Date: Mon, 25 Jan 2016 23:02:32 +0100 (CET)
From: Jacek Lipkowski <sq5bpf@lipkowski.org>
To: rsgb_lf_group@blacksheep.org
cc: rsgb_lf_group@yahoogroups.co.uk
In-Reply-To: <CAA8k23Q=A3dbD_PfpGdL_niWf8cjGzmYV_CH1REqt04H6r5nRQ@mail.gmail.com>
Message-ID: <Pine.LNX.4.64.1601252150380.27152@paranoid.lipkowski.org>
References: <CAA8k23QbbgxAK2o70Q=orhXEjxgEiAB6ptmVX+_zX6zMe--bnw@mail.gmail.com>
 <Pine.LNX.4.64.1601251349010.27152@paranoid.lipkowski.org>
 <CAA8k23Q=A3dbD_PfpGdL_niWf8cjGzmYV_CH1REqt04H6r5nRQ@mail.gmail.com>
MIME-Version: 1.0
Subject: Re: LF: Direct Upconversion from audio
Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed
Sender: owner-rsgb_lf_group@blacksheep.org
Precedence: bulk
Reply-To: rsgb_lf_group@blacksheep.org
Status: O

On Mon, 25 Jan 2016, Andy Talbot wrote:

> Having true baseband as an intermediate step is useful for other bands - in
> particular 24GHz where several types of off the shelf surplus equipment
> already has an image cancelling mixer, just needing I/Q baseband drive.

You can do this easily in software too in a similar way. Many sound cards 
can be dc-coupled (or at least the coupling caps can be increased, so 
there is only a narrow dip near 0Hz) - i'm not sure about the 1/f and 
other noises near 0Hz, but that should not be too bad. Actually i'm not 
sure that you would need to modify anything for ssb, voice should be 
understandable with a few 10s of Hz gap around 1.5kHz, and digital 
transmissions can just avoid the few Hz around 1.5kHz.

The software is simple: in gnuradio use a frequency xlating fir filter 
with a 1.5kHz frequency and float->complex conversion, feed this into a 
complex-> float block and the outputs into a stereo sound card sink. Very 
simple to draw in gnuradio-companion, you don't even have to do any 
programming (and while you're at it, you can also do a receiver which 
would be equally simple). I suspect this could also be trivially done in 
Spectrum Lab.

> Alberto's DSPic solution to generating I/Q baseband signals from voice band
> looks promising, so long as it has analogue output and not just SPI drive
> for a DDS

Yes, this is great, because you don't need a PC for it. Running gnuradio 
on a raspberry pi would also work (and be easier to modify to your needs), 
but the dsppic solution probably has less latency.


Still for LF/MF this seems to be too complex, upconverting in software to 
around 20kHz and then to 136/472kHz and LC filtering seems much simpler, 
and can be debugged just with an audio generator.

Example for 2200m band: a cheap 3.686MHz crystal divided by 32 gives 
115.1875kHz. To transmit on 135.7-137.8kHz you need to generate 
20.5125-22.6125kHz with the sound card (no problem with a cheap 48kHz 
sound card). The mirror frequency is around 94kHz, so it is easy to filter 
out from the 137kHz output.

Example for 630m band: use a 455kHz crystal (or a piezoceramic if you 
don't mind worse frequency stability). To transmit on 472-479kHz you need 
to generate 17-24kHz (cheap sound card). The worse case mirror frequency 
would be at 17kHz, trying to filter 472kHz and reject 432kHz, still doable 
but with a few more LC circuits (see how much an old AM radio uses for 
this selectivity).

VY 73

Jacek / SQ5BPF