Documente Academic
Documente Profesional
Documente Cultură
Cascaded integrator-comb
(CIC) digital filters are computa-
tionally efficient implementations
of narrowband lowpass filters and
are often embedded in hardware
implementations of decimation
and interpolation in modern
communications systems. CIC
filters were introduced to the Figure 1: CIC filter applications
signal-processing community,
by Eugene Hogenauer, more filters, I’ll introduce their struc- Because their frequency- A crucial bonus in using CIC
than two decades ago, but their ture and behaviour, present the magnitude-response envelopes filters, and a characteristic that
application possibilities have frequency-domain performance are sin(x)/x-like, CIC filters are makes them popular in hardware
grown in recent years.1 Im- of CIC filters, and discuss sev- typically either followed or pre- devices, is that they require no
provements in chip technology, eral important practical issues in ceded by higher performance multiplication. The arithmetic
the increased use of poly-phase building these filters. linear-phase lowpass tapped- needed to implement these digi-
filtering techniques, advances delay-line FIR filters whose tasks tal filters is strictly additions and
in delta-sigma converter imple- CIC filter applications are to compensate for the CIC subtractions only. With that said,
mentations, and the significant CIC filters are well-suited for anti- filter’s non-flat passband. That let’s see how CIC filters operate.
growth in wireless communi- aliasing filtering prior to decima- cascaded-filter architecture has
cations have all spurred much tion (sample-rate reduction), as valuable benefits. For example, Recursive running-sum filter
interest in CIC filters. shown in Figure 1a and for anti- with decimation, you can greatly CIC filters originate from the no-
While the behaviour and im- imaging filtering for interpolated reduce computational complex- tion of a recursive running-sum
plementation of these filters isn’t signals (sample-rate increase) as ity of narrowband lowpass filter- filter, which is itself an efficient
complicated, their coverage has in Figure 1b. Both applications are ing compared with if you’d used form of a nonrecursive moving
been scarce in the literature of associated with very high—data- a single lowpass finite impulse averager. Recall the standard D-
embedded systems. This article rate filtering, such as hardware response (FIR) filter. In addition, point moving-average process in
attempts to augment the body quadrature modulation and de- the follow-on FIR filter operates Figure 2a. There we see that D-1
of literature for embedded sys- modulation in modern wireless at reduced clock rates minimising summations (plus one multiply
tems engineers. After describ- systems and delta-sigma A/D and power consumption in high-speed by 1/D) are necessary to compute
ing a few applications for CIC D/A converters. hardware applications. the averager output y(n).
Equation 9
Equation 10
Figure 8: 3rd-order CIC decimation filter structure, and magnitude response before decimation when D = R = 8 The price we pay for improved
anti-alias attenuation is additional
To investigate a CIC filter’s fre- Those B-width shaded spec- on the bandwidth B—the smaller hardware adders and increased
quency-domain behaviour in tral bands centred about mul- B is, the lower the aliased energy CIC filter passband droop. An ad-
more detail, Figure 6a shows the tiples of ƒs,in/R in Figure 6a will after decimation. ditional penalty of increased filter
frequency magnitude response alias directly into our desired Figure 5b shows a CIC filter order comes from the gain of the
of a D = 8 CIC filter prior to passband after decimation by R used for interpolation where filter, which is exponential with
decimation. The spectral band, = 8 as shown in Figure 6b. Notice the ↑R symbol means insert R-1 the order. Because CIC filters gen-
of width B, centred at 0Hz is the how the largest aliased spectral zeros between each x(n) sample, erally must work with full preci-
desired passband of the filter. component, in this example, is yielding a y(n) output sample rate sion to remain stable, the number
A key aspect of CIC filters is the roughly 16dB below the peak of of ƒs,out = Rƒs,in. (In this CIC filter of bits in the adders is Mlog2(D),
spectral folding that takes place the band of interest. Of course discussion, interpolation is de- which means a large data word-
due to decimation. the aliased power levels depend fined as zeros-insertion followed width penalty for higher order