The improved distortion extraction circuit with op-amps and trimmers was used to obtain these oscilloscope traces. The vertical scales are not all the same, but comparison to the noise component gives an indication of the levels. The distortion levels fall at the lower signal level in all cases, and disappear under the noise as the signal is reduced further showing that small signal linearity is excellent. Improving the distortion extraction method of course makes the distortion look worse than with the poorer measurement techniques used to test my previous designs, which added more noise. The amplifier noise level is very low and far below audibility even with an ear held against the speaker, and most of the noise seen is from the amplifier rather than the test equipment. The distortion at the 300mV level now looks a little spikey, as expected for a class-B design, but the level is low enough for this to be unimportant. Even the worst of these distortion waves, the 5kHz at 300mV, when analysed with a pc spectrum analyser, included 2nd harmonic at -94dB and 3rd at -99dB. These are a little above previous measurements because the input stage emitter resistor is now 1R5 instead of 1R to improve stability margin in return for a small increase in distortion. Also the 1n and 390p capacitors were reduced to 560p and 220p to improve high frequency measurement accuracy, so more higher order distortion components are revealed above the audio frequency range where they are not important. Although I checked the harmonic distortion at 20kHz and found it to be around 0.01% this specification is of little interest, all harmonics being at inaudible frequencies. Using 20kHz and 19kHz inputs revealed intermodulation at 1kHz around -90dB, which is more relevant to music reproduction.
1kHz at 300mV and 100mV input:
3kHz at 300mV and 100mV input:
5kHz at 300mV and 100mV input:
HOME.
