Kaben Wireless Silicon

Low Pass Filter

The Magnitude response of a Low Pass SIF filter is shown below. As can be seen, the filter has a 2.7 dB passband ripple (which is dependent on the number of taps and their weights), a SAW like “brick wall” roll off from its corner frequency, and an adjacent channel stop band of -30 dB. Since the SIF is based on a discrete time, Finite Impulse Response (FIR) operation, Image responses occur around the Clock frequency. Interestingly, these Image responses are attenuated by a Sinc function resulting from a time windowing operation in the SIF filter. Consequently, unlike the Image responses in a digital FIR filter, the Image responses in the SIF filter are attenuated by 18 and 20 dB respectively.

Superimposed on the Magnitude response are the Input Clock frequency and the Sampling Clock frequency. It is apparent that further attenuation of the stop band response is present around the Sampling Frequency. This fact can be used advantageously to “notch out” a particularly strong Blocking signal near the band edge of the filter. 

Image Reject Band Pass Filter

The Magnitude response of an Image Reject, Band Pass SIF filter is presented next. Here, the pass band is shown (at 90 MHz) along with an LO frequency (at 100 MHz) which down-converts (down samples) the pass band signal. One result of this down-conversion is that any blocking signal located at the Image frequency (of the desired pass band signal with respect to the LO frequency) will be superimposed onto the desired signal, causing in-band interference.

Here the sampling clock in the SIF has been selected to place a notch onto this Image frequency. The resulting Magnitude response is then a pass band filter with a close in Image Reject notch.

The Sinc roll-off response, resulting from the time windowing operation in the SIF filter, is also superimposed on the Magnitude response.

Bluetooth Filter having Dual Modes of Operation

The Magnitude response of a Bluetooth Filter is shown below. Here, two modes of operation are superimposed, one being a low power mode (Blue trace), and the other a high performance mode (Green trace). In benign interference environments, only a relatively light adjacent channel filtering (-30 dB) is needed, and so power can be saved by powering-down some of the Taps in the SIF filter. On the other hand, in hostile interference environments with strong adjacent channel blockers, robust performance can be achieved by powering-up all of the SIF filter Taps, and dropping the adjacent channel filtering level (-45 dB).

3GPP

This is an example of a simple 3GPP filter where the zeros are fixed by sampling rate. The desired signal is -89 dBm. With this simple filter there is 74 dB headroom required for blockers.

If we now add to the above filter some arbitrarily placed zeros, we can achieve a 29 dB improvement. This means that we now require only 45 dB headroom for blockers. This 29 dB improvement will save 4.8 bits in the subsequent A/D.