channels driven, the result was a loss in level of between 3 and 6dB, depending on the distortion criteria.

A clearer picture of these results can be gained by looking at the plots of distortion vs frequency for a range of power levels into an 8 ohm load (fig.1). At 100W (top trace), the distortion never gets down to the 1%, –40dB line at all. Things only begin to cheer up at 10W and below; by 1W, the distortion has fallen below –50dB or 0.33%, 100Hz–10kHz. A similar if slightly worse performance can be seen in fig.2 for output powers into 4 ohm loads.

From these results, we can see that the Defy-7 is characterized by fairly high distortion at high powers (ie, the top 10dB of its dynamic range), and is poorer still at the edges of the audible bandwidth. One could expect that its tonal quality and clarity might change more than usual when the amplifier is driven to full power, though I have to report that this was not the case even with a relatively unkind 4 ohm speaker loading.

Perhaps the flaw in this reasoning concerns the discontinuous nature of music compared with continuous tone test signals. Here there was a glimmer of hope. On toneburst simulation of music conditions (a 10ms burst), while the amplifier produced no more power into 8 ohms, its 4 ohm delivery (for visible clipping) rose by a substantial 2.5dB, to 141W. Moreover, its delivery into 2 ohms on burst—126.5W, 15dBW—was almost as good as its continuous delivery into 4 ohms. Clearly the Jadis's output stage had some offset effect which limited the maximum output under adverse conditions, but only on continuous tones—not on music. This behavior was also present to a lesser degree at the frequency extremes.

Moreover, the peak current—measured using a load of 1 ohm in parallel with 2.2µF and a 2ms toneburst signal—was quite respectable at +8.5A and –8A, sufficient to dump 115W or so into a 4 ohm load. The Defy-7 is therefore quite capable in a dynamic situation.

At 1W, the total harmonic distortion was typically –50dB, 0.3%. I subjected a 300Hz tone at this level (into 8 ohms) to a high-resolution spectrum analysis (fig.3). Here perhaps is a significant indicator of the Defy-7's musical character. Other than the fundamental at 300Hz, only two harmonics are noticeable above the innocuous noise floor at the –100dB level—the second at 600Hz and third at 900Hz—and significantly, the second harmonic has the correct level relative to the third. Such a distortion characteristic is highly regarded in amplifier circles.

I looked at intermodulation in two different ways. First, the two-tone high-frequency intermodulation performance (using equal-level 19 and 20kHz tones) was truly awful with the amplifier driven to its rated output level (fig.4). Of course, it may be justifiably argued that such a level of a band-edge signal is most unlikely to appear in music. Certainly the Defy-7 was fully slew-limited at this level, generating distortion components of all flavors. However, when I backed off the output level a few dB, it sorted itself out. By 0dBW (1W into 8 ohms), a respectable –61dB was achieved for high-frequency intermodulation (fig.5).

The second test concerned stressing the amplifier by asking it to drive a low-frequency (37Hz) tone into an adverse 4 ohm load at 2/3 of its rated power. The idea is to assess the power-supply breakthrough or modulation. Fig.6 shows the result of this test at a reduced level of 12dBW (32W, 4 ohms) to avoid the effects of waveform clipping. Remembering that UK mains harmonics are at 50Hz intervals, some harmonic products of the 37Hz fundamental can be seen, but there is no significant supply-related breakthrough; just a trace of 50Hz at –70dB.

One aspect which will affect sound quality slightly is the finite output impedance. With solid-state amplifiers it is usually sufficiently low to be considered negligible: 0.1 ohm or under. With a number of tube amplifiers it can be rather higher. For example, the "1000"—a low-feedback triode amplifier made by Audio Innovations in the UK—has an output resistance of several ohms. This is by no means an isolated example. The Defy-7's output or source resistance was quite constant over the audio band, ranging from 0.45 ohms at 20kHz, to 0.4 ohms, midband, and 0.45 ohms at 20kHz. Given 0.1 ohm of speaker cable resistance, one can anticipate that, with a given multi-way dynamic or moving-coil loudspeaker, a slight increase in bass due to lowered electrical damping will be allied to minor variations in broadband frequency response due to the varying load of the loudspeaker's impedance vs frequency curve. Thus it cannot be expected that a tube amplifier like the Defy-7 would replicate the tonal balance of a solid-state amp, even if both were intrinsically neutral on a flat-impedance resistive load.

At low levels into 8 ohms, the Jadis had an extended low-frequency response (fig.7), flat down to 3Hz and peaking +3dB at 0.5Hz—a frequency low enough to be inconsequential. At high frequencies, the –0.5dB point lay at 18.65kHz, with the –3dB point no problem at 55kHz. The high-frequency rolloff did show some sudden changes in slope and phase in the ultrasonic range, notably at 60kHz. It might be worth Jadis's while to direct some design attention to this aspect, as there is evidence to suggest that models with well-ordered ultrasonic rolloffs can sound even better than those without.

This ultrasonic characteristic is reflected in the very minor out-of-band ringing which can be seen in the amplifier's 1kHz squarewave response into a resistive load (fig.8). The amplifier was well-behaved with the usual 2;uF capacitive load addition (fig.9), with well-controlled ringing and overshoot. This performance confirms its good performance with electrostatic speakers.

Channel separation was more than satisfactory but could be improved. The characteristic was odd, with good results for one channel with respect to the other (at 105dB at 1kHz), but not the other way around (65dB). This implied some mild coupling or crosstalk from the output circuit of one channel to the input of the other. This is often a matter of the layout of internal wires—to the speaker terminals, for example.

S/N ratios were fine, channel balance excellent at 0.04dB difference at 1kHz. The input was easy to drive, providing very light source loading, and the sensitivity was a little better than 1V for full power; ideal for use with a passive line controller.

No DC offset was present at the output. (Yes, it is possible to have DC offset with tube designs.) The AC mains transformer ran pretty quietly; I don't expect audible hum to be a problem.—Martin Colloms