Canadian TV, Computing and Home Theatre Forums banner
1 - 9 of 9 Posts

182 Posts
calculating net gain

I have 4nec2 installed and I've been playing around with it and a parameterized SBGH model. I'd like to throw it at the genetic optimizer, but I'm wary of gain/SWR versus net gain. (I'm assuming the gain values aren't net gain.) What's the translation? Or should I just assume that as long as the SWR is close to the un-optimized version that everything is okay?

182 Posts
My SBGH 4nec2 model with surface patch split reflector. This is optimized for GTA ch14-67 reception, although I haven't built it yet. (Need to get the coax in place before I replace my indoor test version.)
CM Single Bay Grey-Hovermann with split reflector
CM 4nec2 model created & optimized by Eric Ball
SY lambda = 0.37758	'794 MHz
SY end=0.135	'end wire length
SY main=0.150, m2 = main / sqr(2)	'main wire length
SY nend = int ( end / lambda * 40 + 0.5 )	'number of segments in end wires
SY nmain = int ( main / lambda * 40 + 0.5 )	'number of segments in main wires
SY gap = 1in, g2 = gap/2	'feedpoint/reflector gap 1"
SY rwidth=34.5in, rw2 = rwidth / 2	'total reflector width 34.5"
SY rheight=48in, rh2 = rheight / 2	'total reflector height 48"
SY nrw = int ( (rw2-g2) / lambda * 5 ) + 1	'number of patches in reflector half width
SY nrh = int ( rheight / lambda * 5 ) + 1	'number of patches in reflector half height
SY rdepth=4in	'reflector offset
SY rgap=0.5in, rg2 = rgap/2	'reflector gap
GW	1	nend	0	end+m2+g2	3*m2	0	m2+g2	3*m2	0.003175
GW	2	nmain	0	m2+g2	3*m2	0	g2	2*m2	0.003175
GW	3	nmain	0	g2	2*m2	0	m2+g2	m2	0.003175
GW	4	nmain	0	m2+g2	m2	0	g2	0	0.003175
GW	5	nmain	0	g2	0	0	m2+g2	-m2	0.003175
GW	6	nmain	0	m2+g2	-m2	0	g2	-2*m2	0.003175
GW	7	nmain	0	g2	-2*m2	0	m2+g2	-3*m2	0.003175
GW	8	nend	0	m2+g2	-3*m2	0	end+m2+g2	-3*m2	0.003175
GW	9	nend	0	-end-m2-g2	-3*m2	0	-m2-g2	-3*m2	0.003175
GW	10	nmain	0	-m2-g2	-3*m2	0	-g2	-2*m2	0.003175
GW	11	nmain	0	-g2	-2*m2	0	-m2-g2	-m2	0.003175
GW	12	nmain	0	-m2-g2	-m2	0	-g2	0	0.003175
GW	13	nmain	0	-g2	0	0	-m2-g2	m2	0.003175
GW	14	nmain	0	-m2-g2	m2	0	-g2	2*m2	0.003175
GW	15	nmain	0	-g2	2*m2	0	-m2-g2	3*m2	0.003175
GW	16	nend	0	-m2-g2	3*m2	0	-end-m2-g2	3*m2	0.003175
GW	100	3	0	-g2	0	0	g2	0	0.003175
SM	nrw	nrh	-rdepth	-rw2	-rh2	-rdepth	-rg2	-rh2
SC	0	0	-rdepth	-rg2	rh2
SM	nrw	nrh	-rdepth	rg2	-rh2	-rdepth	rw2	-rh2
SC	0	0	-rdepth	rw2	rh2
GE	0
LD	5	1	0	0	58000000
LD	5	2	0	0	58000000
LD	5	3	0	0	58000000
LD	5	4	0	0	58000000
LD	5	5	0	0	58000000
LD	5	6	0	0	58000000
LD	5	7	0	0	58000000
LD	5	8	0	0	58000000
LD	5	9	0	0	58000000
LD	5	10	0	0	58000000
LD	5	11	0	0	58000000
LD	5	12	0	0	58000000
LD	5	13	0	0	58000000
LD	5	14	0	0	58000000
LD	5	15	0	0	58000000
LD	5	16	0	0	58000000
GN	-1
EX	0	100	2	0	1	0		
FR	0	55	0	0	464	6
You can do a lot in the SY tags, including calculating number of segments.

182 Posts
4nec2 optimization

In my experience 4nec2's optimization some limitations.

First you need to understand how 4nec2 calculates it's figure of merit from the properties (e.g. gain, SWR). If you specify more than one property, it is important to scale the properties, i.e. an improvement of ? ohms is the same as 1dB. If you use the frequency sweep option, the comparison is done on the average figure of merit across the whole band. (Which doesn't always produce a flat response.) Unfortunately, the properties don't map directly to reception (net) gain.

Second you may need to tweak your variables. For example, take an SBGH with reflecting rods. The driven element has three variables: length of the horizontal wires, length of each of the angled wires, and the distance between the two sides. The reflecting rods add in more variables: distance between the drive element and the reflectors, gap between reflector halves, length of each reflector rods, and vertical placement of each reflector rod.

Whew! The problem comes in when you try to optimize multiple quasi-dependent variables simultaneously. For example, if you are optimizing the driven element, you probably want the vertical placement of the reflector rods to remain relatively the same. So for that you want your variable to be a percentage of the length of the angled wires instead of an absolute distance. (In the same way you should use a variable to calculate the number of segements per wire instead of using a fixed value.)

Finally, the optimizer is based on the assumption the figure of merit is a smooth function with only one maxima. This is kind of like a landscape with only one hill - as long as you always increase in altitude, you will eventually reach the top of the hill. Unfortunately, antenna response curves are much more complex. You may be on top of one hill, but that doesn't mean there isn't a bigger hill somewhere else. (Or, if you overshoot the top of the hill you're on that you won't land on a different hill.)

This is not to say 4nec2's Optimizer isn't usefull. Assuming you can create a reasonable figure of merit, it will happily tweak your design to the micrometer. But if you just feed in your design and click "Optimize" without a little foresight there's a chance you will end up with an even worse antenna.

182 Posts
I am still a little confused about the "Source/Load" concept(s).
Source is a power / frequency source. If you are using one, you are modelling the antenna as a transmitter. However, due to "reciprocity" most of the behaviour of an antenna is equivalent for reception too. (4nec2 doesn't handle modelling reception, so you'd need to use NEC2 directly and do the pre & post-processing yourself.) It's best to stick to one source in your model.

Loads come in multiple types. Spot loads are like soldering a physical resistor, inductor or capacitor to your antenna (LD 0/1). You can also specify the impedance (LD 2/3) or conductance (LD 5) of the materials used to build your antenna.

182 Posts
What specifically would you recommend for a source for a reception only uhf antenna ?
For reception, you need to use EX 1, 2 or 3 cards for the incident wave and a PT card to print out the resulting currents. Unfortunately, I haven't been able to find a lot of info on this kind of modelling, so don't take my examples as gospel.
EK			' extended thin wire kernel
GN -1			' no ground - free space
FR 0 55 0 0 470 6	' 470-794MHz (UHF ch 14-68)
EX 1 1 1 0 90 0 90 0 0	' horizontal (Y) polarized wave from +X
PQ -1			' don't print currents
PT 1 999 2 2		' print current for segment 2 in wire 999
XQ			' execute
EN			' end
This is preceded by the usual geometry and loading cards. Segment 2 in wire 999 is where the source would normally go. Instead I use a LD 4 card to put in a 300ohm resistance. (I'm still not happy with this, but I haven't found a definitive example of the right way to do it.)
This results in the following in the output file (once per frequency step):
                                 - - - RECEIVING PATTERN PARAMETERS - - -
                                           ETA=  90.00 DEGREES
                                           TYPE -LINEAR
                                           AXIAL RATIO= 0.000

           THETA      PHI          -  CURRENT  -         SEG
           (DEG)     (DEG)       MAGNITUDE    PHASE      NO.

            90.00      0.00     1.6998E-03    117.45      246
comparative gain (in dB) = 10*log10(magnitude)
The reason I call it a comparative gain is you use it to compare the results of different runs. If you want the dBd value, you need to model a resonant dipole and calculate the difference. Note: this is actual net gain, no need to compensate for SWR/impedance because you are measuring the current through the 300ohm "source" resistance.

182 Posts
I believe most uhf/vhf antenna comparisons use dBd = dBi minus 2.1 .
That's correct, but the magnitude value isn't dBi either. It's the current generated by the planar wave. So unless you know what current would be generated in a isotropic receiver or a resonant dipole (with the appropriate "source" resistance), then you can't calculate dBd.

However, the comparative gain is still useful. i.e. you can determine that the response at channel 53 is 9dB less than channel 25, you can assume that you're going to have problems picking up CITY-DT. Or if you tweak your SBGH model and you lose 1dB on ch25, while gaining 8dB on ch53, that's a good thing if you're in the GTA.

182 Posts
At this point in time, I am operating in the "Brut Force" mode. I have not explored the Symbols (SY) card yet. I barely have the discipline to stay away from it.
Although the geometry editor is okay to use, you really should use the NEC editor and start using symbols so 4nec2 does all of the calculations. Just be aware that if you go back and change the file using the geometry editor that you will lose all of the symbols.
I think you and ericball were saying the data presented in 4nec2's "Forward Gain" chart is really "Raw Gain" data. It is not necessarily the effective gain delivered by the antenna. There are additional factors (SWR, etc, etc) that could heavily influence what a very nearby tuner sees. I think you are saying, "Net Gain" vs Frequency plots are the proper modeling method to compare different antenna geometries.
4nec2 models transmission, not reception. And although you can use "reciprocity" theories to get from one to the other, there are differences. When transmitting, any energy reflected by impedance mismatches goes back to the transmitter where it is reflected back to the antenna. This is what SWR measures. But in receiving antenna the energy is reflected back through the antenna and radiated.

See for my take on 4nec2 optimization.

182 Posts
Modelling is your first step. (Can you imagine the old days when you had to actually build each idea and take it out to the range?) However, I suspect it will not perform very well. The GH works because acts like a combined antenna and phased feedline. In your design the signal in the outer elements has to flow away from the feedpoint, against the phase of the signal. (Of course, I'd love to be proven wrong. And these days you don't need a Cray to model an antenna.)
1 - 9 of 9 Posts