> An extra 1 error per couple of seconds may bother guys like me, but
> probably won't cause a contester to worry -- most exchanges are fewer
> than 100 characters long.
>
> I venture to add that a contester can probably tolerate an extra
> error rate of even 1 error per 50 characters -- in which case, an
> ideal 180 Hz filter will work.
However, since that error is not predictable and certainly does *not*
occur after 50 "good" characters, one is taking a big chance that the
error changes ("busts") a call. Just a few busted calls (or busted
exchanges) - with the potential penalties and loss of multipliers -
can be a disaster for a contester.
Chen, it would help if we knew how much group delay can be tolerated
in the filters. How much unpredictable delay is there in "the ether"
and how much might be tolerated in a sharp sided filter before the
the demodulation function falls "over the cliff"? In addition, how
much differential gain ("non-flat" passband) can be tolerated -
particularly when the very narrow filter is slightly mis-tuned or
asymmetrical? What does the built-in "selective fading" do to the
demodulation function, particularly if the software lacks effective
ATC?
I don't think your demodulation bandwidth functions consider either
of those real world error contributions from a "too narrow" roofing
filter.
73,
... Joe, W4TV
On 8/25/2013 7:17 PM, Kok Chen wrote:
On Aug 25, 2013, at 2:20 PM, Bill Turner wrote:
Long experience has shown that around 300-350 Hz on the typical ham
receiver works. Less does not.
To paraphrase the other Bill (Clinton), that depends on what you mean
by "works."
If you have built and adjusted demodulators before, you know that
they don't suddenly shut down. They simply degrade.
By how much do they degrade? Again, I recommend that you look
*very, very* carefully (be sure to look at the scales, not just the
shape of the curve) at Figure 2.2 here of what cascading an extra
filter does to RTTY error rates:
http://www.w7ay.net/site/Technical/RTTY%20Transmit%20Filters/index.html
Notice that the error doubles from an ideal 300 Hz filter to an
ideal 210 Hz filter.
Doubling error is never a good thing (especially if you are providing
cell phone service), but look at the scale again.
The ISI from cascading an extra filter (e.g., the filter in your
superhet) adds an extra 1 error per 250 characters for the 300 Hz
filter, and adding 1 error per 100 characters for an ideal 200 Hz
filter.
An extra 1 error per couple of seconds may bother guys like me, but
probably won't cause a contester to worry -- most exchanges are fewer
than 100 characters long.
I venture to add that a contester can probably tolerate an extra
error rate of even 1 error per 50 characters -- in which case, an
ideal 180 Hz filter will work.
So, any practical filter that can pass 180 Hz to 200 Hz worth of
clean passband will probably work more than sufficiently for
contesting. If you want to dig out the weak DXpedition, you will
need every little bit of reduced error as you can, since SNR is not
in your favor to start with, and that is where you will need to widen
the bandwidth.
A filter is flat and has less than 1 ms of group delay for 180 Hz is
probably good enough for contesting.
Us modem designers are trying to squeeze every drop of blood from our
modems (I know that Dave W1HKJ, Stefan DO2SMF and David G3YYD are
constantly trying eek out an extra percent or two fewer errors with
their modems). So we absolutely care about 2x type errors (in the
end that will benefit everybody. But that 2x of errors from ISI is
not going to bother most contesters.
73 Chen, W7AY
_______________________________________________ RTTY mailing list
RTTY@contesting.com
http://lists.contesting.com/mailman/listinfo/rtty
_______________________________________________
RTTY mailing list
RTTY@contesting.com
http://lists.contesting.com/mailman/listinfo/rtty
|