If you like to fiddle around with your crystal sets, then you need some way to do comparisons of before and after when you want to try something new.  If course,  the obvious is to listen.  If it sounds louder, it is.  If you don't hear the other station that is close in frequency any more, then selectivity is better.  If you hear everything you want, then your set is working fine, and you need go no further.  If you are trying to push it a bit, however, some reasonably objective yet quantitative measurements are helpful for you to optimize your set and to compare the performance of different sets, antennas, modifications, etc.  Here are a couple of things I do with my rigs:

    Get a DC voltmeter  which will measure small voltages in the millivolt range.  I use either a digital multimeter or a RS analog multimeter with about 20 thousand ohms per volt sensitivity (less sensitive meters will  load down the circuit unacceptably, and will probably not measure down in the millivolt range, but go ahead and try if you have to).  The RS analog model I have has a 0.6 VDC (600 millivolts) scale which works pretty well.  I then connect the meter across the 47 kohm resistor which is also across the crystal earphone. Using this setup, my closest local BC station, less than a kw about 6 miles away, gives me a reading of about 190 millivolts.  A station with a 50 kw daytime signal about 75 miles distant comes in at about 30 mV.  With the analog meter, there is a little insertion loss, about 10 mV, but not enough to seriously load down the circuit.  The digital multimeter has a virtually imperceptible effect on the signal.  I tried once putting a  galvanometer with  a 50 uA movement in line between the detector and the earphone but the reading, which was only about 4 uA, as expected, wasn't enough to make this useful, being at the very bottom of the scale.  Anyway, for my money, the high impedance voltmeter makes a much more useful tuning indicator. If you are using magnetic headphones, which don't have the 47k resistor across them, you will not get much of an indication on the meter.  Just put a 47k or some other reasonably high value resistor in series with the headphones and measure the voltage across it - yeah, the resistor cuts your signal a tad, but you're only putting it in long enough to sharpen up the rig's tuning anyway.  Incidentally, there is nothing magic about the size of the resistor; I just happen to have several on hand.  If you use a larger resistor, you will get a higher reading on the voltmeter, but will also have a corresponding smaller signal in your magnetic headphones.  Use the smallest resistor that gives you a usable reading on the meter - this is not an absolute signal level you are measuring, but rather a tuning aid.  Just to see what I could do, I tried out a cheap multimeter with a 1000 ohms/volt meter sensitivity.  If I tuned in a reasonably strong station, and just put the meter in place of the 47k resistor, I could indicate enough level on the 2.5V scale, in the neighborhood of 200 mV or  so, to make some use of the meter.  This is pretty low on the scale, but still enough to see the effects of tuning the rig.

   Okay, so now you have a tuning indicator/signal level indicator or whatever you want to call it.  What can you do with it?  First, and obviously, you can use it to tune a station for the best signal.  On a strong station, you will be amazed at how much the signal strength can vary and it not make a heck of a lot of difference in earphone volume.  This is because the ear is a marvelous signal processor itself, and can accommodate a range of loudness, or signal power to the ear of 90 decibels, that's ten to the ninth power, otherwise known as a billion to one between the softest whisper it can detect and the point at which the ear starts to suffer damage with chronic exposure.  Multiply that loudest sound by another thousand, you are now at a trillion to one, and the sound is so loud that it can cause physical pain.  In order to handle this range of signal power, the response of the ear is logarithmic, and the power of a sound has to change by a factor of two  for you to readily detect it.  Since power varies as the square of voltage, this means you usually need to see a forty percent change in voltage, or about a 3 dB change in power  to notice  the difference in loudness.  The voltmeter, being more responsive than the ear to small changes, can help you fine tune to a station, and then compare the difference in signal strengths from different stations using a number rather than an impression.  Once you have played around with this a bit, it's on to phase two:

   Now that you have a signal strength meter, you can start using it to optimize the settings on your rig.  Using it on a receiver similar to Radio number four, which you can find on my project page, I found that the optimum tap for sensitivity and selectivity was at about 35 percent of the coil from the ground end.  The 20 turn tap for the detector (for sharp tuning) showed a 9 dB loss of signal strength compared to the best tap, while giving less than a 2 dB gain in selectivity.  As expected, using an antenna tuner helped, particularly in the low end of the BC band, where the antenna tuner coil alone improved the signal from a 135 foot antenna by as much as 12 dB.   When I put it on my Radio Shack crystal set, after first adding the obligatory 47 kohm resistor across the earphone, I found the sensitivity for it and my selective crystal set to be within a few millivolts of each other.  The list of what you can do from this point is long enough to keep you busy for hours on end.  On to phase 3:

    For this part, you need an rf signal with some audio modulation.  I happened to find an ancient but usable rf signal generator in my school lab which fit the bill.  It has to be variable in frequency, and you need some way to check its output frequency, because we are going to try to measure the selectivity of your set.   I put the output of the generator across the antenna coil of the selective crystal set, and adjusted the signal to a comfortable level on the voltmeter described above, about 50 mV.  Then I adjusted the signal frequency up and down by 25 kHz and 50 kHz, and measured the signal output across the earphone.  From a center frequency of 1100 kHz, the signal strength dropped  by 12 dB at 25 kHz off, and by 20 dB at 50 kHz off.  Running the same test on the Radio Shack set, the change was 2 dB and 5 dB, respectively.  I don't know if my set is really any good since I haven't done much comparison testing, but compared to it, the RS kit selectivity sucks.  For both rigs, selectivity improved some at lower frequencies, and was a bit worse at the top of the band.  At this point, some electronic engineer type is probably reading this and shouting obscenities at his monitor for my sloppy, unsophisticated, and jerry built  methods.  Well, this isn't Nature magazine and I'm not trying to sell cold fusion.  It's a hobby, and if I can have this much fun with a handful of cheap parts and some 20 year old, bottom of the line test equipment,  working on a table set up in the laundry room, so be it.    I have tried to use my "super convertible" as an rf signal generator in the regeneration mode, and it works pretty well.  Just put a 2k resistor in place of the headphones, and you can used it as a nice, unmodulated signal.   Yeah, it drifts some, but as long as you don't have an antenna hooked up to it, it's pretty stable.  Shouldn't be to hard to couple a one transistor audio oscillator to it for some modulation.  Transtronics sells their RF Signal Generator SK-303 kit for about 28 bucks.  Maybe I should give it a try.

    The multimeter connected across your earplug resistor also is most useful in comparing different detectors.  What you are usually seeking is the best sensitivity, so first find a weak but stable station (or detune your set to make one), which gives you about 10 mV or so across the resistor (I use this value with the 47kohm resistor).  Then you can readily compare different detectors by swapping them in and out of the circuit.  With a digital voltmeter with readouts to the nearest 0.1 millivolts, go for a reading of about 2 mV with the station very weak but still clear.  This goes faster if you use a couple of alligator clip leads for the connection.  In a 10-pack of 1N34A diodes, I have found the voltage developed across the diode to vary as much as 30%;  at this level, that equates to about a 2 to 3 dB difference in signal strength, even though the audible difference is small.   For stronger signals, the difference in detector sensitivity is less, but doesn't matter.

    Use the same setup to check out the effectiveness of your wave traps when you're cutting how the high powered stations.

    Another useful piece of test equipment, also excellent for chasing dx stations, is a receiver with a digital frequency readout.  My Radio Shack set, a pocket portable, fits the bill and costs about 50 bucks or so, depending on how the sales are running.    It tunes in 10 kHz increments on the AM band, and I am confident the actual frequency is accurate to within a hundred Hz or less.  You can use this to check the output of your sig gen, and to check its frequency; your less expensive and older signal generators have analog tuning and readout, and their calibration can be suspect.  It also can be used to calibrate the dial of your crystal set in conjunction with the signal generator to at least the accuracy of a standard analog pocket superhet.  You might consider this last to be gilding the lily, but having a known "dial marker " or two can keep you oriented when you're searching the dial.

    A meter capable of inductance and capacitance measurements is nice to have, and I use a friend's whenever I make a new coil or want to check an unknown capacitor.  Haven't seen fit to spring for one myself however.  If I need to, I can always check the inductance of a new coil using a known capacitance with it in a tank circuit and the signal generator.  Q meters?  They're out of my league, so I depend on the kindness of strangers to tell me how to optimize coil Q.

DECIBELS FOR DUMMIES (how to sound like a tekkie)

    The ear is a marvelous device which can process a wide range of noise levels.  At a frequency of 1000 Hz, the weakest sound detectable is at a power of about 1 picowatt per square meter, normal speech is about at 1 microwatt, hearing damage starts at about 1 milliwatt (smoky bar with a loud band), and physical pain is felt at 1 watt.  Since the ear doesn't respond to absolute power changes in a linear manner, and to cut this trillion to one ratio down to size, we use a logarithmic scale instead.  The common logarithm of a number is the exponent or the power to which 10 must be raised in order to obtain the given number.  E.g. the logarithm of 100 (10 to the second power) is 2, of 1000 is 3, and so on.  Power ratios can be expressed as a logarithm, known as a Bel (after Alexander Graham), but more commonly as a deciBel, obtained by multiplying the logarithm of the power ratio by 10.  This gives us a range of hearing loudness from 0 to 120 decibels, relative to the threshhold of hearing (1 picowatt per square meter), a much easier, and more useful set of numbers to bandy about.  When working with your crystal set, you normally want to know how much better (or worse) a change to it is, and can express it in deciBels (dB).  Say that you are taking voltage measurements across a resistor as I do.  If the voltage goes from 20 to 40 milliVolts, the ratio of voltages is 2; but wait.  Power varies as the square of voltage, so the power ratio is 2 squared, or 4.  The logarithm of 4 is 0.6, and 10 times that is 6, or 6 dB.  Incidentally, if the measured voltage had gone down from 20 to 10, you would have gone down by 6 dB - it works in both directions.  So, all you have to do is divide the larger voltage (or current) by the smaller, square the result (this is your power ratio), and multiply the logarithm of that number by 10 to get the change in dB.  Round your answer to the nearest whole number.
    If this seems like too much work, here is a short table of common power ratios, the corresponding dB change, and what it means to your ear:

Power ratio    dB  Effect
        2                 3      barely detectable change
        4                 6       noticeable
         5               7       a little better than 6
       10            10    oh yeah
       20            13     a smidgen better than 10
    100             20    oh yeah again -
     200            23   (just to show you the effect of multiplying two numbers by adding their logs)
note:  to the ear, an increase of 10 dB, followed by an increase to 100 dB above the original power level (another 10 dB increase) would be sensed as two equal step increases in loudness.  Similarly, two 3 dB steps would be sensed as "equal" changes in loudness.