Topband: Noise, DSP and Receivers

[jljarvis]

I'm reading a really great history book, called 
"Guns, Germs and Steel".  So please forgive this 
note's somewhat overambitious title.

Some observations on the various topics...without
reference to the initiator of last-comment, in 
the interest of speed.

1)  The human brain is a wonderful signal processor.
In order to work properly, it needs wideband input.
This is why ssb filters are sometimes better for 
copying weak cw.  The important thing here is, the
brain can sort out the differences between gaussian
distributed white noise, and poisson distributed impulse
noise, and treat each separately.
	Narrow filters tend to distort the waveform of the
gaussian noise, and make it harder for us to 'average out'.
	Credit for this observation goes to W2VJN at INRAD,
by the way.  But it's very much consistent with my
S/NR experience with scientific instruments, and with
my own empirical receiver observations.

2)  Computer DSP can only impact Gaussian noise, not
impulse noise.  So it doesn't work well with receiver
NR schemes.  In general, the reason for this is, DSP
samples signals over some sample period, and determines 
how to treat the NEXT signal event when it arrives.
	If that NEXT signal is significantly different
from the preceeding events, the treatment will be inappropriate.
Thus, DSP Noise Reduction may actually make QRN slightly worse,
because the impulse is a single, uncorrellated event, not a 
continuing noise train.
	This can be mathematically demonstrated, but proper
demonstration requires knowledge of the noise reduction 
algorithm.  Manufacturers are unwilling to disclose this, although
most seem to be 'walking-window averaging' schemes.  

3)  DSP based IF filters are also subject to some limitations.
I have found my 756PRO to be better at narrow bandwidths than
discrete filter based receivers, in terms of ringing.  
    Ringing tends to elongate the noise pulse from lightning,
and makes the AGC-recovery impact worse.
    This is crass speculation...but it is possible that the 
rise times of lightning burst QRN challenge the Nyquist sampling
limit of DSP generated filters, and have some transient effect on 
the S/N ratio.  I have yet to hear it...but expect to have my
1.8 MHz ground plane back up shortly, so will get to explore this
in greater detail over our summer.

4)  The best receiver I've ever used in heavy QRN was my old 781.
Variable AGC is one reason.  The 756PRO (and PRO II) allow you to
set the recovery time constant from the front panel.  I have not
explored this on top band static, yet...but do not expect to
find it any different than the 781.
	HOWEVER....the fundamental architecture of the two receivers
is different.  How the PRO acts, when a two-stage AGC is applied,
first to its 15KHz wide upper IF system, and then to its DSP
controlled lower IF, must be carefully studied.  To my knowledge,
this work has not been done.
	Unlike the MP, incidentally, the PRO's AGC is DSP generated,
and according to ICOM is a two-level system.   It would be speculation
to conclude that the PRO is immune from gain pumping, but I haven't
sensed it so far, and I've had this radio for two years, now.   

5)  To Josep...my personal opinion:  I wouldn't choose a 775DSP.  The 
PRO or PROII are both better radios.  An older 781 or 1000D are still 
very viable choices; with either of which you would be pleased.  
    DSP filters have one advantage over discretes:  cost.  DSP based
'signal processing' has only limited advantage, in my view: The auto-
notch feature is very useful.  NR is virtually useless, except on the
higher bands, without a lot of QRM/QRN.  
    
6) I would like to hear from k5kg and k3ww on this topic.  KG has had
both...and lives in Florida, so has plenty of static to listen to.
WW has both rigs, and can do a side-by-side.   Recommend you contact
them directly for their inputs. (k5kg@arrl.net, k3ww@fast.net)


Jim Jarvis/ N2EA
Essex Vermont
802 872 5830 voice
802 872 5831 fax