Talk:Cascaded integrator–comb filter

From Wikipedia, the free encyclopedia
Jump to: navigation, search

Changes[edit]

I made some changes. I may do some more tweaking. Hope I'm not stepping on anybody's feet. TCMike 03:38, 27 December 2006 (UTC)

Comparison to FIR[edit]

I changed the comparison to FIR filters. CIC filters are a class of FIR filters, but FIR filters have a much wider array of applications. I am not sure this comparison is useful. But to the extent that it is, I think we should also consider comparing to IIR filters. But of course that is a can of worms. TCMike 03:38, 27 December 2006 (UTC)

I agree that these comparisons are not necessarily useful, as we're not really comparing like with like. A CIC is really just an implementation of an FIR and up/down-sampler; FIR is really a mathematical property. I think we should restrict the comparison to other implementations of interpolators/decimators, i.e. pros and cons of:
  • standard (non-)recursive filter + up/down-sampler
  • polyphase
  • CIC
  • CIC cascaded with (non-)recursive filter
Oli Filth 22:27, 31 December 2006 (UTC)
Okay, I think I will work on some frequency response graphs and whatnot. Also, was your objection that the CIC does not have frequency NULLs, or that FIRs can?
TCMike 01:13, 2 January 2007 (UTC)
FIRs can have frequency nulls. e.g.:
\ y[n] = x[n] + x[n-1] + x[n-2]
has nulls at \Omega = \pm 2\pi/3. Oli Filth 07:21, 2 January 2007 (UTC)
The comparisons as such are useful, but the article should probably be reworded to state that it is comparing a conventional or direct-form FIR implementation of equivalent length to a CIC implementation. (I agree with the comparison, but is it original research?) Alinja 16:44, 3 January 2007 (UTC)
What I'm saying is that we don't want to compare the CIC to FIRs in general (as FIRs have far wider applications than just interpolation/decimation). The only meaningful comparison is against a "naive" implementation of interpolation/decimation, i.e. standard filter (which may or may not be FIR) + up/down-sampler.
(What do you mean by original research?) Oli Filth 17:04, 3 January 2007 (UTC)
Exactly. (Wikipedia should not be used to publish your new research, but I guess the comparison issue is fairly common knowlegde so it can be left there) Alinja 17:25, 3 January 2007 (UTC)

Gain is between 0.5 and 1.0, and number of bits must be adjusted, for changes in decimation/interpolation ratios[edit]

# The integrator and comb structure are independent of rate changes (there is no need to reproject the filter on decimation/interpolation rate change).

is partially correct. However, the number of bits at each stage of the filter is affected by the decimation or interpolation rate, as well as the gain of the filter. The gain of the filter for a decimation or interpolation ratio in the form of 2N is 1.0, and for decimation or interpolation ratios not equal to 2N, the gain is between 0.5 and 1.0. Ra2007 (talk) 16:37, 17 December 2007 (UTC)

FIR comparison[edit]

The Hogenauer paper abstract states:

Consequently, comparisons to FIR filters was made in the abstract of the paper that introduced the CIC filter. Ra2007 (talk) 16:37, 17 December 2007 (UTC)

Picture is incorrect[edit]

The picture shows the combs with z-1 delays, when actually they are z-M. -- wjl (talk) 16:19, 14 July 2010 (UTC)

Pronunciation[edit]

is 'CIC' it pronounced as 'sick', or as 'kick' ? —Preceding unsigned comment added by 199.203.84.253 (talk) 12:22, 29 December 2010 (UTC)