Workshop #1
Amplifier Measurement Clinic

How are amplifier measurements done, what do they mean, and do they have any real value? This amplifier testing clinic showed attendees how audio power amplifiers are measured and which measurements are important.

An example power amplifier was measured in a fashion similar to that done for Stereophile equipment reviews. This clinic was conducted by Bob Cordell, who is an amplifier designer experienced in tube, bipolar and MOSFET amplifier designs. Bob has also worked on developing new and more sensitive distortion tests for amplifiers. These include Multitone Intermodulation (MIM) and Phase Intermodulation (PIM) tests, as well as distortion magnifier techniques for measuring ultra-low levels of THD, crossover and TIM distortions. Live measurements on an actual amplifier were made and differences in measurements and design were discussed. In the course of the clinic, Bob shared his views on why amplifier measurements are not always a good predictor of sonic differences.

The clinic began with an explanation of some amplifier review measurement charts kindly supplied by Stereophile's John Atkinson. These charts represented two different amplifiers reviewed by Stereophile. Charts used for illustration included frequency response into different loads, THD as a function of power into different loads, THD as a function of frequency into different loads, and twin-tone (19 kHz + 20 kHz) IM distortion spectra at high power. This part of the clinic gave attendees a chance to get familiar with measurements and what they mean before the actual live measurements were done. There was also ample opportunity here for Q & A on the topic.

Live measurements were then performed on a real Amplifier Under Test (AUT). Equipment used included a THD analyzer, a spectrum analyzer (SA), a PC-based Digital Storage Oscilloscope (DSO), a 19+20 kHz twin-tone source, a high-power 8-ohm dummy load and several other pieces of related equipment.

The Amplifier Under Test (AUT) was a DIY Chip Amp called the Super Gain Clone™. It is a 40-Watt per channel amplifier based on the National LM3886 operating in the inverting mode. It achieves a high input impedance and employs a low-distortion d.c. servo loop to avoid the use of capacitors anywhere in the signal path other than at the very input. The amplifier also includes a passive adaptive soft clipping circuit that can be enabled or disabled. The adaptive soft clip circuit prevents the LM3886 amplifier-proper from ever clipping while preserving the amplifier's dynamic headroom. Amplifiers incorporating negative feedback tend to clip sharply, and the soft clip circuit eliminates this behavior.

THD at 1 kHz was first measured at 36 watts to be about 0.002% - quite good. The AUT was then driven to clipping and the chopped tops of the amplifier's sinewave output were visible. In a second trace on the DSO, the ugly clipping distortion waveform was displayed. The THD analyzer read about 3 percent distortion. On the spectrum analyzer, the numerous harmonic distortion components were evident, along with lots of grunge above the noise floor of the analyzer between the harmonics. This was classic hard clipping.

The soft clip circuit was then turned on. The amplifier's sinewave output was now clipped, but in a more civilized, rounded way. The distortion waveform changed, but its amplitude was still about the same. The THD analyzer still read about 3 percent. The spectrum analyzer still showed plenty of harmonic "pickets", but they were mostly odd-order and they declined with increased frequency more so than with hard clipping. Very interestingly, the grunge between the harmonic pickets was now gone - the space between the harmonics was clean, residing at the noise floor of the spectrum analyzer. Probably mis-using an over-abused phrase, could this be analogous to the "silence between the notes"? Is the grunge between the harmonics that is there with hard clipping more dissonant than the larger harmonic pickets themselves? I don't know.

The 19 kHz + 20 kHz twin-tone IM test was then done on the AUT. Once again, it was performed at near full power, and the spectrum analyzer display was clean of any spectral lines other than the 19+20 kHz test signal down to its noise floor of -90 dB. Again, very good performance, especially for a chip amp. The AUT was then driven to visible clipping. Suddenly a cacophony of spectral lines appeared on the spectrum analyzer, spaced apart by the usual 1 kHz interval. Once again, there was quite a bit of "grunge" that appeared between the IM distortion spectral pickets lying well above the analyzer's noise floor. The soft clip circuit was then turned on. Still lots of IM distortion pickets, but once again the grunge between those pickets was gone. Interesting.

One is tempted to conclude, as some others have, that the audibility of distortion depends not only on its magnitude, but also on its nature. The soft clipping does not reduce the amount of distortion (it actually increases it at levels below clipping), but it gives it a more civilized nature. It is notable that tube amplifiers also tend to naturally clip in a way that is softer than that of a solid state amplifier.