I am not sure the details of all that I write below are
correct, I am not a wsjt expert, but I think the general
ideas are OK.....
(Inline comments below)
(1) First is I'm getting a 140ms burst of noise
every 18.75 seconds.
That is not much: 0.14/18.75 = 0.75%
Any sensible software should ignore what is received
during that time:-)
Analyzing it, it's a short burst of digital data
(coded as brief 5-10 cycle bursts of 30kHz modulation).
I do not know what you mean
by this. Maybe less than
140 ms has to be ignored?
It's spreading all across the band and originates
433.95 MHz, so I assume it's a bad ISM transmitter
somewhere in the area. It appears to be real and
not due to preamp or receiver issue, since it's still
there with no preamap and a 3dB line loss. The fact
that it's as regular as a Pulsar suggests it's some
sort of automatic telemetry device(?) rather than a
garage or car door opener. I see quite a few ISM
signals pop up in the 433.9 region, but only this one
is regular and spreads out all over the 432-435 region.
Means that this one is
faulty and should be fixed.
Probably a lot of work to find it and its owner and
to persuade/force the owner to do something about it...
My question is how resilient digital modes (JT65)
are to this type of QRM.
Should be rather insensitive. JT65 transmits 63 symbols
of 372 ms duration. There are also 63 sync tone intervals.
The total length is 46.8 seconds.
Your QRM, every 18.75 second will happen two or three times
during a transmission. At marginal levels you may loose
detection for maximum 6 of the intervals, minimum 2 intervals.
Probably half of that are sync tones. If you loose up to
25 data tones you are safe even without deep search.
This means that the degradation is something like 20/25
in power, next to nothing in dB. Nevertheless, in marginal
circumstances it is not nice to know one has a degraded
The pulses are quite easily audible (maybe 5-10dB out
the noise)? Noise blanker has no effect since they are
relatively long (140ms).
I have always been interested in interference fighting.
That is actually why I started with SDR in the first
If the qrm is 10 dB above the normal noise floor, it will
destroy tone it falls on under minimum conditions. It lasts
140/372=37% of a tone and would make noise 10 times higher 37%
of the time or 4.33 times higher in the interval. That is
a loss above 6 dB so the interval would be useless. If the
qrm spreads over two intervals, the loss in both would be
half as much, but loosing 3 dB on twice as many intervals is not
really what you would want...
There is another alternative. With an appropriate SDR package
you could blank the signal during the interference pulse.
That would decrease the noise by a factor of 0.63 while
the signal would be degraded somewhat more because its spectrum
would become widened and fall outside the proper fft bin to
some extent. My guess is that the worst case loss would be
0.63 squared, a factor of 0.39. The S/N loss would then be
0.39/0.63=0.63 or about 2 dB.
In case the blanked region would be split equally between two
intervals, the S/N loss would become 0.87 or 0.06 dB on two
A blanker that removes the noise bursts would be somewhat useful.
In case another antenna direction makes the qrm 15 dB above
the noise floor, a blanker would be a bit more useful.
It's annoying but if it's not going to
performance, I can live with it.
I think the dumb blanker in Linrad can be set to
the noise bursts. You would probably need hardware that allows
a bandwidth of at least 25 kHz. If you can make a recording
with a bandwidth of 25 kHz or more I would be interested to
have a look at it. Long ago I made a provision in Linrad to
add a second blanker that would operate on a user-defined
bandwidth, but I have not yet seen any recording with suitable
interference problems. Your qrm could be an interesting case.
I will try to locate the course of course, but
attempts at finding the source of QRM haven't usually been
I think a modest yagi would be fine, but you would need
a receiver that provides true signal level indications.
Maybe not standard equipment since pulses are short.
You might use a dongle (rtlsdr, funcube, or similar) with a
laptop and Linrad. (The S-meter graph would be what to
use.) You might also use other SDR packages - but beware
of the problem that default AGC settings could hide the
true data for you...
(2) The other odd thing is that, based on sun noise,
the beam pattern of the two yagis (horizontal stacked)
seems to be about 5 degres off boresight in the
horizontal plane. I'd never noticed this before, but
I'd never had the ability to remotely sweep the array
across the sun and take measurements. The antennas
are physically aligned well, have identical (within a
few mm) length feed lines of the same coax from the
same spool and the power splitter is symmetrical to
within a couple of mm. I wouldn't expect 100% perfect
boresight alignment, but 5 degrees just seems like a
lot when the beamwitdh of the whole thing is around
Seems you have a phase error in the feed system. Best is to
look att the symmetry of the side lobes. Maximum gain is with
a wide stacking distance that places the sidelobes in the
range 8 to 10 dB. Most people prefer weaker sidelobes than
that (for reasons I can not understand on small antennas)
but you should have significant side lobes. IF you find
that the sidelobes have different levels, you certainly
have a phase error. If cables and splitters are mechanically
equal, then the phase error would be because your antennas are
different. You could compensate for that by changing cable
lengths - or you could look at what is the difference between
Any thoughts? I'll try switching around feed lines
to see what effect that has, but is there something I may
be missing. I'm no antenna expert.
From your next posting I understand you have spur problems.
In case the spurs are
much narrover than the wsjt bin width,
these spurs can be eliminated.
The generally elevated noise floor that you see could turn
out to be impulse noise if you look at a high bandwidth. If so
it can be eliminated. (Powerline noise etc.) It could also
look like white noise also in a high bandwidth and then
(to my knowledge) there is nothing a SDR could do for you.