[TowerTalk] Funniest thing I've seen in weeks

[Jim Lux]

At 09:52 AM 7/30/2004 -0500, Jim Brown wrote:
>On Fri, 30 Jul 2004 07:29:54 -0700, Jim Lux wrote:
>
> >2) Where the antenna is physically large enough that you can coherently
> >combine two paths with the signal, while the noise is uncorrelated.  An
> >array with antennas separated by LARGE distances (many wavelengths)
>might
> >achieve this (imagine combining the receive signals from separate antennas
> >in Los Angeles and San Francisco... the noise probably isn't identical)..
> >The improvement in SNR is sqrt(Nantennas) in this case.
>
>What you describe sounds a lot like a diversity receiving system if you 
>assume
>two receivers at each end. But one really big caveat is that you must account
>for the arrival time and polarity differences when you add the two 
>signals!  If
>there happens to be a polarity difference between the two signals, they 
>cancel
>rather than add. Also, the improvement is SNR assumes that they two
>receivers get equal signals from the transmitter. In the real world, there 
>will be
>differences due to fading, so I would expect diversity to be the more 
>dominant
>factor.

Actually, the accounting for timing and polarity, etc. was implied in the 
"coherent combining".  In practice, you wouldn't need anywhere near that 
spacing, and you could use an algorithm like MUSIC or ESPRIT to do the 
adaptive processing.

post detection diversity combining is part way there and is essentially a 
special case of one technique for doing the general problem of combining 
multiple signals to optimize the combined SNR.

Lots of activity in this area in the wireless business these days under the 
name MIMO (Multiple Input Multiple Output) systems.  The idea being that if 
you can coherently combine multiple paths, you can get higher SNRs, rather 
than just choosing the single best path at any given time.  RAKE receivers 
(or tapped delay equalizers, as used in HDTV receivers) are part of the 
solution (giving you time diversity) and you combine that with multiple 
antennas (giving you spatial diversity).  The real challenge is not in the 
receiving, but in figuring out the optimum transmit (i.e. to invert the 
predicted channel characteristics).

For what it's worth, most hams use traditional SISO schemes.  Those of you 
with multiple receive antennas are doing SIMO.


>If we're talking detected audio, it's quite easy to achieve this result 
>with means
>as simple as running the earlier arriving signal through a digital delay 
>line. If
>there are multiple desired signals arriving from multiple directions and
>locations, this delay line will need to assignable to either receiver 
>output, and
>will need to be very easily and quickly adjustable. DSP products that do this
>are readily available off-the-shelf in the pro audio world. Achieving this 
>variable
>delay at RF may be a lot tougher!

Yes and no... Since in the ham world we deal in narrow band signals, one 
could just translate a small chunk of spectrum to baseband, and then you're 
really at audio rates.  If you wanted to, for instance, do 200-300kHz of a 
band at a crack, though, it would be a bit challenging (since, oddly, 
nobody seems to be selling megasample/second A/Ds with 150 dB dynamic range)

Historically, in the VLBI radio astronomy world, they actually do the 
combining off line in non-real time (as in mailing tapes... because, after 
all, FedEx provides data rates and low prices that telcom carriers can only 
dream of.. just a bit of latency, though)


>Jim Brown  K9YC