Hey guys! I have been trying to design my own simple small signal circuits for using in my condenser microphone. I am pretty new to this, but I was able to get some designs that perform well in regards to THD and Bandwidth. However, when I perform an AC analysis, the bode plot that is displayed also shows the phase relationship within a range of frequencies. When I do a search to understand how much phase shift (or what kind of phase shift) I should see, I can't find an answer. There are tons of videos and blogs that show how to measure the phase shift, but none really saying "how much" I should allow in my design. I saw one guy on reddit saying he allows no more than 10 degree shift from zero. So how much phase shift should be allowed in a well designed circuit? Also, will I be able to hear this as long as the shift is distributed evenly throughout the whole "audible" frequency range? I will post some pics of my bode plot from qspice, and maybe somebody can give some feedback on it.
Here is one where the shift seems to be even up to around 25Khz
Question 1 - is this amount of phase shift considered acceptable in (hifi standards)? Or at least professional standards?
Question 2 - is this going to be audible?
Here is another plot with a more complicated situation going on (180 degrees)
This one has some corner frequencies from some inconvienient bypass caps I had to use.
Question - this shift in the low frequencies, is it normal? Will this cause the low end to be cancelled out or distorted?
Thats it, thanks for reading, any feedback would be really helpful.
Here is one where the shift seems to be even up to around 25Khz
Question 1 - is this amount of phase shift considered acceptable in (hifi standards)? Or at least professional standards?
Question 2 - is this going to be audible?
Here is another plot with a more complicated situation going on (180 degrees)
This one has some corner frequencies from some inconvienient bypass caps I had to use.
Question - this shift in the low frequencies, is it normal? Will this cause the low end to be cancelled out or distorted?
Thats it, thanks for reading, any feedback would be really helpful.
The ideal case is linear phase shift (proportional to frequency), because this is equivalent to a pure time delay,
and does not alter the waveform shape.
More generally, phase shift is linear distortion and cannot create frequencies not present in the input signal,
but it does alter the waveform shape if the phase shift itself is nonlinear. For example, a single pole filter
has an arctangent phase shift function. This is the same phase curve as in your first example.
For less than 5° of phase shift, the pole must be more than a factor of ten from the frequencies of interest.
Your first example of a 1 MHz low pass filter demonstrates this, and has a negative phase shift of about -5° at 100kHz.
The phase curve is quasi-linear for signals with a bandwidth well under 1MHz, and the signal wave shape will be preserved.
Your second example is more complex, because it has two poles (band-pass filter). Ignore the polarity inversion,
and consider only the low frequency pole at about 5Hz. Then the positive phase shift is about +5° at 50 Hz.
The filter is still causal despite the positive phase shift, since this is a sinusoidal steady state plot, and the initial transient
has decayed away. Some can hear this sort of low frequency linear distortion and some cannot, so try it for yourself.
and does not alter the waveform shape.
More generally, phase shift is linear distortion and cannot create frequencies not present in the input signal,
but it does alter the waveform shape if the phase shift itself is nonlinear. For example, a single pole filter
has an arctangent phase shift function. This is the same phase curve as in your first example.
For less than 5° of phase shift, the pole must be more than a factor of ten from the frequencies of interest.
Your first example of a 1 MHz low pass filter demonstrates this, and has a negative phase shift of about -5° at 100kHz.
The phase curve is quasi-linear for signals with a bandwidth well under 1MHz, and the signal wave shape will be preserved.
Your second example is more complex, because it has two poles (band-pass filter). Ignore the polarity inversion,
and consider only the low frequency pole at about 5Hz. Then the positive phase shift is about +5° at 50 Hz.
The filter is still causal despite the positive phase shift, since this is a sinusoidal steady state plot, and the initial transient
has decayed away. Some can hear this sort of low frequency linear distortion and some cannot, so try it for yourself.
The second example appears to invert polarity: the phase shift is around 180 degrees instead of around 0 degrees. That can be audible in some rare cases, although it easily gets masked by reverberation.
(Ignoring 180 shift due to inversion) phase shift and frequency response are both entirely dictated by the positions of the poles and zeroes (for a time-invariant linear system). For a digital system there is also an overall delay which acts as a phase shift proportional to frequency in addition to that from the poles and zeroes. The maths of these dependencies on poles and zeroes is subtly different for analog and digital systems.