[TowerTalk] 1 or 2 dB

[Lux, Jim]

On 5/18/22 5:53 PM, Dave Sublette wrote:
> I've hesitated to chime in here because I thought the problem was being
> discussed very well and perhaps I don't have anything to add.  But I do.
>
>  From December 1983 to April 1989 I operated 160 through 6 meters from
> Kwajalein, Marshall Islands.  The pileups were big and were there every
> time I pressed a key or opened a mic.  I made 83,000 QSOs, filled 19
> logbooks. There were no home computers for me at that time.
>
> One night things were a bit slow on the bands and I had three stations in
> San Diego on 20 meters to keep me company.  All were running comparable
> stations.  KW and tribanders at 70 feet.  They all lived within a couple of
> miles of each other.  So I proposed an experiment.  I had them each say a
> sentence in turn, perhaps no more than 6 seconds.  And keep it going
> around, one sentence after another.  I watched my S-meter.  At any
> given moment, one of them would be 10 to 20 dB stronger than the other
> two.  And it was a different station each time.  We did this for maybe ten
> minutes and the results were as I have described.

This is exactly what the models predict.  It turns out that if you model 
two paths, separated by 2 milliseconds, with random phase, you get 
Rayleigh fading with 10-20 dB fades.  I forget what speed the 
"ionosphere clouds" move at, but it's related to that - think of the 
ionosphere as a very lumpy medium with blobs that are several hundred 
meters in size all moving around.




>
> I learned some things from this that I wish I had known while chasing DX
> from the States before I went to Kwaj.  Primarily, propagation is king.  No
> amount of power will guarantee your signal to be highest at all times.  I
> came to think of propagation as a searchlight, constantly scanning.
> Whichever station is shined on at the moment would be the one that got
> through the pileup. Had I realized this, my anxiety level about working the
> DX over the years would have been much lower. Sooner or later, it will be
> your turn.

I think that's exactly it - on commercial and military links, they just 
factor that in, with ARQ repeats of dropped blocks, etc. And this is 
where diversity really shines (Ralph Wallio, W0RPK was a big advocate of 
diversity).  Because if one path fades, another is likely to be good. 
For hams, it really only works for receive - I don't know how the 
legalities would work transmitting the same signal from two places at 
once.  I guess you could divide your power into two different 
polarizations, otherwise you'd need to have your endpoints far enough 
away that they're decorrelated, and I'm thinking that's on the order of 
several km because OTH HF radar has arrays  on that size - the  JORN 
array in Alice Springs is 2.8km long, ROTHR in Virginia is 2.58 km long.

Clearly, they could build bigger OTH radar arrays if it work better.

But you know, here we are talking 1 dB and 2 dB.  Diversity, done right, 
could get you 10 or 20 dB. The problem is that on Tx, say you 
transmitted two signals - you could wind up fading your own signal if 
they arrive out of phase at the DX.

But at 1kHz different frequencies?




>
> The second thing that occurred to me is that there is no substitute for
> crisp operating technique.  Call only when the DX is listening. Hear the DX
> station before you call.
>
> In any pileup I could hear at least five levels of signals, even when
> everyone one was transmitting at the same time.
>
> Based on all of the above, I would tend to disregard any worry over one or
> to dB.
>
> But, as was pointed out, over the long haul in a contest it evidently makes
> a difference.
>
> There are two other circumstances where 1 or 2 dB makes a difference.  EME
> communications.  We don't strive for that last tenth of a dB noise figure
> for no good reason.
>
> The other circumstance in my mind would be digital modes on a "dead" band.
> It could make a difference of whether or not the report was decoded.

digital modes have the huge advantage that fades are short duration - so 
the average path is there, and using repeats, or long interleave blocks 
works.