[TowerTalk] Fw: Feeding single band HF yagis 500+ ft from the shack

[Lux, Jim]

On 5/18/22 7:51 AM, Kim Elmore wrote:
> I don't mistrust contesters that tell me they're certain of this, 
> but... I'm wondering how they *know* that the difference is 1-2 dB on 
> the *receiving* end? If they can increase the smoke by 1-2 dB and all 
> of a sudden make the QSO, how do they *know* that's what did it? HF 
> propagation is funny stuff. I do a LOT of statistics in my job as a 
> research meteorologist/scientist, even though I'm not a formally 
> trained statistician, and this is always my first question when 
> presented with statements like this. Show me the data and how it was 
> analyzed.
>
> I'm pretty sure that there is no such data set in existence and I 
> don't know of a good way to collect one. However, I suspect that the 
> innate variability of ionospherically propagated HF signal strength is 
> far larger than 1-2 dB and that any p-values we'd find at the 1-2 dB 
> thresholds would be pretty large and so deemed statistically 
> insignificant. I suspect there are too many degrees of freedom to ever 
> pin this down.
>
> I have no intention of starting a fight or creating discord. I 
> *deeply* respect Jim's judgement and experience. Even so, I'd love to 
> craft an experiment that would allow us to statistically determine the 
> dB threshold that truly makes a difference on each band. While I'm 
> spitballing, I might as well include different geomagnetic conditions 
> as well.
>
> That said, there's also a good argument for not wasting a dB if you 
> can affordably avoid it. 



The variability of an ionospheric path is way more than 1 dB. But as you 
say, this is really complex to tease out, because on the Rx end, it's 
your signal competing with both near and distant noise, and the other 
signals.

If I'm transmitting in the middle of a thunderstorm - the Rx is seeing 
both my signal, and my local noise over essentially the same 
(scintillating) path.

If I'm trying to break a pileup, then I'm really competing with the 
"other guys" who also have varying paths.   So what you have is multiple 
signals, each with varying power - the distribution isn't gaussian, but 
if you think about it, what you might have is a competition between two 
gaussian variables with slightly different means, but similar variances. 
And getting through depends on the instant where yours happens to be 
bigger than the other.

This kind of problem has been studied a lot in two contexts, where some 
actual math and models might be available: one is detection of targets 
with radar in clutter; the other is jamming and multi signal channels.

Finally, a LOT also depends on the "skill" of the receiver. Some 
receivers (the person or machine) are better at isolating a single 
signal from the interference.