[Date Prev][Date Next][Thread Prev][Thread Next] - [Date Index][Thread Index][Author Index]

RE: DSP noise reduction--now I'm confused


The files sound like an excellent idea (pun clearly intended).
What you have described in your return comments clearly indicates
that the authors of the Clearspeech algorithm have chosen an
extremely long time constant with the intent being to notch tones.
I have to admit I prefer the manual notch for tones (unless there are
more than one) and the faster time constant for other things.

For example,  the Automatic NF in the IC756 works very well against
some EMI from monitors and that ilk by taking out those things
that are tone like, but which change a bit more rapidly than this
long time constant in Clearspeech could track.  The automatic
noise reduction filter makes morse code usable that I can no
longer copy even in a 250 Hz but the thing that is most amazing
to me about using a well implemented one, is how much less the
fatigue factor is.  Constantly trying to parse tones out of a
channel of noise is tiring to say the least and this implementation
mitigates that to a great extent.  On the other hand, I would
never apply this to a bauded modem signal.   The implementation
in this, and in Clearspeech,  are not symmetric FIR filters
and would not have linear group delay so it would distort the
received waveform.  Before I believed this actually helped
with (say) the AO-40 beacon, I would like to see some tests
run.  While you are making files, can you make beacon with and
without the filtering?

Bob, N4HY

-----Original Message-----
From: Jon Ogden [mailto:na9d-2@speakeasy.net]
Sent: Thursday, February 20, 2003 9:55 AM
To: Robert McGwier; Amsat Bulletin Board
Subject: Re: [amsat-bb] DSP noise reduction--now I'm confused

on 2/20/03 7:57 AM, Robert McGwier at rwmcgwier@comcast.net wrote:

> The Widrow algorithm has
> no method for distinguishing between good tones and bad tones.
> The time constant or gain might make it work in Morse mode but it
> will nevertheless depress the power in the Morse tone.  The
> slower the Morse, the deeper the depression given a fixed adaptation
> constant.

That might be correct.  However, not knowing the software of the ClearSpeech
unit and how it works, I can't comment on it's algorithms.  This I do know:

1.) For normal CW, there is NO depression or at least none that is

2.) In order for the tone to start being suppressed, the tone must be there
for several seconds.  I've never counted how long before it starts notching
it out, but it is much longer than what is used in even 5 wpm CW.  It's a
notch designed to filter out tuner-uppers and carriers and heterodynes and
it works well.

Unless someone is sending extremely slow CW where the length of a dash is
several seconds there is no degradation.

When I get time and a good sat pass, I am going to make some recordings of
all this and put them up on my web site.


Jon Ogden
NA9D (ex: KE9NA)

Citizen of the People's Democratic Republik of Illinois

Life Member: ARRL, NRA
Member:  AMSAT, DXCC

http://www.qsl.net/na9d   <- Updated on 1/22/03!!!

"A life lived in fear is a life half lived."

Sent via amsat-bb@amsat.org. Opinions expressed are those of the author.
Not an AMSAT member? Join now to support the amateur satellite program!
To unsubscribe, send "unsubscribe amsat-bb" to Majordomo@amsat.org