Finally got around to making some measurements on the Sure Electronics 4 x 100 W amplifier kit. It throws a bit of light on “T-Amps” and other class D amplifiers.

Frequency response at 4 ohms, 1 watt. Less than 1dB down at 20 and 20K Hz. Not ruler-flat like a top shelf class-AB amplifier - a bit reminiscent of a mid-fi 1960s tube amp.

Frequency response at 4 ohms, 25 watts. Essentially the same as 1 watt. Good; response doesn’t get worse as power increases.

Frequency response at 8 ohms, 1 watt. Whoa... the extreme top rises, instead of falls. Looks as if the output chokes were chosen for a compromise impedance of around 6 ohms. Interesting...

Frequency response at 2 ohms, 1 watt.  Uh huh... the top end falls off more than twice as fast as at 4 ohms. Just what you’d expect as the inductive reactance becomes larger with respect to the load impedance.

The conclusion is that switching amps have an inherent problem in the output filter. A choke and capacitor filter the high switching frequency out of the speaker outputs. The choke has to be selected to match the speaker’s impedance. An LC filter depends on the load for damping, so Q varies with load impedance. Chokes saturate, and inductance varies with current below saturation, leading to distortion. Capacitors, especially ceramics, change capacity as voltage varies, again contributing distortion. What happens with a real “8 ohm” speaker whose measured impedance ranges from 6 to 45 ohms over the 20 - 20,000 Hz audio band? This amp sounded very unhappy on my Quad electrostatics for example.

These and other chips take feedback from the “digital” side of the output filter. Taking feedback directly at the speaker terminals could correct for the filter, but delays and phase shifts make that impossible, at least at the current state of the art.

In all, T-Amp chips come off well with typical speakers. Lotsa watts per dollar, commendably low distortion, and decent frequency response. Easy to listen to, if not up there with the best class-A and AB amps available. The switching frequency is higher than all but the very latest chips, which helps keep modulation and filter distortion low.

For an excellent technical analysis of Class-D output filters, see the TI website here:

Too bad Tripath is out of business and the supply of chips is drying up.