Halcro dm38 power amplifier Measurements
Not having any convenient current-source test gear, I assessed the Halcro dm38's measured performance only via its conventional balanced and unbalanced voltage-mode inputs. I ran it for an hour at 60Wpc into 8 ohms before performing any measurements. The pillars in the centers of the aluminum side moldings were just too hot to touch, implying a temperature around 65 degrees C, while the central boxes were cooler, at an estimated 50 degrees C.
The voltage gain into 8 ohms was to specification at 29.5dB via the unbalanced input, but 6dB lower via the balanced input, which is unusual. Both inputs preserved absolute polarity, the XLR jack being wired with pin 2 hot. The input impedance, assessed at 1kHz, was just above twice the specified value, at 21.6k ohms unbalanced, 44.5k ohms balanced. This increase will make the amplifier compatible with a wider range of preamplifiers, however.
The output impedance was very low, at below 0.1 ohm in the bass and midrange, increasing slightly to 0.1 ohm at 20kHz. As a result, the modification of the dm38's response by the usual Ohm's Law interaction between the amp's source impedance and the manner in which the speaker's impedance changes with frequency will be minimal. With our simulated speaker load, the response variation was around ±0.1dB across the audioband. This can be seen in fig.1, which also shows the dm38's response into resistive loads of 8, 4, and 2 ohms, driven from its balanced input. Although a very small degree of infrasonic rolloff is revealed by fig.1, the Halcro amp otherwise has a wide bandwidth, with a high-frequency -3dB point at 122kHz. Driven by the unbalanced jack, the bandwidth was even greater, at -3dB at 172kHz (not shown), resulting in a nicely square 10kHz squarewave response (fig.2).
Fig.1 Halcro dm38, frequency response at 2.83V into (from top to bottom at 2kHz): simulated loudspeaker load, 8 ohms, 4 ohms, 2 ohms (0.5dB/vertical div., right channel dashed).
Fig.2 Halcro dm38, small-signal 10kHz squarewave into 8 ohms.
Channel separation via both inputs was superb, at 110dB or better at 1kHz. Although the crosstalk increased to -90dB at the band extremes, this is still excellent performance. The measured background noise (with the input shorted) was also superbly low, at -107.5dBA ref. 1W into 8 ohms. The unweighted audioband S/N ratio was only slightly worse, at 105dB, though extending the measurement bandwidth to 10Hz-500kHz degraded the measured figure to 86.5dB.
It is in the area of distortion, of course, that Halcro's reputation of being a maker of extremely linear amplifiers was made. When I measured the dm58 monoblock, its performance was at the limits of my test equipment, so I approached the task of measuring the dm38's distortion with some trepidation. Fig.3 shows how the percentage of distortion and noise in the amplifier's output varies with power and load impedance. The downward slope below 10W of the three traces indicates that the measured percentage is actually dominated by noise. (As the output power drops, the identical level of background noise increases as a percentage of that output power.)
Fig.3 Halcro dm38, distortion (%) vs 1kHz continuous output power into (from bottom to top at 100W): 8 ohms, 4 ohms, 2 ohms.
Between 10W and 30W into 8 ohms—the lowest trace—the distortion starts to rise out of the noise floor, at around 0.001% (-100dB), but the sawtooth shape of the trace indicates that this is at the limit of my Audio Precision System One's resolving power. From 30W into 8 ohms to the "knee" in the trace at 190W, the slight rise in the THD+N percentage is due to an increasing level of distortion harmonics. This is different from the dm58, where there was no rise below the knee (see fig.8 in my October 2002 measurements). However, even at 190W into 8 ohms, the actual THD figure is still very low, at 0.025% (-91dB).
There is a more clearly marked rise in the measured distortion percentage above 10W into 4 ohms, and even more so into 2 ohms. But note that the Halcro comfortably exceeds its specified output power both into 8 ohms—205W (23.1dBW) was available at 1% THD—and into 4 ohms, where it gave out 370W (22.7dBW) at the same THD figure. It stumbled slightly into 2 ohms, giving out 182W (16.6dBW), which implies significantly lower voltage delivery. This was not unexpected; Halcro's Bruce Candy has made no secret of his opinion that speakers with impedances below 4 ohms are unnecessarily demanding on amplifiers.
Fig.4 shows how the measured THD+noise percentage changes with frequency into 8, 4, and 2 ohms. (The output level chosen for this graph was 10V, which is where the amplifier's distortion starts to become distinguishable from the background noise.) The measured figure into 8 and 4 ohms hovers around 0.001% over the entire audioband, though a very slight rise is apparent above 20kHz—the dm58's performance was very similar in this respect—and the right channel was slightly less linear than the left. But into 2 ohms, there is both a fourfold increase in THD and a rise toward the top of the audioband.
Fig.4 Halcro dm38, THD+N (%) vs frequency at 10V into (from bottom to top): 8 ohms, 4 ohms, 2 ohms (right channel dashed).
Figs.5, 6, and 7 show the dm38's linearity in a different manner. The bottom trace in fig.5 shows the waveform of the amplifier's output when the 1kHz driving signal has been removed with a sharp notch filter. The level was 1W into 8 ohms. A small degree of second-harmonic distortion can be seen, overlaid by noise. In fact, as implied by fig.3, the actual noise at this power level was much higher in level than the residual distortion—I averaged 32 'scope readings to drop the relative level of the random noise compared with that of the repetitive signal (footnote 1).
Fig.5 Halcro dm38, 1kHz waveform at 1W into 8 ohms (top), 0.002% THD+N; distortion and noise waveform with fundamental notched out (bottom, not to scale).
Fig.6 Halcro dm38, 1kHz waveform at 61W into 8 ohms (top), 0.0015% THD+N; distortion and noise waveform with fundamental notched out (bottom, not to scale).
Fig.7 Halcro dm38, 1kHz waveform at 120W into 4 ohms (top), 0.0055% THD+N; distortion and noise waveform with fundamental notched out (bottom, not to scale).
Increasing the output power to 61W makes the distortion waveform more evident (fig.6), with some higher harmonic content starting to make an appearance. It is important to note, however, that the actual THD+N level was just 0.0015% (-96dB). Finally, fig.7 shows that halving the load to 4 ohms without altering the output level introduces a fairly pure-looking third harmonic. Again, it must be noted that the absolute level of this distortion is still very low: 0.0055%, or -85dB.
At lower frequencies and higher powers into both 8 ohms (fig.8) and 4 ohms (fig.9), the third harmonic remains predominant. (This behavior is higher than the residual harmonic distortion level in my Audio Precision signal generator, so is presumably real.) With the Halcro's switching power supply and what must be excellent internal grounding arrangements, there is not a trace of AC-related spuriae to be seen in these spectra.
Fig.8 Halcro dm38, spectrum of 50Hz sinewave, DC-1kHz, at 110W into 8 ohms (linear frequency scale).
Fig.9 Halcro dm38, spectrum of 50Hz sinewave, DC-1kHz, at 215W into 4 ohms (linear frequency scale).
Finally, fig.10 shows a spectrum of the dm38's output while it drove an equal mix of 19kHz and 20kHz tones into 4 ohms at a level a dB or so below visible waveform clipping on the oscilloscope screen. The source was 24-bit data decoded by a Benchmark DAC 1, which produces a second-order difference component below -100dB with this signal (see fig.8, May 2004, p.119). The difference component in the Halcro's output lies at -94dB (0.002%), which, allowing for the Benchmark's contribution, means the dm38 is misbehaving very slightly on this test. Very slightly (footnote 2).
Fig.10 Halcro dm38, HF intermodulation spectrum, DC-24kHz, 19+20kHz at 300W peak into 4 ohms (linear frequency scale).
Its measured behavior indicates that the dm38 is best not used with amplifiers that have impedances lower than 4 ohms. But even though its circuit is not quite as breathtakingly linear as that of the dm58 monoblock—assuming my dm38 sample was representative—it still offers superb measured performance in line with Halcro's reputation.—John Atkinson
Footnote 1: Each doubling of the number of captures increases the uncorrelated noise level by 3dB but the correlated distortion by 6dB, thus dropping the noise contribution by 3dB. (I trigger the 'scope with the unfiltered waveform so that it starts its capture at exactly the same point for the averaging.)—John Atkinson
Footnote 2: It is fair to note that the sample of the dm38 reviewed by Martin Colloms in the June 2004 issue of British magazine Hi-Fi News was significantly more linear than my sample. Of course, that sample was set for 240V operation, though that should not be a factor with an amplifier using a switch-mode power supply.—John Atkinson