Some Things I'd Like to See in NEC-5

L. B. Cebik, W4RNL

nless you work directly with either of the NEC cores (-2 or -4), you may not realize just how many of the functions upon which you rely are simply not there. Instead, those who privately or commercial provide NEC with input and output modules embellish the core's calculations. As a result, the average user is rarely sure of what NEC is doing and what the implementing software additions are doing. For example, EZNEC and NEC-Win Plus provide the user with tabular entry screens for entering the wires that make up the antenna structure, thereby obscuring the fact that the NEC core responds to ASCII inputs, like the sample lines in Fig. 1. Based on Fortran, NEC requires an input sequence that follows Fortran rules, although between NEC-2 and NEC-4, the number of floating decimal places allowed on a Fortran line increased.

NEC-2 emerged in the early 1980s, and NEC-4 followed in the early 1990s. However, the early 2000s saw no further development of the cores. So perhaps it is useful to think about what one might wish to see in the core of the future, a core that I shall call NEC-5. My list of druthers is idiosyncratic, based on my experiences with using the cores, as well as on a good bit of beta testing and technical support. But the simple fact that I do not catch bugs that others have found obvious is reason enough to consider my list of desires personal rather than perfectly general.

Some Basics

Let's start with a basic property of NEC's calculations. Some users have wondered why the calculations are limited in the most general case to a little over 11,000 segments and in some extended cases, to about 20,000 segments when using NEC-4 in its double-precision option. Here, the core is not at fault. It can handle much larger matrices. The limitation is the operating system on which folks run NEC most often: Windows. The virtual memory system in XP has some limits, although there are ways of re-adjusting those limits. As well, some programmers have used various maneuvers, such as going to single precision and pre-segmenting the matrix for bulk transfer. All of these programming techniques try to overcome the operating system's limits, not the core's limits.

However, there is one major area in which NEC is deficient. It calculates only the axial currents, that is, those along the length of a wire. NEC-4 increased the standard order to which the core calculates and developed a much more accurate direct calculation of stepped-diameter elements. However, the calculations remain well shy of accurately handling large diameter steps, and Leeson substitute uniform-diameter elements remain the standard of accuracy. However, Leeson substitute elements remain limited to linear elements without loads except at their center, and the elements are reliably usable only within about +/-15% of the operating frequency.

As well, NEC will not calculate the effects of a boom at right angles to an element, when the boom is either in close proximity to the elements or is in contact with them. Experiments that I have performed show the same output reports with no boom, with a near boom, and with a contact boom for a series of Yagis that should have shown some differences.

The problem may lie in the restriction of current calculations to only the axial currents, with no calculation of the transverse currents. Even when NEC-4 was under development, computers were far slower than today. Some of those limitations went into the development of the calculation scheme used by NEC. Although it might involve complexities in the algorithms used by the core, the addition of transverse current calculations might expand the utility of NEC for handling a large collection of antenna types.

MININEC is not bothered by the stepped-diameter problem that is endemic to NEC. However, as suggested by Fig. 2, its segmentation and current calculation scheme differs from NEC. NEC places a current locus in the center region of each segment, while MININEC uses pulses that occur at the junctions of segments. As a consequence, when I wish great accuracy for various types of antenna, I often turn to highly corrected versions of MININEC. For example, when I designed a number of quad loops (square, diamond, or rectangular) with fat horizontal sections and wire vertical sections, only MININEC yielded results that corresponded tightly with prototypes. Whether NEC-5 needs to adopt/adapt the pulse system is uncertain, for it is not clear what effects the change might have on all of the other calculations that occur in the core. Indeed, there has been talk from time to time that the next generation of NEC may use Eigen functions for its calculations. Such a shift might well overcome all of the basic calculation limitations of NEC-2 and NEC-4. But the development effort might also require thorough-going changes throughout the core's system of calculations.

The key point of this note is not the technique used to achieve the goals, but the goals themselves.

Some More Specialized Improvements

NEC has a number of functions--some old, some new--that can stand some improvement.

1. Transmission Lines The transmission line or TL function within NEC is a special re-formulation of the network or NT function that allows the placement of transmission lines between any two designated wire segments within the model. Although the user can elect to use the straight-line distance between the segments that serve as terminations, the actual distance is a user option. Within NEC, the user selects the electrical length of the line. However, many implementations have given the user an interface that allows the entry of the physical length along with a velocity factor.

Transmission lines do not function like the geometry or wire entries. They play no role in the matrix and current calculations. Rather, they function as mathematical "add-ons." One major limitation is that NEC transmission lines use equations for (or that are equivalent to) lossless lines. One major improvement that might be feasible is the option to employ equations for (or equivalent to) lossy lines. Private and commercial implementers of NEC can then add any number of selection tables for common (or even rare) lines. For short runs, typical of phasing lines within a horizontal array, lossless lines yield no significant errors relative to prototypes. However, in numerous cases, transmission lines may be long enough that basic match line losses are significant. As well, the absence of a match between the load and the line creates a multiplier on the basic matched-line loss. This information can be valuable, not only for HF stations with long lines, but as well for VHF and UHF installations, where even shorter lines may create notable losses.

2. Insulation NEC-4 has two primary means of introducing insulation or a second medium. One of those methods allows a second medium to replace the normal vacuum used in the absence of this command. The UM command replaces the normal antenna environment (in free space or above a ground plane) with a medium for which the user specifies the conductivity and permittivity. This command has two limits. First, it extends everywhere except below a ground plane. Second, it cannot be used with the Sommerfeld-Norton (S-N) ground calculation system.

The second method of introducing a more limited second medium is through the use of the insulated sheath or IS command. Specifying a sheath also requires the input of values for conductivity and permittivity. However, the sheath always applies to round wires, and the sheath begins where the wire ends--from a radial perspective. The user provides a sheath radius to set the outer limit of the second medium.

Both commands share a common advantage. The user may specify conductivity and permittivity values over a very wide range. Hence, the second medium may have the properties of a second conductor, an insulator, or a semi-conductive material. Theoretically, one might model the structure of copperweld wire as well as the usual non-conductive insulation on wires.

The insulated sheath command may be more flexible than many users presently know. Although the sheath is round, the user can approximate the velocity factor of covered parallel transmission lines by some judicious preliminary line models to simulate actual lines that do not have the same shape. Fig. 3 suggests the means and the need for preliminary models in this connection. The tubular parallel line has mostly air between the actual conductors, but the sheath model may select combinations of permittivity and thickness (outer radius) that yield very close approximations of actual line functioning.

In addition, the user can approximate the properties of support tubes and cores around which one might wind a helix. One can begin with a wire to which he assigns a very low (1e-20 S/m) conductivity. Around this wire, one may then order up an insulated sheath with the correct thickness, conductivity, and permittivity to simulate a tube of any particular insulating material used to support the helix. The technique may not be applicable in all cases, but it may prove useful in some. See Fig. 4.

Nevertheless, there are numerous other types of cases that call for a limited region of an insulating material for which NEC in its present forms is helpless. One common case is the support of antenna wires by a substrate of known conductivity and permittivity, that is, the antenna on a circuit board. Setting aside the matter of round vs. flat conductors for the moment, NEC has no effective way to establish below the antenna conductors a second medium that has a limited area and thickness. In principle, it might be possible to construct such areas from the SC or SM surface patch commands, but these commands at present have no provision for the assignment of something equivalent to permittivity values. These commands had their origin in trying to simplify the construction of and calculations for structures only indirectly associated with the basic operation of an antenna (such as ships or vehicles). However, some adaptation of patch techniques to create insulating substrates might extend the utility of NEC, especially in the UHF region.

3. Circular Polarization Radiation Tables The RP far-field output radiation table offers a variety of calculations of use to the modeler. Many of the variations appear in the options offered by the XNDA entry on the RP0 command line. Rather than being a true floating decimal entry, the 4 numbers of XNDA stand for 4 different selections of desired outputs. Perhaps the most common entry is 1000, which provides the standard radiation table listing the vertical and horizontal components of the total field as well as the total field itself. Alternatively, 0000 provides the major and minor axis components of the total field, as well as the total field itself.

It is possible to use the data from the radiation table under XNDA = 1000 to produce tables listing the left-hand and right-hand components of the total circularly polarized field. In fact, more than one implementation of NEC produces such tables and associated graphics. It would be useful if these calculations were a part of the core itself, produced under additional options within the XNDA entry. Left-hand and right-hand would replace in the table the present listings for major and minor axis or the vertical and horizontal components for standard outputs. Fig. 5 shows the difference between the components of standard and circular patterns for a representative helical antenna.

NT with Impedance The present network or NT function is little used by most modelers, except for its specialized TL version. Perhaps one of the major reason for its disuse is the need for entering y-parameter (short-circuit admittance parameter) values as series and shunt components of the network. For simple networks, there are a number of calculation shortcuts that one might take, although more complex networks may require longer analytical conversion.

Modelers appear to come in two varieties: those who are comfortable with y-parameters and those who are not. Those who are not comfortable--or even familiar--with y-parameters never use the NT facility. However, the NT facility does allow the formulation within the model of a variety of networks, especially ones used at the antenna feedpoint, to simulate matching networks. As well, the requirement for admittance values limits the utility of NT itself by making it frequency specific, similarly to using series resistance and reactance values for loads rather than using inductance and capacitance values.

Therefore, it would be useful in the hypothetical NEC-5 to reformulate the NT command to allow options for using either series resistance-reactance entries or for using components within prescribed basic networks. There are presently work-arounds that allow such formulations. These formulations use small wire matrices, with appropriate network elements directly inserted on the proper wires. The geometric location of the wire matrix is remote from the antenna so as to prevent unwanted interactions between the antenna proper and the network wires. However, the antenna and the wire grid use a near-zero-electrical-length transmission line (TL) for connection, which electrically bonds the network to the former feedpoint segment. Reformulating the NT command would largely simplify this process.

Some Very Specialized Utilities

There are a number of needs that may not require core modification as much as external input or output modifications by those who develop such modules. I shall pause to list only one. In connection with the suggestion for developing within the core a means to simulate substrates, users would also benefit from being able to use something other than round wires that lie at the heart of NEC calculations. Ideally, a user should be able to specify a rectangular cross section of a non-round wire (and possible even an oval cross-section) and arrive at an appropriate entry in the wire diameter/radius portion of the wire entry lines. In theory, and in conjunction with the ability to model substrates, one might then reliably model antennas affixed to circuit board materials.

What seems simple on the surface does not always work out in detail with the same level of simplicity. By some simple experiments, we may derive round-wire equivalents to rectangular and other odd-shaped materials (such as U-channel and L-stock) when we use these material as elements in the air. However, it is not clear that a seemingly equivalent round wire adjacent to a substrate will show the same properties as a flat strip bonded to the substrate if we only use the equivalencies developed in air, that is, without the substrate.

The development of suitable conversion tables and equations requires rather extensive bench top experimentation before we can trust in them enough to go from model to prototype, the way that we presently do with round-wire models and round-wire prototypes. In fact, the entire arena of finding conductivity and permittivity values for insulating materials is still a search through many references for values that often appear as ranges rather than specific figures.

Because the implementation of the suggested conversions from non-round to round wires demands so much practical effort, it is unlikely ever to be an integral part of the NEC cores. Hence, if conveniences such as this suggested measure and others of its ilk are ever to reach practical NEC programs, the private and commercial programmers will have a major role to play.

Some of this type of work has already appeared in some commercial programs. For example, Expert MININEC has developed a broadcast industry version that welds certain calculations of high importance to licensure to the basic MININEC framework.

The list of special needs that may best be left to input/output programmers is extensive. NEC calculates using peak voltages and currents; RMS conversion is an input/output matter. SWR is a post-core-run calculation provided by the input/output programmer. Resetting output data for a specific antenna power level also makes use of the core output data and revises the report numbers accordingly. The NEC core operates with Cartesian coordinate inputs, not compass-rose inputs. As well, the native output is in terms of phi and theta patterns, both of which use opposing directions of increasing value relative to the azimuth and elevation systems used by most modelers. Few programs have addressed all of these conventions and conversions in a thorough-going manner. Even though many, if not all, of them might be addressed within the core, the likelihood of such an occurrence is so slim as to be invisible.

Conclusion

It is very likely that NEC was formulated by those in the basic sciences for others in the basic sciences. However, NEC has developed a user-base that continues to grow. Some individuals who offer training in the use of NEC still harbor a prejudice against new users unless they have been through appropriate EE or equivalent training. Hence, they tend not to offer their training at a level appropriate to those re-assigned to modeling duties from other engineering fields, field engineers and technicians, and amateur radio operators. My e-mail has clearly established that all 3 groups are important parts of the NEC user base.

Nevertheless, since the earliest days at which modeling programs became public domain, first with MININEC and later with NEC-2, programmers developed entry level software incorporating the modeling cores. It has fallen on the shoulders of the programmer--whether commercial or private--to develop both input and output modules that ease the burdens on the modeler and provide maximum information in the most usable forms. These entry-level programming efforts have been abetted by the appearance of basic courses and self-study tutorials in the use of NEC.

To a large extent, these same programmers have been in the forefront of developing aspects of the NEC cores that allow users access to some of the aspects of NEC's calculations not deemed necessary or suitable within the original cores. The problem is exacerbated to some degree in NEC-4 due to limitations in the alterations that a programmer may perform on the core when permitted under license to incorporate the core into a commercial program.

Should there ever be a NEC-5, perhaps the core development will include due acknowledgement of the entire range of users. Although we shall always need the independent programmer who creates input and output interfaces to render easy the creation of models and render usable the output data, much user-oriented modification may occur within the cores themselves. In fact, NEC-2 has seen modifications to incorporate some NEC-4 features, such as the calculation of the maximum near-field voltage or current and the application of insulated wires.

My list of suggested changes and additions is simply a partial list out of my own experience. As such, it certainly omits many possibilities that might emerge from the experiences of others whose work has taken them in different directions. If ever a NEC-5 or its equivalent should emerge, I can only hope that it will take the full spectrum of user needs into account during its development.

-30-

~ antenneX ~ May 2006 Online Issue #109 ~