Carol gives the following table of measurements and calculated results.

Table 1. Transverter Measurements | |||||||
---|---|---|---|---|---|---|---|

Freq MHz |
Noise Source ENR (dB) |
Noise/10 kHz |
Conversion Gain (dB) |
Noise Figure (db) |
|||

50 Ω expected | Noise On | Noise Off | On-Off (Y) | ||||

144 | 15.2 | -134 dBm | -118.8 dBm | -132.1 dBm | 13.3 dB | 26.5 | 2.1 |

432 | 15.3 | -134 dBm | -118.7 dBm | -131.7 dBm | 13 dB | 24.1 | 2.5 |

Lets focus on the 144MHz measurements.

These are a measurement of the system (ie Flex 1500 transceiver + Electraft XV144 transverter) and Y factor based calculation of Noise Figure (NF), and note that the Noise ON and Noise Off figures are from the Flex 1500 which has 26.5dB subtracted (a calibration adjustment for expected transverter gain).

This article presents a simulation of a two stage measurement which establishes the NF and Gain of the transverter. The two stage technique is described at Noise Figure Y factor method calculator.

To achieve that, we must determine the NF of the Flex 1500, the ‘instrument’. Since the necessary measurements were not made, for the purposes of this article I will assume that the NF of the instrument is 10dB, that the noise source has an ENR=16dB at the transverter output frequency, and calculate the Noise Off and Noise On powers in the same measurement bandwidth for that scenario. Instrument NoiseOn=-117.0dBm and NoiseOff=-124.0dBm.

For the system measurements, I will add the gain adjustment back in to the Table 1 figures, so DUT NoiseOn=-92.3dBm and NoiseOff=-105.6dBm, and the ENR was given as 15.2dB.

Entering them into Noise Figure Y factor method calculator we have:

and the calculated results:

The process calculates the Gain and NF of the transverter itself to be 26.3dB and 2.05dB respectively, and the system NF=2.11dB.

These figures depend on my assumption of the NF of the ‘instrument’, and different figures will flow into some differences in the calculated results.

- Milazzo, C. Aug 2015. Signal level measurement with PowerSDR and external transverters.

]]>

Let’s explore his second option, as unlike the first, it does work reliably.

Above is an NEC-4.2 model with current shown (magnitude and phase). The stubs conductors are all defined from top to bottom.

The S/C stubs are capable of supporting both common mode and differential mode currents.

Let’s take the current in each conductor of one of the stubs and calculate the magnitude of the common and differential mode transmission line currents.

Above is a plot of the common mode and differential mode current calculated from the NEC model. Note that |Ic| is not zero, so any analysis of how the stub works that has ignored this element of its characteristic, and as it happens a really important element, is incomplete.

Let’s look at a device that provides the differential mode behavior, but without clearly simulating the common mode behavior.

The figure above shows such a coaxial implementation from (King et al, 1945) Fig 22.3(b). (King et al, 1945) notes “their operation is, however, much less satisfactory than with the open wire stubs”. They also state “It is important to note that the inside diameter of the coaxial sleeves must be large compared with the diameter of the antenna and their length considerably less than λ/4 if a sufficient phase shift is to be achieved”.

This is all a bit nebulous, and they are struggling with this phase shift explanation, giving little guidance for effective designs. Nevertheless, these designs are not uncommon and rarely provide evidence of cophase operation.

Lets consider this firstly from the desired charge distribution on the horizontal wires of the cophased antenna.

Taking another figure from (King et al, 1945), we can see that between each half wave section, we need some device that:

- accommodates a large charge difference between the conductors on each side of it; and
- minimises the common mode component of charge on its terminals.

If the antenna was not cophased, there would be a large common mode element of charge at the device terminals, and so spoiling that condition is a path to preventing that undesired mode exciting successfully.

The quarter wave S/C stub (as in the diagram above) fulfills these functions as:

- the differential input impedance of the quarter wave section is high, so permitting the high charge difference required for cophase mode; and
- the stub in common mode acts as a common mode charge sink, the charge maximum in common mode being at the outboard common mode open end of the stub.

There may be other devices that work effectively, for example forming the quarter wave stub into a spiral about the horizontal conductors can make a more compact form which has advantage in a vertical collinear.

But there are many devices employed in published designs that are unlikely to work, and either have not been measure to prove co phasing, or the inventor’s measurements demonstrate they are not cophased (eg the W5GI Mystery Antenna).

- Franklin C. Aug 1924. Improvements in wireless telegraph and telephone aerials GB Patent 242342.
- King, Minmo and Wing 1945. Transmission Lines Antennas and Waveguides. New York: McGraw-Hill.

The original transformer above comprised a 32t of 0.65mm enamelled copper winding on a FT240-43 ferrite core, tapped at 4t to be used as an autotransformer to step down a load impedance of around 3300Ω to around 50Ω.

The FT114 core has a quite low ΣA/l value (0.000505), essentially a poor magnetic geometry.

A better choice for his enclosure is the locally available LO1238 core from Jaycar (2 for $5) with ΣA/l=0.0009756/m which is comparable with the FT240 form (though smaller in size) and nearly double that of the FT114. The LO1238 is a toroid of size 35x21x13 mm, and medium µ (L15 material).

A more detailed analysis of a 3t primary winding of the effects of magnetising impedance on InsertionVSWR and system loss when it is in shunt with a 50Ω load was performed.

Above is the expected core loss.

Above is the expected InsertionVSWR.

These both look encouraging, and the next step would be to build and measure some prototypes.

Above, VK4MQ’s prototype in development. (I do not recommend the pink tape.)

]]>The original transformer above comprised a 32t of 0.65mm enamelled copper winding on a FT240-43 ferrite core, tapped at 4t to be used as an autotransformer to step down a load impedance of around 3300Ω to around 50Ω.

A very rough approximation would be that with two stacked cores, the number of turns would be around the inverse of square root of two, so 70% of the original.

A more detailed analysis of the effects of magnetising impedance on InsertionVSWR and system loss when it is in shunt with a 50Ω load was performed.

Above is the expected core loss.

Above is the expected InsertionVSWR.

These both look encouraging, and the next step would be to build and measure some prototypes.

To the original question, would half the turns be enough? No. Notwithstanding that, you are likely to find such being used, being sold.

]]>Some of the data is derived from manufacturer’s published complex permeability curves. The plot above shows the Ferroxcube’s published curve for 3C81 material, and points at which it was digitised to extract a table of µ’ and µ”.

The calculator result is a cubic spline interpolation of the tabulated data.

Some data is derived from manufacturer’s published complex permeability tables, most or all of the Fair-rite data is sourced in this way.

Above is a plot of Fair-rite’s published complex permeability table for #77 mix. Note that µ’ has negative values from 7-1140MHz which raises concern for the quality of the data, it may be the result of blind extrapolation.

Whilst I prefer to use data traceable to the manufacturer, I favor a solution that discards data that is on the face of it suspect.

So, in processing these tables data is discarded where µ'<0.95 or µ”<0. Only interpolation is done (ie no extrapolation). Note that this would include some data from the plot above that is probably extrapolated by Fair-rite, but not obviously wrong. It would be better if they published just what they measured. This change took effect on 07/09/2019.

The calculator result is a cubic spline interpolation of the tabulated data.

The calculator includes data for some cores for which published data could not be found. They are mostly characterised by my own measurement.

I took a baseline measurement with an AA-600 after some refurbishment work in Jan 2018, and was able to compare a current sweep to that baseline.

Above, a wide Return Loss sweep of the Diamond X-50N with feed line compared to the baseline (the thin blue line).

By and large they are almost identical, save small departure around 435MHz.

Above is a comparison of the Return Loss at low values. Antscope does not display mathematically correct plots when the data goes off scale (as in this case), this plot is mathematically correct and allows better comparison of the important out of band Return Loss.

It is worth remembering that the AA-600 operates on second harmonic above 200MHz, and third harmonic above 400MHz, so the measurements become a little noisier.

Importantly, the out of band Return Loss is almost identical and this would not be so if feed line loss had degraded (eg due to water ingress), so there is no evidence to suggest that the feed line had degraded.

Above is a narrower sweep around the normal operating frequencies. There is a small degradation in Return Loss which is probably attributable to temperature differences of more than 20° between measurements.

So, the comparison with the archived baseline gives no cause of concern, the antenna system is probably unchanged.

Well, in fact I have done just that at 80 frequencies over a wide range, in-band and out-of-band, if you like. That captures much more information than VSWR measurement at one or a few frequencies.

The traditional ham approach is the measure VSWR at the operating frequency and focus on that, but that is unlikely to be very sensitive to some types of transmission line degradation (eg increased loss).

Analysis of the derived Return Loss figures in-band and especially out-of-band gives much more insight.

I have a clear window that shows if water has leaked down the cable. It should not leak down the inside because it is closed cell foam, but it should indicate if the birds pick a hole in the jacket… and possibly the copper.

]]>Note that the measurements are of a particular implementation and should not be taken to imply generally to 5/8λ verticals, but the solution method can be applied more generally. Lets assume that the measurement is not affected by common mode current.

The answer to the last question first is that a series inductor will not bring the VSWR much below 3. It is a common belief that a 5/8λ vertical can be matched simply with a series inductor.

There are many ways to match the measured antenna, and there are articles on this site describing some of them, but a simple and effective method in this case is the single stub tuner.

Above is a graphical solution using Simsmith. The section of line nearest the measurement load is -ve length, it is to back out the effect of the line section into which measurements were made (antenna feed point is at the cursor, 139-j191Ω). The next line section is the series section, followed by the S/C stub. In this case the series section and stub use RG213 to reduce loss. Total matching system loss is a little under 0.3dB, and the stub can easily be weatherproofed with hot glue and heat shrink tube.

One could use RG58, an exercise for the reader is to assess the loss of that option.

Obviously the length of the measurement section plays into the solution, and using its length to the mm in the model gives a more accurate result.

]]>A coax trap (before cross connection).

The whole subject of trapped antennas elicits a lot of online discussion that is often more about semantics than understanding.

Some key points:

- It is impossible to wind a coil that does not exhibit self resonance at some frequency, and the assertion that it is adequately characterised as an ideal inductance in series with some resistance is quite wrong at frequency higher than say 10% of its self resonance frequency (SRF). So the distinction between a coil and a parallel resonant circuit is often misguided.
- A simplistic explanation of a trap is that it is designed to be resonant at the higher desired band of operation, and at that frequency it acts like an open switch disconnecting the outboard wire sections. You can make a trap that way but it has some significant disadvantages.
- The coax trap is a little more complicated that a fixed inductor tuned by a fixed capacitor.

A trapped dipole for 80m and 40m using bootstrap coax traps used a coax trap that is resonant at about 6.5MHz.

Zooming in a little, we can see that the traps have a modest inductive reactance at 80m (400 – 500Ω) and a largish -ve reactance around 40m (-4000 – -3000Ω). In concert with the wire length and configuration, this results in VSWR minima in the 80m and 40m band, fairly low structure loss, and a pattern largely like an Inverted V dipole of half wave length.

You could fret about the series resistance component of trap Z, but at the end of the day, what is important is how much power is lost in the structures (dipole wire, traps) and in this case dipole + trap loss is less than 0.5dB at 40m, comparable to the coax feed line loss. The situation is a little worse on 80m where it is a shortened loaded dipole, but loss is still quite low (<1dB in antenna structure).

This design was built and tested, and worked pretty much to the model predictions. It demonstrates that a lot of notions about trapped dipoles are flawed.

Distortionless Linesfrom time to time, often in the vein of

they don’t exist, so why discuss them?

The concept derives from the work of Heaviside and others in seeking a solution to distortion in long telegraph lines.

The problem was that digital telegraph pulses were distorted due to different attenuation and propagation time for different components of the square waves.

Heaviside proposed that transmission lines could be modelled as distributed resistance (R), inductance (L), conductance (G) and capacitance (C) elements.

In each incremental length Δx, there is incremental R, L, G and C.

Characteristic impedance Zo and complex propagation coefficient γ can be derived from the model, and it becomes apparent that only under certain conditions is the attenuation and phase velocity independent of frequency.

Heaviside determined that condition was that G/C=R/L, so this is the condition for a Distortionless Line.

That same condition means that Zo is a purely real number, and if Zo is purely real, then the line is Distortionless.

Note that Lossless Line is a special case of Distortionless Line, and necessarily Zo is purely real.

It is true that fabrication of a Distortionless Line is a considerable challenge, though some techniques might deliver an approximately Distortionless Line over a limited frequency range.

Whilst some might dismiss the concept of Distortionless Line as impractical, they probably apply Distortionless Lines freely in their analytical techniques without understanding what they are doing.

Let us look at some common operations.

Waves add, the E and H fields add vectorially, as do their V and I equivalents.

It happens that in the special case of purely real Zo (ie Distortionless Line) that the directional powers add, but they add by virtue of expansion of the expression for adding V and I (which simplify when Zo is real).

Most use of a Smith chart involves normalising plotted values to some real Zref. In so doing, an assumption is made that constant VSWR circles are wrt that Zref, and solutions (eg matching) based on that are conditional on the assumed real (Distortionless) Zref.

The clean separation of real and imaginary Z and Y components and the arcs of constant R, X, B, G on a Smith chart are conditional on real Zo, as is the assumption that all practical values of Z map inside the ρ=1 circle. (The assumption that ρ<=1 depends on real Zo.) The loss of this property quite restricts its usefulness.

None of this is to suggest that a Smith chart with Zref including a significant imaginary part is invalid… just many things we have learned to do with a ‘real’ Smith chart may not be valid (ie within acceptable error limits), and the scales are no longer as simple and useful as the chart devised by Philip Smith.

ρ, Return Loss, and VSWR are derived from the complex reflection coefficient Γ. Calculating Γ wrt some nominal real Zo involves some error, and raises the question of whether application of those to a real scenario (where Zo is different, eg not real) is within acceptable error limits.

S parameter techniques assume additive power, and that may not be appropriate for some problem solutions. The analysis tools effectively coerce a Distortionless Lines treatment.

Practitioners use analytical techniques that depend on an assumption that Zo is real (ie Distortionless Line) widely, and often without acknowledging it or even thinking about the implications.

Virtual Distortionless Lines are everywhere, and worth understanding when Distortionless Line analysis is being applied and whether it is appropriate.

]]>In fact, I did explore #73 as an option, this article presents some key comparisons. The two key statistics shown in this article provided the basis for selecting the design.

Note that the scales are different from plot to plot.

Where the magnetising impedance appears in shunt with an ideal transformer with Zin=50+j0Ω, Insertion VSWR can be calculated.

Where the magnetising impedance is a shunt component of total Zin=50+j0Ω, Core loss VSWR can be calculated.

Predictions of Insertion VSWR and core loss of the popular BN73-202 transformer with 2t primary for MF and my BN43-202 with 5t primary reveals that the BN43-202+5t is a better transformer for my intended application.

There is no need to follow the traditional design which anecdotally works, but is published without performance predictions or measurements… well that IS traditional.

]]>