Re: AIC23 _ High Pass Filter_BIG_FUNDAMENTAL_PROBLEM

Akash_DSP wrote:
Akash_DSP wrote:
Akash_DSP wrote:
Akash_DSP wrote:

...

Hey Jerry,

thanks for the reply.

Yes, the periodic vairation is sinusoidal. This means that if I
give
an
audio input with a frequenct spectrum (<20 KHz), frequencies that
will
have
come out of the D/A will have different delays.
Different delays for different frequencies, of different delays for
the
same frequency at different times?

...

Jerry
--
Engineering is the art of making what you want from things you can
get.
Hey Jerry,

it is different delays for different frequencies at the same time.
It seems that you have discovered the dispersion of the anti-alias filter. That is normal.

Jerry
--
Engineering is the art of making what you want from things you can
get.
Hey Jerry,

you mean to say that if I give an audio input and observe at the
output
that each frequency comes out with a different delay. That is normal?

If you program your C6416 or c6713 do you see similar behaviour?
In order to sample successfully, there must be no components of the sampled signal as high as half the sampling frequency. The use of an anti-alias low-pass filter in the signal path before the sampler is so nearly universal that many codecs have a filter built in. Being analog, the a-a filter will affect the phase. The designer is responsible for correcting any phase distortion that interferes with proper the function

of his design.

Examining the data *** should show you how much phase shift to expect.

Once you plot the filter's phase/frequency characteristic, you will know

if what you see is normal.

Oversampling is an effective way to minimize phase shift in the band of actual interest. After sampling, the signal can be further filtered digitally, then decimated.

All real design requires balancing competing objectives; that is an art.

Jerry
--
Engineering is the art of making what you want from things you can get.
???????????????????????????????????????????????????????????????????????


Hey Jerry,

thanks for the descriptive mail. I really appreciate it.

I understand aliasing thats why in the first message I wrote that my
sampling frequeny is 44.1KHz and I am looking at frequency spectrum 0
-20Khz (Following the Nyquist Criteria). I also understand the use of anti
aliasing filter.

But I am seeing phase delay just moving +- 1ooHz . e.g. there is a big
phase difference b/w the input frequencies 16.0 KHz and 16.10 Khz. 16.0 Khz
and 17.7Khz will then have the same delay. As it is periodic delay.

I am unable to find the reason for this as this is what we normally dont
see in a DSP chip.

You're right: that isn't normal. The problem might be with the equipment used to measure the phase or the way you use it. I can more readily imagine that than I can something funny happening digitally.

Do you have a two-channel oscilloscope available? If you do, sync on the codec's input and observe input and output simultaneously as you slowly vary the frequency.

You can also use a single-channel scope. Apply the input as the horizontal signal and the output as the vertical. You will see a straight line when the signals are in phase and an ellipse when they're not. You can actually measure the phase angle with this setup, but I won't get into those details unless someone asks.

Jerry
--
Engineering is the art of making what you want from things you can get.
.

Quantcast