The color difference channels are modulated in quadrature, but this has
nothing to do with the comb filter.
The comb filter works exactly as explained in my other post. Look at
this diagram:
http://www-ccrma.stanford.edu/~jos/p...b_Filters.html
The Z raised to the minus M is a generic delay of M samples. Basically
you sum the real time signal with a delayed signal. It's that simple.
matt weber wrote:
On Sat, 03 Dec 2005 18:20:58 GMT, "FDR"
wrote:
"Carl" wrote in message
...
I was looking at a receiver catalog, and some of the receivers were
advertised as containing a "DSP comb filter" which could be turned on
when needed.
Can anyone tell me what a comb filter does and when it would be used?
It's supposed to clean up the signal. Basically, the frequency response has
many peaks and valleys that together resemble a hair comb.
The Comb filter was developed for Television. The colour information
in NTSC is transmitted in 3 parts. A B&W receiver only decodes one
part. The other two called Chromiance and Luminance are not entirely
seperated in the frequency domain, but are in the time domain. In
lower cost Color TV's, there is simply a 'hard' filter and the
chromiance and lumiance signals are processed as though they were
fully seperated in the frequency domain, resulting in loss of colour
detail and resolution. A comb filter performs the decoding effectively
in the time domain, where full seperation is possible, resulting in
higher detail and resolution.
An example of such of a system is AM stereo, and some Modems (Milgo
4800 baud comes to mind). It is possible to generate independent
sidebands, so a Double Sideband, Supressed Carrier, or Un-Suppressed
carrier (AM stereo) overlap in the frequency domain, but the carriers
used to produce them in the balanced modulator are at quadrature
(i.e. one sideband was built with Sin (wt) and the other with
Cos(wt)), the sidebands will also be a quadrature. This is in fact
how the Telephone company used to transmit long distance calls.
If you use an envelope detector with AM stereo you get L+R, however
if you use two product detectors that use phase locked, and 90
degrees out of phase carriers, it is possible to get for example only
Left Channel out of one, and only Right Channel out of the other
detector.
The Intergral over time (power in this case) of sin(wt)cos(wt) dt will
always be zero.
A comb filter generally involves some sort of phase locked, synhronous
detection even if carried out digitally.