Hafler XL-280 power amplifier Page 2
Interestingly, more conventional measurements on the 280 yield results which aren't all that different from the SWDT results. Hafler's literature states that the XL-280 has "conventional harmonic and intermodulation distortion specifications comparable to the DH-220 amplifier." At 0.05% (as compared with the SWDT's 0.03%) and 0.005% respectably, those are very good but not fantastic specs. So what makes the 280 any better than the 220? Presumably, it is those elusive qualities for which we have no measurements, but which are improved by such circuit refinements as:
No input, DC-blocking capacitor, and only one other capacitor in the entire signal path. Low-frequency phase shift should thus be very low, although not as low as it would be with an all-DC coupled signal path, and the amplifier should be free from capacitive aberrations (as outlined by Walt Jung and Richard Marsh in their Audio Amateur articles on getting the most from the DH-200).
In common with amplifiers from Krell, Threshold, and Naim, the XL-280 has no inductors in series with the outputs. These inductors are included in nearly all designs to define the HF rolloff with capacitive loads, and thereby avoid high-frequency (ultrasonic) oscillation. Without them, the amplifier must be very much more intrinsically stable; with them, sound quality undoubtedly suffers.
30dB less negative feedback than the DH-220, for reduced TIM.
Two completely isolated power supplies, making the amp essentially a dual-mono design, apart from a common line transformer. Each supply consists of two separate sub-supplies for the output and voltage amplifier stages.
31,200µF of power-supply storage capacitance56% more than in the DH-220.
A rated current capability of 18 amperes, as compared with the 220's 10 amps. And a short-term power rating at 1 ohm of 325 watts!
While these are excellent particulars, there are any number of (admittedly much costlier) amplifiers which can easily out-spec and out-rate the XL-280. We have a few of them on hand, but so as not to prejudice myself unnecessarily, I made a point not to subject them to the SWDT until I had finished all my listening tests on the XL-280. I did, however, run the 280 through Hafler's SWDT, and tweaked it for maximum audible HF null, using the amplifier's built-in "Excelinear" adjustment, and with the speakers with which it was to be auditioned.
It was while trying to null out the white-noise input signal that I noticed two odd things. First, the residual (un-nulled) sound differs markedly in spectral content on both sides of the null point. To one side of maximum null, the entire spectrum above the lower midrange increases in level; to the other side of null, the residual sounded primarily like midrange energy. The Excelinear trim adjustment seemed asymmetrical in its action, producing a different quality of residual sound on each side of the null point. Additionally, instead of nulling only the high end, as it is supposed to, adjusting the trim seemed also to change the pitch of the upper midrange and middle high range. Thus, getting maximal null involves a little judgment, as well as a sensitive ear.
I wondered if these nulling oddities might be due to unmatched loads, due to the much longer wire running to the distant load speaker. But shortening that cable made the sound from the load speaker so loud at the monitor location that there was no possibility of hearing what was going on. It occurred to me then that I might get a better idea of what was what by substituting a heavy-duty resistor, close to the amplifier, for the faraway loudspeaker. This did not work. The moment the amplifier under test was switched on, it generated a loud and ominous hum and the 10-watt resistor became very hot almost instantly. Something was causing oscillation. (I am still trying to figure out what.) I reconnected the remote load speaker.