Mystère ia21 integrated amplifier Measurements
Sidebar 3: Measurements
To perform the measurements on the Mystère ia21, I mostly used Stereophile's loan sample of the top-of-the-line Audio Precision SYS2722 system (see the January 2008 "As We See It" and www.ap.com); for some tests, I also used my vintage Audio Precision System One Dual Domain.
As Erick Lichte mentioned, he sent me the first sample of the ia21 (S/N 210213) after his preliminary auditioning, as there seemed to be something wrong. I carried out a full set of tests with the amplifier, and yes, one channel was offering around ten times the distortion of the other. Swapping the KT88 output tubes from one channel to the other didn't change each channel's behavior; however, swapping the 6SN7 input tubes did move the higher distortion to the other channel, suggesting that one of these small-signal tubes had gone bad (a very rare occurrence, in my opinion). I asked distributor Kevin Deal for another set of 6SN7 tubes. Unfortunately, these didn't fix the problem, so we decided to restart the review with another sample of the ia21 (S/N 210221).
I performed a complete set of tests on the new sample from both the 4 and 8 ohm output transformer taps, with both the KT88 output tubes and the EL34 output tubes. In each case, the bias switch on the amplifier's left side panel was correctly set for the tube type. The maximum voltage gain into 8 ohms was very similar with either output tube and on the low side for an integrated amplifier, at 32.8dB from the 8 ohm tap and, surprisingly, slightly higher from the 4 ohm tap: 33.0dB. Both taps preserved absolute polarity (ie, were non-inverting), and the input impedance was very high, at 100k ohms or greater at all audio frequencies.
It was when I measured the ia21's output impedance that I ran into one of the amplifier's idiosyncrasies. The conventional wisdom in hi-fi amplifier design for the past 50 years has been that the amplifier should behave close to a voltage source, with a sufficiently low output impedance that a loudspeaker's changing demands for current with frequency don't affect the amplifier's response. By contrast, the Mystère offered very high output impedances: 14 ohms from the 8 ohm tap and 7.5 ohms from the 4 ohm tap. (Both figures changed only slightly with frequency, and were not significantly affected by the type of output tube used.)
In a different view of amplifier design, one that stems from telephony, you match the amplifier's output impedance to the load, which results in the optimal transfer of power to the load. Tubed instrument amplifiers tend to be designed according to this paradigm, but the Mystère's output impedance is almost twice as high as needed when viewed in this light. A third way of designing an amplifier is to make it have a very high output impedance, so that it acts as a current source; ie, it maintains the same output current into all load impedances. As high as it is, the Mystère's output impedance is too low for that paradigm to be in operation, however, so I am left mystified by the ia21's design philosophy.
The downside of this high impedance is that the response variations due to the Ohm's Law interaction between the impedance and the manner in which the partnering loudspeaker's impedance changes with frequency will be extreme. The gray trace in fig.1, for example, shows the amplifier's response from its 8 ohm tap into our standard simulated speaker. The response changes by ±4.1dB, which will be very audible; in a sense, the Mystère ia21 will sound different with every speaker with which it is used. Even from the 4 ohm tap, with its lower output impedance, the variations are only slightly reduced in magnitude (fig.2, gray trace). However, by comparing fig.2 with fig.1, you can see that the lower-impedance transformer tap has a flatter response in the octave and a half above the audioband and that, in fact, the Mystère has a superbly flat, wide bandwidth from this tap. These two graphs were taken with EL34 output tubes; the behavior with KT88s was almost identical, other than the fact that neither the ultrasonic rise in response from the 8 ohm tap nor the small peak around 150kHz were quite as pronounced.
Fig.1 Mystère ia21, EL34s, 8 ohm output tap, frequency response at 2.83V into: simulated loudspeaker load (gray), 8 ohms (left channel blue, right red), 4 ohms (left cyan, right magenta), 2 ohms (green). (2dB/vertical div.)
Fig.2 Mystère ia21, EL34s, 4 ohm output tap, frequency response at 2.83V into: simulated loudspeaker load (gray), 8 ohms (left channel blue, right red), 4 ohms (left cyan, right magenta), 2 ohms (green). (2dB/vertical div.)
These response graphs were taken with the ia21's volume control set to its maximum. Commendably, reducing the setting to 12:00 didn't change the ultrasonic bandwidth. The rise in response above the audioband from the 8 ohm tap results in a small degree of overshoot and a single cycle of ringing with a 10kHz squarewave (fig.3), which was only slightly reduced in amplitude from the 4 ohm tap. The small-signal 1kHz squarewave was superbly square (fig.4), the flat tops of the waveform confirming the excellent LF extension see in the frequency-response graphs. Channel separation was good rather than great, at about 80dB in both directions at 1kHz. This increased to 90dB at low frequencies, but decreased to 4854dB at 20kHz, depending on direction. The ia21 was moderately quiet, its wideband, unweighted signal/noise ratio (ref.1W into 8 ohms with the input shorted but the volume control set to its maximum) measuring 68.7dB in the left channel, 71.6dB in the right. These figures improved to 89.4 and 80.2dB, respectively, when A-weighted.
Fig.3 Mystère ia21, EL34s, 8 ohm output tap, small-signal 10kHz squarewave into 8 ohms.
Fig.4 Mystère ia21, KT88s, 4 ohm output tap, small-signal 1kHz squarewave into 8 ohms.
Figs.58 show how the THD+noise percentage in the Mystère ia21's output varied with power into 16, 8, and 4 ohms from the 8 and 4 ohm taps and with KT88 and EL34 output tubes. General points to be gleaned from these graphs: 1) the maximum clipping power is obtained when the load is matched to the nominal output transformer tap; 2) the distortion is low below a few hundred milliwatts into higher impedances; 3) the rise in distortion with power is worse with EL34s than with KT88s; and 4) the KT88s clip in a more abrupt fashion than the EL34s. With KT88s, the amplifier easily meets its specified power of 50Wpc into 8 ohms (17dBW) at 1% THD+N, but offers less power at that level of distortion with EL34s. However, the less sharp onset of clipping with EL34s means that, at a relaxed 3% THD definition of clipping, the ia21 delivers 67.5Wpc from the 8 ohm tap into 8 ohms (18.3dBW), or from the 4 ohm tap into 4 ohms (15.3dBW).
Fig.5 Mystère ia21, KT88s, 8 ohm output tap, distortion (%) vs 1kHz continuous output power into (from bottom to top): 16, 8, 4 ohms.
Fig.6 Mystère ia21, KT88s, 4 ohm output tap, distortion (%) vs 1kHz continuous output power into (from bottom to top): 16, 8, 4 ohms.
Fig.7 Mystère ia21, EL34s, 8 ohm output tap, distortion (%) vs 1kHz continuous output power into (from bottom to top): 16, 8, 4 ohms.
Fig.8 Mystère ia21, EL34s, 4 ohm output tap, distortion (%) vs 1kHz continuous output power into (from bottom to top): 16, 8, 4 ohms.
I have shown only one circumstance where I plotted the THD+N percentage against frequency into different impedances, from the 4 ohm tap with the amplifier fitted with KT88 output tubes (fig.9). The behavior is broadly similar from both taps with both kinds of tube: the ia21 is at its most linear into higher impedances in the midrange and low treble, with the right channel (red and magenta traces) not quite as well behaved as the left (blue and cyan). I had assumed that the rise in THD below 50Hz was due to the onset of core saturation in the output transformer, but the 2 ohm behavior (green) is better than that into higher impedances. The distortion also starts to rise a little in the top two audio octaves, due to the circuit's limited open-loop bandwidth.
Fig.9 Mystère ia21, KT88s, 4 ohm output tap, THD+N (%) vs frequency at 2.83V into: 8 ohms (left channel blue, right red), 4 ohms (left cyan, right magenta), 2 ohms (green).
With both EL34 (fig.10) and KT88 (fig.11) tubes, the predominant distortion harmonic at low powers is the subjectively benign second, though FFT analysis indicates that other harmonics, decreasing in level with increasing order, are also present (fig.12). This graph also indicates that AC-supplyrelated spuriae are present; repeating the analysis with a 1kHz tone at one-third power shows that both the fundamental tone and its harmonics have sidebands at the supply-related frequencies of ±60 and ±120Hz (fig.13). The slightly worse linearity in the right channel can be seen, in fig.13, to be associated with a higher level of third harmonic than in the left channel.
Fig.10 Mystère ia21, EL34s, 8 ohm output tap, 1kHz waveform at 1W into 8 ohms (top), 0.09% THD+N; distortion and noise waveform with fundamental notched out (bottom, not to scale).
Fig.11 Mystère ia21, KT88s, 8 ohm output tap, 1kHz waveform at 1W into 8 ohms (top), 0.17% THD+N; distortion and noise waveform with fundamental notched out (bottom, not to scale).
Fig.12 Mystère ia21, EL34s, 8 ohm output tap, spectrum of 50Hz sinewave, DC1kHz, at 1W into 8 ohms (left channel blue, right red; linear frequency scale).
Fig.13 Mystère ia21, KT88s, 8 ohm output tap, spectrum of 1kHz sinewave, DC10kHz, at 16W into 8 ohms (left channel blue, right red; linear frequency scale).
Finally, the increase in linearity in the top two octaves seen in fig.9 doesn't result in a poor result on the high-frequency intermodulation test. The difference product at 1kHz resulting from asking the ia21 to drive an equal mix of 19 and 20kHz tones at 10Wpc into 8 ohms from its 8 ohm tap lay at 66dB (0.05%) in the right channel and74dB (0.02%) in the left (fig.14).
Fig.14 Mystère ia21, KT88s, 4 ohm output tap, HF intermodulation spectrum, DC24kHz, 19+20kHz at 10W peak into 8 ohms (linear frequency scale).
In most respects the Mystère ia21's measured performance is typical of a classic tube design using only a limited amount of global loop feedback, though its wide bandwidth is better than that of such designs. As to why Erick Lichte preferred the EL34 tubes to the KT88s, I must assume that it was due to the combination of a slightly greater degree of second harmonic coupled with the softer clipping characteristic. But I just can't come to grips with why (or how?) the ia21's output impedance is so extraordinarily high.John Atkinson