Dear Stefan, LF Group,
I did some experiments on single-turn loop Q when designing the "bandpass
loops" (BPloops2.pdf at
https://sites.google.com/site/uk500khz/members-files/files ). With a 1m x 1m
square loop made of 15mm copper tubing, I got unloaded Q well over 100 with
C about 400nF at 137k. This was one unusual example of where the capacitor
dominates the loss in a tuned circuit - I got a much lower Q using a single,
lower voltage metallised polypropylene capacitor than when using 4 x 100n,
1kV capacitors in parallel. I assume this is because higher voltage =
greater electrode area connected in parallel = less resistance, and more
capacitors = less resistance from the connecting leads and interface to the
metallisation. I have noticed similar effects with high-current capacitors
in QRO PAs. DK7FC's Q of 8 with 470n may be due to this - it implies Rloss
of about 0.3 ohms, which seems much too high for a few metres of copper
tubing, so something is definitely wrong here. Obviously, it is also very
important to have excellent connections between tubing and capacitors in
order to realise milliohms of resistance. Another possibility is that the
loop is inductively coupled to something that has high losses - you need to
keep it a metre or so away from other conducting objects when doing the
measurements.
I assume there are different paths the signal comes from, this is why i
cannot eliminate it completely, right?
Apart from what others have already said, the null can be degraded by RF
currents induced in nearby conductors acting as parasitic antenna elements -
this includes connecting leads, building structures, cables etc etc. 25dB is
fairly typical I think.
I used a high mu toroid (ferroxube, blue material) to transform the
primary side (=the loop) to 50 Ohm. I have done this by varying the
secondary turn number until i achieved a maximum voltage at a 50 Ohm load
at a given input signal.
Whatever the value of unloaded Q, the "maximum power transfer" theorem
applies, ie. the source resistance of the antenna at resonance should be
transformed to equal the RX input resistance to achieve the maximum signal
power delivered to the receiver - this is what you did empirically. In this
condition, the loaded Q should be half the unloaded Q , so in Stefan's case
tuning should be very flat with a loaded Q of 4.
Is it important to terminate the loops transformer output with a 50 Ohm
load in that case? On an oscilloscope i found that the level of DCF39 is
higher when having no R connected to the output but the waveform looks
much better / cleaner!
In my loop designs, I have made a trade-off by increasing the loading
(reduced loaded Q ), which increases the bandwidth, but reduces the signal
power delivered to the receiver. If you wanted to obtain maximum loop
selectivity, you could reduce loading of the loop, which would also reduce
the power delivered to the RX, so it is your choice... Whatever you do, the
band noise at the output of a loop like this is only a fraction of a uV, so
you either need a RX with a low noise figure at 137k (very rare!) or a
preamp with a low noise figure. Actually, with the preamp in the BPloops
article, simply connecting the 50ohm input directly across the parallel
tuned single turn loop is quite a good combination (loaded Q about 15 for my
loop, a little higher for Stefan's slightly smaller loop, but only assuming
he can improve the unloaded Q to say 50 or more). The preamp has low enough
noise to easily hear the band noise with these loops, but beware - some
receivers have such bad sensitivity at 137k that further gain will still be
needed.
I am quite pleased with the way the single-turn loops have worked out - they
work just as well as multi-turn loops, but are almost impossible to de-tune,
are less susceptible to unwanted capacitive coupling, easy to make
weatherproof, are mechanically simple and robust.
Cheers, Jim Moritz
73 de M0BMU
|
