|
|
Class T-Amppeteraczel | 09 October, 2006 18:33
8-Channel Digital Power Amplifier Designed by AudioDigit (www.audiodigit.com) in Italy. Manufactured by Autocostruire di Melani Antonella, Via A. Modigliani 27/B, 51100 Pistoia, Italy. Voice: +39 335 290925. Fax: +39 0573 31018. E-mail: info@autocostruire.com. Web: www.autocostruire.com. AudioDigit MC8x100 eight-channel “digital” power amplifier, 480 euros (c. $610) plus shipping and import duty. Tested sample on loan from manufacturer.
As I have stated before, I’m not really interested anymore in testing parity products (such as, let us say, the 217th multichannel receiver to come on the market) but would greatly prefer to explore new and different, or at least unconventional, designs that point to the future. I was very excited, therefore, to discover on the Internet this strictly 21st-century low-cost amplifier from Italy. It uses special audio-amplifier IC’s manufactured by Tripath Technology Inc. of San Jose, California, which employ a technology Tripath calls Digital Power Processing. DPP-based amplifiers are “Class-T” designs, which are not quite like class D pulse-width-modulation amps but still have similar modulated switching carrier outputs. The claimed benefit is tremendous power efficiency without any tradeoffs in audio quality plus unequaled power/size ratio. I asked for and obtained a review sample of the amplifier, possibly the first Italian MC8x100 to reach the United States. The Design The amplifier uses two Tripath TAA4100A four-channel Class-T audio amplifier IC’s, for a total of eight channels. This IC was originally designed for automotive head-unit applications, and pressing it into service to drive domestic high-fidelity speakers is a bit of a stretch, but that was the Italians’ idea. A single toroidal transformer feeds all circuits. The entire eight-channel unit is 17½" wide, which is standard, but only 2" high and 9½" deep. It weighs just a little over 13 pounds. Not even Bob Carver’s famous magnetic field amplifiers of the early 1980s, considered disturbingly small and light by the high-end audiophile press, achieved this kind of size reduction. The amplifier has no controls of any kind, just an on/off switch. There are eight unbalanced RCA-jack inputs and eight speaker connectors of the type that accepts banana plugs, bare wire, or spade lugs. Unfortunately, the binding posts are just a tad too widely spaced to accept dual banana plugs; this is some kind of European hang-up about safety. (They have AC plugs whose dimensions fit the standard dual banana jacks—only an idiot would plug one of these into an amplifier’s output.). Class-T operation isn’t really digital in the sense that 0’s and 1’s are processed in the signal path, but the waveform before the output filter is a complex “digital” waveform (i.e., pulses) of varying frequency. By contrast, the corresponding waveform of a class-D PWM amplifier is fixed in frequency (generally between 100 kHz and 200 kHz). The difference lies in the architecture of the Tripath IC, which we won’t go into here (it involves DSP and “predictive processing,” among other things). The chief benefit is claimed to be power-conversion efficiencies of 80% to more than 90%, while yielding audio quality approaching class A and class AB. (I said, “claimed to be.”) The Measurements The frequency response of two of the eight channels, at 1 watt into 8Ω, is shown in Fig. 1. The bass rolloff starts at a higher frequency than is normal in standard high-fidelity amplifiers. Down by more than 0.4 dB at 20 Hz is not a very good spec. As for the 0.36 dB rise at 20 kHz, it is almost certainly inaudible but also abnormal. Our analog expectations may be too sanguine in the case of this “digital” amplifier. Let’s not exaggerate, though; this is still a very acceptable frequency response.
Fig. 1: Frequency response of channel 1 (cyan) and channel 5 (red) at 1 watt into 8Ω. The distortion curves depart significantly from the one in the instruction manual. That one bottoms out at 0.009% (–81 dB) with a 1 kHz input into a 4Ω load and clips at 45 watts, with very high-distortion output available up to 110 watts. My measurement of the same frequency into the same load shows an absolute minimum at –72 dB (0.025%), clipping at 40 watts (in the better of two channels), and maximum available output of 61 watts at sky-high distortion. Big difference. This is basically a low-powered amplifier. (The word I have from Italy is that the power supply in my test sample was set at a lower voltage, for safety reasons, than the one belonging to the amplifier spec’d in the manual.) Fig. 2 and Fig. 3 show the THD+N into 8Ω and 4Ω, respectively, at three different frequencies each, for two of the eight channels. Radical filtering above 20 kHz was applied to remove all out-of-band noise, since the out-of-band switching carrier components invariably affect the accuracy of the measurements. That is the reason for the change to 6 kHz from the usual 20 kHz as the highest test frequency; the second harmonic (12 kHz) and third harmonic (18 kHz) are included that way, whereas the use of a 20 kHz fundamental would be meaningless. As can be seen, the curves are pretty much clustered together and stay below –60 dB (0.1%) at all levels, except the 20 Hz curves, which are quite horrible. This amplifier is incapable of genuinely clean 20 Hz output beyond a couple of watts. Maybe the assumption is that a multichannel system driven by the amplifier will have a separately powered subwoofer.
Fig. 2: THD+N vs. power of two channels into 8Ω. Channel 1: 20 Hz (red), 1 kHz (cyan), 6 kHz (magenta). Channel 5: 20 Hz (yellow), 1 kHz (green), 6 kHz (blue).
Fig. 3: THD+N vs. power of two channels into 4Ω. Channel 1: 20 Hz (red), 1 kHz (cyan), 6 kHz (magenta). Channel 5: 20 Hz (yellow), 1 kHz (green), 6 kHz (blue). I wanted to investigate the 20 Hz distortion further. Fig. 4 is the FFT spectrum of a 20 Hz tone at 10 watts out, the level of maximum distortion, into a 4Ω load. Two of the eight channels are shown. The curves prove that the high distortion consists largely of second harmonic (40 Hz) and third harmonic (60 Hz), with a little fourth harmonic (80 Hz) thrown in. Nothing mysterious or exotic there; this is absolutely classic low-frequency distortion.
Fig. 4: Spectrum of a 20 Hz tone at 10 watts into 4Ω, two channels (channel 1, cyan; channel 5, red). Channel separation, on the other hand, is very good indeed, as shown in Fig. 5. Even at 20 kHz, the crosstalk is of the order of –68 dB, and over most of the audio spectrum it is –80 to –98 dB. That’s right up there with the best. It needs to be pointed out, however, that I measured channels 1 and 5, which are on two separate IC’s. My test setup was wired that way, and I saw no compelling reason to change it. If I had measured two channels on the same IC, the results may not have been quite as impressive.
Fig. 5: Separation between channels 1 and 5 at 1 watt into 8Ω. Finally, the PowerCube test was a total bust—not because of bad results but because it could not be performed. As I’ve explained many times before, the PowerCube test measures the ability of an amplifier to drive widely fluctuating load impedances. As far as I know, The Audio Critic is the only American audio journal to publish PowerCube measurements. The instrument for the test is made in Sweden; it produces repeated 1 kHz tone bursts of 20 ms duration into 20 different complex load impedances across the amplifier (magnitudes of 8Ω/4Ω/2Ω/1Ω and phase angles of –60°/–30°/0°/+30°/+60°). The graphic output of the instrument shows the 20 data point connected to form a more or less cubelike polyhedron. The test shows up the differences between otherwise similar amplifiers when it comes to real-world loudspeaker loads rather than just resistances. The AudioDigit amplifier went into protection in the earliest part of the test, on the 3rd of the 20 loads (–60°/2Ω). Obviously, the protection thresholds are set all wrong, in the expectation of 8Ω to 4Ω nonreactive, or mildly reactive, loads only. It is possible that the amplifier could draw a good PowerCube if the protection circuits were set differently. The Sound As I have pointed out innumerable times, a properly designed amplifier has no sound of its own. It is impossible for two amplifiers to sound different at matched levels if each has high input impedance, low output impedance, flat frequency response, low distortion, low noise floor, and is not clipped. The MC8x100 is a special case because of its peculiar THD+N vs. power curves, allowing considerable high-distortion output beyond the clipping point. The expectation of some sonic anomalies is therefore not altogether unreasonable. For a quick check, I connected the amplifier to a pair of floor-standing wide-range speakers of decent quality (Sony SS-K90ED’s), with channels 1 and 5 feeding left and right. I thought I heard a few subtle, momentary sounds I didn’t like. An ABX comparison with a conventional amplifier of comparable power would definitely be in order. That takes time, and I want to post this already delayed review forthwith. I’ll do the ABX tests later and append the results here when I am done. (Don’t expect anything revelatory.) Conclusion The AudioDigit Class T-Amp MC8x100 is a minor technological tour de force—with warts. I decided not to use it between the electronic crossover and the eight drivers of my pair of Linkwitz Lab “Orion” speakers, which was my original plan before I did the measurements. There just isn’t enough power before clipping, and after clipping the bizarrely available power appears to be unacceptable because of high distortion. Too bad—I would have liked to tell the high-end fanatics that I am using a really cheap amplifier to drive a state-of-the-art loudspeaker. The amplifier may still be perfectly adequate in a surround-sound system using conventional speakers, but we’d better wait for the ABX tests to confirm that. PS: No ABX Test Well, I set up an ABX test. Very carefully. (The identity of the other amplifier is at this point irrelevant, as you will see.) When I switched on the AudioDigit amplifier, it started to smoke and smell, and blew its power-supply fuse. Removing the cover, I identified the cluster of power-supply capacitors as the culprit. They smelled smoky; the rest of the amplifier did not. I think what happened was that my line voltage was a little high that afternoon, and the amplifier could not take the turn-on surge. Of course, one or more capacitors could have been faulty to begin with. Maybe it was pure luck that the measurements could be completed. The rest of the equipment plugged into the same power strip—and that includes the other amplifier—did not bat an eyelash. Nor have I had a similar problem with any piece of gear in the 18 years I have lived in my house. Draw your own conclusions. The amplifier is going back to Italy. Maybe they’ll send me an improved version. I still think it’s a very intriguing design that breaks with the past and looks ahead.
MenuSearchRecently... |
