I used a circuit simulator to model a halfwave balun. T1 is a transmission line whose length is an electrical halfwave.
Here are the model results for a halfwave of 75-ohm coax. The Y-axis is mismatch loss in dB and the X-axis is frequency in MHz.
I decided to build a balun to validate the model. I used 93-ohm coax because it should have only 0.24 dB loss at the band edges and I just happen to have some around. In fact, I located an 8.5" piece, which is what I calculate the length should be for 585 MHz, the arithmetic center of the 470-700 MHz band.
I cut 1/4" of vinyl off the ends of the cable, mashed the shields together, and soldered them. I stripped 1/8" of dielectric off the center conductors and soldered a half-watt resistor that measured (before heating it!) within an ohm of 300 across the tips. The tips touch the resistor body, with 3/8" of excess resistor lead length hanging in space (I may want to use this selected part for another project). I soldered a few inches of 75-ohm RG-6 to the balun using the shortest leads I could manage. I plugged the male F-connector on other end into my power splitter.
Minimum return loss for the balun occurred at 555 MHz so I centered my 200-MHz spectrum analyzer passband there. The power splitter return loss (open-circuit reference only--too hard to short the female F-connector on the power splitter) was 32 dB at 455 MHz, 33 dB at 555 MHz, and 29 dB at 655 MHz. The balun return loss was 15, 19, and 12 dB for the same frequencies. This is equivalent to SWRs of 1.43, 1.25, and 1.67, and mismatch losses of 0.14, 0.06, and 0.28 dB. Although a bit lopsided because I didn't quite center the response, this isn't all that far from the 0.17 dB that the circuit model predicts 200 MHz away from bandcenter for 93-ohm cable. I don't have an independent way of assessing what the 300-ohm load looks like over this frequency range.
I think this is a nifty balun for UHF-TV, particularly if you can get your hands on a few inches of 93-ohm coax (RG-62).