DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws;
  s=s1024; d=btopenworld.com;
  h=DKIM-Signature:Received:X-Yahoo-SMTP:X-YMail-OSG:X-Yahoo-Newman-Property:Message-ID:From:To:References:In-Reply-To:Subject:Date:MIME-Version:Content-Type:Content-Transfer-Encoding:X-Priority:X-MSMail-Priority:X-Mailer:X-MimeOLE;
  b=1tBXe+1LSRmeSwrAqnzcVhaglCofVO2eolwxg0YFSPzqrGM8/i3GaMlg8xhsZN66xCXH2YkhQzrEZ5WaeYEqFAHSmNLj5Q2vBsbnhSgGM7G5t7BITwcOkqjJeKa+FCN0OoYnTb6rKNu37aB64odCYN2b6ttvrYSsH71l5lNuiGc=  ;
Message-ID: <E0FACAB32B6841339C28D99E9ACFA95B@JimPC>
From: "James Moritz" <james.moritz@btopenworld.com>
To: <rsgb_lf_group@blacksheep.org>
References: <4C58579F.30406@telus.net> <8CD01F14619A000-1C48-3D75@webmail-d073.sysops.aol.com> <8CD02ACE55ECBAC-1CF4-9FFF@webmail-m086.sysops.aol.com> <4C5ABD7D.2080301@iup.uni-heidelberg.de>
In-Reply-To: <4C5ABD7D.2080301@iup.uni-heidelberg.de>
Date: Thu, 5 Aug 2010 23:22:30 +0100
MIME-Version: 1.0
DomainKey-Status: good (testing) 
Subject: LF: Re: Bandpass filter design
Content-Type: text/plain;
	format=flowed;
	charset="utf-8";
	reply-type=original
Content-Transfer-Encoding: 7bit
Sender: owner-rsgb_lf_group@blacksheep.org
Precedence: bulk
Reply-To: rsgb_lf_group@blacksheep.org


Dear Stefan, LF Group,

The AADE utility will do the design calculations for you... the limitation 
of filters like this is that, in order to achieve a particular shape of 
frequency response, bandwidth, and frequency, there is a certain minimum 
value of component Q required, whatever type of circuit you use. Generally, 
things that increase "selectivity" require more Q. A narrower passband (as a 
fraction of the centre frequency) requires a higher Q. A more rapid the 
transition from passband to stopband requires higher Q. A higher order 
filter (i.e. a larger number of LC tuned circuits in the bandpass case, to 
give higher attenuation in the stop-band) requires higher Q. At 136k, Q of 
100 or so is easy to achieve; with special pot-cores, or very big coils, a Q 
of 1000 or more can be obtained. This practically means filters with 
moderately sharp cut-off can be fairly easily made with a bandwidth of 
several kHz, or with much more difficulty a bandwidth of a few hundred Hz. 
So a high rejection of DLF will easily be obtained by the input filter if it 
is designed to give adequate image rejection, but that would be difficult to 
achieve with DCF39, only about 1kHz above the top of the band. This is why 
crystal filters became popular ;-)

As Michel says, the coupled-resonator type of filter is best for narrow-band 
designs (like the "top-coupled" circuit in your drawing). It is nice because 
you usually end up with inductors that are the same value. The 3kHz ladder 
filter used in my 9kHz preamp circuit is a type more suitable for a large 
bandwidth / centre frequency ratio - if you try to design such a filter for 
a small ratio, say 5kHz BW at 137kHz, you will end up with an impractical 
design with very small shunt inductances and very large series inductances. 
Cauer/elliptic bandpass filters tend to get a bit complicated... and are 
still limited by inductor Q anyway.

For the 12kHz IF filter, if you wish the whole 2.1kHz of the 136kHz band  to 
pass through the filter, a coupled resonator filter is again probably the 
most practical - but you will end up with a somewhat asymmetrical response 
due to the nature of this type of design. Due to the lower centre frequency, 
Q requirements are reduced, and you will probably be able to get better 
rejection of DCF39. But you will need bigger inductors to do it. You could 
instead design a filter with a narrower bandwidth - say a few hundred Hz 
just to pass the QRSS segment at the top of the band. In this case, you 
could get quite high rejection of DCF39 with a fairly simple filter.

A better approach may be to have a rejection notch filter to attenuate the 
DCF39 carrier. This could be done at the input frequency or the IF. With 
bridge-type circuits, rather high rejection can be achieved at a spot 
frequency with a single tuned circuit, at the expense of some attenuation of 
nearby frequencies.

But as Alan suggests, you may not need much filtering - sound cards do vary, 
but quite often they have suprisingly good linearity. Provided you use the 
minimum possible gain to raise the band noise above the sound card noise 
level, and the level of DCF39 is not high enough to actually saturate the 
sound card input, it may work OK. Intermodulation may not be serious, since 
there is essentially only one strong signal reaching the sound card input, 
and so not much for it to intermodulate with. Having aother signal 60dB 
above the wanted signal may not be an issue. I guess the sidebands from 
DCF39 that actually fall within the 136kHz band may be more of a limiting 
factor. They do noticeably raise the noise level at my QTH - You are 
obviously much closer!

BTW- are you really getting 60dB attenuation of 153kHz? The level of DLF on 
your spectrogram seems a bit hard to believe.

Cheers, Jim Moritz
73 de M0BMU