I am working on prediction of the Antenna Factor of a small square
loop, and have created and run NEC-2 Models at a range of frequencies.
The models are documented in the draft article at
http://www.vk1od.net/SmallUntunedSquareLoop/ssulNEC.htm .

I would appreciate comments on potential pitfalls in the models that I
have created. Should I expect an antenna built to the design to
perform with perhaps tenths of a dB of the model?

Owen
--

On Tue, 10 Oct 2006 23:29:22 GMT, Owen Duffy wrote:

I would appreciate comments

Hi Owen,

Maybe use a log-log scale?

73's
Richard Clark, KB7QHC

On Tue, 10 Oct 2006 17:17:33 -0700, Richard Clark
wrote:

On Tue, 10 Oct 2006 23:29:22 GMT, Owen Duffy wrote:

I would appreciate comments

Hi Owen,

Maybe use a log-log scale?

It is if you think about it, -20log(Vout/Vin) vs log(Freq).

I am really interested in modelling accuracy issues like segment
length, right angle corners, load placement etc.

Owen
--

On Wed, 11 Oct 2006 00:33:58 GMT, Owen Duffy wrote:

On Tue, 10 Oct 2006 17:17:33 -0700, Richard Clark
wrote:

On Tue, 10 Oct 2006 23:29:22 GMT, Owen Duffy wrote:

I would appreciate comments

Hi Owen,

Maybe use a log-log scale?

It is if you think about it, -20log(Vout/Vin) vs log(Freq).

Hi Owen,

I suppose so, however, when I see a sigmoidal curve, it begs a
different representation which, to me, reveals an underlying concept
(whatever that might be). Even with the redundancy of log
representation of a log value, its straight line representation may
worth consideration. I don't know why at this pass. This is just a
preference I would consider.

I am really interested in modelling accuracy issues like
segment length,

Generally self-driven by the results of successive modeling of
increased segmentation approaching an asymptotic level. The implicit
question would be: "Does the addition computational load warrant the
increased count?"

right angle corners,

Area dominates such considerations, practicality is also significant,
results generally varies by little. As Reggie would offer, it is a
trade-off in the amount of wire, and spacing that would tend to
increase loss, and give rise to self resonance. He offered very good
advice on the size of coils, and this may have occurred before you
joined the group. I'm sure a review of his web page would reveal an
"unzipped" executable, but may lack the discussion he offered here.

load placement etc.

An issue of balance (physical placement mimicking literal field
balance).

73's
Richard Clark, KB7QHC

On Tue, 10 Oct 2006 18:36:23 -0700, Richard Clark
wrote:

On Wed, 11 Oct 2006 00:33:58 GMT, Owen Duffy wrote:

On Tue, 10 Oct 2006 17:17:33 -0700, Richard Clark
wrote:

On Tue, 10 Oct 2006 23:29:22 GMT, Owen Duffy wrote:

I would appreciate comments

Hi Owen,

Maybe use a log-log scale?

It is if you think about it, -20log(Vout/Vin) vs log(Freq).

Hi Owen,

I suppose so, however, when I see a sigmoidal curve, it begs a
different representation which, to me, reveals an underlying concept
(whatever that might be). Even with the redundancy of log
representation of a log value, its straight line representation may
worth consideration. I don't know why at this pass. This is just a
preference I would consider.

I know we like to see straight lines in things, leads to simple
explanations.

I think a piecewise explanation of what happens here is:
- below the second knee, current is uniform and:
- below the first knee, loop inductive reactance is small, and
induced voltage is dominated by the changing freuency;
- above the first knee, induced voltage still varies with frequency,
but the loop inductive reactance is compensating that to a fair extent
and current (or loaded terminal voltage) is almost constant with
change in frequency;
-above the second knee, current is not uniform, reactance changes more
rapidly with change in frequency as resonance is approached, induced
voltage increases with increase in frequency.


I am really interested in modelling accuracy issues like
segment length,

Generally self-driven by the results of successive modeling of
increased segmentation approaching an asymptotic level. The implicit
question would be: "Does the addition computational load warrant the
increased count?"

The computation load isn't a big issue in these case, and I have
changed the segmentation with minutest change in results around the
current segmentation strategy.


right angle corners,

Area dominates such considerations, practicality is also significant,
results generally varies by little. As Reggie would offer, it is a
trade-off in the amount of wire, and spacing that would tend to
increase loss, and give rise to self resonance. He offered very good
advice on the size of coils, and this may have occurred before you
joined the group. I'm sure a review of his web page would reveal an
"unzipped" executable, but may lack the discussion he offered here.

Certainly, it can be solved from first principles at lower frequencies
where current is uniform, and as you say, area and self inductance are
the critical quantities. It is at the right hand end where current is
non-uniform that the NEC model becomes most relevant, and where I
depend on its accuracy... if I got things right!


load placement etc.

An issue of balance (physical placement mimicking literal field
balance).

Thanks Richard,

Owen
--

"Owen Duffy" wrote in message
...

I am working on prediction of the Antenna Factor of a small square
loop, and have created and run NEC-2 Models at a range of frequencies.
The models are documented in the draft article at
http://www.vk1od.net/SmallUntunedSquareLoop/ssulNEC.htm .

I would appreciate comments on potential pitfalls in the models that I
have created. Should I expect an antenna built to the design to
perform with perhaps tenths of a dB of the model?

Owen
--

One thing that has bothered me about small loop modeling. When
driven with a voltage source; loops of length 0.1 wavelengths
exhibit a radiated power ,and efficiency, of 0. In other words
the structure loss is than the input power. Removing
the wire conductivity loading corrects the problem. In your
particular case, at 5 MHz, the real input impedance is
more than three orders of magnitude greater than the radiation resistance.
I guess it could just be numerical rounding errors, but have gotten
identical results when using double precision.

Frank

On Wed, 11 Oct 2006 16:32:50 GMT, "Frank's"
wrote:

....

One thing that has bothered me about small loop modeling. When
driven with a voltage source; loops of length 0.1 wavelengths
exhibit a radiated power ,and efficiency, of 0. In other words
the structure loss is than the input power. Removing
the wire conductivity loading corrects the problem. In your
particular case, at 5 MHz, the real input impedance is
more than three orders of magnitude greater than the radiation resistance.
I guess it could just be numerical rounding errors, but have gotten
identical results when using double precision.

Hi Frank,

That is interesting.

I haven't pursued modelling the loop being excited directly, but I
accept your observations.

I have no idea what causes the problem. Perhaps a NEC guru may have an
explanation?

I thinking about the impact of error in the estimate of either
radiation resistance or loss resistance of the conductor in the
scenario that I did model, it should be low, probably insignificant.
The magnitude of both components is small relative to the load
impedance and the loop's own reactance, and does not influence the
loop current (and therefore the load voltage) much.

Whilst it would be interesting to know if it is a fault in modelling,
or just numerical stability in the computation engine, I suspect the
effect you observed probable does not give great concern for the
accuracy of my receive loop models.

I would like to have modelled insulated wire in the loops, but don't
have NEC-4... so I will build the loops from bare wire, although I
think it would make very little difference, even more so at the lower
frequency side of the range.

Thanks, appreciate the thoughts.

Owen
--

Hi All,

Well, it seems enough problems have accumulated to warrant further
discussion.

On Thu, 12 Oct 2006 22:08:49 GMT, Owen Duffy wrote:

On Wed, 11 Oct 2006 16:32:50 GMT, "Frank's"
wrote:

...

One thing that has bothered me about small loop modeling. When
driven with a voltage source; loops of length 0.1 wavelengths
exhibit a radiated power ,and efficiency, of 0. In other words
the structure loss is than the input power.

To say the least. For a loop with sides of 1 meter, the radiation
resistance at the low end of the chart verges on 150 nanoOhms. If
this were copper wire, it would exhibit about 2.5 Ohms per 1000 feet.
Instead it is steel wire (about 5 to 10 times the resistance) over a
length of 13 feet or so. I will be generous and call it half an ohm
total.

Removing
the wire conductivity loading corrects the problem.

I hope you don't pack your own parachutes!

In your
particular case, at 5 MHz, the real input impedance is
more than three orders of magnitude greater than the radiation resistance.

Examining the radiation resistance again 0.0015 Ohm; copper loss
remains the same for intents and purposes.

I guess it could just be numerical rounding errors, but have gotten
identical results when using double precision.

Hi Frank,

That is interesting.

I haven't pursued modelling the loop being excited directly,

I had presumed you performed that exercise.

but I
accept your observations.

I have no idea what causes the problem. Perhaps a NEC guru may have an
explanation?

Hmmmm.

I thinking about the impact of error in the estimate of either
radiation resistance or loss resistance of the conductor in the
scenario that I did model, it should be low, probably insignificant.

You might want to revisit those thoughts.

The magnitude of both components is small relative to the load
impedance and the loop's own reactance, and does not influence the
loop current (and therefore the load voltage) much.

What you are saying is the antenna is a voltage source, and power
efficiency can go to the devil. For receivers, they forgive such
blasphemy. Just how much appears to be unknown at this point (we will
say things are in limbo). However, if you add a capacitor, you should
be able to observe the loss in the expressed Q. This is an untuned
loop, is it not? Somewhere there's the devil to pay otherwise. }:-)

Whilst it would be interesting to know if it is a fault in modelling,
or just numerical stability in the computation engine, I suspect the
effect you observed probable does not give great concern for the
accuracy of my receive loop models.

Turning the antenna into a voltage source gives rise to many nuances
that would otherwise be smothered.

I would like to have modelled insulated wire in the loops, but don't
have NEC-4... so I will build the loops from bare wire, although I
think it would make very little difference, even more so at the lower
frequency side of the range.

This alone appears to be a safe conclusion.

73's
Richard Clark, KB7QHC

On Thu, 12 Oct 2006 17:17:27 -0700, Richard Clark
wrote:

Hi All,

Well, it seems enough problems have accumulated to warrant further
discussion.

On Thu, 12 Oct 2006 22:08:49 GMT, Owen Duffy wrote:

On Wed, 11 Oct 2006 16:32:50 GMT, "Frank's"
wrote:

...

One thing that has bothered me about small loop modeling. When
driven with a voltage source; loops of length 0.1 wavelengths
exhibit a radiated power ,and efficiency, of 0. In other words
the structure loss is than the input power.

To say the least. For a loop with sides of 1 meter, the radiation
resistance at the low end of the chart verges on 150 nanoOhms. If
this were copper wire, it would exhibit about 2.5 Ohms per 1000 feet.
Instead it is steel wire (about 5 to 10 times the resistance) over a
length of 13 feet or so. I will be generous and call it half an ohm
total.

Removing
the wire conductivity loading corrects the problem.

I hope you don't pack your own parachutes!

In your
particular case, at 5 MHz, the real input impedance is
more than three orders of magnitude greater than the radiation resistance.

Examining the radiation resistance again 0.0015 Ohm; copper loss
remains the same for intents and purposes.

I guess it could just be numerical rounding errors, but have gotten
identical results when using double precision.

Hi Frank,

That is interesting.

I haven't pursued modelling the loop being excited directly,

I had presumed you performed that exercise.

but I
accept your observations.

I have no idea what causes the problem. Perhaps a NEC guru may have an
explanation?

Hmmmm.

I thinking about the impact of error in the estimate of either
radiation resistance or loss resistance of the conductor in the
scenario that I did model, it should be low, probably insignificant.

You might want to revisit those thoughts.

I have also created a (non NEC) model that depends on the assumption
that current is uniform. Frank has commented on a NEC model at 5MHz,
where the loop side is 0.016 wavelengths, so current is close to
uniform. Using that frequency in my non NEC model suggests that the
radiation resistance is 0.0026, conductor loss resistance is 3.07,
loop inductive reactance is 163, and the load is 50. The circuit
impedance is 53.07+j163, and load voltage is relatively insensitive to
small changes in radiation resistance or conductor loss resistance.

The model is at http://www.vk1od.net/SmallUntunedSquareLoop/temp/1.htm
..

Changing the NEC model to be driven by a voltage source, I get loop Z
of 3.63+j173, so NEC predicts a higher loop Z. The efficiency
calculated by NEC is 1.54%, much lower than the 0.085% implied in the
simple model. The NEC efficiency implies that radiation resistance is
0.055. Doubling the number of segments in the NEC model does not
markedly change loop Z, but increases efficiency to 2.3%, so there is
something strange happening in the way it determines radiation
resistance. With a single segment per side, NEC suggests efficiency of
0.8% which is close to my simple model.

One of the data points from my NEC2 run is at 5.091MHz, and Antenna
Factor for the loop and 10m of RG58CU is 30.624dB/m. That includes an
amount of the coax loss, and that amount should be 0.309dB, leaving
the loop itself at 30.316dB/m.



Owen
--