I hate to recommend the 9:1 balun with end fed random length wire ....
That is one of my favorite general-purpose wideband HF antennas, although for the best results one should be aware of the following:
1) Either a 4:1 or 9:1 ratio works well, but the transformer should be wired as an UN-UN, and not a Balun. I wonder if the Chameleon is wired as an un-un?
2) Failure to choke or isolate the feedline! 6 scramble-wound turns of coax doesn't cut it. Heavily choke it with a ferrite-based isolator at the feedpoint, or just after the Un-Un, or if purposely incorporating the common mode of the feedline as another-radial, choke it about a quarter wave away. Do not take into account the velocity factor as we are dealing with common mode, not transmission line mode for the dual-purpose radial. Lug the best low-loss feedline that you can especially when operating higher than 20/30 meters.
3) If operating in a purposely non-resonant fashion, where the tuner may be close to the antenna (less than about 50 feet in a portable scenario), try to be on the high-side of resonance (antenna system resonant lower than the bands you want to operate on) on all the rest of the bands. This means that your tuner will be using mostly capacitance, which is more efficient than a lot of inductance. This can be especially true when using a rigs internal tuner, or other automatic tuner that relies on small toroid-wound inductors rather than low-loss coils. In some cases, it may make sense to use a hard-core external tuner with low-loss inductance, even when using qrp.
4) Failure to provide a decent ground system, either on ground, or elevated. This makes choking the feedline even more important than ever, unless you are purposely incorporating it some length away from the feedline heavily choked there.
5) Check / adjust the system for natural resonance *without* the un-un transformer, and then place the un-un back inline. See #3. I use an analyzer to make quick work of this task.
6) If you run with an inverted L, strive for the least amount of multiple current cancelling nodes in the vertical portion if vertical radiation is desired. That means calculating / visualizing the half-waves of current counting from the END of the wire backwards to the feedpoint. Ideally one would want the major part of the current node in the middle of the vertical section. This changes according to frequency of course.
By doing this - even though it sounds like a lot to keep track of, satisfies my need to be a little bit better than random.
