ALL PASS Filter-Why the name filter [closed]

Question

As per my knowledge the purpose of All Pass filter is to add phase shift (delay) to the response of the circuit. The amplitude of an allpass is unity for all frequencies.Then why it is called as a filter??

Answer 1

It is designed using filter design techniques and implemented using filter components. Though it has ideally no effect on the signal amplitudes (hence all-pass) it affects phase, or group delay.

Combining an all-pass filtered signal with its original input will modify the amplitude response - e.g. where gain = 1 and phase = 180 degrees, adding input and output will produce a notch in the frequency response, so you could argue that the frequency selective aspect of a filter is latent, even if not directly visible.

More importantly, once you have designed a low pass filter with specific characteristics, there are techniques to transform it into a high-pass, bandpass, bandstop, or all-pass network with closely related properties. At which point it appears logical and consistent to label the all-pass network a "filter" like the others, even though it appears not to be frequency selective.

However if you are still not convinced, "all-pass network" is probably an acceptable alternative term in most uses.

Answer 2

The term "all pass" is something of a misnomer. An all pass filter has a pass band in much the same way a low pass filter does. These two plots are from this Texas Instruments App Note.

Notice the similarities in the plots. The low pass filter has a pass band defined by its magnitude response, while the all pass filter's pass band is defined by its group delay response, and it is important to understand the signicance of the all pass filters passband.

We typically use a low pass filter to control the maximum frequency content of a signal. When using an all pass filter however, we need to ensure that the signal applied to the filter is band limited to the it's pass band.

Here I show two examples of passing a pulse through an "all pass" filter. The first example is exaggerated to make a point. Because the square pulse has significant frequency content well beyond the filter's pass band (i.e. where filter's group delay is constant), the all pass filter destroys the signal.

In the second example, I pre-filter the square wave with a 10 kHz low pass and use an all pass filter with constant group delay out to 4 kHz. Since the input signal contains much less energy outside the "all pass" filter's pass band, much less distortion takes place.

It is worth noting that the Bessel low pass is considered a good alternative to the all pass filter because of its group delay characteristics. When using a Bessel low pass, high frequency content will be attenuated thus eliminating the distortion that an all pass filter can cause. The all pass is easier to design for a specific time delay however.

Also, in a manner similar to the maxiamally flat magnitude response of the Butterworth, Budak showed that a modified Bessel polynomial will generate a maximally flat group delay all pass filter.