Tube Power Amp Reviews

As I mentioned in my listening report, the Baron offers 32 different combinations of its settings, meaning that for measurement purposes it is 32 different amplifiers. To supply a full set of measurements for every combination wouldn't leave room in this issue for any other reviews! Accordingly, I only thoroughly measured the Baron's all-triode and all-pentode modes, from both the 8 and 4 ohm transformer taps, with the negative feedback set to none or all (labeled Stage III). Other than spot checks on other settings, I'll leave you to infer what happens for the mixed pentode/triode settings. Before I performed any measurements, I took care to optimize the bias and balance for each channel.

First, don't let the XLR input socket mislead you. Pin 2 is wired as "hot," as is usual, but pin 3 was shunted to ground on my sample. This is an unbalanced input. If you use a preamp with balanced output, its "cold" driver will see a short circuit to ground, to its possible distress. (Aware of this loading problem, Mesa now ties this pin to ground with a 60k resistor.) I measured the input impedance as 42k ohms at 1kHz. The amp is wired to be non-inverting.

Given the Baron's dual-mono construction, it was a little disappointing to find the amplifier's channel separation falling to 70dB above 10kHz (not shown). I assume that this is due to the proximity of the two input circuits and the negative feedback switching resulting in some capacitive coupling. Noise levels were low, at typically 76dB (unweighted ref. 1W/8 ohms), this improving to around 86dB when A-weighted, meaning that what noise there was was dominated by very high- and/or very low-frequency components (principally hum).

Voltage gain varied significantly with setting and load, ranging from a very high 39.4dB (all-pentode, zero feedback, 4 ohm tap loaded with 8 ohms) to a more normal 29.2dB (all-triode, Stage III feedback, 4 ohm tap loaded with 8 ohms).

Looking at the output impedance revealed the Baron's musical-instrument amplifier heritage: Whereas a modern hi-fi amplifier is intended to act as a voltage source—its source impedance is so low that its output voltage doesn't drop significantly as the load impedance falls—the Baron's source impedance is intended to be equal to the load. (In this respect, the Baron's output transformers perform very similarly to that of my much-beloved 1970 Fender Bassman 50.) A quick workout with Ohm's Law should convince you that making source and load impedances equal results in maximum power transfer to the load, as paradoxical as that might sound. However, it does mean that half the total power delivered by the amplifier is dissipated in the output transformer—which, as a result, runs very hot. Second, with a real-world loudspeaker, the necessarily high source impedance results in a frequency response that varies significantly according to the speaker's modulus of impedance.

Fig.1, for example, shows the Baron's amplitude response in triode mode (8 ohm tap, zero feedback) into 8 ohms, 4 ohms, and into our dummy loudspeaker load, designed for us by NHT's Ken Kantor. (Described in the August 1995 issue of Stereophile, p.168, this is a collection of resistors, capacitors, and inductors intended to model the impedance signature of a typical two-way, Zobel-compensated, sealed-box loudspeaker.) In triode mode, the HF response rolls off early, being –5dB at 20kHz—there's the mellow top octave I noted in my auditioning. But more important, note the 3dB drop in output level into the lower impedance and the severe modification of the amplifier's response by the dummy load. To a large extent, this amplifier will sound different with every loudspeaker with which it is used. Its 1kHz source impedance varied from 18 ohms (pentode, 8 ohm tap) through 7.75 ohms (triode, 8 ohm tap), to 4 ohms (triode, 4 ohm tap). While the 20kHz impedance was generally a little lower than at 1kHz, at the low-frequency extreme (20Hz) the impedance in the pentode mode was even higher, reaching a maximum of 23 ohms.

Fig.1 Mesa Baron, triode mode, 8 ohm tap, 0dB feedback, frequency response at (from top to bottom at 1kHz): 2.83V into simulated loudspeaker load; 1W into 8 ohms; 2W into 4 ohms (right channel dashed, 2dB/vertical div.).

Loop negative feedback lowers an amplifier's source impedance; how does changing the Baron's feedback level reduce this interaction? The answer can be seen in fig.2—not much. A swing of ±3.4dB with no feedback is reduced to ±2.8dB with Stage III feedback. And in pentode mode with maximum negative feedback (fig.3—note the different vertical scale on this graph), the swing is even greater, reaching a maximum boost of almost 5dB in the low treble. Here's part of the reason the Baron tended to sound bright with my preferred speakers. Note, however, that the amplifier offers a more extended HF response in pentode mode, only falling by 1dB at 20kHz. This is confirmed by the 10kHz squarewave response (fig.4), which has a reasonable risetime. Note the very slight and well-damped ringing in fig.4, this correlating with the small response peak just above 100kHz in fig.3.

Fig.2 Mesa Baron, triode mode, 8 ohm tap, 0dB feedback, frequency response at 2.83V into simulated loudspeaker load with (from top to bottom): 0dB, Stage I, Stage II, and Stage III feedback (2dB/vertical div.).

Fig.3 Mesa Baron, pentode mode, 8 ohm tap, Stage III feedback, frequency response at (from top to bottom at 1kHz): 2.83V into simulated loudspeaker load; 1W into 8 ohms; 2W into 4 ohms (right channel dashed, 5dB/vertical div.).

Fig.4 Mesa Baron, pentode mode, 8 ohm tap, Stage III feedback, small-signal 10kHz squarewave into 8 ohms.

When it comes to describing the Baron's distortion performance, there are two opposed ways of looking at it. First, the Baron is overall a relatively non-linear device: Its distortion is typically 10 times higher than you'd find in a modern solid-state design, or even a modern tube design. Second, when you consider its absence of overall loop feedback—even in its highest feedback setting, it only offers a few dB of gain and distortion reduction—its transfer function is actually remarkably linear. The Baron also resembles the human hearing mechanism in that its distortion increases in proportion to level.

This can be seen from fig.5, which shows the THD+Noise percentage on a 1kHz tone plotted against output power into 8 and 4 ohms, with the amplifier in triode mode with 0dB feedback. Below 50mW (0.05W), the curve starts to rise, meaning that the figure is actually dominated by noise below this power level. The minimum value of 0.1% at the 50mW level is actually the amplifier's basic level of distortion at these minimal power levels, which is acceptably low. But as the output power rises, so does the THD, but in what appears to be straightforward manner. The clipping point in triode mode (defined as 1% THD+N) into both 4 and 8 ohms is around 20W. But note that the "knee" in the amplifier's output plot lies at the 2% level. Relaxing our definition of clipping to 3% THD+N therefore results in a maximum power output of 55W into 8 ohms (17.4dBW), agreeing with the specification.

Fig.5 Mesa Baron, triode mode, 8 ohm tap, 0dB feedback, distortion (%) vs output power into (from bottom to top): 8 ohms and 4 ohms.

I suspect that it was this increase in distortion level with increasing level that caused me to find the Baron's dynamics limited in triode mode. Remember that I found about 20W to be the practical limit. Remember also that the emphasis of the low-treble region with the two-way speakers I use (see fig.1) will help unmask distortion products. The effect of changing the feedback level on the output power curve is shown in fig.6. The clipping point doesn't change, but each increase in negative feedback results in slightly lower distortion at lower power levels.

Fig.6 Mesa Baron, triode mode, 8 ohm tap, 0dB feedback, distortion (%) vs output power into 8 ohms with (from bottom to top): Stage III, Stage II, Stage I, and 0dB feedback.

In pentode mode, the amplifier behaves somewhat differently. Fig.7 shows the Baron's THD+N plotted against output power driving 8, 4, and 2 ohm loads from its 4 ohm tap. (The feedback level was Stage III.) First, note that, even allowing for the effect of the feedback, the overall distortion level is lower than in triode mode. Second, the output power is considerably higher, the 4 ohm power from the 4 ohm tap equaling 135W at the 3% THD point (18.3dBW). Into 8 ohms, the pentode Baron's 8 ohm tap raised 140W (21.5dBW), the 4 ohm tap 100W (20dBW).

Fig.7 Mesa Baron, pentode mode, 4 ohm tap, Stage III feedback, distortion (%) vs output power into (from bottom to top at 10W): 8 ohms, 4 ohms, and 2 ohms.

However, this was for a 1kHz signal. Even with the maximum feedback, the pentode mode's distortion rises for both higher and lower frequencies (fig.8). While this was also true for the triode mode (fig.9), the differences were less extreme. And what about the influence of feedback? Fig.10 shows the difference between the THD+N vs frequency plot in triode mode for the two extreme settings, driving 8 ohms from the 8 ohm tap. While the Stage III feedback basically halves the level of distortion, note that the right channel is already much better than the left above 1kHz. Even though I carefully set up the output-tube bias and balance, there will always be some variability in performance.

Fig.8 Mesa Baron, pentode mode, 8 ohm tap, Stage III feedback, THD+noise vs frequency at (from top to bottom at 5kHz): 4W into 2 ohms, 2W into 4 ohms, and 1W into 8 ohms (right channel dashed).

Fig.9 Mesa Baron, triode mode, 4 ohm tap, 0dB feedback, THD+noise vs frequency at (from top to bottom at 10kHz): 4W into 2 ohms, 2W into 4 ohms, and 1W into 8 ohms (right channel dashed).

Fig.10 Mesa Baron, triode mode, 8 ohm tap, THD+noise vs frequency at 1W into 8 ohms with (from top to bottom): 0dB and Stage III feedback (right channel dashed).

As Scott Frankland discusses elsewhere in this issue, more important than the level of distortion is its nature. Remember that I much preferred triode mode? Fig.11 indicates that the main harmonic present is the benign second, with a touch of fourth. But in pentode mode under the same conditions (fig.12), some jagged-looking spikes appear in the distortion waveform, suggesting the presence of higher-order harmonics. Even though the level of THD is still quite low—around 0.14%—there is plenty of evidence in the pyschoacoustic literature that this kind of distortion spectrum is subjectively more dangerous. Add this to the more severe modification of the amplifier's frequency response in pentode mode (fig.3), and you probably have the explanation for why I didn't much like pentode operation, and why Chip's wife commented that triode operation was "gentler."—John Atkinson

Fig.11 Mesa Baron, triode mode, 4 ohm tap, 0dB feedback, 1kHz waveform at 2W into 4 ohms (top); distortion and noise waveform with fundamental notched out (bottom, not to scale).

Fig.12 Mesa Baron, pentode mode, 4 ohm tap, 0dB feedback, 1kHz waveform at 2W into 4 ohms (top); distortion and noise waveform with fundamental notched out (bottom, not to scale).