Modeling a multiband Yagi in 4NEC2?

I tried to model a multiband yagi. I added radiators for 10-15-20M, and reflectors spaced .15 wavelenghts away. This gave a nice result with three sharp dips in SWR in the 10-15-20M center frequencies.

For some limitation of NEC2, I had to make equal wires for all the bands and then extend them with more wires (NEC2 wanted parallel wires to have segments aligned)

I then tried shortening the elements with coils. I shortened these extension wires, and put a series L at the first segment, in the 15 and 20m driven elements and reflectors.

When doing a frequency sweep, the 20 and 15M bands show acceptable dips, but the 10M band does a strange thing: it has a wide SWR dip from 27 to 28mhz, then a spike at 28.1mhz, and then comes back down at 28.2mhz

Is this spike an artifact of simulation, or is it real interaction between the elements altering the frequency?

Here's the NEC file:

CM Back yard dipole, converted with 4nec2 on 26-mar-20 9:20
CE
SY DIEZmdrlen=5.03  '2.66
SY QUINmdrlen=7.07  '7.07
SY VEINmdrlen=10.63 '10.63
SY DIEZmdrpos=0.8   '0
SY QUINmdrpos=1.06  '1.06
SY VEINmdrpos=1.7   '1.7
SY h=8  '8
SY d=.019/2 '.019/2
GW  2121    2   .03 QUINmdrpos  h   -.03    QUINmdrpos  h   d
GW  1   20  .03 DIEZmdrpos  h   DIEZmdrlen/2    DIEZmdrpos  h   d
GW  2   20  .03 QUINmdrpos  h   DIEZmdrlen/2    QUINmdrpos  h   d
GW  3   20  .03 VEINmdrpos  h   DIEZmdrlen/2    VEINmdrpos  h   d
GW  111 20  -.03    DIEZmdrpos  h   -(DIEZmdrlen/2) DIEZmdrpos  h   d
GW  112 20  -.03    QUINmdrpos  h   -(DIEZmdrlen/2) QUINmdrpos  h   d
GW  113 20  -.03    VEINmdrpos  h   -(DIEZmdrlen/2) VEINmdrpos  h   d
GW  1111    5   -.03    DIEZmdrpos  h   -.03    QUINmdrpos  h   d/2
GW  1112    5   -.03    QUINmdrpos  h   -.03    VEINmdrpos  h   d/2
GW  2111    5   .03 DIEZmdrpos  h   .03 QUINmdrpos  h   d/2
GW  2112    5   .03 QUINmdrpos  h   .03 VEINmdrpos  h   d/2
GW  4   4   DIEZmdrlen/2    QUINmdrpos  h   DIEZmdrlen/2+.5 QUINmdrpos  h   d
GW  5   4   -DIEZmdrlen/2   QUINmdrpos  h   -(DIEZmdrlen/2+.5)  QUINmdrpos  h   d
GW  6   4   DIEZmdrlen/2    VEINmdrpos  h   (DIEZmdrlen/2+.8)   VEINmdrpos  h   d
GW  7   4   -DIEZmdrlen/2   VEINmdrpos  h   -(DIEZmdrlen/2+.8)  VEINmdrpos  h   d
GW  31  15  (DIEZmdrlen*1.05/2) -DIEZmdrpos h   -((DIEZmdrlen*1.05/2))  -DIEZmdrpos h   d
GW  51  15  (DIEZmdrlen*1.05/2) -QUINmdrpos h   -((DIEZmdrlen*1.05/2))  -QUINmdrpos h   d
GW  61  15  (DIEZmdrlen*1.05/2) -VEINmdrpos h   -((DIEZmdrlen*1.05/2))  -VEINmdrpos h   d
GW  32  5   (DIEZmdrlen*1.05/2) -QUINmdrpos h   (DIEZmdrlen/2+.5)   -QUINmdrpos h   d
GW  33  5   -(DIEZmdrlen*1.05/2)    -QUINmdrpos h   -(DIEZmdrlen/2+.5)  -QUINmdrpos h   d
GW  34  5   (DIEZmdrlen*1.05/2) -VEINmdrpos h   (DIEZmdrlen/2+.8)   -VEINmdrpos h   d
GW  35  5   -(DIEZmdrlen*1.05/2)    -VEINmdrpos h   -(DIEZmdrlen/2+.8)  -VEINmdrpos h   d
GE  1
LD  0   6   1   1   0   15.5e-6
LD  0   7   1   1   0   15.5e-6
LD  0   4   1   1   0   3e-6
LD  0   5   1   1   0   3e-6
LD  0   34  1   1   0   15.5e-6
LD  0   35  1   1   0   15.5e-6
LD  0   32  1   1   0   3.5e-6
LD  0   33  1   1   0   3.5e-6

GN  2   0   0   0   13  0.005
EK
EX  0   2121    2   0   0   0   0

FR  0   0   0   0   14.1    0
EN

$$$$
`
• Please consider editing your question to add the frequency sweep graphs. Apr 24 '20 at 16:41
• To answer your question, it would be interesting to see if you get the same results after making two changes to the model: 1) "Construct" each element from a single wire and place the loads accordingly; 2) Replace the parallel feed wires with transmission lines of the appropriate characteristic impedance. Apr 24 '20 at 16:54
• @BrianK1LI I don't know how to use the transmission line feature, and I haven't found a good explanation of how to. But I changed something: I made the parallel wires shorter, and "extended" each of them with extra wires to the proper length. This seemed to tame the spike, which again leads me to believe this is a simulation error.
– hjf
Apr 24 '20 at 17:20
• Use "Settings > Geometry edit (Ctrl+F3)" to see the drag'n'drop editor. Click on the button that looks like a ladder, then click on Add, then click on the two segments that you want to be joined with a t-line. You can then select the actual length and Z in the panel on the right side of the screen. Apr 24 '20 at 18:02
• When I sweep your .NEC file, I see a different SWR profile. The 1.2:1 SWR minimum occurs at 27.25MHz, the 5:1 peak occurs at 27.6MHz and a 5:1 local minimum occurs at 27.9MHz. It seems odd that we should observe different results. Apr 24 '20 at 18:55

From the information presented below, I believe the "problem" with your model arises from the parallel wires that couple the forward elements.

I made a reasonable facsimile of the antenna in your NEC model with the changes I described in my comment:

• Each element consists of a single, segmented "wire"
• Loads are placed as close to the positions you specified as possible
• A 50-ohm transmission line replaces the parallel wires that couple the forward elements

Here are the element currents at the resonant frequency of 28.2-MHz:

These produce useful directivity at the maximum radiation angle of 18-degrees above the horizon:

and the SWR plot:

Changing the impedance of the transmission line to 200-$$\Omega$$ produced an SWR curve quite similar to your original NEC model, but I believe it's still advisable to use NEC's native transmission line facility.

• Thank you! It's a thing that I'm also wondering for practical construction: when feeding multiple feed points, should I use coax between feed points? Or parallel wires at a distance that will keep the impedance maybe?
– hjf
Apr 25 '20 at 5:07
• As long as the "parallel wires" are meant to act as a transmission line, I recommend the transmission line model. I suspect it simulates faster and has less chance of interacting with the rest of your model in "unexpected" ways. Also, remember that NEC doesn't know anything about the TL you add; it's a purely mathematical construct, though quite a good approximation. Apr 25 '20 at 10:15
• If you concur with the analysis, please mark the question as "Answered" so it comes off the "Unanswered" list. Apr 25 '20 at 10:16