0
\$\begingroup\$

I came across the following amplifier for an audio ADC input stage in an application note for the CSxxxx codecs (appnote, figure 2. Codec datasheet. VQ is approximately 2.7V for biasing.):

enter image description here

I did an AC simulation of this circuit. These are magnitude and phase (OUT+ and OUT- are AIN+/AIN-, respectively): enter image description here

I was wondering about the importance of linear phase in this situation. On one hand, I learned that linear phase is important to minimize distortions due to the different group delay. On the other hand, the phase difference is very small for frequencies up to 20 kHz (relevant for audio) leading to small delays which will probably not be noticeable. Is this reasoning correct?

Close-up of the phase vs frequency from 20 Hz to 10 kHz on a linear scale:

enter image description here

Will a Bessel filter with a highly linear phase like the one shown at the beginning of this document make more sense as an input stage?

\$\endgroup\$
2
  • \$\begingroup\$ It depends on what your application requires more: linear phase, or attenuation. You can use Cauer/elliptic for maximum attenuation and then equalize it with an allpass for a relatively flat group delay (within prescribed ripple) but it will not be a flat group delay like Bessel. OTOH, you can also use a Bessel with zeroes in the stopband, which will only affect the phase after the zeroes, but the passband will still be Bessel (lazy droop). Or anything in-between. Flip a coin, throw a dice, etc. \$\endgroup\$
    – a concerned citizen
    Commented Sep 16, 2022 at 9:30
  • \$\begingroup\$ Also, that filter is not a linear phase: the peaking in the group delay is huge but, it is after ~100 kHz. Still, for audio it's just fine (especially the phase stays below 1 degree @20 kHz). \$\endgroup\$
    – a concerned citizen
    Commented Sep 16, 2022 at 9:37

2 Answers 2

Reset to default
1
\$\begingroup\$

On one hand, I learned that linear phase is important to minimize distortions due to the different group delay

Yes, important for applications where signal-shape needs to be maintained such as high-speed transducer monitoring in control systems but, fairly irrelevant for audio.

On the other hand, the phase difference is very small for frequencies up to 20 kHz (relevant for audio) leading to small delays which will probably not be noticeable.

Think about what tone controls do; they mess around with the phase of some signals in the spectrum big-time. Of course they increase/reduce amplitude in certain parts of the spectrum too but, is it a disaster for audio? No.

The tiny bit of phase-shift you see in your top graph (and that's at 100 kHz) is irrelevant.

Will a Bessel filter with a highly linear phase like the one shown at the beginning of this document make more sense as an input stage?

  • Sensible Audio = no
  • Audiophiles = an absolute tragedy if not used
  • Signal processing = maybe to yes
  • Radio modulation and demodulation = yes
  • Control system = possibly to yes
  • Any general type of filtering - normally no
\$\endgroup\$
0
\$\begingroup\$

The filter presented in the datasheet is should be good enough.

The ADC is sampled at 6.144 MHz so you only need to filter extremely high frequencies away, like the datasheet says.

The filter is there also for reason of providing low enough impedance and high enough current drive to the ADC input.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.