2010-03-27
Phantom-Powered Active Loop Receive Antenna for 30 Metres
This loop antenna was built for 30 metre QRSS reception, but tunes beyond 30 metres and might be useful for general narrow-band (fix-tuned) HF work.
The Antenna
The loop is two turns of ~3 mm multi-strand hook-up wire, wound on a large (465 mm diameter) embroidery frame. The coil is centre-tapped, and referenced to "ground" at that point. A polyvaricon tunes the coil to resonance at the frequency of interest, and a push-pull JFET buffer amplifier transforms the very high impedance of the parallel resonant circuit down to something suitable for what is seen through the coax from the receiver (and also offers some power gain at the same time to offset feed-line loss). The buffer can deliver a relatively large amount of power, in excess of 0 dBm. This should offer good strong-signal handling, but I have not measured its IP3.
Each J310 stands about 2.5 mA. I hand-picked a pair of well matched J310s (by Vpp and Idss), this is likely unnecessary but it can't hurt to ensure each arm has similar biasing and gain to optimise the distortion cancelling effects. The 33 Ohm resistors in the drains help eliminate any tendency for spurious oscillation. The supply current feeds into the bifilar drain loads from the decoupled "cold"-end of the output winding, using it as a choke. This seems to work well in practice, with no visible distortion asymmetry (when over-driven) from the DC bias on the magnetics. (The standing currents in the bifilar winding cancel, but those in the output secondary do not, fortunately the supply current is only about 5 mA so this should be of no consequence. As long as the transistors saturate well before our ferrite core - an FT50-43 - we are in good shape.)
More gain is available by bypassing the 100 Ohm source resistors. Oscillation can occur with excessive gain. You can set the gain at a higher, but stable level by adding resistance in series with the bypass capacitors.
The Bias-Tee
To feed the DC supply to the antenna at the shack-end of the coax a simple bias-tee was constructed. As the frequency of operation is only mid-HF a simplistic ferrite toroidal choke and two capacitor affair was constructed in a small die-cast box. I was confident this construction was sufficient at the frequency of operation, but curious about its performance elsewhere, as such it became one of the first test subjects for my experimental scalar network analyser.
Despite having very poor "design hygiene" for high frequency response, sweeps of the bias-tee with the analyser suggest it offers acceptable performance across HF. There are some oddities in these measurements however. Fortunately they appear to be measurement equipment problems rather than excessively nasty behaviour of the device under test:
Note how the measured "return-loss" exceeds the bridge directivity below 20 MHz. This is of course an illusion, caused by the small stray reactances of the bias-T conjugate matching the bridge for better than calibration reference balance. Ideally I should change the graphing software to compute error bars based on the reflection signal magnitude compared to the directivity established by Open-Short-Load calibration at the same frequency. As the directivity of the bridge exceeds 30 dB above a few MHz all the way to 60 MHz we can safely say the bias-Tee is an reasonable match over the same range displaying a return loss exceeding 20 dB the entire way.
Transmission measurements similarly have some weirdness. Despite the test setup having 20 dB of attenuation in the signal path the bias-T performs the miracle of over-unity performance. Of course this is not real! Again it is just conjugate matching something in the system to make the reference calibration invalid. The variation is only a fraction of a dB so for all intents the bias-T is near-perfect across HF and into low VHF.
Lab Testing
It is rather difficult to lab test a loop antenna in a reasonable way. In particular it is extremely difficult to immerse it in a RF field of sufficiently controlled spatial uniformity and consistent amplitude with frequency to make absolute and repeatable measurements. For my initial testing a 100 mm diameter coupling loop was connected to the signal generator and loosely coupled to the antenna loop. Sufficient drive was applied to achieve a few dBm out of the loop buffer and into the power meter at peak of resonance. This allowed crude measurements of bandwidth and amplifier compression.
More detailed investigations were made by sweeping the unit with the same experimental scalar analyser used to test the bias-tee. Exact amplitude measurements made in this way are fairly meaningless, as the coupling loop is not well matched to the generator. 10 dB of padding was placed between the coupling loop and the generator but as the loop was placed almost orthogonally to the antenna loop to provide weak coupling (for Q estimation) the coupling loop does not see much of the antenna loop loss resistance.
The loop tunes 7.23 MHz to 12.32 MHz with the polyvaricon used. The apparent loop Q drops with frequency, being 55.6 at 7.23 MHz, then 33.3 at 10.33 MHz and 28 at 12.32 MHz. The loop inductance is about 4.1 uH so the input impedances which match these Qs are 10 kΩ, 8.8 kΩ and 8.9 kΩ respectively. Wider sweeps show problems with the test set-up, in particular generator harmonic energy when tuned below the loop resonances. Neither sweeps have me feeling very comfortable about the quality of test set-up (or the loop construction for that matter). The HF feed-through is probably due to the unshielded housing of the buffer amplifier and stray circuit capacitances. Maybe a LPF should be added to the output to suppress these responses? A HF receiver should reject them with no dramas, but the number of hints of internal resonances and general "complexity" of the baseline above and around resonance doesn't make me too comfortable.
When the amplifier is left unpowered the loop leaks through RF at a lower amplitude, and the resonance is shifted down in frequency somewhat. This is immediately apparent when you connect a receiver to the loop, even without powering it on you can peak-up the background noise level by tuning the polyvaricon. However once power is applied the loop must be retuned (up somewhat) for maximum background noise. Loop Q is degraded quite significantly in the leak-through mode, with Qs of 39.3, 17.1 and 9.1 for the test frequencies discussed above. Loop Q in general could be improved by weaker coupling to the JFET gates, some simulation could optimise the values required if the resonator usable Q was measured and the FETs well characterised. Noise figure would be compromised by resistive DC gate biasing, maybe use chokes?
On thing I initially found rather disturbing about this particular sweep is the 2nd harmonic peak is absent from the "off" run. The 3rd harmonic peak is 15 dB different just as the fundamental is - but the 2nd harmonic "on" signal is smaller and 15 dB down from its level is into the -41 dB "leakage floor" we appear to be observing... All was revealed when I considered that the reference level here is about -12 dBm meaning the leakage is around -53 dBm, quite likely considering the unshielded nature of the amplifier in close proximity to the coupling loop.
Field Testing
For real RX testing I lashed-up the antenna on the balcony, mounted a little above railing height, and A/B compared it against the base-loaded vertical I use for QRSS transmission. This is *not* a very fair test, as the vertical is 3 metres tall, while the loop is less than half a metre in diameter and was mounted at the base of the vertical. Also, interaction between the antennas was *not* controlled during the experiment, and experience has taught me this is critical for meaningful results.
The loop is largely limited by its aperture (cross-section). Compared to my loaded vertical it is about 6 dB down, making it ineffective for its original design purpose (improving my QRSS RX noise floor). However, unlike the omnidirectional vertical it has well defined nulls which are useful for avoiding local interference. The nulls do help dodge some noise, but unless I make it larger or mount it higher the loop is simply not as good for QRSS RX.
For comparison purposes I have built a much larger passive tuned loop antenna. The square loop is root-2 metres on a side in a diamond configuration (for mechanical simplicity) and is matched to the coax using a ferrite transformer placed near the polyvaricon that tunes the loop winding to resonance. It was not possible to fit the loop in the shack for Q-determination and hence the matching is at best an educated guess. Its Q is likely dominated by whatever the receiver input impedance reflects into the loop tank through the matching transformer. Experiments continue, comparing the three antennas for relative performance. More experience and additional test instrumentation is required before fair comparisons can be made.
2 comments.
Attachments
| title | type | size |
|---|---|---|
| Active Loop Circuit Diagram Source | application/postscript | 13.671 kbytes |