[opendtv] Re: DTT tuner design
- From: Richard Hollandsworth <holl_ands@xxxxxxxxx>
- To: opendtv@xxxxxxxxxxxxx
- Date: Sat, 7 Jul 2007 13:54:20 -0700 (PDT)
Yes, multipath can (sometimes) be a GOOD THING.
When performing lab tests with the Brazil Ensembles, et. al., the C/N is
established
for one and only one path. Additional paths are then turned on one at a time to
ensure that they all have the desired C/N (SAME if equal strength case).
With uncorrelated noise on each path, total C/N actually improves when the
signals
are in-phase....which is the job of the adaptive correlator.
This explains why the reported C/N values under simulated multipath conditions
are less than for just a single path.
In HF communications, adaptive equalizers coherently combine different
propagation paths (e.g. F1 and F2) so that when the signal fades away
on one path, the other path is usually still usable.
In microwave communications, a similar effect occurs with antenna space
diversity....
holl_ands
======================================
Albert Manfredi <bert22306@xxxxxxxxxxx> wrote: Dale Kelly wrote:
>I misspoke, either modulation system requires increased C/N
>to handle multipath.
I this strictly true? I don't think so. And I believe that's why the Chinese
standard adopted a training sequence in the GI (same one they use in their
single carrier mode), rather than using the end of the previous symbol, as
standard COFDM does. And that's also my bet for DVB-T2. (Just speculation.)
The ideal equalizer should not require a higher C/N to twist multipath
distortion back into shape.
Here is an example. I'm assuming here that the numbers provided are C/(N+I),
and that when this value is less than 15 dB, what you're actually seeing is
a (C+I)/N that is close to 15 dB.
(I also believe that eventually we'll be seeing (C+I)/N better than 15 dB,
when the combined trellis and R-S FEC schemes are more cleverly exploited.)
These are C/N numbers that apply to the Samsung Gemini chip.
No multipath 15.2 dB
Brazil A 15.0
Brazil B 17.0
Brazil C 13.5
Brazil D 13.8
Brazil E 19.9
Compare this to the CSA values for 64-QAM and 3/4 FEC (same spectral
efficiency if GI is 1/16).
No multipath is 18 dB, Ricean is 18.6 dB, and their Rayleigh, which is less
strenuous than Brazil E, is 21.7 dB. (Brazil E has 3 dB echoes, vs two.)
So basically, the planning factors used by the CSA do assume a deficit of ~3
dB or more in C/N margin, compared with a well executed equalizer demod. And
that's reflected in the field strength minima. If good demods are used, the
FCC and CSA planning factors should provide very close to the same coverage,
is my thinking.
Bert
_________________________________________________________________
http://liveearth.msn.com
----------------------------------------------------------------------
You can UNSUBSCRIBE from the OpenDTV list in two ways:
- Using the UNSUBSCRIBE command in your user configuration settings at
FreeLists.org
- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word
unsubscribe in the subject line.
---------------------------------
Ready for the edge of your seat? Check out tonight's top picks on Yahoo! TV.
Other related posts:
