Cecil Moore wrote:
Ian White GM3SEK wrote:
It's a persistent ham myth that an RF choke has specially good
properties when the total length of wire is a quarter-wavelength, and
specially bad properties at twice that frequency. When the wire is
wound into any kind of coil, neither of those claims is true (except
maybe by some rare coincidence).

Please don't imply that I said anything about the total
length of wire - I didn't.

In that case, I suggest you stop making constant references to "1/4WL
self-resonance" and "1/2WL self-resonance". If you don't mean it
literally, it's a very misleading metaphor.

What you say is true and I
never said otherwise. Well-designed coils can be modeled
as rough approximations to transmission lines.




The choke acts essentially as a parallel-tuned circuit, with its
inductance tuned by its own self-capacitance. There will be a series
resonance at some higher frequency, but not at twice the parallel-
resonant frequency (except, again, perhaps by a rare coincidence).

I didn't say exactly twice the frequency and I said it
was an approximation. The chokes at:

http://www.k1ttt.net/technote/airbalun.html

average close to double the frequency.

We're actually looking at exactly the same data (except that the
original reference quoted on K1TTT's site also includes a ferrite bead
choke for comparison).

I have graphed the |Z| data for all the chokes (see link to spreadsheet
below) and there is no consistent trend. In the following table, Fmax is
the frequency of maximum impedance, and Fmin is the frequency of any
minimum observable within the frequency range (the 8t 1 layer choke has
two very small minima). Ratio is Fmax/Fmin.

Choke Fmax Fmin Ratio
6t 1 layer 24 none -
12t 1 layer 15 31 2.1
4t 1 layer 21 34 1.6
8t 1 layer 12 19 1.6
12 32 2.7
8t bunched 6 36 6.0

Judging from the shapes of the graphs and the table above, I would say
that "twice the frequency" is not even valid as an approximation.


No one would expect a bunched coil to be very well behaved.
Everything I have said applies to a coax choke wound on
some kind of coil form with some care given to its design.


Across the whole 1-30MHz band, the bunched choke behaves as an almost
perfect L-C circuit, free from any unwanted resonances. The only problem
with that design is to reproduce the exact parallel-resonant frequency
from one example to the next.


There is NO series resonance at twice the parallel-resonant
frequency - that would be about 12MHz, and nothing at all "special"
is happening there. At 18MHz, where the total winding length is 0.25
wavelengths, there is a very small wobble in the data, but nothing
more.
The series resonance, where the phase angle flips from negative to
positive again, is at 31.5MHz, which is totally unrelated to any of the
other frequencies above. The winding length is 0.5 wavelengths at
35MHz (where the data runs out) but again nothing "special" is
happening there.

Again, no one would expect a bunched coil to be well behaved.

Thus there is no evidence whatever for the myth of the "resonant
length of wire in a choke".

You keep saying that as if I said otherwise. I didn't. The
length of the wire is irrelevant to this discussion.

Turning now to the solenoid-wound choke, the different method of
winding has increased the parallel resonance of the same length of
cable from 6MHz to 9MHz. This is consistent with simple L-C
behaviour, and with the solenoid having less distributed capacitance
than the bunched winding.
Once again, this choke behaves almost entirely as a parallel-tuned
circuit. There are slightly larger wobbles in the data at the
frequencies where the total winding lengths are a quarter-wave and a
half-wave, but these "transmission-line" effects are still very minor,
and completely dominated by the simple L-C behaviour.

The point is that there is a 1/4WL high impedance resonance
and a 1/2WL low impedance resonance that are roughly where
they should be. The 1/2WL low impedance resonance should
be avoided.

As shown above, "1/2wl self resonance" ceases to be a valid concept
once a length of wire is wound into a coil...

The 1/2WL self-resonance has little to do with the length
of wire. It is where the phase angle flips at a point of
low impedance. The 1/4WL self-resonance is where the phase
angle flips at a point of high impedance. The length of wire
is irrelevant, a moot point. I don't know why you brought
it up in the first place.

If you say "the length of wire is irrelevant to this discussion" - with
which I most strongly agree - why do you persist in using these terms
"1/4WL" and "1/2WL" - what dimension of the choke are they referring to?

The Excel workbook at

www.ifwtech.co.uk/g3sek/misc/chokes.xls

contains three spreadsheets.

1. Original data
For all the coiled chokes (same data in the ARRL Antenna Book and on
K1TT's site) with graphs of |Z|. There are minor dips at higher
frequencies, but they are *minor*, and always in a region where the
impedance is so low that you wouldn't be using that choke anyway.

These graphs simply don't support the assertion of a series resonance at
"twice the parallel-resonant frequency" - not even as an approximation.

2. Three chokes compared
The solenoid-wound 8-turn choke, the bunched 8-turn choke, and the
ferrite choke for comparison. The graphs give details of the Z magnitude
and phase.

3. LC model
For the 8-turn solenoid choke. The inductance is calculated from the
physical dimensions of the choke, using the standard ARRL formula
(winding length assumes close-wound RG213). The self-capacitance is
calculated from the inductance and the choke's parallel-resonant
frequency. The dynamic resistance is the peak value from 12MHz, and is
assumed constant at all frequencies.

Those simple assumptions - a fixed L, C and R, all connected in parallel
- give a very good fit to the measured data at all frequencies (only one
point has been forced to fit, namely the peak at 12MHz). This shows that
the dominant behaviour of the choke is like a simple LC circuit, damped
by some loss resistance.

Much of the loss resistance is probably due to losses in the PVC jacket
of the RG213. If these losses are actually increasing with frequency
(rather than being constant, as assumed) then the fit at all frequencies
would be improved.

This very simple LCR model predicts almost everything that was measured.
However, it cannot predict any series resonance at some higher
frequency. If Cecil cares to produce a transmission-line model of the
same choke that can do better, I'm sure we'd all be interested to see
it.



--
73 from Ian GM3SEK 'In Practice' columnist for RadCom (RSGB)
http://www.ifwtech.co.uk/g3sek