Impedance was measured with an AA-600 looking into 500mm of RG400 then the Guanella 1:1 balun, then 9m of fabricated transmission line as described in earlier articles in the series.

Above is the impedance measurement plotted on a Smith chart. This is more useful and very meaningful as an interactive display in Antscope where are you move the cursor, the frequency and key data are displayed.

The impedances on some bands are towards extreme, they would be well beyond the capability of most internal ATUs, but can be matched easily and reasonably efficiently with a good T match.

Above is a plot of R,X,|Z| for the same data. One of the shortcomings of Antscope is that it is not possible to disable the |Z| trace which seems to be there for appeal to so many hams who think of impedance as a DC like quantity (ie scalar).

The impedance measurement was at the ATU terminals essentially, and the voltage on the load side of the balun is of greatest interest in this scenario. The impedance measurement was adjusted to back out the 500mm RG-400 tail and 0.995m, vf=0.75 of 113Ω line in the balun.

Above is a calculation of voltage and current at the load side of the balun at 3.6MHz. It can be seen that at 100W, peak voltage is fairly low.

Above is the same calculation for 7.15MHz, voltage is a little higher but still quite manageable.

On the lower bands, the line sections are electrically short and the ATU voltage will not be greatly different.

]]>

Many times I have tried to validate it and run into problems. At one time I reported them to the author, but they were never acknowledged, much less fixed.

The specific problem on this occasion relates to the receiver performance tab.

Above is a screenshot (with my annotations) where I have basically stripped the configuration down to a *receiver* attached to a noiseless antenna with lossless line.

The *receiver* is specified in the box at (1), the up down arrows used to adjust the value. In this case, the *Noise Figure* is 2dB and the untitled quantity to the right is the equivalent noise temperature in kelvins (correctly calculated).

Note that the equivalent noise temperature at the receiver input (169.62K) is the ONLY noise source in the whole model, everything else is noiseless.

If we look at (2) we see two boxes entitled *Rx Noise Temp °K* (the ° symbol is an error) with value 329.4K. Despite the apparently similar title, this is clearly a different quantity to that at (1)… it is almost double (1).

But then the *Rx Noise Figure* box at (2) shows the *Rx Noise Figure* to be 2.00dB. But this does not reconcile with the *Rx Noise Temp °K* value at (2).

Now look at (3) to find *System Noise Temp* to have yet another value of 335.4K. Since there is no other noise in the system, no loss or gain elements other than the *receiver* at (1), one might have expected the value to be the same as (1) (169.62K).

Also at (3) is *Sys Noise*, presumably the system noise figure corresponding the *System Noise Temp*, and it is exactly that.

So whilst the values at (3) are consistent with each other, they don’t seem to relate to anything else detailed in the model on this tab. The term *receiver* or abbreviation *rx* seems to be used rather loosely to refer to at least two different entities (though in the simplified lossless model two should be equivalent).

This is only the start of difficulties I had reconciling / understanding the calculated values on this tab. The problems here were exposed by trying to simplify the model to explain issues with more complex scenarios.

This is not the first time I have had difficulty understanding or reconciling what are fairly basic calculations and I have come to the conclusion that I cannot depend on the tool and am not surprised that it does not reconcile with other tools.

Mind you EME Calc is apparently widely used by hams, but possibly by people who have not validated its results.

]]>Above is a view of the steel mast with the Inverted V G5RV rigged from the top of the 11m mast using a halyard though a purchase on a small gibbet to offset the antenna and feed line from the mast. There are lateral guys at 7m height, and the left hand one is non-conductive synthetic fibre rope. Atop the mast is a 2m/70cm vertical.

Above is close up view of the dipole feed point.

The open wire feed attaches to the dipole legs using stainless steel clamps shown above in mockup. the actual construction was liberally coated with marine grease to exclude oxygen and water from the electrical contact.

Above is a view from one support post to the feed point. A winch type agricultural fence strainer is used to adjust wire tension for the desired sag / tension (4.4N) to withstand 144km/h windspeed with safety factor 3.5.

The feedline option used was the Sotabeams 110mm ABS/PC spacers.

Above is the Sotabeams 110mm spacer threaded with the 1.6mm aluminium wire.

Above is a view looking up the feed line which is twisted and supported away from conductors (the lateral guy to the right is non-conductive).

Above is a view of the feed line approach to the cable entry panel. Again, liberal use of marine grease to exclude oxygen and water from the balun terminals.

Above is the balun and cable entry panel. The balun is described in a series of articles starting at Design / build project: Guanella 1:1 ‘tuner balun’ for HF – #1.

Above is detail of the insulator termination.

The technique for the right hand connection is shown in a video. The tail is 7×19 galvanised FSWR swaged at the insulator, but almost anything could be used.

On the inside of the feed entrance panel is a 300mm RG-400 jumper to an ATR-30 ATU. This is the so-called tuned feeder implementation described by Varney in his papers detailing the G5RV.

This antenna system is as described by (Varney 1958), the famous G5RV. Above is Fig 2 from (Varney 1958) where he details the purely open wire feed (or “tuned feeders”) configuration. (The scan is poor, the right hand side of L2 connects to the right hand side of the variable capacitor at the bottom of the schematic.) The ATU used though is a T match with an effective 1:1 current balun and 50Ω coax to the transceiver

On test, it works just fine… but the true test is its survival and performance over time.

- Varney, Louis. July 1958. An effective multi-band aerial of simple construction In RSGB Bulletin July 1958.

This article is a walk through of the expected WSPR receive S/N for the case of the 20mW tx on a quarter wave vertical.

Ground wave suffers attenuation due to two key components:

- dispersion of energy as the wave spreads out from the source; and
- absorption of energy in heating the soil.

Item (1) is simply inverse square law effect, and Norton provides us with several approximations for estimating (2) from Sommerfields work.

Calculate efficiency of vertically polarised antenna from far field strength uses Norton’s f5 approximation for ground wave attenuation.

Above is a calculation for a 100% efficient transmitter. (The trick to getting this is to leave the *measured field strength* field empty and the calculator will insert the value that gives 100% efficiency.)

So the next question is what ambient noise level might we expect in a rural setting on 40m.

ITU-R P.372-12 gives us guidance, suggesting the median ambient noise figure (NF) at 7MHz in a rural precinct is 44.6dB. This will dwarf the NF of a good HF receiver which will be better than 15dB without preamp, so we will run with total NF of 44.6dB.

Let us now calculate the signal to noise ratio (S/N) in 2.5kHz noise bandwidth of the rx signal of -7.1dBµV/m with NF of 44.6dB. We will assume that the rx antenna is a 100% efficient QW vertical with directivity 3 (gain 4.77dBi).

Again a handy online calculator can be used, Field strength / receive power converter.

Above is the input data for the problem. The use of 2500Hz noise bandwidth is to obtain results in terms of WSPR’s reference for S/N reports.

Above the results contains at the end, the rx S/N which is 2.46dB.

Because the external noise is much greater than the rx internal noise in this scenario, small reduction in rx antenna efficiency and gain will not significantly alter the rx S/N ratio, though it will reduce the received power.

Rx ambient noise varies with location, time, polarisation, neighbourhood effects, atmospherics etc. The ambient noise scenario modelled here is for a rural precinct, and city noise could easily be 10dB or more worse. Rx ambient noise in this scenario dominates total noise, and any difference would roll directly into rx S/N.

A further source of noise is unwanted signals. Because of the uncoordinated way in which WSPR works, there is a significant risk of interference from stations that might reduce the S/N ratio observed on the desired signal.

The tx antenna used for Richard’s experiment is likely to have an efficiency of perhaps -1dB give or take, and that would directly roll into observed rx S/N ratio.

Soil is not homogenous, and ground type for Richard’s test is unknown. Soil type has a significant effect on path loss over such a long ground wave path, and that rolls directly into observed S/N ratio.

Richard collected 169 paired observations of two transmitters received at one receiver 20km distant on 40m.

The observed median S/N ratio for the transmitter discussed in this article was 1dB, which is quite in keeping with the prediction of 2.5-1=1.5dB for a typical antenna over an ‘average ground’ path with typical rural ambient noise.

There is a great deal of uncertainty in the estimate, and the largest element would be the soil type assumed, next would be the ambient noise at the receiver. Nevertheless, the article describes a process for estimating the performance and the estimate is improved by better knowledge of all the influencing parameters.

- ITU-R. Jul 2015. Recommendation ITU-R P.372-12 (7/2015) Radio noise.
- Norton, KA. Dec 1941. The calculation of ground wave field intensity over a finite conducting spherical earth. Proc IRE, vol 29, pp 623-639.
- ———. Oct 1936. The propagation of radio waves over the surface of the earth and in the upper atmosphere Part I. Proc IRE, vol 24, pp 1367-1387.
- ———. Sep 1937. The propagation of radio waves over the surface of the earth and in the upper atmosphere Part II. Proc IRE, vol 25, pp 1203-1236.

Exploiting your antenna analyser #28 gave an example of use of one method to resolve the sign of reactance comparing measurements made with a slightly longer known transmission line.

One way to predict the input impedance to the longer line is using a Smith chart. This article presents a Smith chart prediction of the expected input impedance of a 8′ section of RG8 at 14.17Mhz (vf=0.66, length=0.175λ) for the cases of Zload being 60.3+j26.9Ω and 60.3-j26.9Ω.

The impedance is normalised to 50Ω and plotted on the Smith chart, point 1 above. A radial from the centre through point 1 is drawn to the edge of the chart. Another radial is drawn a distance towards the generator of 0.175λ and using a pair of dividers or ruler, point 2 is plotted on that radial at the same distance from the centre (same VSWR) as point 1.

These points are on a constant VSWR arc but the arc has not been draw because the two arcs would overlap and might be confusing to some readers.

The impedance is normalised to 50Ω and plotted on the Smith chart, point 3 above. A radial from the centre through point 3 is drawn to the edge of the chart. Another radial is drawn a distance towards the generator of 0.175λ and using a pair of dividers or ruler, point 4 is plotted on that radial at the same distance from the centre as point 3.

It is clear that the predicted R value of the points 2 and 4 are very different and will be sufficient to differentiate the +ve and -ve X cases.

Reading of the normalised R value for each and multiplying it, we get R=52Ω for point 2 (the +ve X case) and R=30Ω for point 4 (the +ve X case).

The predicted R values of 52Ω and 30Ω are very slightly different to the more exact values using in the original article a good transmission line calculator because the Smith chart solution above is for a lossless line, and being a graphic solution is it not as precise as the mathematical one.

Nevertheless, the Smith chart solution is quite good enough to differentiate the two possibilities and to show that the sign of X in the original measurement is in fact -ve.

Watch the blog for continuing postings in the series Exploiting your antenna analyser. See also Exploiting your antenna analyser – contents.

]]>Above is the internals of one, it is a 51Ω 5% metal film resistor.

They often fail a DC test and tapping them gives erratic resistance readings up to hundreds of ohms, and of course they can be unreliable at RF.

They rely upon the resistor pigtail to make a spring contact with the inside of the barrel, and give that the pigtail is soft copper with little spring the contact is not very reliable.

The second major problem is that even with good contact, the VSWR is significantly worse than specification.

Above is a typical VSWR sweep to 600MHz, VSWR exceeds the specification above 230MHz and they are not really useful above 100MHz but for the crudest applications.

Why is the VSWR so bad?

Above is a resistor from one with the epoxy coating abraded away to reveal that it is a metal film resistor with a helical resistance element. It is an inductor with resistance. The Chinese copyists may have seen a design using a carbon composition resistor, or low inductance metal film resistor and substituted the cheapest thing they could procure, copied the specs and flog them to unwary buyers online.

]]>Many analysers do not measure the sign of reactance, and display the magnitude of reactance, and likewise for magnitude of phase and magnitude of impedance… though they are often incorrectly and misleadingly labelled otherwise.

The article The sign of reactance explains the problem and dismisses common recipes for resolving the sign of reactance as not general and not reliable.

This article gives an example of one method that may be useful for resolving the sign of reactance.

My correspondent has measured VSWR=1.68 and |Z|=66 and needs to know R and X. From those values we can calculate R=60.3 and |X|=26.9.

The method involves adding a short series section of known line, short enough to provide a measurement difference in R, and that R would be different for the case of =ve and -ve X, all of these measured at the same frequency.

There is a risk when the measurement is of an antenna system that significant common mode current may alter the measurement, so lack of consistency with expectation flags a potential problem that needs to be investigated.

An additional 8′ section of RG-8U was inserted and the impedance looking into that section was measured at 14.17MHz.

Now let us predict the input Z to an 8′ section of RG8 with loads of 60.3+j26.9 and 60.2-j26.9 using a good transmission line calculator.

For the +ve X case, the R component of Zin is predicted to be 57Ω, for the negative case it is predicted to be 32Ω. These are sufficiently different for the test to be conclusive.

In the event, the measured R was close to the 32Ω predicted for the case of X being -ve.

On the basis of Rin being close to prediction for the -ve X case, it can be reliably concluded that at the first point Zin=60.3-j26.9Ω.

The importance of this was that the known value allowed calculation of the feed point impedance of the antenna which was at the end of 6′ of RG8X to be 79.2+j14.6Ω. That informs the design of a matching scheme.

Watch the blog for continuing postings in the series Exploiting your antenna analyser. See also Exploiting your antenna analyser – contents.

]]>

I have constructed NEC-4.2 models of a 52MHz ground plane with four 45° inclined radials at 10m height above ‘average ground’ (0.005,13) on and connected to a conductive support pole which is bonded to ground at the lower end, and one with horizontal radials.

Comparing the patterns at low angles shows there is not much in it, but below 32° elevation which tends to be of greater interest at VHF, the winner is actually the inclined radials though the difference is less than 1 dB.

Now the patterns are dependent on height, soil type, common mode current on the masting, and radial inclination… but trials at a range of practical heights did not reveal a significant difference between the configurations.

In this case, the radials needed to be inclined down 58° for a 50Ω match, and the patterns are very slightly different.

Above, a trivial difference between the 45° and 58° patterns at low elevation.

It is one of the many cases of accumulated ham wisdom that does not stand scrutiny with modern analytical tools. This stuff is stated as fact with evidence, and part of the ham tradition is to learn it without question and repeat it like the pros.

Small amounts of lumped shunt capacitance at the feed point will ‘detune’ the antenna and require lengthening of the radiator to achieve minimum VSWR at the desired frequency, and in the process it will raise the feed point R component a little (by virtue of the L network created).

Choosing sufficient C could be used to match a GP with horizontal radials, but even inadvertent small amount of C in an antenna mount might account for a GP matching with lesser inclination of radials that the models above indicate.

The model included a conductive mast from radials to ground. Most analyses of GP antennas ignore common mode current, indeed it seems hams by and large insist that there is no common mode current by virtue of the radials and it is unusual to see counter measured deployed. The common mode current affects pattern to some extent, may contributed to EMC problems, and in the case of HF RX may degrade noise floor.

]]>An important attribute of such devices is their ability or not to measure the sign of reactance.

Review of the MFJ-225 manual lists features like “Complex impedance (R and X)” which might trick buyers into thinking that it can measure X, but in fact it can only measure the magnitude of X (which is usually written as |X|), and “Phase (0-180°)” which is wrong, it measures the magnitude of Phase |Phase|.

In fairness, deep in the manual is the statement:

8.5 Sign Ambiguity (±j): Most handheld analyzers, including the MFJ-225, lack the processing capability to directly calculate the reactance sign for complex impedance (Z = Rs ±j).

A cleverly crafted statement to hide behind a claim about “most handheld analysers” which is arguably untrue. It doesn’t matter when some or most cannot display sign of reactance, the simple fact is that if you want a handheld instrument that does correctly measure the sign of reactance, there are plenty available. The one unequivocally true part of the statement is simply:

**The MFJ-225 does not display the sign of reactance or the sign of Phase.**

So you might ask why then the front panel annotation is not |X| instead of X, |Z| instead of Z, and |Phase| instead of Phase. It is misleading, either dishonestly so or just honest incompetence.

The idea behind the 5/8λ ground plane popularity is that claim that it has higher gain at low angles than a simple 1/4λ ground plane.

The 5/8λ ground plane is not resonant, and the feed point impedance is hardly suited to direct coax feed.

The chart above is for a 5/8λ ground plane elevated to 5m height above average ground (0.005,13). The feed point impedance in this case at 5/8λ radiator height (14.2MHz) is about 110-j485Ω.

The most common feed described is a series inductor to cancel the large capacitive reactance, but in this case that leaves a feed point resistance of 110+Ω (+ to include inductor loss), give a VSWR(50) of more than 2, hardly a ‘match’.

Is there a simple solution?

The solution is to add a little capacitance between the bottom of the radiator and the ground plane, it is the yellow element on the plot above, 11pF. This transforms the impedance to have R=50, and that can be transformed to 50Ω using the 3.73µH series inductor.

More on this shunt capacitance later…

The tapped inductor match is often proposed and most hams who do not have basic circuit analysis skills and an understanding of coupled inductors might follow W5DXP’s guidance:

Seems to me, if the inductive reactance of the coil equals the capacitive reactance of the antenna, then any tap point on the coil will be purely resistive and somewhere it will equal 50 ohms.

In fact this match is not the no-brainer the W5DXP and others think.

The impedance at the base of the radiator in this example is 110-j485Ω. It is a challenge as it lies on the ‘wrong side’ of the R=50 circle to match it using the tapped inductor.

The tapped coil is in fact two coupled inductors (flux coupling factor 0.35) and is modelled as a classic T network (L1, L2, L3 in Winsmith model below). The low flux coupling factor means that any analysis based on 100% flux coupling are incredibly naive and hopelessly inaccurate.

Above is a lossless match that solves the problem. The chart is done as a lossless match because practical losses are very small and of little consequence to explaining the technique… other than complicating the explanation.

The solution is to add a little capacitance between the bottom of the radiator and the ground plane, it is the yellow element on the plot above, 20.6pF. This transforms the impedance to have R<50, and that can be transformed to 50Ω using the tapped inductor as shown on the chart.

More on this shunt capacitance later.

Above is a table of the key parameters of the match as optimised using Excel’s solver. Ql has been made very large to simulate a lossless match to give a simpler Smith chart.

Both solutions above used a small shunt capacitance at the base of the radiator. In many implementations, the structure supporting the vertical radiator introduces or can introduce some small amount of capacitance in shunt with the base of the radiator. This is why some constructors have not had a problem, their imperfect support structure supplies the necessary element to making matching work.

Of course any shunt capacitance required is reduced by that supplied by the support structure.

Let us look at two simple alternatives to a shunt capacitance in the case of the series inductor match. The same technique could be applied to the tapped inductor match, but it is an unnecessarily complicated solution driven by a lack of understanding of how the system actually works.

Another method is to use a short section of transmission line to achieve the same outcome.

Above, the match using 2.8° (eg 108mm of RG213) series section (yellow) and 3.7µuH series inductor to obtain a match.

Of course any series line section required is reduced by the capacitance supplied by the support structure.

Alternatively, a shunt o/c stub could be used.

Above, the match using 2.8° (eg 108mm of RG213) shunt stub (yellow) and 3.72µH series inductor to obtain a match.

Of course any shunt section required is reduced by the capacitance supplied by the support structure.

Some of the ham theory of matching 5/8λ ground planes involve some pretty wooly thinking, mostly driven by shallow understanding of basic circuit analysis.

For some constructors, it all just works somewhat simply as a result of some essential small shunt capacitance provided by the support structure for the vertical radiator.

Others have no end of problem trying to find a solution, but there are simple solutions. The solutions that are best suited to particular implementations depend on the structures involved, weatherproofing considerations etc and are left to the implementor to design.

The numerical examples above relate to a specific example scenario, individual implementations will vary a little but the techniques explained can be applied to them. Clearly accurate measurement of the base impedance of the radiator as built is a great start to tailoring the matching system to suit.

These are not the only ways of matching a 5/8λ ground planes, but they are probably the most popular.

]]>