help with navtex/medium wave receiver sensitivity and decoding

Does anyone have any tips on increasing sensitivity for a medium wave
receiver trying to distinguish navtex 170Hz phase shifts on a low
power carrier at 518kHz?!

Ive been on and off trying to do this project for my boat for a year.
I say off and on, its become most frustrating! I can't use a loop
antenna because its just too big and too directional for use on a
boat. So I'm currently using a Nasa Marine active aerial (it is just
a small plate attached to a standard fet common source driver
circuit). I have that going down some coax through capacitive
impedance transformer and inductor creating some extra front end
selectivity, to the input of a 612 mixer, getting 6kHz IF (after being
mixed with a lo). I'm driving the 612 differentially, as its a slight
improvement over single ended. From here Ive tried using several op
amp filter circuits, and they all perform adequately.

Heres my problem, the receiver just picks up too much noise.
Everything interferes with it, even the scan on my oscilloscope. Even
without this, taking the aerial out of the room, the actual noise in
the system is too high to pick up anything but the strongest navtex
signals (which are decoded). Of course AM radio broadcasts are way
above the background noise, and navtex signals are much lower power.
And AM radio has the help of our ears to tune out the noise.

It works better when closer to the transmitter of course, but the
navtex spec says you should be able to pick up stations 400 miles
away, and i'm only just decoding the local one 100 miles away.

So my question is, does anyone know of any special techniques on the
RF side that can improve my noise handling/signal integrity? (my
expensive sony worldband receiver seems pretty good at it on SSB).
And....does anyone know how the upper market receivers get such a good
signal? My receiver just samples the input frequency (as does the
Nasa low end I think), so is very susceptible to any interference.

Grateful and very interested to hear views on this.

Many Thanks
Andy

 

It seems to me you could perhaps used a tuned circuit there, with
sufficient bandwidth.
In fact the active antenne you describe perhaps needs an LC to ground.
Else indeed any spike will overload the FET, easy with a 'metal plate'.

 

In spite of the rich detail you've given us, this is still a problem
with many interesting dimensions.

Atmospheric noise is very strong at MF, so even with a capacitive probe
antenna such as yours it should still dominate. I would start by
measuring your signal strength at IF or baseband with the antenna
connected, then with the antenna replaced by an equivalent impedance.
Ideally you'd see an increase in noise level from 3 to 6dB -- much less
than that and your system's inherent noise is your limitation, much
higher than that and you've got too much gain ahead of the mixer, which
will make it susceptible to distortion.

If its practical I'd do a similar test with the antenna. I'd calculate
the antenna's effective capacitance to ground, then replace it with a
capacitor of that value from ground to the FET gate. You should see the
same 3-6dB increase in noise with the antenna vs. the noise with just
the cap connected -- if you don't, then your performance is being
limited by the antenna.

After noise, the next thing you need to worry about is distortion. If
your active antenna or your mixer are being overwhelmed by strong
signals (such as AM broadcast signals) then your desired signal will
either be blacked out, or the strong signals could combine in your
electronics to generate a signal at your desired frequency. Building
low distortion front ends is worth several pages if not a whole book,
but the gist of it is that you want to limit your signal's bandwidth as
much as you can, and you want to use intelligent design practices. Your
active antenna can't do much before the signal hits the FET gate, but
you can take Jan's suggestion and put a filter between the active
antenna and the mixer. I'd put in an elliptical filter that had good
attenuation in the AM broadcast band, to give your little Navtex signal
a fighting chance.

Assuming that you're limited by atmospheric noise at the output of your
IF, the only thing you can do is to improve your signal processing. I'm
not familiar with Navtex signaling, so I can't give you more than
general guidance, but here goes:

The nature of the atmospheric noise at MF is _not_ Gaussian. Noise in
that band is predominantly from electrostatic discharges both near and
far. It has a noise density that is more like a Cauer density, in that
for all practical purposes it has an infinite variance. Any sort of
signal processing scheme which is based on an assumption of Gaussian
noise will be compromised by that assumption.

The best way to deal with MF atmospheric noise is to take advantage of
the fact that it tends to occur as 'crackles' or 'hits' that have a
high-amplitude, wide spectrum characteristic. In general what one does
is to use 'side information' to detect when a discharge event has
happened, then discount the data that's collected during that event.
This is what noise blankers in communications radios do.

If the Communications Gods are smiling on you then there are clear
channels adjacent to your signal, or Navtex uses forward error
correction. In the case of the clear adjacent channels, you can use an
IF that's wider than your signal and use a noise blanker. Your signal
will go quiet during the crashes, but that non-information is
demonstrably better than having huge signals get into your demodulator's
filters to mess things up. In the case of FEC, you can demodulate the
signal as usual, then look for unusually large filter outputs. When
this happens you replace the demodulated bit with an erasure and forward
the whole shebang to an error correction algorithm that can benefit from
the erasures (and some do, to a significant degree).

I hope this helps.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Posting from Google? See http://cfaj.freeshell.org/google/

"Applied Control Theory for Embedded Systems" came out in April.
See details at http://www.wescottdesign.com/actfes/actfes.html

 

In what sense too big? A ferrite rod is not that big.

 

I'm at a different location of course but a speccie analyser fed by a high
impedance buffer with a 18" telescopic aerial plugged in, just shows noise
and more noise. Biggest signal was a national radio station about 900kc at
8mV. 0-1Mc background about 0.5mV. Only starts cooling off above about 10Mc.
Using 1kc bandwidth gave a few carriers (modulation=?) around the 518kc area
but they were at the 20uV level and sitting barely above a buzzsaw of noise
filling all the frequencies and sourced from the aerial, local equipment
and PC.
Don't know 'Navtek' but I'd try very strong bandpass filtering at 518kc
after the FET. (you've probably not really enough wideband signal voltage to
overload it and may as well take benefit from the increased signal level).
The mixer is also going to give image frequency 6kc responses, these are
unfilterable. I'm guessing you're running the LO at 512kc from a 4060 type
xtal divider. The 3rd 5th 7th and 9th harmonics will easily mix with
up-spectrum buzzsaw noise and shift it down also to the 6kc slot.
Commercial stuff will filter, mix up to maybe a 60Mc IF to lose image
responses, filter and then mix back down again to whatever, (and then filter
again and once more for luck .

 

The active antenna has back to back diodes to protect it in this case

 

This is brilliant...thanks for taking the time to put in this level of
detail....I will spend this eve looking through your points..

 

A ferrite rod is small but very directional...in a boat you never know
which direction you're going to be facing...and the receiver needs to
work on a fixed antenna on its own!

 

Interesting... yes the incoming signal is 518k and I'm mixing it with
512k generated from an 18.430M crystal divided down. This generates a
bit of noise, and there is some chance of that getting into the
system. However, that was the first thing i checked..I disabled it
and the noise level visible on a scope drop by perhaps 10 to 20%, pre-
filter, however there is very little evidence of a drop after
filtering. The signal is a freq shift of 170Hz so detecting it is
difficult even with a clean signal, and I need to do it digitally
using the lo generator as a reference, otherwise theres no hope of
tracking the 170Hz as a crystal is not accurate enough to use absolute
frequencies.
Basically you're looking at an intermittant signal (perhaps a 10min
burst every 3 hours in peak times that are close enough to see), which
is from a transmitter putting out 100W to 1kW, 100 miles away, which
is substantially less than a commercial AM station.
Also, the inverter on my low voltage kitchen lights screws the signal
audible on my multiband sony recever, as does the TV set, my
oscilloscope if it gets too close, any PC and the LCD display that
shows the Navtex information. So its a bit of a harsh environment for
it.
The FET front end already has a 60Q bandpass filter, which ive tuned
for peak (need some lee way as the next step is to dual band and pick
up 490khz which is the local navtex), i then go into a tuned
circuit(est Q80) with another can again peaked, and output this into
the differential inputs of the 612. So two filters pre 612, then band
pass filter with a Q of 30 before trying to detect the signal.
I would have run the LO higher to make filtering a little easier, but
I need a low LO for my software to decode.
I'd love to try it on an HP, but so far Ive been outbid on ebay! I
think I'm being a cheap on my bids...! I'll see if I can borrow one to
check where the noise might be coming from.
You can hear navtex by tuning near it on SSB, its sounds a lot like
RTTY.
Andy

 

There are some fundamental differences between the way an E field
antenna (like a probe) and an H field antenna (like a loop) respond to
locally generated noise.
Your may want to investigate the Ryan storm scope that can determine
the distance of a lighting strike by comparing the relative amplitude
of the e and H field. You may want to inquire over at
rec.amateur.antenna. And you may want to consider using two
perpendicular loops.

Mark

 

F
Further to this...I think I will need to get my hands on a spectrum
analyser. My scope is just not set up to read the small signals pre-
mixer and is probably just showing its own interaction. Its on max
sensitivity reading the output of the mixer, so pretty sure its not
being overwealmed. I did once try amping pre-mixer as Ive heard the
612 is not the best noise performer, but i just ended up creating a
feedback oscillation randomly! Navtex does have error correction , in
that it interleaves the transmitted bytes to detect burst errors, and
the spec requires you to receive both before you commit that character
to the screen. The problem is the shift is so small 170Hz that you
have to have your detection in synch with your local oscillation,
which means trying to measure the waveform period (as I do) or using
some heavy DSP (which has a theoretical solution for this solution Ive
heard). All you need is a little noise, or even a miniscule bit of
mains humm to raise and lower the signal and the period measurement
goes out of the window. This easily detects burst noise, by presenting
periods outside of the expected measurement window.

What is interesting is that the noise as appearing on the scope in
time domain at any rate looks the same but bigger when I hold the
plate antenna. Of course that could be a red herring as I have no
idea what other frequency components there are in there, but the
general shape of the background noise in the time domain remains the
same.

Ive heard of someone using a large hoop antenna to clear up the
signal, though thats too directional for my application.
Presumably with the dominance of the background noise making the
antenna bigger wouldnt work, but narrowing the bandwidth could.
However if I narrow it any more I wont be able to pick up the adjacent
490kHz local navtex signal (the active antenna is supposed to be dual
frequency, but only by virtue of having a loose enough Q).

I wonder if there is a better active antenna circuit than the simple
fet circuit in the active antenna.

Really appreciate your input
Andy

 

I will ask the question about the antenna in rec.amateur.antenna
assuming Tim above doesnt get uppity about cross posting again! (I
posted this question in the radio amateurs site also)

 

I found speccies to be unbelievably useful. If you enjoy the radio
stuff (obviously!) should be No1 on the kit list. ('scope a near 2nd)

Seems you've covered most stuff worth doing on the Rx. Looks like an
electrically less harsh environment is the order of the day,
(something a distance from civilisation, something say like a boat
trailing a kite flown long wire).
Failing this, my parlance would say you're buggered
john

 

Demodulating FSK using center crossing really isn't very optimal. If
that's what you're doing you'll pick up quite a bit of range just using
the two filters method or a discriminator.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Posting from Google? See http://cfaj.freeshell.org/google/

"Applied Control Theory for Embedded Systems" came out in April.
See details at http://www.wescottdesign.com/actfes/actfes.html

 

ISTR sets which used a combination of a ferrite rod and a collapsible whip
but maybe that wouldn't work in your case.

I also recall two loops set at right angles to each other. Remember that the
highly directional point is the null signal not the peak signal.

 

Would 2 90 degrees crossed ferrite rods work?
Sum the signals from the coils?


|
|

 

1. Go with a ferrite loopstick antenna. Try pointing it vertically
so the null is at the zenith, and straight down into the ground.
A lot of the noise at that band is electrostatic, which your active
antenna seems to be designed to pick up.

2. Try for the narrowest bandwidth at the front end. Tune the antenna,
or have a high Q LC filter, or even build a crystal or ceramic resonator
filter at 518 (and 490) kHz. It only needs to be 500-1000 Hz wide.
That applies to your IF bandwidth, too.

2a. Since your IF is 6 kHz, watch out for images. That could double
your noise if the front end doesn't reject at the image frequency.

3. Look at "RTTY Terminal Unit"s, ie. fsk demodulators. The standard
design was two filters as narrow as possible (determined by the bit
rate) that were compared. One side could drop out and the other
half could still compare signal at that shift with the "no signal"
voltage.

Mark Zenier
Googleproofaddress(account:mzenier provider:eskimo domain:com)

 

Regarding using a vertical ferrite loopstick antenna (I presume this
is a ferrite bar would with a coil as usual), if you put it
vertically, and the transmitter is a vertical mast..does this cause a
sensitivity issue? Normally ferrite rods are held horizontally. But
my antenna theory is basic so I'm probably missing something...

My problem with using the RTTY model is that lo drift will take me
right off the sweet spot (only have a 170Hz to play with between mark
and space).

IThanks Mark, this is very useful

 

no, by just summing you'd still get two nulls. (at 135 and 215 degrees)

If you can phase them so that the signal from one is (ideally) 90 degrees
different from the other signal you'd get 360 degrees range with no nulls

 

OK, should be no problem to make 90 degrees at _one_ frequency,
either slightly detune it, or just a phase shift network before the add.
I think the idea from the other poster to put the rod vertical is good too,
I just tried it with the local radio station on MW 1007kHz, and the directional
effect is gone (I can zero it horizontally), and the signal does not attenuate
a lot.