From: "James Moritz" <J.R.Moritz@herts.ac.uk>
To: rsgb_lf_group@blacksheep.org
Date: Thu, 26 Feb 2004 13:38:32 -0000
Organization: University of Hertfordshire
Message-ID: <000001c3fc6d$d667fd80$fce8c593@RD40004>
MIME-Version: 1.0
Importance: Normal
In-reply-to: <001101c3f91c$82ba8c10$687a37c0@w2ksn>
Subject: LF: RE: feedback requested for new WOLF
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Transfer-Encoding: 8bit
Sender: <owner-rsgb_lf_group@blacksheep.org>
Precedence: bulk
Reply-To: rsgb_lf_group@blacksheep.org

Dear Stewart, LF Group,

I'm sending this again - it seems not to have made it through the system
the first time!

Here are some comments on your new WOLF proposals from my point of view.
The LF gear here is basically an RA1792 receiver, plus the TX "Exciter"
is an old Racal 9084 signal generator. Both these have PLL synthesisers
with internal OCXO references, which give them much better frequency
stability than amateur-type rigs; additionally, they can use an external
reference, which for last year's PSK tests was a Z3801 GPS
receiver/clock, which has both a 1pps output, and a 10MHz OCXO which is
continuously calibrated against the GPS signal. 

For reception, R3 is closest to what I am currently using anyway. For
the R1 and R2 options, as others have already said, using an SSB filter
for LF receive around here is problematic because of the strong signals
that would tend to be present in the passband - I guess these will
frequently be 60dB or more over the wanted signal. While a sound card
with a halfway decent A/D converter can probably cope with this, as can
the RX front end, the RX IF and audio sections were never designed with
this in mind. It would be possible to re-design the IF/audio circuits
for this purpose, but that would be at least as complicated as providing
the external frequency locking for the receiver. An exception to this
would be if you were designing a specialised homebrew receiver for this
mode - then it could be a quite simple TRF design with a
crystal-controlled, fixed BFO frequency. A further problem with R1
concerns using a loop antenna for receive. At my QTH, the optimum
orientation of the loop to receive North American signals puts the local
Loran station in the loop null, making it difficult to use the Loran
signal for synchronisation without degrading the noise rejection of the
loop. 

Using either LORAN or the GPS 1pps to calibrate the receive frequency
and the sound card sampling rate would certainly be useful for other
purposes too, as well as WOLF reception.

On the transmission side, T1 is essentially the method I have used in
the past for PSK generation anyway, excepting the use of the GPS 1pps
for synchronisation, which was done with hardware. It makes the carrier
frequency totally independent of sound card sampling rate error. I have
also used method T2 (again without the GPS synchronisation) to generate
"Jason" signals. The sound card initial accuracy is not particularly
good, so certainly some form of calibration would be needed. 

T3 sounds like quite a good way of achieving  the required results
without requiring high quality references in the TX - if the frequency
generation for RX used the same sources as TX (e.g. an HF transceiver
with a linear transverter to 136k), couldn't this also be used to
calibrate the RX frequency?

Considering the clock oscillator of a sound card can be quite hard to
get at, and may well not be at a particularly convenient frequency, I
would think T4 would be no easier than making a frequency source that
was locked directly to the 1pps signal. I suspect T5 would run into
similar problems as R1 and R2 re bandwidth, unless multiple LF receivers
were available.

How precise would the frequency of TX and RX need to be? It occurs to me
that, even if using precise frequency references, the DDS synthesisers
widely used now have a kind of "built-in" frequency error. The tuning
increments of the DDS are very small, but are not generally convenient
fractions of 1Hz, whereas the tuning display normally does display the
frequency as a whole number of Hz, and presumably selects the nearest
rounded-off DDS tuning step. This will result in a frequency error which
depends on the DDS clock frequency, the phase accumulator length, any
PLL multiplying factor, the set frequency, and the way the tuning
microcontroller rounds off the frequency display. It would be difficult
for the operator to figure out what actual error this would cause at a
given frequency. I guess this might come to 10s of millihertz with
typical HF receivers.

Cheers, Jim Moritz
73 de M0BMU