Building a G3JKF Magnetic Loop Array

Despite having a plethora of conventional and well-functioning antennas outside, I keep a constant interest in magnetic loops and their various designs.

The attraction of a loop is clear: small, not very prone to electrical noise, wide band coverage, and eminently suited to indoor use.

For a couple of years, I've been in touch with Ken Franklin, G3JKF, who has very carefully developed and tested, over a long period of time, a three loop array.  If you ever catch him on WSPR mode, you'll be amazed at how well his loops do.  A characteristic of Ken's signals is that he reaches just about everyone that any other, full-sized antenna operator reaches, often with comparable (or even better!) signals.  Rarely are signals from Ken significantly down on full antennas.

So, having messed about for ages, I decided I'd make use of a day off, and put together a G3JKF loop for 40m-10m. A 7.5-350pF Russian vacuum cap has just enough low-side range to tune 10m to a perfect match at the digital end of the band.

Before I start detailing the construction, remember that soldering with a blowtorch can be pretty darned dangerous! This is especially true of this build because you may find yourself soldering 'upside down' fittings, which may drip hot flux and/or solder onto your skin.  Eye protection and some decent heat-resistant gloves are advised.  Flux is also very corrosive, and generates noxious fumes that can give you a belter of a headache.  So plenty of open windows if you are working indoors.

You need pre-soldered, 15mm fittings as follows:
14x 90 degree elbows, 4x 45 degree elbows and 6 tees.  

You need a total of about 13 metres of copper pipe, but there is scope to recycle some you may already have. 

You need a pipe cutter, soldering flux (this is essential), and a gas blowtorch.  You can also use an electrical pipe heater, but these are too slow for me!  Also, some coarse wire (steel) wool.

Remember to clean each joint immaculately with wire wool before soldering, and add a liberal amount of flux to the pipe to be fixed. 

Start by making the complex-looking top joint for one half, as per the image.  This first one has to be assembled in a vice, all the fittings made square, and all of them soldered together in one go.

The first soldering is the most complex.  The arms are cut to 130mm each.

When soldering, heat the pipe ahead of the joint first, then move on to the joint itself.  If you heat just the joint, solder may run, but the underlying pipe may not be quite hot enough.  When you see a complete ring of solder appear at the joint, stop heating.  

When soldering pieces of tube, you may need to support them to make sure they're reasonably square and not sagging - there is some play in the soldered fittings.

With care, you should be able to control the heating well enough to melt one side of the connectors whilst the other just about remains solid.  When one joint is done, move straight to the nearest next connection to be made, until completed.  Give a good 15 minutes or more to cool down, as copper retains heat for a remarkably long time.

If, for some reason, a joint looks a bit suspect, you should first heat it to get rid of any blobs of solder, then clean it thoroughly with wire wool.  Add some flux at the joint, and heat the joint until solder melts into the join.  This should fix most failed joints, which will happen sometimes.

You then need to add the 45 degree elbows, and two pieces of copper 525mm long.  Solder these to the arms coming off the central connecting section.

The completed top joint for one half of the loop, plus the 45 degree bends.  The longest, topmost tube is just a connection to the capacitor, however you configure it.

At the end of these pipes, you add a further 90 degree elbow, and a 1m long piece of copper pipe.  Continue adding elbows and 1m copper pipe until you get to the point where it needs to join the second top joint section.  Repeat for the other side, at which point you will have nearly finished two complete loops of the array.

To join the bottom elbows, which you can't really access at times, use a large wood saw's blade near the handle, to act as a reflector of heat, a hint sink with an air gap to stop the floor being burned(!) and  a catcher for hot solder or flux.

You then need to make a second top joint, exactly like the first, and eventually join its arms to the outer loops.


The central loop vertical tubes have to be a bit bigger - 1100mm - to allow for the 'up and over' as a result of using a tee at the top joint.  The arm from the elbow at the tee to the second, down-going 90 degree elbow is 395mm long, of which you need two, of course.  Complete the loop all the way round.


My support is a simple lattice of light timber, held aloft at about 7 feet with a timber stand.  PVC clamps hold the loop firm. Details as per the photo.

You now simply need an air spaced or vacuum variable capacitor.  Anything in the range 10 - 1000pF should do, but my vacuum cap is 7.5-350pF, and tunes 40m-10m. The feed point is a simple SO239 connector to a gamma match.  A very short wire runs from the outer to the centre of one of the outer loops. The centre pin is connected to a piece of drawn copper or similar, which is soldered about 80% of the way up one of the sides. There are many matching variants, and you may prefer another.

 
Et, voila!  You have what must rank as the pinnacle of magnetic loop design.  Being 1m cubed, more or less, I can't actually get mine out of the kitchen where I built it without chopping the loops in half!  Eventually, I plan to house the loop in a timber frame with fibreglass panels and 'Coraline' corrugated roof; that' much easier than weatherproofing the capacitor, and avoids the effects of our heavy winds.

The completed loop, now with vacuum cap 7.5-350pF.  Braid straps to the cap ends are kept to the bare minimum - about 1 inch.

Very short, wide copper braid to the loop, kept in place with steel clamps. You should not use this method other than for testing, because soldering is better to reduce losses.  Cap is held in place with a ring of cable ties, and another to the copper tube each side.

There are very few of these loops that I know about, so if you have the time, the inclination to do some soldering, and a small amount of space to spare, why not have a go and put one to work on WSPR?  Total cost, even if you bought all the materials and tools, is only about £120, and much less if you already have tools and some spare lengths of copper around.



UPDATE (i):  Today (22/12/2014), 19 pairs of 21MHz WSPR reports heard by a combination of the respected stations, K9AN and N2NOM yielded, for both stations, a median difference of 0dB between my indoor three loop array and, in most cases, the sloping dipole operated by the well-run station at G8VDQ.  M1AVV and M0XDC added a couple of data points to the total.  Spots excluded the time during and leading up to the grey line, as there are significant differences in signal enhancement due to even quite small longitude separation between stations.

UPDATE (ii): Following discussions with G3JKF, the feed system has been modified to a 'parallel gamma' match.  Firstly, the coax goes into a homebrew or commercial 2:1 or 4:1 balun, rather than directly to the loop. In my case, a 4:1 balun yielded a perfect match at the upper end of the array's range, but a 2:1 didn't.

The centre pin from the balun is connected to a triangular section of microbore tube or thicker copper wire, which attaches to the loop at the same, 80% point up the side of one of the outer loops.  This returns to the centre pin parallel to the side of the loop, spaced by about 2cm, and then parallel to the bottom of the loop.  Adjusting the spacing can lower the SWR somewhat, and I found it was better to be at 2cm than 1cm, for example.

The 'parallel gamma' feed arrangement.  4:1 balun centre pin to triangle, attached to loop at 80% up left side; outer to short wire connecting to the centre of the lower leg of the loop. 

The outer conductor goes from the balun to the centre of the lower part of the loop, as it did under direct coax connection.

These changes bring the loop to a perfect match at 10m, whereas previously, it was nearer 1.5:1 or greater.  WSPR tests show the loop works very well at 10m.

~5:30-9UT 40m results, 2015 Jan 18.