[BC] Listening t ests... how valid?

[Michael Bergman]

> Let consumers compare typical levels of analog noise to the BER 
> dropout and/or blends back to the noisy signal that caused the BER 
> dropout, and not some sterile imaginary-world lab example of how 
> things would be if each radio set had a shielded coax connection to a 
> pure signal source, and I suspect the IBUZ train would never leave 
> the station...

I think you're missing one key element of a digital delivery system.
Because of interleaving and FEC, inside the coverage area, the digital
delivery closely approaches the ideal situation in the testing.  

That is, you're right that the raw bit delivery suffers from the problems
you mention.  However, FEC (forward error correction) and interleaving clean
up short burst errors.  Long enough errors still happen, but inside the
coverage area you will find them to be relatively rare.  Check FEC and
Interleaving on wiki for more info.

Next, all codecs I'm familiar with have error concealment algorithms.  A
short hit that gets through the FEC and interleaver will be smoothed over,
similar to copying lines or frames in NTSC error concealment.

You're absolutely right that there will also be times, particularly at the
edge of coverage, when there are blends and mutes.  If you're driving
radially out of town on a commute, expect the EOC section to be on the rough
order of 10km in length.  Before hitting the EOC section, you will see the
performance of the codec to be largely as it would in the ideal testing.
For WETA on the Tomorrow Radio testing, that was about 40km of driving
before EOC (again, going radially out of town).

Bottom line: the listening test represents a big part of the real-world
experience.  Please take a look at the Tomorrow Radio test report to get a
feel for the area where this is true--the large green route areas should act
like the listening tests.  The shorter segments where there is a mix of
green and red dots shows the EOC region.

When someone gets into the EOC region, the receiver has hysteresis to
prevent fast blending back and forth.  IIRC, we disabled that hysteresis to
force the actual digital performance, in that testing.   So the Tomorrow
Radio testing shows more blends than a real receiver would present to the
audio codec--it would drop to analog more quickly and stay there, rather
than stressing the codec as you mention.

Intuitively, you're right that there could be such a problem as you
describe.  Recognizing this, the designers put these tools in the technology
to deal with it, and this is true of IBOC, Sirius, XM and E-147.

> Let's stop  ...  you have vested interest in the switch to digital; I 
> do not.  I'm a listener and lover of radio, not a publicly-traded 
> manufacturer hoping to capitalize on IBUZ in order to prop up a 
> marginal stock price, or even a privately-held manufacturer hoping to 
> bloat the books and get rich with a hype-based IPO.  

This is unfair to Bob, who is one of the people on this forum who posts
extremely accurate technical information.  I think I've seen all his posts
on this topic, and I've seen only excellent engineering discussion from his
side.  He knows what he's talking about and I hope he keeps posting.

Mike Bergman