Newbie question: is it safe to assume that a flat THD profile over the typical 20HZ-20kHZ range is sonically preferable to one that rises at high frequencies? I'm referring to the flat versus rising profiles shown below. These are of 2 commercial amps (with thanks to Stereophile.com).
Follow up question: what causes the rising distortion profile at upper frequencies and how is it avoided? Again, I have to assume a flat distortion profile is preferable sonically but it seems to be rare in typical class AB amplifiers
Follow up question: what causes the rising distortion profile at upper frequencies and how is it avoided? Again, I have to assume a flat distortion profile is preferable sonically but it seems to be rare in typical class AB amplifiers
Attachments
ALL NFB amps generally. No amp has gain to infinity. They all fall-off sometime. With small parts this might be very supersonic (but it does rise somewhere). When you must use BIG (slow) parts to make power yet use NFB to control distortion, you must pick your compromise.
Because speech/music power falls-off in the upper octaves there is some excuse for NFB falling and THD rising in the upper octaves.
A flat distortion profile is not, in itself, of any advantage. What is important is the distortion level in a frequency band.
As an example, what is to be preferred:
Amp A: THD @ 1kHz = 0.001%, THD @ 20kHz = 0.05%
Amp B: THD @ 1kHz = 0.05%, THD @ 20kHz = 0.05%.
For me the choice is a no-brainer: amp A is to be preferred, although amp B has a flat THD curve.
Jan
Normally its the dominant pole limiting the open loop gain at higher frequencies, which is itself constrained by the speed of the output devices or the VAS (which are usually the limiting factor). FET outputs can fare a lot better than BJT due the inherently higher bandwidth. Current feedback amps have an intrisically better bandwidth too as the compensation requirements for stability are less severe.
With less open loop gain at HF there's less linearization to be had at higher frequencies, so distortions become more apparent.
Also with BJT CFP outputs switching distortion in the output stage rises precipitously with frequency from perhaps 5 to 10kHz and above.