Originally I was going to base the design on that found at http://www.qsl.net/m0ayf/active-loop-receiving-antenna.html, because it seemed like a great idea having something that would work over the whole HF spectrum. Basically, it's a 1.5m diameter loop connected directly across the inputs of a NPN differential pair, with shunt feedback in an attempt to reduce the input impedance, although looking at the feedback resistor value I'm skeptical there is any benefit accrued by this. Conventional loop antennas are resonated with a capacitor across the loop terminals and the signal is taken as the voltage across the terminals. This design does not resonate loop, instead it attempts to present as low an impedance across the terminals as possible and thus senses the loop current as opposed to the voltage. Take a look at http://sidstation.loudet.org/antenna-theory-en.xhtml, which uses a Spice simulation of an N-turn loop antenna to show that this can result in a flat response over a wide frequency range.
A better solution would be to use the classic op-amp transimpedance amplifier topology to convert current to voltage, but this would require a fully differential op-amp, and the noise performance is unlikely to be as good as a simple diff-pair. However, a better solution seemed to me to be the approach detailed in http://files.radioscanner.ru/files/download/file6511/loop_wide.pdf. This uses a ferrite current transformer with the antenna loop as the primary turn and a single secondary turn connects to a pair of augmented common-base stages with the output taken differentially. I liked this design because with enough bias current it makes the emitter impedance vanishingly small, getting close to the ideal current-sensed active loop antenna.
To cut a long story short, I was very disappointed with the results. I simply couldn't get enough output level. As a sanity check, I tried the general design approach at http://www.qsl.net/kc2tx/, using the same loop of wire I'd been experimenting with. Even without any additional amplification this was hugely better than the transimpedance approach, and I decided to pursue this route.
Let's take a closer look at the KC2TX design:
The electrostatic screen formed by the coax outer is claimed (in fairness, not by KC2TX) to make the loop sensitive only to the magnetic field, thereby rejecting local noise sources that tend to be dominated by their electric field. This in fact is nonsense, and I recommend you read W8JI's treatment of the subject if you want to delve into this deeper. The main take-away is to recognise that balance with respect to the coax outer potential at the feedpoint is the key to rejecting unwanted coupling to the feeder and common-mode pick-up on the two halves of the loop. Nevertheless, my design retains the electrostatic screen simply because it's simpler to fabricate the loop out of coax and use coax connectors at the feedpoint enclosure.
Ignoring the 'electrostatic screen', we have an equivalent circuit that looks like this:
The 50R load is transformed by a factor 2^2 (owing to the centre-tap), resulting in an equivalent load resistance of 200R across the tuned circuit. Provided the Q is high enough, the current in the two halves will be balanced, despite the 50R only being connected across half the tuned circuit. So what is the Q? First let's calculate the loop's inductance. There's a good overview of appropriate formulas at http://www.qsl.net/in3otd/rlsim.html. I used the formula attributed to Grover for a rectangular loop (FW Grover, "Inductance Calculations: Working Formulas and Tables", D Van Nostrand Co., New York, 1946):
Where a and b are the length of the sides and d is the wire diameter (all in meters).
I'm using UR43 for the loop, which has a 0.9mm centre conductor. With 0.76m (2.5ft) sides, this gives an inductance of 4.1uH.
Since the Q will be dominated by the inductance, we can work out the Q from the equivalent parallel load resistance and the inductance using:
Using 3.65MHz, which is approximately the geometric mean frequency
for the UK 80m band limits, this gives a Q of just over 2. In my
opinion this is barely enough to ensure balance, although it obviously
makes it a 'no-tune' solution.
Turning the problem on its head, what is the maximum Q we could realise without the need to alter the tuning across the band? Again using the UK band limits, if we allow the signal level to drop by 3dB at the band edges, we have:
Using our previous equation, the equivalent parallel resistance would be about 1.1k.
We could achieve this using a 22:1 transformer (a turns ratio of about 4.7), but since we are likely to require a pre-amp anyway, we could combine the impedance transformation into the amplifier function.
Before continuing, it's worth working out what capacitance is required to tune the loop. From:
We can work out that C = 464pF at 3.65MHz. But the UR43 used for the loop has a capacitance of 100pF/m, which adds 152pF. This is slightly simplistic because the capacitance is distributed around the loop, but it's good enough as a first approximation. So we will need an additional 312pF to resonate the loop. This is a healthy margin, and tells us the loop is not too big.
A schematic of the pre-amp and its associated power-over-coax unit is shown below:
The power-over-coax unit uses PTT to gate the supply voltage to the
pre-amp. When the supply is present at the pre-amp the loop is
connected to the pre-amp input. This provides some protection from the
transmit antenna. I use the PTT output from my Orion that is sequenced
and has 'hang', to ensure the relays have time to do their thing.
The pre-amp itself uses a U430 JFET matched pair. These are getting a bit old in the tooth now, but they're still available. I used them because I happened to have several in the junk box. They are nevertheless ideal for this application. U431 types could also be used, but run double the current. Matched pairs are a good idea, since we're very interested in maintaining circuit balance.
The U430 gates are at ground potential by virtue of R3/R4 and L6, so each JFET runs at IDSS (30mA max). Restricting the supply to 5V keeps the power dissipation well below the maximum of 500mW. Normally a differential pair would have a 'tail' resistor, but we want a large tail resistance for good common-mode rejection. Increasing the source lead resistance reduces the quiescent current, which can only be offset by jacking up the gate voltage. It's not possible to get an adequate compromise for sensible supply voltages. Another possibility would be to use a transistor current source. For simplicities sake I chose to go a different route, and the 'tail' is realised using an RF choke.
The output transformer originally used 15/15/15 turns giving a 4:1 ratio. Later on I carefully removed 8 turns on the secondary to make it approximately 18:1, which gives more gain. The 50R load is therefore transformed to look like about 900R across the Q2/Q3 drains.
The gm for the U430 is about 10ms (=10mA/V), so the differential gain
is 20*log(10e-3*900/2) = 13dB. In addition, we can calculate the gain
associated with raising the loop's terminating resistance from 200R to
1120R (2*560R on each gate). That gives approximately 15dB, so the
total pre-amp gain is theoretically 28dB.
The photo below shows the pre-amp construction. I used an MTCC-13
500pF compression trimmer from JAB
for C8. Mounting is a challenge; in the end I cut some islands in the
copper-clad board and soldered the bent connection tags of the
trimmer to those islands. In retrospect I think finding the right
value with a spare variable, measuring it and fitting a fixed
capacitor would have been just as good, because the tuning is still
The power-over-coax unit is shown below:
The frame for the loop, which can be seen at the top of this page, was constructed from black 40mm PVC pipe for the vertical section and 21.5mm plastic overflow pipe for the horizontal section. The latter could only be obtained in white, so it was sprayed matt black. I found that it was possible to carefully ream-out a hole in the 40mm pipe such that the 21.5mm pipe was an interference fit. Notches are made in the ends of pipes for the coax to sit in. This makes it possible to remove the loop without de-soldering the PL259 connectors.
According to ON4UN in his book "Low-band DXing" (3rd Ed), small loop antennas do not exhibit the pseudo-Brewster angle effect normally seen with vertical antennas over real ground. I wondered whether the DX I couldn't hear was because none of my antennas had a low enough angle of radiation. My conclusion so far is that either the small loop antenna does suffer the pseudo-Brewster angle effect, or my main problem is simply the noise present in a suburban environment.
ON4UN doesn't offer an explanation as to why the pseudo-Brewster
angle might not be applicable to the small loop antenna, and I've not
been able to dig anything up on the internet to back up these claims.
Having since thought about this a bit more, I can perhaps see that if
you were to consider the small loop antenna as predominantly sensitive
to the H-field that you might not see the effects of slow phase
reversal of the E-field as it gets back up off the ground and adds to
the directly received wave (the cause of the pseudo-Brewster angle).
However, the small loop is only predominantly sensitive to the H-field
component in the induction field (i.e. very close to the antenna), and
the induction field is not responsible for electromagnetic wave
transmission (or reception). The Fresnel zone for very low radiation
angles will extend a long way out from the antenna, in fact far enough
that the E and H-field intensity has become equal. Therefore the angle
of radiation will be determined by ground conductivity in exactly the
same way as for any vertically polarised antenna.
Something else to consider about the loop is that there is no overhead notch in the vertical radiation pattern, like verticals possess (note they are both vertically polarised antennas). Verticals are often cited as being helpful for reducing high angle QRM, so possibly what you gain on the swings you lose on the roundabouts (is that just a British expression?).
The loop does help with local noise sources. At least part of this is because the antenna aperture is small and it is located as far as I can get from my own and any other house. It is possible to null local noise sources by rotating the antenna. But in the early hours of the morning when the local man-made noise level is at its lowest, I don't see much difference between it and my vertical. In these instances it is really directivity that is needed to improve SNR in the favoured direction.
There is little directivity on most skywave signals. This is to be expected because the depth of the azimuth null is only useful at very low angles (e.g. local noise sources via ground wave). We might therefore expect more directivity on DX signals compared with high angle QRM - not the ideal way round!
Building this antenna certainly hasn't been a waste of time, but its benefits are marginal if you already have a low noise floor. In that case you might be better off with something like a K9AY array. However, if you live in a high man-made noise area, it could really help. When I get QRV on 160m again, it will be worth re-evaluating the benefits of a small loop antenna on that band, because I've read the loop's benefits are more noticeable.
Something I want to try in the future is a double-size version using
wire rather than coax. Without the capacitance to the coax shield, it
should be possible to make it resonate on 160m & 80m by switching
the resonating capacitance. Capacitive step-down matching could also
be incorporated and a 1:1 balun ensures loop balance. This might not
need a pre-amp on 80m owing to the increased size.