Folks,
Recently I decided to examine the so-called 'digital', Class D amplifiers. There are now chips from TaCT (Danish company recently acquired by TI), UcD (Philips), ICEpower (OEM only, from Bang and Olafsen), Tripath (Tripathi Corporation), and, amongst the Japanese, a Sharp offering. Most of you have heard of Bruce Candy's Australian Class D amplifier, the Halcro, which is emerging as one of the best amplifiers in the world. There is a lot of buzz at Audioasylum and DIYAudio with comment ranging from the rhapsodic to the ho-hum; so it deserved a good look. I was captivated by this very efficient concept in my youth from a UK entrepreneur called Clive Sinclair, and was interested to see how it had progressed in forty years.
The amps operate at around 300-400 KHz, depending on implementation. By varying the on and off period of a mosfet output stage operating in push pull - by altering the mark/space ratio -and by using a (second order) LCR filter at the output, an analog waveform can be constituted from an integrated train of very fast on/off pulses, allowing a near perfect waveform with low distortion and very good transient response.
I bought a well known ready-built amp module, connected it to a Nirvana Plus 55W power supply, and began the listening tests.
Strengths:
1. Very good transient/impulse control. Slam and impact is amazing, and probably a tad better than the 100W AKSA.
2. Good detail in midrange, with some coloration in the top end. The sound is arresting, but not overly natural.
3. Astonishing bass. Tight, powerful, even a little 'wet', I suspect there's a little H2/H3 in there. Very appealing for subwoofers!
4. Good spatially. I hear a clear soundstage. In Class AB design (as you probably realize!) soundstage is incredibly difficult to achieve. This amp has good imaging with width and depth but it's not quite up there with the AKSA.
5. High efficiency but a surprising amount of heat generated in the output filter, which runs too hot to touch for more than a few seconds. At idle, I'd say the PWM amps are not much more efficient than low bias Class AB.
6. Good vocals; but a lack of warmth, not much engagement.
7. Outstanding load independence.
Weaknesses:
8. There is a rawness to the sound, almost abrasive, which is not quite fatigueing, but certainly not smooth, and lacking refinement. It's really only apparent on soft chamber music and acoustic vocal, and shows more as an edginess to the voice rather than obvious sibilance. Immediate impressions are very good, but quiet, low level listening reveals refinement levels not quite to very good Class AB levels.
9. On some electronic music ('Song for Olabi', by a Danish group called 'Bliss'), there is some sort of digital noise, low frequency. I don't know, but it's quite unnerving. It only occurs on some electronic music, NOTHING else. I will check this more, of course.
10. Sometimes there is an impression of a lack of 'body', as though a singer is not pushing hard through the diaphragm when holding a note. This creates a lack of conviction in the performance; one is not involved.
I had been concerned that PWM amps might now be surpassing good analog designs. However, after extensive listening I've come to the conclusion that they cannot match good Class AB or A analog designs at present until they increase their resolution; this probably means they must increase their operating frequency to beyond 1MHz. I'm informed that only at 2.7MHz operating frequency can they give true 20 bit resolution; this is probably more resolution than is required, but it indicates a clear evolutionary path. The other area of deficiency appears to be emotional engagement; this cannot be easily measured or engineered into these designs. So for perhaps three or four years to come, we can expect the better analog amps to hold their position!
Ben and I plan to write a white paper on this topic for the website in due course.
Cheers,
Hugh
