Halcro dm38 power amplifier Follow-Up, January 2007
When I reviewed the 180Wpc, $18,790 Halcro dm38 stereo power amplifier in October 2004, I had found a rather higher level of distortion just before the amplifier reached its clipping point than was specified by the manufacturer. I also concluded that while the solid-state dm38 sounded superb, particularly with respect to its effortless dynamics and its astonishingly clean presentation, the amplifier is balanced toward the cool side of the spectrum, and will work best with speakers and source components that don't themselves sound lean.
After the review sample (serial no.110015) had been returned to Halcro in Australia, designer Bruce Candy informed me that they had found that a small capacitor in the amplifier's circuit was out of tolerance, and that they would send me a second sample. This sample (serial no.1100130) duly arrived, and after a delay while I reviewed other products, I unpacked it and used it for much of my usual review auditioning. The front end for my listening was the Mark Levinson No.30.6 D/A processor driven by the digital output of a Classé cdp-202 CD/DVD-Audio player and the Ayre C-5xe universal player. Preamp was the Mark Levinson No.326S, and cables were generally AudioQuest and Crystalconnect. I used the new dm38 over several months, with Revel Ultima Studio, Era Design 4, Stirling LS3/5a Mk.2, Harbeth HL-P3ES 2, and Snell LCR7 XL loudspeakers. My comments on the amplifier's sound are based on my experience with all of those speakers.
While I had some doubts about the universal applicability of the first dm38 sample, I had no such misgivings about the second. Far from sounding cool, its character—if an amplifier as neutral as this one could be said to have a character—was, if anything, slightly warm. "Rich-sounding," "weighty," "authoritative," read my listening notes—the dm38 seemed to offer a superbly clean window into the recorded soundstage. That soundstage was also wide and deep where appropriate, with superbly well-defined and stable instrumental images.
I used the Halcro while applying the final touches to my two most recent recording projects, Variations (Stereophile STPH017-2) and There Lies the Home (Cantus CTS-1206), feeding the 24-bit/88.2kHz files from my laptop to the Levinson DAC via the AES/EBU data output of a Metric Halo MIO2882 FireWire interface. Dynamics seemed unrestrained with the full-range Revel speakers, and while I still feel my reference Mark Levinson No.33H monoblocks have a little more low-frequency authority, the dm38 is no slouch in this area. If anything, the Halcro's higher-frequency presentation was even freer of grain, judging by the sound of my own master files.
But it was the conflict between my measurements of the dm38 and the specified performance that was the primary motive for my doing a follow-up. Fig.1 shows how the first sample of the dm38's percentage of THD+noise varied with increasing output power into 8, 4, and 2 ohms. The downward slope of the traces in this graph below 20W or so indicates that if there is any distortion present, it is buried in the amplifier's noise floor. Above that level, it does appear that the THD is higher than specified, up to the point where the amplifier starts to clip. Fig.2 plots the THD+N percentage against frequency, again into 8, 4, and 2 ohms, and while the distortion is still very low, again it appears that it is higher than specified, with fig.3 clearly indicating a predominance of third-harmonic content.
Fig.1 Halcro dm38, sample 1, distortion (%)vs 1kHz continuous output power into (from bottom to top at 10W): 8, 4, 2 ohms.
Fig.2 Halcro dm38, sample 1, THD+N (%)vs frequency at 10V into (from bottom to top): 8, 4, 2 ohms (right channel dashed).
Fig.3 Halcro dm38, sample 1, 1kHz waveform at 120W into 4 ohms (top), 0.0055% THD+N; distortion and noise waveform with fundamental notched out (bottom, not to scale).
Turning to the new sample of the dm38, my initial measurements of the second sample were identical to those of the first. Before I published them, however, I needed to check out my measurement equipment.
The first suspect was the signal generator of my 17-year-old Audio Precision System One DSP. This had been checked out by the factory a few years back, and at the output required to drive the Halcro to a level where I had been finding third-harmonic content, the THD+N level was just 0.00055%, or one-tenth that of the amplifier. The waveform of the generator's distortion content is shown in fig.4, plotted to the same vertical scale as fig.3. To examine the harmonic content of the generator's output, I first fed it to the National Instruments measurement card in my PC, which is the platform for the Miller Audio Research software suite, but the only harmonic apparent above the noise floor in the graph (not shown) was the second.
Fig.4 Audio Precision System One signal generator, spectrum of balanced 1kHz sinewave at 1V RMS.
I therefore fed the residual signal after the 1kHz fundamental had been notched out to an external 24-bit A/D converter (a Benchmark ADC1) and fed that unit's AES/EBU output to the input of a PrismSound DScope 2 to perform FFT analysis. The resultant spectrum, with 128 samples averaged to further reduce the noise content, is shown in fig.5. The combination of the gain applied by the Audio Precision's analyzer and the Benchmark ADC1 was 60dB, meaning that 60dB must be added to the numbers in this graph's vertical scale to give the true level of the harmonics.
Fig.5 Audio Precision System One signal generator, spectrum of balanced 1kHz sinewave at 1V RMS (+60dB gain).
The AP's monitor output is unbalanced, and the electrical connection to the A/D converter, then to the DScope card in a second computer, introduced some grounding spuriae in this graph, most noticeably at 60Hz, 180Hz, and 300Hz. These should be ignored. The primary distortion harmonic in fig.5 is the second at 2kHz, which lies at –52dB; ie, –112dB (0.0005%), once the 60dB of gain is taken into account. The third, fifth, and seventh harmonics can also be seen just above the –60dB (–120dB) line. But the point to note is that the Audio Precision's own distortion is well below what I measured in the dm38.
The next suspect was the test load. My standard load is a matrix of noninductively wound 8 ohm, 1% Dale power resistors, each with a power handling of 250W and mounted to a heatsink. Double-pole, double-throw switches allow these to be configured with values of 8, 4, or 2 ohms. I had last checked the switches and wiring in 2003, but my next test was to acquire new power resistors and hook them up without any switches in the signal path, paralleling them as required using 12-gauge jumper cables. These gave identical results to my usual load.
In an e-mail he had sent me following my tests of the Halcro dm88 (published in the August 2006 issue), in which I had also found slightly higher levels of distortion than Halcro specified, Bruce Candy had suggested an additional test, to examine the linearity of my test load: "Connect, say, ten 1 ohm, 2W metal-film resistors in parallel to give 0.1 ohm in total. Connect one end of this resistance in series with your 4 ohm loudspeaker earth [ground] and the other to the amplifier's positive terminal. Adjust the Audio Precision's output signal to a value that results in 450–500W [with the dm88], then measure the THD across the 0.1 ohm resistor. This should be low; if not, the load is nonlinear and thus the source of the problem."
The dm88 had been returned to the distributor by this time, but I performed Candy's test on the dm38, at a level close to clipping into 4 ohms. To my surprise, while the test load appeared to be linear with the amplifier's right channel, I found some third harmonic present with the left. I then found that the third harmonic moved to the right channel when I swapped the channel wiring—but, to my even greater surprise, not when I swapped back the loads but not the wiring. Only one explanation correlates with this behavior: The observed distortion was coming from the wire (with its connectors), not the load resistors or the Halcro amplifier!
To facilitate the hooking-up of amplifiers in my test lab, I use 6' lengths of 14-gauge multistrand cable fitted with stackable dual banana (4mm) plugs and short adapter cables fitted with a dual banana plug at one end and, at the other, single bananas, spade lugs, or bare ends. It turned out that the connection of one of the dual bananas to one of the 6' cables was not as conductive as it should have been, due to oxidation of the cable conductors under the banana plug's grub screw. Mystery solved—though I don't have a clue why the nonlinearity manifested itself only at high currents.
I removed all the banana plugs from the cables, restripped the cable ends, reattached the plugs, and redid the distortion tests on the Halcro dm38 using the new load. Fig.6 shows how the THD+N percentage changes with output power into 8 and 4 ohms. By comparison with fig.1, it can be seen that the THD+N figure does not increase until the amplifier clips, with the sawtooth nature of the traces due to the fact that the dm38's distortion is at the residual level of the Audio Precision's resolution. Fig.7 plots the dm38's THD+N percentage against frequency at the same level as fig.2. The audioband distortion is now just 0.00072% into both 8 and 4 ohms, and fig.8 reveals that while it is almost as low in harmonic content as the Audio Precision's generator output (fig.4), there is still a hint of third harmonic.
Fig.6 Halcro dm38, sample 2, distortion (%)vs 1kHz continuous output power into (from bottom to top at 10W): 8, 4 ohms.
Fig.7 Halcro dm38, sample 2, THD+N (%)vs frequency at 10V into (from bottom to top): 8, 4 ohms.
Fig.8 Halcro dm38, sample 2, 1kHz waveform at 120W into 4 ohms (top), 0.00072% THD+N; distortion and noise waveform with fundamental notched out (bottom, not to scale).
But overall, the Halcro dm38 now conforms to its specified low distortion at high powers and, possibly, the dm88 as well.
With its measured performance exonerated by my final series of tests and a sound quality even better than I remember of the first sample, the Halcro dm38 continues to occupy its position in the first rank of recommended amplifiers.—John Atkinson