[akis]

Hi!
Everybody talks about the frequency response of the amps.
And almost everybody means how the voltage gain of the amp varies over a range of frequencies.
But frequency response is not only that. It is also the phase shift of the output signal in comparison to the input signal over a range of frequencies. I have hardly heard anything about this aspect of frequency response. Shoudn’t this phase shift be ideally constant (0 or Pi rad) over the audio frequencies, just like the amplitude of the ouput voltage should ideally be the same at all frequencies? Are phase shifts no so “bad”? What values are acceptable?

 

[Tomo]

Hello,

Phase shift between input signal to output signal means nothing to us. It merely means "all the signal" is shifted a fixed amount in time. This will not change the signal characteristics.

Note phase shift may be frequency dependent. However, the difference in phase shift is significantly small that you may consider it constant.

Tomo

 

[Joobu]

Phase shift can never be 0. This would imply a non-causal system. Even a piece of wire has some delay.

The criteria for zero-distortion in a band of frequencies is linear phase shift. Filters and cutoff frequencies can be designed for linear phase in the audible range (20-20kHz). Non-linear phase shift can be the cause of distortion, overshoot and other characteristics of imperfect reproduction.

 

[akis]

Yes Tomo, that’s what I said. Phase shift should be constant over the audio frequencies, in other words, it should not vary with frequency.
So, I did a spice simulation with the original Szekeres amp (only the bias point was changed). It had a 1uF cap in the input and one 1000uF at the output, bypassed with a 10uF. The phase shift starts at a maximum of +53deg at 10Hz and falls off exponentially until +2deg at 300Hz. After that it stays almost constant 0deg up to 20kHz(-2deg). This was probably due to the small input cap, so I changed it to 10uF, and at 10Hz it is still +30deg. (The choice of the capacitor values is the most common for the ac coupled szekeres amp, i beleive).
Is that ok? Are these considered small values? Where is it “better” to have phase shifts, at low or at high frequencies?

 

[Joobu]

Phase shift doesn't need to be constant, it should be linear. This is because group delay is the derivative of phase shift. Group delay is the measure of how long it takes for a certain frequency to propogate. Therefore, group delay is the important measure, not phase shift. As long as the group delay is constant over your frequency band then the phase response is fine.

In regards to simulations (assuming you're using Orcad): With special Pspice markers you can plot the group delay of the output.