Canadian TV, Computing and Home Theatre Forums banner

"Top Hat" GH with NARODs for VHF-HI: Major GH Improvement

388761 Views 939 Replies 87 Participants Last post by  Scorpio_333
this is a modification of the JED's GH10 (rev2),
with bent NARODs and 3 narod reflectors.


Numerical gain data (VHF-hi):
Ch		7	8	9	10	11	12	13
swr		1.63	1.69	2.24	2.2	1.84	1.34	1.17
Net Gain	9.06	9.33	8.9	8.78	8.83	8.81	8.29
Numerical gain data (UHF):
Ch		14	15	16	17	18	19	20	21	22	23	24	25	26	27	28	29	30	31	32	33	34	35	36	37	38	39	40	41	42	43	44	45	46	47	48	49	50	51	52	53	54
swr		1.22	1.27	1.36	1.45	1.52	1.55	1.47	1.38	1.35	1.35	1.36	1.37	1.38	1.38	1.39	1.39	1.4	1.4	1.41	1.42	1.44	1.48	1.53	1.57	1.6	1.62	1.64	1.67	1.69	1.72	1.75	1.78	1.8	1.83	1.85	1.87	1.91	1.95	2.01	2.09	2.18
Net Gain	13.59	13.9	14.14	14.29	14.35	14.33	14.43	14.59	14.65	14.66	14.65	14.61	14.58	14.55	14.51	14.5	14.49	14.47	14.43	14.38	14.27	14.13	14.05	14.1	14.27	14.46	14.61	14.76	14.88	14.99	15.11	15.22	15.32	15.41	15.47	15.47	15.39	15.18	14.86	14.39	13.74
nec code
SY radius=0.003175
SY hat_rad=0.003175	'0.0010265
SY l1=0.381
SY g1=0.019
SY z1=0.037
SY l2=0.280
SY g2=0.009
SY z2=0.137
SY l3=0.556
SY g3=0.0055
SY z3=0.262
SY l4=0.511
SY g4=4e-3
SY z4=0.414
SY l5=0.270
SY g5=4e-3
SY z5=0.592
SY x=-0.081
SY a1=0.196
SY a2=0.192
SY a3=0.190
SY b=0.120
SY feed=0.0548348
SY a1ma2=a1-a2
SY a1pa2=a1+a2
SY a1ma2pa3=a1ma2+a3
SY a1pa2pa3=a1pa2+a3
SY n_l=0.449369
SY n_top_l=0.223443
SY n_b_l=(n_l - n_top_l)/2
SY n_z_sp=0.012
SY n_h=0.04795
SY n_b_z=a1pa2pa3*.7071+n_z_sp+hat_rad-radius
SY n_top_z=n_b_z+n_h
SY n_back=-0.25325
SY n_refl_size=0.404294
SY n_refl_z=0.367714
SY n_refl_size1=0.447082
GW	1	23	0	-feed	0	0	-feed-0.70711*a1	0.70711*a1	radius
GW	2	23	0	-feed-0.70711*a1	0.70711*a1	0	-feed-0.70711*a1ma2	0.70711*a1pa2	radius
GW	3	23	0	-feed-0.70711*a1ma2	0.70711*a1pa2	0	-feed-0.70711*a1ma2pa3	0.70711*a1pa2pa3	radius
GW	4	15	0	-feed-0.70711*a1ma2pa3	0.70711*a1pa2pa3	0	-feed-0.70711*a1ma2pa3-b	0.70711*a1pa2pa3	radius
GW	5	45	x	g1	z1	x	l1+g1	z1	radius
GW	6	33	x	g2	z2	x	l2+g2	z2	radius
GW	7	65	x	g3	z3	x	l3+g3	z3	radius
GW	8	61	x	g4	z4	x	l4+g4	z4	radius
GW	8	33	x	g5	z5	x	l5+g5	z5	radius
GW	13	7	0	-n_top_l	n_top_z	0	-n_top_l	n_b_z	hat_rad
GW	14	13	0	-n_top_l	n_b_z	0	-n_top_l-n_b_l	n_b_z	hat_rad
GX	20	010
GW	10	53	0	n_top_l	n_top_z	0	-n_top_l	n_top_z	hat_rad
GW	15	95	n_back	-n_refl_size	n_refl_z	n_back	n_refl_size	n_refl_z	radius
GX	40	001
GW	100	13	0	-feed	0	0	feed	0	radius*.8675
GW	16	105	n_back	-n_refl_size1	0	n_back	n_refl_size1	0	radius
GE	0
LD	5	0	0	0	24900000
GN	-1
EX	0	100	7	0	1			
FR	0	0	0	0	800	0
RP	0	60	73	1001	0	0	3	5		
See less See more
1 - 16 of 940 Posts

Ken Nist sent me some interesting comments & suggestions re the modeling of your GH3.8 Combo design (detailed in post #1 of this thread).

Ken's Comments & Suggestions
In regards to the SBGH Combo design:

1. Wire 4 and 11 are very close. Most moments method programs handle such cases poorly. Experimentation might show that a different spacing is optimal.
2. The positions of the 3 VHF reflectors look optimal. But moving them closer to the other elements probably costs very little gain while making construction easier.

1. 4NEC2 Modeling limitation for closely spaced Parallel Wires
4nec2 modeling program has this limitation. In the case of the top-hat Narod design, the wires that Ken is speaking about, is the close proximity of the array's horizontal stub close to the "c" dimension of your top-hat Narod. Real world in-field experimentation is needed to investigate the effect of this modeling limitation.

Perhaps there are some DIY builders that could carry out some in-field testing with varied spacings using the "digital-cliff" measurement technique.
( Explanation on "digital cliff" measurement at .. )

AFAIK DIY builders have only reported building 300Ohm's straight Narod design, mainly with DBGH designs, but I am not aware of any in-field measurement data.

I would encourage any forum member in the "die-hard DIY builders category" to consider the digital-cliff measurement and report on their build results, ( if they can afford the modest cost involved and of course that most precious commodity ... time).
That would provide very useful in-field real-world measurement data for those that spend the time modeling and designing new antenna variants.

2. Investigation of Narod reflector Spacing
Ken's other suggestion concerning the spacing of the Narod reflectors is worth a look.

Description of GH3.8 ez file sent to Ken

                      EZNEC-M ver. 3.0

SBGH GH10 with TopHat Narods &               6/5/2010     12:19:12 PM

         --------------- ANTENNA DESCRIPTION ---------------

Frequency = 195 MHz
Wire Loss: Aluminum (6061-T6) -- Resistivity = 4E-08 ohm-m, Rel. Perm. = 1

              --------------- WIRES ---------------

No.            End 1     Coord. (m)              End 2     Coord. (m)       Dia (mm)  Segs
          Conn.      X       Y       Z       Conn.      X       Y       Z
1         W25E1        0,-.05483,      0      W2E1        0,-.19343,.138594      6.35   23
2          W1E2        0,-.19343,.138594      W3E1        0,-.05766,.274359      6.35   23
3          W2E2        0,-.05766,.274359      W4E1        0,-.19201, .40871      6.35   23
4          W3E2        0,-.19201, .40871                  0,-.31201, .40871      6.35   15
5                 -0.081,  0.019,  0.037             -0.081,    0.4,  0.037      6.35   45
6                 -0.081,  0.009,  0.137             -0.081,  0.289,  0.137      6.35   33
7                 -0.081, 0.0055,  0.262             -0.081, 0.5615,  0.262      6.35   65
8                 -0.081,  0.004,  0.414             -0.081,  0.515,  0.414      6.35   61
9                 -0.081,  0.004,  0.592             -0.081,  0.274,  0.592      6.35   33
10        W23E2        0,-.22344,.468654     W11E1        0,-.22344,.420704      6.35   7
11        W10E2        0,-.22344,.420704                  0,-.33641,.420704      6.35   13
12        W36E1        0,.054835,      0     W13E1        0,.193428,.138594      6.35   23
13        W12E2        0,.193428,.138594     W14E1        0,.057663,.274359      6.35   23
14        W13E2        0,.057663,.274359     W15E1        0,.192014, .40871      6.35   23
15        W14E2        0,.192014, .40871                  0,.312014, .40871      6.35   15
16                -0.081, -0.019,  0.037             -0.081,   -0.4,  0.037      6.35   45
17                -0.081, -0.009,  0.137             -0.081, -0.289,  0.137      6.35   33
18                -0.081,-0.0055,  0.262             -0.081,-0.5615,  0.262      6.35   65
19                -0.081, -0.004,  0.414             -0.081, -0.515,  0.414      6.35   61
20                -0.081, -0.004,  0.592             -0.081, -0.274,  0.592      6.35   33
21        W23E1        0,.223443,.468654     W22E1        0,.223443,.420704      6.35   7
22        W21E2        0,.223443,.420704                  0,.336406,.420704      6.35   13
23        W21E1        0,.223443,.468654     W10E1        0,-.22344,.468654      6.35   53
24               -.25325,-.40429,.367714            -.25325,.404294,.367714      6.35   95
25        W49E1        0,-.05483,      0     W26E1        0,-.19343,-.13859      6.35   23
26        W25E2        0,-.19343,-.13859     W27E1        0,-.05766,-.27436      6.35   23
27        W26E2        0,-.05766,-.27436     W28E1        0,-.19201,-.40871      6.35   23
28        W27E2        0,-.19201,-.40871                  0,-.31201,-.40871      6.35   15
29                -0.081,  0.019, -0.037             -0.081,    0.4, -0.037      6.35   45
30                -0.081,  0.009, -0.137             -0.081,  0.289, -0.137      6.35   33
31                -0.081, 0.0055, -0.262             -0.081, 0.5615, -0.262      6.35   65
32                -0.081,  0.004, -0.414             -0.081,  0.515, -0.414      6.35   61
33                -0.081,  0.004, -0.592             -0.081,  0.274, -0.592      6.35   33
34        W47E2        0,-.22344,-.46865     W35E1        0,-.22344, -.4207      6.35   7
35        W34E2        0,-.22344, -.4207                  0,-.33641, -.4207      6.35   13
36        W49E2        0,.054835,      0     W37E1        0,.193428,-.13859      6.35   23
37        W36E2        0,.193428,-.13859     W38E1        0,.057663,-.27436      6.35   23
38        W37E2        0,.057663,-.27436     W39E1        0,.192014,-.40871      6.35   23
39        W38E2        0,.192014,-.40871                  0,.312014,-.40871      6.35   15
40                -0.081, -0.019, -0.037             -0.081,   -0.4, -0.037      6.35   45
41                -0.081, -0.009, -0.137             -0.081, -0.289, -0.137      6.35   33
42                -0.081,-0.0055, -0.262             -0.081,-0.5615, -0.262      6.35   65
43                -0.081, -0.004, -0.414             -0.081, -0.515, -0.414      6.35   61
44                -0.081, -0.004, -0.592             -0.081, -0.274, -0.592      6.35   33
45        W47E1        0,.223443,-.46865     W46E1        0,.223443, -.4207      6.35   7
46        W45E2        0,.223443, -.4207                  0,.336406, -.4207      6.35   13
47        W45E1        0,.223443,-.46865     W34E1        0,-.22344,-.46865      6.35   53
48               -.25325,-.40429,-.36771            -.25325,.404294,-.36771      6.35   95
49         W1E1        0,-.05483,      0     W12E1        0,.054835,      0    5.5086   13
50               -.25325,-.44708,      0            -.25325,.447082,      0      6.35   105

Total Segments: 1778
See less See more
Re 300Ohm's nec file of GH6.3 TopHat Narod with mesh Screen


In post #17 of the YA-1713 thread you made the following comment:
300Ohm said:

A couple of notes on 4nec2 modeling you may not be aware of while you were gone.

1) per Arie Voors, when running an AGT test, set the resolution to 2 degrees instead of the default 5. This is to take care of possible tiny lobes that could distort the test.

2) Also, set Auto-Segmentation to 21. If Auto-Segmentation is set to around 15, the gain could be significantly overstated. Theres a thread on this somewhere.
Yes, I have been away for about 15 months and I appreciate the heads-up about the latest 4nec2 modeling notes.

In reference to your 2nd comment about 4nec2 modeling notes (quoted above), am I to assume that your nec file in post #31 of this thread is an example of the new 4nec2 auto-segmentation note? [Reference link for nec file: ]

Nikiml's Top-Hat Narod design ( GH6.3 ) is based on the j3d GH6_rev2 which was designed for a bandwidth of Ch 14-51 (470 to 698Mhz).

The comment note in your nec file of post #16 states:
CM model segmentation is derived from autosegmentation 22 at 800Mhz
CM for VHF AGT of 1 set the radius multiplier of the souce wire to 1

I am somewhat puzzled as to why you set the auto-seg for 800Mhz, when the upper freq of the design objective for UHF is 698 mHz (a difference of 102mHz above the upper design limit)

I ran the AGT test at 800mHz and got this result: 1.0 (0.1dB), which is an excellent AGT result, but it is 102mHz above the upper design limit.
The validation check after running the AGT test just showed the typical warnings len/rad at junction (...x...) below 6 ( nothing to be overly concerned about)

For AGT tests, I always follow Ken Nist's advice when running AGT and run it at the mid freq of the desired bandwidth. For this design, the desired bandwidth is Ch 14-51 and the mid-freq is 584 mHz.

When I set the freq for the AGT test at 584mhz, instead of 800mHz as you did, I get the warnings that Ken Nist spoke about in my post # 30 on this thread and these warnings pop up before the AGT test is run. Reference link to Post 30: [ ]

AGT test warnings at 585mHz
GH 6.3 Combo by Nikiml with 300Ohm Mesh narod wavelength=0.513 mtr.

Note : parallel Wires 4 (tag 4) and 9 (tag 14) have unequal segmentation
Note : parallel Wires 13 (tag 24) and 18 (tag 34) have unequal segmentation
Note : parallel Wires 23 (tag 44) and 28 (tag 54) have unequal segmentation
Note : parallel Wires 32 (tag 64) and 37 (tag 74) have unequal segmentation
If you check your model's geometry you will find that these warnings are because of the close proximity of the "c"section of the Narod to the array's horizontal stub.
Wires 4 and 9 are the top-left
Wires 13 and 18 are top right
Wires 23 and 28 are left bottom
Wires 32 and 37 are right bottom

Personally, I will continue to run AGT tests at the mid freq of the design bandwidth. In some cases I also check at the upper and lower design bandwidth limits.

I don't use auto-seg, partly because Arie Voors previously had bugs with that code in the past and I prefer to run a "convergence test" at higher segmentation density as well as the AGT test, as outlined in the Beginner's guide to NEC modeling.

On another point about your mesh screen UHF results: I suspect that you would see a better UHF gain-freq response if you eliminated the UHF collinear reflectors. I think you mesh screen is causing severe disruption to the normal gain pattern that the coliinears normally provide. You are probably aware that this can easily be done with 4nec2 by using the geometry editor to delete those elements and then run a freq sweep using "this current modified geometry" If you like what you see, you can save as a new file name from the geometry editor. Just remember that such saves will discard any SY variable data you had in the model.

See less See more
300Ohm said:
Ah ha, I found the auto-segmentation controversy ! Before and after Snowman53 post #42 here:
If its a bug, I think its a NEC engine bug, not just 4nec2.

Thanks for the link re the discussion on auto-seg. I have only quickly scanned from post #422 to 432, but I'm not at all sure, I can identify any clear consensus. It appears that there was agreement to use auto-segmentation, but I am left wondering if the original cause of different modeling results was clearly understood.

A modeler gives up control of his model to this auto-segmentation algorithm that Arie has coded for 4nec2. I'm from the old school, where you manually set the segmentation. Then as I said before, besides running an AGT test at the mid designed bandwidth freq, (or at the band limits, if required) you also run convergence tests by increasing the manually set segmentation to ensure there is no significant change in the gain/swr results if more segments are defined in your model. Now for TV UHF, you also run up against that wire diam /seg_len problem, so you do have to be careful and really understand what these tests do and what you are doing as well ... life is not always simple and blind accepance on automatic computer segmentation may have some unexpected consequences.

I would refer folks to the ARRL Antenna Book Chp 4 "Antenna Modeling & System Planing ( I have 21st edition). It discusses the Accuracy Tests for nec models. There are only two: The Convergence test and the Average gain test (AGT)

Here is an important quote from pg 4-16: Like the convergence test, the average gain test is a necessary but not a sufficient condition of model reliability. Pass both tests, however and you can be pretty well sure that your model represents reality.

For any modeler that posts on this forum, that does not understand this quote, I highly recommend that you beg, borrow or steal a copy of the ARRL Antenna Book and start reading. I don't really advise breaking the law :eek: .... It should be available in just about any public library in North America.

You should also read all 4 parts of "A Beginner's Guide to Modeling with NEC". Part 4 has good discussion on the convergence and AGT tests under the heading "Testing Models". These guides are available on Arie's site.

I hope this helps.
See less See more
300Ohm said:
But I can also now see a problem arising where I manually segment something X, someone else has auto-segmentation on or uses/changes a different manual segmentation, and we get results that differ on the order of a couple of db.

Based on my own experience, I can tell you that I find a difference of a couple of dB, very surprising for a model with the min of 10segs per half wavelength vs say 22 segs per half-wavelength. Were the segments per element of the two models apportioned the same by element length? If you are tweaking the antenna design, you always need to recalculate the segmentation to ensure this equal segmentation is maintained for all elements.

If you are certain that all elements were equally segmented by length, then that couple of dB difference would have been easily detected by a convergence test. You do not need auto-segmentation to do that. So I am still left wondering what were the reasons that lead to the auto-seg consensus.

Anyway, not to worry... I'm a old foggy that's set in his ways and if anyone wants to read-up on this subject in more detail, I provided some good reference links in my previous post on this subject.

I really need to get back working on my web-site. Spending far too much time on this forum lately.

See less See more
Zapperaman said:
Yes, you do. Some of us are eagerly awaiting that, as well as your Rev 3 GH10 antennas... back to work!
The launch is looking more like mid July now.
Some of my time presently is spent regarding the latest legal twist for current Canadian Nortel LTDs. This is truly a very sad story, and for most people it is totally unbelievable that it could actually happen in Canada... but it has !!

Normally I don't reference such news stories on the forum, and I suspect that Stampeder will remove this link, but until he does, here is a short article by a Nortel pensioner in Atlanta. Tom tells this Canadian Nortel LTD story "Straight and True" and I thought you should know.
All we can do is write & phone our elected politicians to get them to act and bring justice for those Nortel LTD employees, and that is what many of us are doing.
Some Final Thoughts on Autosegmentation

Note to Stampeder: This discussion should really be moved to the Software 4nec2 modeler's thread, but I leave that to your discretion


In reference to your post #39

I can tell you that I find a difference of a couple of dB, very surprising for a model with the min of 10segs per half wavelength vs say 22 segs per half-wavelength.

OK, 1 1/4 dbi, I exaggerated a bit,heh.
A couple of comments on the plots by Snowman53:

1. The plots of Gain vs auto-seg per 1/2 wave for SBGH, show a significant drop in gain over a narrow range from 10 to 13 segments per1/2 wave.
Above 13, the gain results are reasonably constant (ie good convergence). In the case of DBGH plots, I wonder if those models were collinear or mesh screen or both? There was no discussion in the thread that I could find, that discussed how auto-segmentation actually calculates the segments for each element and what that implies for larger models.

As all modelers who post results on this forum should know, the minimum segmentation recommended by L. B. Cebik in his "Beginner's Guide to NEC Modeling" is 10 segments per 1/2 wave at the design frequency of the antenna. So I ignore the Snowman53's plot results for < 10 segments per 1/2 wave.

2. As I mentioned previously, my experience with convergence tests on manually segmented models at 10 -20 segments per 1/2 wave, have never shown anything close to a 1 1/4dB difference.

I suspect that Snowman53's auto-segmentation results might actually be characterizing the auto-segmentation algorithm, rather than antenna model differences at the 10 -13 auto-seg range. It is unfortunate, that a manual segmented model was not compared with an auto-seg model, since that would answer my suspicion. However your GH6.3Narod_with_Screen model presents an opportunity for any modeler to investigate that difference !

I propose this Modeling Challenge:
Take the GH6.3-Narod_with Screen model by 300Ohm in post #31 of this thread and make a manual segment version using the design freq of 584mHz with 10 segments per 1/2 wave. You can use a smaller segmentation density for modeling the mesh screen reflector. If you are really keen, you could show the effect of varying the segmentation of the mesh screen from a very small density up to 10 segments per 1/2 wave.

(1) Run a sweep 461 to 707; step 12 and compare the gains. What is the max difference between this manual segmented model vs the auto-segmentation 22 at 800Mhz model?

A final thought:
In his post summary, Snowman53 stated:
Snowman53 said:
In summary I think NEC is OK for comparing variations of the same antenna only if a consistent segment sizes are used.

I am not sure one can trust NEC to provide consistent values between different types of antennas which may use different segment sizes.
I would welcome a discourse on this topic with the resident modeling wizards.
Reference post:

"I am not sure one can trust NEC to provide consistent values between different types of antennas which may use different segment sizes."

Now if the segmentation is not done correctly due to modeler's in-experience, or other cause, this statement is true, but if the models follow all the the segmentation rules correctly, I don't believe this statement to be true. If it was, it would mean that all the excellent work by Ken Nist comparing commercial antennas would be in doubt as well.

I think the trust that Snowman53 is unclear about concerning different types of antennas needs an answer and I totally agree with his statement "I would welcome a discourse on this topic with the resident modeling wizards."

Perhaps the "resident modeling wizards" will seriously take a look at my modeling challenge.
I am reminded of that old expression .... "You can lead a horse to water, but you can't make him drink" The challenge has been presented .... any takers?


[ Oh,Oh .... I hear Zapperman calling gotta go and get back to work ....Cya later]
See less See more
With reference to 300Ohm's post #49 ->

300Ohm said:
... its best to get more modelers into this discussion , rather than just you and me alone. holl_ands will probably be back, and maybe Snowman53 too. mclapp, since he's very good with EZNEC can provide some insight too.
That would be great. I did mention "Perhaps the "resident modeling wizards" will seriously take a look at my modeling challenge."

I would suggest that you enlarge the scope of the discussion to think about collectively generating a more detailed modeler's guideline that Stampeder could make into a "sticky"

For me, there are really two important questions that snowman53's auto-segmentation analysis raises. I had mentioned them in my previous posts on this subject, but I guess I wasn't that clear. So let me have another go and provide some points that might start the discussion with a wider viewpoint.

Question 1. How to determine an antenna model's "reliability" for posted results of Gain/Swr etc

The late L. B. Cebik (W4RNL) pioneered the use of MoM nec programs in the ARRL as a better way to compare commercial antennas for Radio Hams, and to provide a software tool to better design "Home-Brew" Radio Ham antennas covering a wide range from 80m up to "Moon-Bounce" frequencies.

His recommendation for determining a model's reliability is to run two tests: not only AGT at at least one frequency (within the designed bandwidth), BUT one should also run Convergence Tests, to determine how increased segmentation affects the results of a specific nec model.
The value of these two tests is simply this: It is applied to a specific antenna model, hence comparisons of models of different antennas that have been found to be "reliable" via these two tests, provides the best and only assurance that the comparison is valid.
The AGT test should be done at at least one frequency inside the designed bandwidth and Convergence tests should be done from the minimum recommended density of 10 segments per 1/2 wave up to the max limit of segmentation that meets the segmentation rules.

In 4nec2, Arie has developed a Convergence test that does exactly that. It is the "Conv-test" function in the Calculate->Start optimizer (F12) tab of the main window of 4nec2. I would recommend that you ensure your Auto-segmentation setting is disabled before running the Convergence and AGT tests. If you have never run the Conv-test function in 4nec2, take a close look at it. It is an extremely powerful tool that Arie has built into 4nec2 !

I really believe that running both tests will provide the answer to Snowman53's query "I am not sure one can trust NEC to provide consistent values between different types of antennas which may use different segment sizes."

Question 2. Why is Auto-segmentation used ? Is it ever Disabled ? ...and other questions

Auto-segmenation is a useful computer aid for 4nec2 modelers, especially when the model is of an antenna with complex geometry. Snowman53's analysis shows that setting auto-seg to >13 gives good convergence on the specific SBGH and DBGH models that he considered in his analysis. It does not necessarily imply that all models will be reliable if you use auto-seg at a setting >13 (say 21 or 22). Instead, you should test each specific model using AGT and Covergence tests. That will give a measure of the model's reliability; using auto-seg does not give such assurances.

My modeling challenge was geared to explore what the auto-seg algorithm in 4nec2 actually does for complex antenna geometries, including a mesh screen. I was not familiar with the details of the snowman53 models, hence my suggestion for the GH6.3-narod with Screen. Feel free to explore with any other model.

Personally I don't use auto-seg. From what I understand, it is very useful when first setting up your model segmentation and should be used when running sweep optimizations. I suspect that most other times your should disable auto-segmentation. There might be some other strange effects that others have noticed, using auto-seg. You probably have the most experience of anyone on this forum with mesh-screen models; would you recommend using auto-seg with mesh screens? From Arie's help file, it's interesting to note that you can selectively turn-off auto-segs for specific wire elements in your model via a preceding + <sign>. This would allow the modeler some control over the auto-seg algorithm.

I see more and more new comers to 4nec2 modeling on this forum. I know from my journey that there is lots to learn and even more to continue to learn - for all of us. I really believe a more detailed 4nec2 guideline "sticky" that provides some explanation on these subjects would be very useful and would further grow the reputation for as the world's best modeling forum.

Anyway...time to go... I hope folks find this helpful in better defining my thoughts. I'm way over my allotted time... you know ...Zapperman is calling it's back to "kompozer" and more html stuff.

4NEC2 Auto-segmentation Help Notes

When optimizing and/or sweeping antenna performance, it is likely thatdimensions and/or frequencies are changing, so it is possible that certain NEC requirements, such as the minimum number of segments per half-wave are not met any more. To overcome this problem, Auto-Segmentation is used. 

If this feature is enabled (See Settingssettings on Main window), each calculation/optimization/sweeping step, the wire length for each wire is checked against the actual wave-length and the specified number of segments per wire, and the position for voltage/current sources, RLC-loading and transmission lines is updated.  

 If a certain number of segments is already specified in the input file, at least this number of segments is set with auto-segmentation. This to avoid auto-segmentation decrease the number of segments specified for critical regions.  If you do not want to specify a minimum, set the number of segments to one.

If the number of segments is preceded by a “+” sign the number of segments is fixed, meaning, the segments for this wire are not affected by the auto- segmentation process.

 When enabling the Auto-Segmentation feature, you are asked to specify the number of segments per half-wave. Although, mostly accuracy improves if more segments are specified, also the computation time is increased. For ‘simple’ structures, without sharp angles, a value from 10 to 20 will be sufficient. If precise impedance values are required or if wires joining at sharp angles are used more segments could be required. If you are not sure what amount to specify, please use increasing values and compare the results. (You could do this using the Sweepereval function). Also abrupt current changes on the structure might indicate an insufficient number of segments. One place more segments aren’t better is if wires of different diameters are connected.

If computer performance is an issue, you could consider not to the use the auto-segmentation feature. If this feature is disabled, the processing especially for large input files with many wires is speed-up.
See less See more
  • Like
Reactions: 1
Re post 59:
Nikiml said:
I tried convergency test on the original Jed GH8 at freq of 590Mhz and it did not converge. I am starting to think that maybe the AGT does not stay at 1 when 4nec2 does the convergence test.

j3d's nec files have a high segment count with a segment per 1/2 wave >>10, so that doesn't leave a lot of room for increased segmentation before you exceed the segmentation rules for len/rad parameter. He does this because he believes adjusting the wire radius of the Vsource to fine tune the AGT breaks nec sgmenation guidelines, but .... that's another topic for later discussion.

Don't be overly concerned about the AGT test. Since it is calculated at a specific frequency, it will change at other frequencies.
The 4nec2 help gives this guideline:
To check the reliability of your model, you can run an ‘average gain test(AGT). 
To run such a test, use the ‘Generate (F7)’ command, select ‘Far field pattern’ and check ‘Aver-Gain test’. Then click the ‘Generate’ button.

When the ‘expert’ button is selected and using the lowest option-box, 
one should select the ‘1 – average gain’ option to perform this same test. (The other option-boxes could be left to 1:Ver/Hor/Tot field, 0: no norm, 0: power-gain). 

To get usefull test results 4nec2 automatically (temporary) removes all 
wire- and other resistive loading and (if specified) changes the gound-conditions to perfect-ground when doing the required calculations.  

The test-results are presented on the main (F2) form. 
The following gradation can be made for the Average-Gain-Test results:

> 0.95   and <    1.05      Model is likely to be accurate
> 0.90   and <    1.10      Model is usable for most purposes.
> 0.80   and <    1.20      Model may be useful, but can be improved.
< 0.80   or    >    1.20      Model is questionable and should be refined.

When the resulting value is below 0.8 or above 1.2 the value is displayed in red color. The db value between brackets can be used to correct the far-field pattern gain.
To see the conv-test in a better light, start the model with segmentation based on the minimum guideline of 10 segs per 1/2 wave. Remember to disable your auto-segmentation option before running the conv_test.

The bandwidth of your UHF is Ch14-51 (470 to 698) mHz. The mid-freq is 584mHz. Set your frequency in the freq tab of the NEC Editor (new) to 584mHz. You will see that 4nec2 reports that freq and it's wavelength in the main window. The 1/2 wave for 584mHz is half of that value or 513/2 = 256.5mm or 0.2565m.

At 10 segs per 1/2 wave, your seg-len (segment length is 25.65mm)
The number of segs for the Vsource (tag 100) is it's (length / seg-len)
In meters that's 0.15/0.02565 = 5.84 segs, which you round off to 5, because the Vsource must have an odd number of segments to ensure the Vsource is exactly at the center of Tag 100.

Repeat the same process for all other tags, except round up the number of segs. (these can be even or odd rounded up integers) and that is your segmentation for 10segs per 1/2 wave at the mid UHF freq of your design.

Now run conv-test and post your results of Gain vs total segments. Then we can discuss the degree of convergence.
See less See more

Re your conv-test results in post #61, I sent you an email with some tips and suggested investigations you could do to obtain a better understanding of these results. I don't have the time to dig into now.

I would simply point out that runs 1-0 to 1-3 are for segmentation at 10 segs per 1/2 wave to ~23 segs per 1/2 wave and the gain varies from 13.13 to 13.02. The next run 1-4 has a total seg count of 1004, which is ~31 segs per 1/2 wave.

I suspect that for run1-4 (and 1-5 to 1-12), the normal 4nec2 segment check would report errors. If this turns out to be the case, conv-test results for these runs are not valid.

If you have the time, it would be very interesting to investigate a manual nec file based on run 1-4 at ~ 30 segs per 1/2 wave. Does it pass the normal 4nec2 segment check with no errors at 585mHz? What about runs 1-5 to 1-12 ?
mclapp said:
..I was going to run that nec model (post 54) in EZNEC and run the auto seg on it but I couldn't import it in because EZNEC likes to import actual x,y,z cooridinates and those are not.

You can easily convert the 4nec2 nec file into compatible (x,y.z) co-ordinates for EZNEC, by resaving the nec file in 4nec2's geometry editor, under the settings tab in 4nec2's main window. That save will strip out all the SY cards.

re your other comment:
Many times after auto segmentating certain models I had to adjust the source wire diameter very close to the size of the wires it was attached to to get the AGT back in line. I would think that is a good thing because I'm still on the fence about adjusting the source wire much to get a good AGT.
I have been absent from the forum for the past 15 months or so, but when I started my trolling on the Antenna R&D forums to get caught up on past events, the comments on AGT by j3d really caught my attention. The public domain nec2 core has well documented limitations, one of which is connection of wire elements having different diameters. If the AGT test is adjusted for 1.0 by changing the wire diameter of the Vsource, then we are "adding" to this nec2 core limitation in our model, rather than always trying to minimize the nec2 core limitations to get the "best" model for representing the antenna design. This is the point that j3d was making.
Reference post by j3d 5 May 2009 -->

From my reading in this thread, I don't see that much discussion took place on this subject. For those that have a desire to better understand the limitations of the nec2 core and how best to deal with nec2 limitations in their models, I see the question to be answered as this:

If the adjustment to the wire diameter of the Vsource is only a very small percentage, does this "procedure" have merit or should it be abandoned?
In post 49 of this thread, 300Ohm mentioned the following:
300Ohn said:
...And maybe Arie Voors too. Hes posted on this website in your absence, and likes us too. So has Dr. Natan Cohen, the Fractal Guy, the originator of the fractal antenna.
I would suggest that this subject should be discussed in more detail.
I'm sure that all, would like to get the views from Arie and Dr Cohen.

300Ohm said:
Ive always looked at it, because the Vsource wire isnt a real wire in the build anyway, that its sole purpose is for AGT adjustment. (or maybe possibly as part of a balun wire)
... Heh, yeah. I think some of those warnings may be built in cost overrun warnings.

Re your comment "the Vsource wire isnt a real wire in the build anyway": Are you stating an opinion that you believe the "AGT wire adjustment procedure" is valid or it is not valid, since the Vsource wire element is not a part of the antenna design?

I'm also not exactly clear on what you mean by "those warnings may be built in cost overrun". By cost overrun, are you implying that 4nec2 warnings are not really all that important and thus can be ignored? Only errors are really important?

I'm not being critical, I just feel that clear understandings are needed to have a good discussion.

More "Food for Thought" re ConvergenceTest & AGT Adjust

My thanks to 300Ohm and mclapp for their comments:
Reference links:
300Ohm (post 68) -->
mclapp (post 70) -->

Like 300Ohm, I have used the AGT adjust "procedure" in all my past posted 4nec2 nec files. I never really questioned the effect of different wire diameter, on this AGT adjustment, because the results never indicated any significant wrong results.

After recently reading j3d's post on the subject, and with a re-read Cebik's modeling guide, part 4 ( the last 2 pages under the heading "Testing Models"), I thought it would provide a better understanding of 4nec2 limitations, if jed's question could be answered with some detail and if possible some reference links.
Reference link to j3d's post -->

For me, the questions that need answers is simply this: ... in the case of the AGT adjust procedure, do two wrongs actually may a right" and along with that, do other modelers, besides forum members, use this AGT adjust procedure?

The recent posting (post 61) by Nikiml re Convergence Tests on his model, provides some "food -for-thought" and I invite comments. If the majority of readers on this subject feel, that this is somewhat of a waste of time, I will stop posting on this subject, as I have plenty of other activities to keep me busy. The only way I can "measure" the interest in this subject is from the posts that are made.

Food - for - Thought Comments
Nikiml's post 61, Convergence-Tests showed, that up until segment-check errors occurred, the raw gain varied from 13.13 dBi (total_segs=328) to 12.99 (total segs=1004). Reference link (post 61): -->

By extrapolation of Nikiml's data, 21 segments per 1/2 wave raw gain would be 13.03 dBi. For this specific model, the Convergence Test reveals the "most accurate" raw gain convergence is 12.99dBi. So does the recommended model at 21 segments per 1/2 wave using the AGT adjust procedure, give a more accurate result, not only for this model, but for any model, regardless of the complexity? The beauty of the Non-adjusted AGT Test and the Convergence Test is: you can depend on the results, regardless of antenna model differences (GH vs Yagi or BowTie etc)and the antenna's complexity.

In an effort to continue this discussion, here are some more "food-for-thought" links.

1. Arie's 4nec2 Forum
The AGT is off - Why? A bug?

Auto Segment

2. Practical Antenna Modeling - WX7S

3. Antenna Modeling for Radio Amateurs
Here is an excellent overview about some of the history of antenna modeling with references to mathematical models. There is a brief discussion about limitations of the various modeling approaches and a summary of EZNEC and 4nec2 is included. This pdf doc was written by Steve Stearns, K6OIK.

See less See more

In reference to your post 72 -->

That is an excellent summary of your GH10n3 - GH4.3 variants.

Most on this forum prefer the tablular data, but I believe a picture is worth a thousand words, so I put together a graphical comparison plot for the UHF and VHF-Hi Net Gains.

To get an appreciation for how good your models are, I compared your net gain data with two commercial antenna models done by Ken Nist (hdtvprimer website)
UHF - Channel master CM-4221 4bay Bowtie, (no longer in production, but all antenna buffs know this antenna as a top performer.)
VHF-Hi - Wade wideband yagi YA-1713 which is for VHF-Hi only and recognized by antenna buffs as an excellent antenna.

Net Gain - UHF

Net gain VHF-Hi

These plots clearly show the GH-xNy is in a "best-of class" category for a DIY Combo (VHF_Hi & UHF) antenna. Well done !!

Here is the link for Ken Nist's models for some of the newer commercial antennas. Take a look at the net gain for VHF-Hi commercial models; HD-7698P and CM-2020 near the bottom of the page.

Added Footnote re YA-1713 comparison:

Ken Nist's model of the Wade YA-1713 antenna shows a rapid drop-off at the upper 3 Mhz of the 6Mhz Ch 13 bandwidth.
The model results for Ch 7-12 are accepted as close to real-world, but Ch 13 model results are not. I used the exact data that was published on the hdtvprimer website for YA-1713, so this comparison shows the rapid fall-off for Ch 13.

300Ohm posted this comment concerning the model results of YA-1713:
Yep, but my point is, when using the YA-1713 graph in comparison, at least a footnote should be made about the channel 13 drop off, given what we now know.

(Winegard claiming 10.3 dbd on channel 13 on their specs )

Regarding a possible explanation of the rapid Ch 13 drop-off of the YA-1713 model by Ken Nist, I offered this possible explanation in the YA-1713 thread -->
The winegard claim of 10.3 dBd or 12.45dBi is very close to the gain in the reference post above.
See less See more
300Ohm said:
I thought we werent going to use that YA-1713 channel 13 drop off. I would just level the graph for it as the same as channel 12 and call it done, heh.

I was using Ken Nist's data from his YA-1713 model, exactly as he published the info, for a comparison to Nikiml's GH-Combo designs. The fact that there seems to be a small discrepancy in Ken's model re the Ch13 roll-off does not in any way make this comparison invalid. And as you pointed out re your Ch12 comment, most readers accept the YA-1713 model results for Ch 7-12.

If you or anyone else, are interested in my "take" on the sharp drop-off of Ken's YA-1713 model, I posted a possible explanation here -->

300Ohm said:
Yeah, but also in the thread investigating holl_ands YA-1713, his antenna was physically a little different than the model, which made me speculate Ken had a much older version than what has been sold over the last few years.

I think you your comment (quoted above) should really go in the YA-1713 thread. My posting re the Ch 13 drop-off showed that just a 4mm decease in the total lengths of the reflector and all directors, would provide excellent model results for the YA-1713 antenna over the Ch7-13 bandwidth. 4mm is only 0.157" on a reflector total length of 32.14" and the two director total lengths of 25.14" and 23.86". In percentage terms - the reflector decreases by 0.488% , director D1 by 0.625% and directors D2-D6 decrease by 0.658%.

It is clear that slight differences in the lengths of the parasitic elements could make a major difference in the free-space model results for Ch13.

300Ohm said:
Yep, but my point is, when using the YA-1713 graph in comparison, at least a footnote should be made about the channel 13 drop off, given what we now know.

Thanks for the suggestion. A footnote has been added to the comparison post -->
1 - 16 of 940 Posts
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.