EAR/Yoshino M100A monoblock power amplifier Measurements

Sidebar 3: Measurements

The first round of measurements were performed on the sample that had not developed audible distortion. Noting that the M100A is described as operating in class-A, my usual practice before measuring an amplifier is to run it at one-third power into 8 ohms for one hour. This severely stresses a design with a class-B output stage, as it results in the maximum heat being dissipated in the output transistors. My first test for any amplifier is therefore to quickly check its clipping power at 1kHz (defined as 1% THD+noise) with my Audio Precision System One.

I was therefore surprised to find that, despite its specified 100W output at 2% THD, the EAR/Yoshino Paravicini M100A amplifier clipped at less than 10W into 8 ohms from the 8 ohm output transformer tap. I thus ran the amplifier at 3W into 8 ohms for an hour. Following this preconditioning period, the chassis was very hot, so much so that I couldn't keep my hand on it for more than a couple of seconds. I also noted a quite audible 1kHz tone coming from the amplifier's innards.

One thing alarmed me: If I turned the amplifier on without an input or load connected—something the manual warns against, I was later to note—both front-panel meters pegged, even with the M100A's volume control turned all the way down, and the rear-panel 6.3A fuse blew. Everything was fine when I turned the amplifier on with connections made, however, the lower meter's needle just to the right of its midpoint and the upper meter accurately indicating power with the resistive load matched to the output.

Getting the basics out of the way: The M100A inverted absolute polarity via its unbalanced input; the balanced input was polarity-correct with pin 2 of the XLR wired as hot. The balanced input impedance was around 26k ohms at 1kHz; the unbalanced input impedance at the same frequency was a low 8.5k ohms. (Both figures were measured with the volume control all the way up.) Voltage gain into 8 ohms depended both on the output transformer tap and the input used. The balanced gain at 1kHz was on the low side at 28.1dB, 26.2dB, and 23.3dB from the 16 ohm, 8 ohm, and 4 ohm taps, respectively. Unusually but usefully, the gain from the single-ended inputs was higher, at 31.6dB, 29.8dB, and 26.8dB, respectively.

Due to the use of an output transformer, the M100A's source impedance was higher than we usually see with a solid-state design. It was lowest from the 4 ohm output, at 0.55 ohm across most of the audioband—equivalent to a damping factor of just under 8—rising to 1.05 ohms at 20kHz. The 8 ohm figures were 1.1 ohms, rising to 2.3 ohms at 20kHz, while from the 16 ohm transformer I measured 1.9 ohms at 20Hz and 1kHz, and a very high 4 ohms at 20kHz.

As a result of these high values, there will be significant interaction between the M100A and the load impedance of the speaker with which it is used. This can be seen in fig.1, which shows the amplifier's frequency response from the 8 ohm tap, measured into 8, 4, and 2 ohm loads, as well as into our standard dummy loudspeaker load. Very audible response variations of +0.8dB, –1dB occur with the dummy load, these figures halving from the 4 ohm tap and almost doubling with the 16 ohm tap (not shown). Presumably due to its increasing source impedance at the top of the audioband, the amplifier's ultrasonic response rolls off a little early, with the level at 20kHz down by 1dB from the 8 ohm tap into 8 ohms, –1.8dB from the same tap into 4 ohms. This was using the balanced input; the 8 ohm response from the 8 ohm tap was 1.5dB at 20kHz using the single-ended input jack.

Fig.1 Paravicini M100A, 8 ohm tap, frequency response at (from top to bottom at 2kHz): 2.83V into dummy loudspeaker load, 1W into 8 ohms, 2W into 4 ohms, 4W into 2 ohms (0.5dB/vertical div.).

Despite this ultrasonic rolloff, there was the beginning of a slight overshoot apparent on the 10kHz squarewave with the output transformer matched to the load (fig.2). This disappeared when the load impedance was set to half the transformer tap rating (fig.3), due to the earlier UHF rolloff.

Fig.2 Paravicini M100A, 8 ohm tap, small-signal 10kHz squarewave into 8 ohms.

Fig.3 Paravicini M100A, 8 ohm tap, small-signal 10kHz squarewave into 4 ohms.

Even though it uses a single-ended topology and has low loop negative feedback, the M100A is very linear at normal levels. Figs.4–6 plot the percentage of distortion and noise present in the amplifier's output against frequency with the generator set to produce 1W with the transformer tap matched to the load impedance. The THD figure hardly changes in these graphs as the load is dropped from 16 to 4 ohms in the midrange and treble, though the 2 ohm load leads to increasing distress from the higher-impedance taps. Note, however, the rise in THD below 100Hz, presumably due to output-transformer core saturation. And, surprisingly, the impedance peak between 4kHz and 6kHz with our dummy speaker load gives rise to increased distortion compared with the pure resistive loads, this—also to my surprise—greater in degree with the higher-impedance taps.

Fig.4 Paravicini M100A, 4 ohm tap, THD+noise (%) vs frequency at (from top to bottom at 5kHz): 2V into simulated loudspeaker load, 2W into 2 ohms, 1W into 4 ohms, 0.5W into 8 ohms, 0.25W into 16 ohms.

Fig.5 Paravicini M100A, 8 ohm tap, THD+noise (%) vs frequency at (from top to bottom at 5kHz): 2.83V into simulated loudspeaker load, 4W into 2 ohms, 2W into 4 ohms, 1W into 8 ohms, 0.5W into 16 ohms.

Fig.6 Paravicini M100A, 16 ohm tap, THD+noise (%) vs frequency at (from top to bottom at 5kHz): 1.4V into simulated loudspeaker load, 8W into 2 ohms, 4W into 4 ohms, 2W into 8 ohms, 1W into 16 ohms.

I was surprised yet again when I looked at the small-signal distortion waveform under various conditions. The bottom trace in fig.7 shows the residual waveform with the M100A driving 7.2W into 4 ohms from its 8 ohm output transformer tap. Almost pure third-harmonic can be seen, overlaid with noise at this respectable (0.05%) THD level. But look what happens when I kept the input level the same but matched the load to the transformer tap (fig.8). Even though the distortion level remains low at 0.05%, the subjectively benign third harmonic has been replaced by a spike representing high-order and therefore more objectionable harmonics, this coincident with the positive sinewave peaks. This perverse behavior was the same whether I used the balanced or unbalanced inputs.

Fig.7 Paravicini M100A, 8 ohm tap, 1kHz waveform at 7.2W into 4 ohms (top), distortion and noise waveform with fundamental notched out (bottom, not to scale).

Fig.8 Paravicini M100A, 8 ohm tap, 1kHz waveform at 4.6W into 8 ohms (top), distortion and noise waveform with fundamental notched out (bottom, not to scale).

As suggested by figs.4–6, the distortion at low frequencies is higher than at higher frequencies. This can be seen in fig.9, which shows the spectrum of the M100A's output while it drives a 50Hz tone at 2.6W into 8 ohms from its 16 ohm output. To my surprise, given my conjecture that the low-frequency nonlinearity is due to the output transformer, the highest-level harmonic apparent in this graph is the second, at –62dB (0.08%), with the third harmonic lower in level at –66dB (0.05%). Though the noise floor is appreciably low in level, a number of power-supply–related spikes are also apparent.

Fig.9 Paravicini M100A, 16 ohm tap, spectrum of 50Hz sinewave, DC–1kHz, at 2.6W into 8 ohms (linear frequency scale).

The only bright picture to be found in the distortion measurements is that intermodulation distortion at small-signal levels was vanishingly low, even with the load matched to the transformer tap (fig.10). Perhaps this good behavior below clip correlates with Jonathan's description of the M100A's "extreme You Are There transparency."

Fig.10 Paravicini M100A, 8 ohm tap, HF intermodulation spectrum, DC–24kHz, 19+20kHz at 1W into 8 ohms (linear frequency scale).

I always wait until the end of the test session to assess in detail how an amplifier's distortion changes with output power—my experience has been that that is when some amplifiers will break. However, I wished that I had done more than a cursory check of the M100A's clipping behavior before the other tests. This can be seen in fig.11, which plots the THD+noise percentage against output power into 16, 8, 4, and 2 ohms from the 8 ohm tap.

Fig.11 Paravicini M100A, 8 ohm tap, distortion (%) vs continuous output power into (from bottom to top at 0.1W): 16 ohms, 8 ohms, 4 ohms, 2 ohms.

Basically, the Paravicini amplifier only gets close to achieving respectably low distortion over all its dynamic range and delivering its specified output power only when the load impedance is half the value of the transformer tap. Go to a load a quarter of the rating and the clipping power drops, though to a still usable amount. The 8 ohm tap gives 100W into 4 ohms (20dBW) and 55W into 2 ohms (11.4dBW), for example.

I was also concerned that once the amplifier clipped, it gave even less power. Observing the waveform on a 'scope revealed that this behavior was due to the amplifier going into ultrasonic oscillations on every positive sinewave. No wonder Jonathan was bothered by hearing clipping during his auditioning of the M100A. It's only because Jonathan's speakers are more sensitive than the norm and he used them from the amplifier's 8 ohm outputs, coupled with the fact that at normal playback levels, the amplifier is not being asked to deliver much more than 1W, that the M100As survived the auditioning with only occasional audible clipping.

It was understandable that I became concerned that our samples of the M100A were broken. And, as US distributor Dan Meinwald confirmed when he saw the preprint of the review, this was indeed the case. We therefore put the review on hold—we had originally planned to publish it in our October 2001 issue—until we could get samples that were working as intended by the designer.

When we received the reworked amplifiers, I ran a second series of tests. The highish output impedance was the same, as was the voltage gain. However, there was now a slight peak in the treble when the load impedance was equal to or higher than than the output tap (fig.12). Though the small-signal distortion still increased with decreasing frequency below 100Hz, the overall THD was considerably lower (fig.13), even into low impedances. And the distortion spectrum of the new sample showed fewer high-order harmonics and a regular series of low-order harmonics (fig.14). Note that this graph was taken at 42W into 8 ohms from the 8 ohm tap, compared with 2.6W for the earlier sample (fig.9). The 120Hz AC supply components is also very much lower in level, at almost –90dB (0.003%).

Fig.12 Paravicini M100A, 8 ohm tap, frequency response at (from top to bottom at 2kHz): 2.83V into dummy loudspeaker load, 0.5W into 16 ohms, 1W into 8 ohms, 2W into 4 ohms, 4W into 2 ohms (0.5dB/vertical div.).

Fig.13 Paravicini M100A, 4 ohm tap, THD+noise (%) vs frequency at (from top to bottom at 5kHz): 2.83V into simulated loudspeaker load, 4W into 2 ohms, 2W into 4 ohms, 1W into 8 ohms, 0.5W into 16 ohms.

Fig.14 Paravicini M100A, 8 ohm tap, spectrum of 50Hz sinewave, DC–1kHz, at 42W into 8 ohms (linear frequency scale).

The clipping behavior (fig.15) was now what we had expected, with just over 100W available when the M100A's output tap was matched to the load, and respectively low distortion below the clipping point.

Fig.15 Paravicini M100A, 8 ohm tap, distortion (%) vs continuous output power into (from bottom to top at 0.1W): 16 ohms, 8 ohms, 4 ohms, 2 ohms.

Fig.15 was assessed using continuous 1kHz sinewaves. I used the Miller Amplifier Profiler to examine how the amplifier manages with a signal more typical of a musical signal: 10 cycles of 1kHz followed by 400 cycles of silence. Fig.16 shows the result for the M100A's 8 ohm tap. With this transient-rich signal, 120.4W (20.8dBW) was available into 8 ohms at 1% THD (black trace), compared with 112W using a continuous sinewave. As expected, less power was available when the load impedance dropped below the output transformer tap, though the amplifier would still give raise 47W into 2 ohms from the 8 ohm output tap (blue trace). From the 4 ohm tap, this more than doubled to 107W.

Fig.16 Paravicini M100A, 8 ohm tap, distortion (%) vs 1kHz burst output power into 16 ohms (red trace), 8 ohms (black), 4 ohms (green), 2 ohms (blue), and 1 ohm (magenta).

Judging by the second sample of the M100A, this is an amplifier that delivers excellent measured performance, especially since it's a single-ended design. Regarding the first sample, the question must be asked why didn't we immediately suspect it of being broken? Remember—there is no way for a reviewer to actually know what the device under test's intrinsic behavior should be. Unless no signal at all comes out of the unit, the point at which a product can be deemed "broken" is a subjective judgment call. And in the case of the M100A, the original samples had been used at Home Entertainment 2001 and the importer had subsequently auditioned them at Jonathan's with the JMlab speakers without raising the alarm.—John Atkinson

X