Since I have only been playing with crystal sets seriously for a few years,  and much of my effort has been expended in developing sets that my students could build with solid results, going for optimum results has not always been my goal; rather, I have concentrated on reliability, ease of construction and operation, and cost.  Still, I do experiment from time to time to see what can be achieved with a set that for years was relegated to the "toy" category.  Here is a collection of things I have tried recently, and the results which may apply only to my type of location, but which I offer in the way of second hand experience for you.

ANTENNA STUFF:
    a.  I have a 180 foot or so antenna up in the air with a decent cold water pipe  ground.  My daughter's set uses a 60 foot antenna grounded to the house electrical system.  When using my antenna tuner with the big antenna, I rarely have need to change the coil tap or capacitor position in the BC band and usually use only the first couple of taps ( 10 to 20 turns).  With her setup, changing the coil tap over the band is necessary, as would be expected from a short antenna.  When I tried to use the electrical ground with my antenna, I really got disappointing  results, except with the two locals at the top of the band - guess the length of wire in the house wiring made the antenna/ ground system act like a dipole.  I intend to get daughter a better ground and see how that works.  The antenna tuner is configured in an L pattern, and is connected to the set via a small coupling coil.
    b.  The L antenna tuner doesn't help much on signal strength with the long antenna, but it does cut down on the 6 MHz stations from Radio Havana and World Harvest Radio - they still get through, and I plan to try a wave trap to cut them down further.
    c.  There it was; 1200 feet of vertical tower, just waiting for me to hook up to.  Too bad the tower was solidly grounded.  Since I didn't have free reign to experiment, I am now looking at some way to tap into that hunk of metal with some type of unintrusive matching network.
    d.  Okay, this time I have a tower, up in the Blue Ridge Mountains.  Fire tower  that is unused but open to public.  Tower is only 30 feet high, but on top of a 2000 foot mountain.  This time I have 200 feet of antenna, some ground wire, and a couple of rigs, tuner, wave trap etc.  Results?  Nada.  Couldn't get a good ground connection, and didn't have time, materials, or inclination to try to rig a counterpoise, drive a metal rod into the granite,  erect a dipole... I could have cried.   Lessons learned:   take a friend, scout the site before you try it, and come loaded for bear.  Having read the exploits of several ham dxpeditions, I should have known better.  With crystal sets, you just can't take anything for granted, 'cuz you just can't turn up the gain, or  settle for a little less efficiency from a short antenna, and all the things that hams can sometimes get away with.  Maybe my next dxpedition should be to the "big city" with a box loop antenna - hi.
    e.  Back to home base.  I have been working hard to optimize my rig to get more dx stations, and have been quite successful lately.  Just a year or so ago, I got only 3 or 4 locals, and that was it; today, I get all 4 locals solid, a 50 kw station about 80 miles away solid, and routinely get 3-6 daytime 5 kw stations about 70 miles away.  At night, I get a grunch of very distant stations, some so often that I use them for "dial markers" when I am looking for new ones.  I have experimented with different coils, detectors, headphones, etc., and just recently took a Peanut special and got dx with it.  This last puzzled me, since the coil I was using one was one from an old oatbox special, and the "rig", to use the term loosely, was strung out all over the table and hooked together with alligator clips.  I was positive I had tried this long before, but without the good results.  Then, I decided to do some investigating, and so I started trying different combinations to see what didn't work.  I very quickly degraded to my situation of a year ago.  Bottom line; the antenna tuner,hooked up as shown on my antenna page, was the single most important factor in getting good signals across the BC band and in cutting out the "ghost" hf stations.  I had been using it routinely for several months, with little thought, and, once I found the best setting, seldom had reason to adjust it as I tuned across the band.  Even when I fiddled with it, I kept coming back to the same settings.  As I noted earlier, with a short antenna, more adjustment was needed.  Also, a good earth ground is essential as well.  Here I was experimenting to get the "ultimate" set, and the antenna and the tuner were doing all the heavy lifting.  My recommendation to you out there in cyberspace, after you get an ok set, and have a good antenna and ground in place, your very next project should be an antenna tuner.  I see some more elegant rigs on the web these days, a vast improvement over a short year ago, and the better ones make provision for antenna tuning.  Look on the Crystal Set Radio Club pictures page for a couple of them (you have to join to see the pictures).
    f.  I put up a 50 foot long double-wire sloping antenna, about 5 feet up to 40 feet off the ground, to see what a shorter antenna than my 200 foot random wire antenna would do.  Actually worked pretty well, but I had to work a bit harder with tuning the antenna to get all my daytime stations, and lost a few dB of sensitivity.  On summertime dx at night, it got some stations the big antenna didn't, and vice versa.  Still, it was a pretty good antenna.
    g.  On to the next step - took down the 50 footer, and put in its place a 30 foot single wire antenna - this is wire "out the window", it you will.  At this point I really had to get serious about antenna tuning.  Compared to the big antenna, I lost up to 12 dB of signal in the low part of the band, and had to retune the antenna every hundred kHz or so, particularly at the low frequency part of the band.  I also had to abandon my standard L-antenna tuner, eliminate the small antenna coupling coil,  and go with one where the antenna tuner coil itself is loose coupled to the tuning coil.  With my absolute best sound powered headphones I could just barely hear all the 11 daytime stations I can routinely get ( I call them the big 5 and the small 6).  The big 5 all were still very clear and strong enough with my Mouser earplug, but with the plug I lost a few of the small 6.  Then I cheated, and hooked up my Radio Shack amplifier, and got them all back.  Incidentally, this is a nice little amp for about 12 bucks, and has a little speaker; you can also plug phones into it, and I  found that a dollar store set of stereo phones, modified to be parallel and to a mono plug worked fine with it.  I don't plan to try a shorter antenna anytime soon, but personally feel that anything shorter is a waste of my time for my location.  If you are one of those unfortunates that has to suffer with a poor antenna setup, remember that you (1) really need to use an antenna tuner and (2)  a simple audio amp comes in handy.  I think my next antenna experiment will be with a length of pvc pipe wound with a hundred feet or so of wire.
    h.  The big antenna fell down, and, not caring to break out the fishing weights and nylon line in the heat, I replaced it with a 60 foot section flattop of two parallel wires, about 28 feet up, fed with a single 60 foot wire out the window.  Works pretty well, but it's summer, and the real test will come in the winter dx season.

DETECTORS:
    a.  Sometimes, just sometimes, a second 1N34 in parallel with the first gave me almost 2 dB  of increase in signal strength.  More was not better.  Don't know why it worked, but it even sounded slightly louder (2 dB is tough to detect aurally).  Diodes in series cut the signal strength, as expected.  Of note is that even when the second diode was known to be less sensitive, I still got the increase in signal when I put it in parallel with the first.
    b.  Made a full wave bridge of 1N34s.  Using the normal optimum tap, I got a 3 dB decrease  in signal strength.  Just in case I had unleveled the playing field by not tapping higher on the coil to match impedance better, I put it across the whole coil - got a whopping 6 dB increase in strength, but only for the two loudest locals at the top of the band, and the rest of the band disappeared with the attendant decrease in selectivity.
    c.  Okay, time to get serious - used two diodes with a center tapped coil for a full wave network again.  No improvement over a single diode with a conventional optimum tap.
    d.  Got another earplug, resistor, and diode, and put in parallel with the first, but with second diode going in opposite direction - good signals in both plugs, but signal in either was 3 dB less than with a single plug - might be worth the signal loss to be able to use both ears - try this one again sometime on the weak dx stations.  Less trouble, if using crystal earplugs, is to just put a second one in parallel with the first - you get about a 3 dB decrease in signal, but you get to use both ears.  Dave Cripes uses two plugs under shooters muffs to cut the ambient noise and gets excellent results.
    e.  Biased a 1N34 with a 1.5 volt battery through a 1 Megohm pot.  With the crystal earplug noticed a change in the tonal quality of the stations I was hearing - less bass and more treble- right before I saturated the diode.  No change with a 2k magnetic headset.  For my ear's frequency response, it helped me find a previously unheard station, but intelligibility wasn't as good as without the bias.  Going to visit this again.  Volume one of the XSS collection indicates I should get better results even with a 1N34.
    f.  Heard of some guys using a light emitting diode (LED), with bias if needed, as a detector - even used the brightness of the diode as a tuning indicator.  So I tried the red LED in my RS 130 in 1 kit, and yes it detects with bias, but not as well as a diode or even a BC or BE junction on a transistor, which I also tried.  My take is that if you can use a LED as a detector and tuning indicator, you don't need much of an antenna.  These same guys probably find the 1N34 is being driven into saturation on the locals.  See you in the next DX contest, guys.
    g.  Tried biasing the B-E junction of a 2N2222 transistor, just putting a 1 1/2 volt battery and a 1 Megohm pot across it.  Just before saturation, it had detection comparable to a 1N34.  Dave Cripe gets better results, but he uses a different transistor - I have one of his super detectors, courtesy of Larry, and sometimes it really shines, particularly on the really weak signals, for which it was designed in the first place.  His secret is that he reverse biases the base-emitter junction of a small signal transistor, and at about 6.8 volts the transistor goes into avalanche (whatever that is);  adjusting the bias to a current of about 2 uA, he claims it is about 10 dB more sensistive than a 1N34A.
    h.  Philip Miller Tate in Old Blighty graciously sent me a few OA47 gold band diodes in exchange for some 1N34As.  They all tested out as slightly more sensitive, from 1 to 2 dB better, than an assortment of 1N34As I had on hand.  Not exactly a magic bullet, but every dB counts when you're going after dx.  Incidentally, they look suspiciously (exactly) like the 1N34A.
    i.  Philip sent me some more diodes; AAZ 118 variety.  These seemed to outperform the OA47.  All in all, however, the 1N34A "type" from Radio Shack is a best buy for availability, reliability, and cost.  Tests by me and others shows that not all diodes are alike, and it definitely pays to try out several until you get one you like.
    j.  Tapping the detector down on the coil is a common design technique for best impedance matching between detector/audio and tank circuit.  Ben Tongue and Al Klase have done some nice research on getting more out of the tank circuit by using transformers to match the lower impedance of the detector and audio to the higher impedance of the tank circuit, thereby getting a higher voltage signal from the top of the tank circuit without a loss of circuit Q.  Ben also goes into great detail about selecting the proper diode to make the best use of this technique.  If you're not a techhie, just go to his pages, get the recommended transformer, and do it.  I found that it works fine with most diodes, but a pair of Hewlett Packard Schottky diodes in parallel gave me the best selectivity, particularly at night when the dx stations crowd the band.  During the day, the lowly 1N34A is a bit more sensitive, and is the preferred detector then.  The AAZ118 was somewhere in between. Also tried a 2N404 transistor, but it was much less sensitive.  BTW, my station count went to 98 in the last DX contest, so something is working.
    k.  Bill Wilson from Jacksonville, AL scored 117 stations in the last contest - Bill biases everything he uses for a detector, and claims good results even when biasing a '34.  I'll get back to you on this one.

COILS:
    a.  I just haven't been able to make a ferrite core coil give me the results I get with a tp core.  I have gotten some good dx with a modified RS crystal set coil, adding a coupling coil from the antenna instead of tapping the tuning coil, and using about 1/3 of the tuning coil for the detector tap.  Of course, I also had to use the tuner and a wave trap, but it was nice dx.
    b.  On my links page, find the G3XRT crystal set.  I made the basket coil per the specifications at that site, winding 75 turns of #26 enameled in a 2 inch diameter  coil.  I used over 2 and under 1 instead of just over/under, in order to further reduce distributed coil capacitance.  I also used hot glue to hold the coil together - worked pretty well.  I only made the one 25 turn tap for the detector, intending to use a loose coupler for antenna transfer - this worked pretty well too, since I wound the primary coil on a toilet paper core, and it slipped right into the basket coil.  The coil was sensitive, had more inductance (about 350 uH) than I was expecting, and was pretty sensitive.  Selectivity didn't seem as good as my standard selective coil wound on a tp core, and suppression of the HF intruders, if anything, was only marginally better.  Perhaps I need to look for a better tap to improve that selectivity.
      c.  Made a second basket coil using #24 enameled wire, this time using 80 turns with taps at 20, 25, and 30 turns.  Signal strength seems slightly better with this coil, and all 3 taps work pretty well.  Need to get some quantitative data on both these coils.
    d.  For my latest contest rig, I took some poster board and made some 4" diameter coil forms, then wound them with #20 solid plastic coated hookup wire.  Worked much better than last year's rig, but I sense that the experts have more up their sleeves yet.  Mike Tuggle used large Litz wire basket coils on his latest contest rig, and scored a bunch of west coast stations from Hawaii on a 50 foot antenna - that's over 2500 miles!  My best distance was just over a 1000.  I know his location, over water, helped a lot, but still...

LOOP ANTENNAS:  Just about given up on this concept in my location - not a peep.  Put this one in the the same category with "crystal set in a headset" and "pocket crystal radio".  Where I live, you need to take this antenna business somewhat seriously.  If you want to convince me otherwise, you are welcome to bring your pet set for a visit and a demo.  Personally, I have a friend with a 700 foot dock over the marsh that might accommodate a beverage antenna...I finally got a regenerative receiver working, and may try this with a loop antenna - it works pretty well with just 6 feet or so of antenna as it is, but works better with regular antenna/tuner/traps etc.

MISCELLANEOUS:
    Scott Balderston put me onto substituting an audio frequency choke for the standard 47k resistor across the terminals of a crystal earplug.  Using the primary (high impedance) side of a matching transformer I use with my low impedance (60 ohms dc resistance on the phones), I got a marked improvement in signal level.   Also tried a power supply transformer - no luck, a Radio shack 1000:8 ohm audio transformer - worked, but lost some of my bass, a 10 mH radio frequency choke - nada, and an audio choke that came with a Radio shack am/fm kit - really lost my bass, but got louder.  If you like the earplugs, and you should because they are really tough to beat for the price, sensitivity-wise, you should look into this audio choke business.  If you are just a casual listener, the 47k resistor works fine, and is a very cost effective way to eliminate earplug distortion (10 cents).  I got my choke from Scott, and it's an oldie.  I think he has found a supply of newer ones, and might be willing to help if you aren't having luck finding one.  He's on my links page under Scott's Crystal Radios.
    In the same line, I tried a 100k pot in place of the 47 k resistor.   Results were inconclusive.  Going to try this again with a meter on the set to see if I can find a spot which works best.  Incidentally, putting a second earplug in parallel with the first gives you binaural capability with little if any loss of sensitivity - don't need to use a second resistor either.

    Tried emulating the effect of a double-tuned rig using two sets, capacitively coupled.  Too many knobs to turn.  Guess I'll have to bite the bullet and use my  "getting harder to find" smooth operating double 365 air variable.  Heck, I just wanted to see if the double tuned rig would make a significant difference up where my local bandmasters dominate the band starting from 1270 kHz all the way to the top.  The rig I am describing is similar to the Miller 565/575, with both tuned circuits acting as a passband tuner.  I have had some success with both triple and double tuning to get above the 1490 kHz barrier, but it takes skill and patience to do so.  Think I will try to assemble a set which only tunes the top 400 Khz of the band and see if I can spread it out a bit.  also need to get some higher Q coils while I'm at it.  Anybody want to donate some Litz wire to the "cause"?