Here's an excerpt from my high-performance crystal set presentation at the 1999 Antique Wireless Association convention:

Q IS GOOD: Q, the quality factor of a coil, is the ratio of the inductive reactance (i.e. "AC resistance")of a coil at a particular frequency to the resistive losses of the coil. This resistive component is the sum of the DC resistance and other losses such a dielectric losses in non-conductive material close to the coil and eddy-current losses in nearby conductors or the coil wire itself.  High Q not only yields better selectivity, but also raises the impedance of the tuned circuit, which can allow more voltage to be applied to the detector.

MAKE THE COIL LARGE: The inductance increases as the square of the coil diameter while the wire length and associated losses increase linearly.  So for a given inductance you'll use less wire and have less loss.

MAKE THE COIL "SQUARE": With all other things being equal, the Q of a coil is effected by it's aspect ratio (length to diameter). There is a sweet spot where the length approximates the diameter. Avoid very long (>2 D) or very narrow (<0.5 D) coils.

TURN SPACING: Try to space turns a minimum of 1 wire diameter. Close wound coils suffer from eddy-current losses because adjacent turns are in the strong magnetic field close to the surface of the wire.

BIG WIRE: Use the largest practical wire diameter. Litz wire has a larger effective diameter because of it's greater surface area.  However, it's difficult to find in large sizes, and the various combinations of wire size and number of strands are tailored to a particular frequency range.  For broadcast band use, the strands should be #44 or #46.

WIRE TYPE: Insulated wire makes it easy to achieve spacing between turns.  Silver-plated Teflon-insulated hookup wire is almost ideal.  (I understand that all Teflon wire is silver plated for chemical compatibility reasons.)  Bare copper is OK until it oxidizes.  Insulation or enamel help on this account. Tinned PVC-insulated wire is lossy, as tin and lead are poor conductors.  You can make usable coils with tinned wire, but probably won't achieve Q's much above 200.

PVC PIPE: This stuff work just fine for our purposes.  I recently made some coils on 6.25" green PVC using #16 silver-teflon wire.  Both had measured Q's well above 400.

TRADE OFFS: As you increase the diameter of the coil, the number of turns is reduced, so choose a wire size/spacing that will give you an, approximately, 1:1 length to diameter ratio.  Also keep in mind that the coil needs to be separated 2 to 3 diameters for other major parts of the radio.  To realize the advantages of ultra-high-Q coils, you will also need low-loss variable capacitors.  The common units with phenolic insulators are suspect, as are the polyethelene units in most transistor radios.  Look for caps with ceramic or glass insulators.  The only real solution, short of spending big money, is to stalk the hamfests.

FERRITE CORES: Ferrite lets you make physically small coils with reasonable Q values (200+) at low cost.  Torriod coils are self-shielding, so set layout is non-critical.  The various ferrite materials are compounded for specific applications and frequency ranges. Some of this stuff is actually designed to eat RF.  Junk-box ferrites are almost sure too be the wrong mix.  Order Amidon mix 61 cores from  FT-82-61’s are a good place to start.  Get some low-temperature magnet wire from the same source: No scraping! You just solder it.  Antenna rods from AM radios are also a good bet.