TopBand: DSP comments

[km1h @ juno.com]

I am forwarding this from another list. It is just about the best
explanation of IF and AF DSP I have read. 

73   Carl  KM1H



--------- Begin forwarded message ----------
From: Ham-Equip Mailing List and Newsgroup <Ham-Equip@ucsd.edu>
To: Ham-Equip@ucsd.edu
Subject: Ham-Equip Digest V97 #1255
Date: Tue, 21 Oct 97 10:58:29 PDT
Message-ID: <199710211758.KAA12428@mail.ucsd.edu>
----------------------------------------------------------------------

Date: Tue, 21 Oct 1997 16:43:55 GMT
From: ke4zv@bellsouth.net (Gary Coffman)
Subject: AF vs. IF - DSP - Why?

On Mon, 20 Oct 1997 19:20:31 -0700, Pete McGreevy <PMCGREEV@MAIL.IDT.NET>
wrote:
>
>What are the actual benefits of having the DSP of a tranceiver in the IF

>stages, as opposed to the AF stages?  In what types of situations would 
>an IF DSP system really shine, while AF might not?  And finally, oculd 
>you actually tell a difference between the two systems if you were 
>blindfolded... and what would that difference be?

There are two reasons to prefer that the DSP be at IF rather than at 
baseband AF. The first reason is that it will be (or at least can be)
inside 
the AGC loop of the receiver. Since the DSP will typically be the
narrowest
filter in the system, putting it inside the AGC loop will prevent
"pumping" of
the AGC by out of passband signals. 

This can be very important if there are strong signals near the one you
are
trying to copy because the pumping can make the signal you want to
receive
disappear as receiver gain is varied by the AGC. But there is a downside
to this
too. Because DSP filtering introduces a delay (sometimes a signficant
delay if
extensive processing is done), AGC action can become "squirrelly", with
large
overshoots and undershoots due to the lag in response. So there are
pluses and
minuses to having the DSP inside the AGC loop.

The second reason may or may not be important. By doing the filtering 
*before* demodulation of the signal, effective SNR can be improved, 
sometimes dramatically. Some systems actually implement the demodulation
function in DSP too, so near ideal demodulation can be done. This can
greatly
improve the ultimate weak signal ability of the receiver (provided it
isn't
already limited by other factors such as a poor front end noise figure).

Once the signal is demodulated, certain information about the signal
is lost and cannot be recovered by post-demodulation processing. So
a baseband (AF)  processor is faced with a garbage in--garbage out 
situation. This is mainly significant when working near the noise floor, 
because that is where the output of the demodulator will be most
uncertain.
For the relatively strong signals most often encountered on HF, this is a

moot issue. But if you care about working the weakest signals, you'll 
want to achieve as close to a matched filtering situation as possible 
*before* demodulation.

Note that for some forms of modulation, use of a product "detector"
isn't really demodulation. It is just another mixer doing frequency
translation to a very low frequency IF which happens to be in the
audible range. Thus use of audio DSP on this sort of signal still 
conveys the effective SNR improvement benefit of IF DSP. An 
example of this sort of signal is a SSB signal. 

An example where actual demodulatin is required is DSB AM. 
AM has to be envelope detected (or synchronously coherently 
demodulated), rather than being product "detected",  for the 
baseband  audio to be recovered. So you want the DSP applied 
before the demodulator for this type of signal. (This is mainly
of interest to the SWL today.)

Note, if you are trying to decode CW, it is also DSB AM, and
must be envelope detected to recover the baseband modulation,
IE the pulses of the transmitting key whose timing encodes Morse. 
This is done with a demodulator called a *slicer*, so you want the
DSP matched filtering to occur before slicing to get maximum
effective SNR benefit from the DSP.

We don't often think about this because, when receiving CW
manually, we do the slicing in our heads rather than with a 
hardware diode detector as we do for AM voice (in this case 
the receiver product "detector" is just acting as mixer converting 
the modulated signal to a lower "IF" which we have conditioned 
our brains to *both* demodulate and decode). But if you are trying 
to machine decode Morse (or any other discrete code transmitted by 
OOK), this can become painfully obvious if you try to demodulate 
(slice) before match filtering. 

In summary, an audio DSP does the same thing for us as an IF DSP
if we are using it for SSB or *manual* OOK Morse reception, both
of which are "demodulated" in the brain. For other types of signals,
IE various types of DSB AM or FM modulated signals, we need the 
DSP to be before the demodulator for best effective SNR. Having the 
DSP inside or outside the AGC loop must also be considered (though 
note that if you use audio derived AGC, an AF DSP still can qualify).
Probably the best reason for wanting the DSP at IF is that it can also
serve as a near ideal demodulator for the types of signals which require
demodulation in the receiver (as opposed to in the brain).

Gary
Gary Coffman KE4ZV          | You Make It  | Email:
Destructive Testing Systems | We Break It  | ke4zv@bellsouth.net
534 Shannon Way             | Guaranteed   |         or
Lawrenceville, GA 30244     |              | ke4zv@radio.org

------------------------------

--
FAQ on WWW:               http://www.contesting.com/topband.html
Submissions:              topband@contesting.com
Administrative requests:  topband-REQUEST@contesting.com
Problems:                 owner-topband@contesting.com