Stefan wrote:

> And for some reason, the signal was much stronger last night
> at SQ5BPF. Did you see it? What could be the reason?

I think this is typical of short ranges.

When I model the propagation using ray tracing, it needs the
first dozen or more rays (ground wave, 1-hop, 2-hop, 3-hop and
so on) to produce something like a realistic diurnal.  At some
ranges, 2 or 3 rays have significant contribution for much of
the time amd small shifts of relative phase of the rays can make
large changes to amplitude.   Maybe we are seeing that.

It gets simpler at longer range: the contribution from higher-hop
paths is much reduced.

> i'm changing to:
> Coding 8K19A

OK.

The table shows also that for longer messages - convolutional
payload of 36 bits or more such as 4chars+CRC12 or 3chars+CRC18,
it is always worth using a rate 1/16 code, ie 16K.

8K or 4K codes have symbols twice and four times the duration,
so would only be better if symbol timing was a potential problem,
eg with fast signaling at LF or MF.

You can see also that longer constraint lengths are more
successful with larger payloads.  Eg no point in using more
than K21 with payloads between 18 and 36 bits, and limit to K23
which payloads up to 72 bits.

The convolutional payload is always 6 * number_of_chars + CRC size.

--
Paul Nicholson
--