The sterba antenna is sometimes misspelled sturba curtain. The sterba curtain is in the same antenna family as Bruce Arrays. 

You can read about distributed feed curtains in Jasik's "Antenna Engineering Handbook". 

Please note: There is a large difference between distributed feed or branched feed curtain arrays like USIA arrays and Lazy H antennas and narrow or single band curtains like Sterba, Bobtail, and Bruce arrays. The Bobtail isn't even a true curtain array, it's just a simple three-element vertical broadside-array with a unique feed system that produces binomial current distribution.

Many people incorrectly call USIA or distribution / branch-fed curtains "Sterba curtains". Factually there are HUGE differences in performance and construction of broadcast curtain arrays and Sterba arrays. 

Because the Bobtail isn't a true curtain (and doesn't have the potential for significant gain), I won't include them in this discussion. We are actually installing a large USIA or Distributed Feed Curtain here at W8JI. You can see the new tower that will eventually support one end of the curtain at this link.

The planned curtain at W8JI will be aimed at Europe, it will have a reflector, and will cover both 80 and 40 meters. The upper element will be 300 feet high! This antenna will have about 18dBi gain on 80 meters, and almost 23dBi gain on 40 meters.

Sterba Curtains and Bruce Arrays 

Sterba Curtains are modest-gain single-band antennas. They provide a very limited gain-bandwidth product and are critical to construct.

Let's look at a Sterba array so we can see why Sterbas have narrow bandwidth and limited gain:

Let's walk through the current distribution of Ant 1 above: 

1. The feedline connects to the middle of a lower 1/2 wl "dipole" section. As with any dipole, high voltage appears at the ends. 
2. Vertical sections D are 1/2 wl transmission lines. One terminal is excited by the voltage at the end of the lower dipole.
3. Section A top is excited by the high voltage on the wire of transmission line D that connects to the dipole. Maximum current in section A is at the bent end where it transitions to vertical sections C.
4. Vertical Section C has highest current (maximum radiation) at the bends. The current gradually transitions to a high voltage at the large black "dot" in the middle of section C. 
5. The bottom of vertical section C again has maximum current at the bend to lower horizontal section A. 
6. The inner area of lower section A has high voltage, that excites the second conductor of vertical transmission line D.
7. The upper end of the second conductor of vertical transmission line D voltage feeds the outer ends of the upper middle 1/2 wl dipole.

This poor method of excitation produces three very undesirable circumstances: 

• The overall path through conductors that supplies current to the center upper current maxima is through 2-1/2 wavelengths of wire on each side! On very high frequencies the physical length and the series resistance of that wire length might not be significant. On lower frequencies the long physical length means appreciable series resistance is added to the current path. On 20 meters, for example, the upper center half-wave is excited through 160 feet of wire, the entire current path being over 300 feet long! On 80 meters it would be over 650 feet of conductor length to the current maximum, with a total current loop distance of 1/4 mile!

• The phase of current in the upper 1/2 wl section depends very heavily on the accuracy of wire length in terms of wavelength. There are 2-1/2  360-degree long segments, or 900 degrees conductor distance, in series with the feed to the upper current maxima. Even if we ignore other effects and consider the phase solely dependent on conductor length, an error of 5% in electrical dimensions will result in a phase error of 45 degrees! That means if we could somehow build a perfect small Sterba that is the equivalent of two half waves stacked 1/2 wl over two more half-waves (~5dB maximum gain) on 7.2 MHz, the system would start to lose gain with a move to the CW band of 40 meters! We are fortunate when a Sterba covers a single band. (A six-section Sterba I used was only a few dB over a dipole at the design frequency, and fell equal to the dipole at the extremes of the band.)

• Collinear and stacking distances are limited by the necessity of maintaining 1/2 wl long transmission lines. Gain is further limited by the requirement of zero distance end-to-end element spacings. This means the antenna sacrifices potential gain by using spacings less than optimum.

The amateur radio fascination with Sterba and Bruce arrays probably stems from confusion. I've noticed most amateurs incorrectly call large distributed-feed or branched-feed curtains used by short wave stations "Sterba" curtains, but they are definitely not!! As a matter of fact I can't recall ever seeing a Sterba Curtain used in any commercial SWBC array.  

USIA or Distributed/Branched feed-system Curtains

Distributed feed curtains use a series of common points, each fed from low loss transmission lines, to distribute power. Conductor loss is less, phase error is significantly reduced, and all elements receive equal currents. 

This places conductor resistances in parallel, and makes array patterns stable over very wide frequency excursions. In addition to having more gain, a distributed feed curtain (such as USIA arrays used at VOA sites) can be used over a 2:1 or broader frequency range with minimal gain and pattern change. It is very easy to make a distributed feed curtain operate on 80 and 40 meters with full gain and no pattern distortion. 

A distributed or branched feed curtain also allows designers to use optimum element spacing, both in collinear and broadside (stacking) distances. This means a 4-element branch fed curtain can provide the highest gain per acre of any antenna design.

VOA's USIA Curtains, despite occupying a tiny fraction of the space required for their Rhombics, are the highest gain arrays at VOA's International Broadcast sites.

Note: The Lazy H, when center fed, is a distribution or branched-feed curtain.  

The antennas below are similar to a planned 80/40 meter antenna. The basic change will be I'll use single wire elements instead of wire cages and operate open wire lines with standing waves. This simplification is acceptable for amateur use because voltages and currents are much lower with 1500 watts CW or SSB rather than 100 kW or more carrier power (400 kW PEP on AM) used by VOA. 

My Curtain will have the upper element at 300 feet, and the lowest element at 100 feet. It will be three layers high and three sections wide. My variation models just over 20 dBi on 80 and over 23 dBi on 40 meters, with ability to steer the beam in azimuth and elevation. 

I've had similar scaled down distributed feed curtains for 20 meters and up in the past, and they worked quite well. Unfortunately I don't like 20 meters and up, so I only used them as a brief experiment lasting a few months.

By the way, curtains like this have considerably more gain than Rhombic antennas. The curtain would also occupy a tiny fraction of the space required by a Rhombic producing significantly less gain. For example, my planned 80/40 curtain is only 450 feet wide and 300 feet high yet has over 16 dB gain over a dipole on 40 meters.  A 1500 watt transmitter into the curtain would produce the equivalent of 60 kW to an optimum height  dipole.

Of course this all hinges on getting new 340-foot Rohn 55G and 360-foot Rohn 65G towers installed, something that unfortunately seems to keep getting placed on the back burner. 

Here are a few pages my planned system is derived from: 

Basic curtain with reflector. In amateur applications it is possible to use a system like this over a 2.6:1 frequency range! The system can also be scaled down to use fewer elements, and the reflector can be omitted.