Here's a tid-bit you won't find everywhere: the Orion S-meter calibration curve.
#define TT565_STR_CAL { 15, {The first column is the Orion's response to the "?S" command in computer units. The second column is indicated decibels (dB) as read on the actual S-meter.
{ 10, -45 }, /* S 1.5 min meter indication */
{ 13, -42 },
{ 18, -36 },
{ 22, -30 },
{ 27, -24 },
{ 30, -18 },
{ 34, -12 },
{ 38, -6 },
{ 43, 0 }, /* S9 */
{ 48, 10 },
{ 55, 20 },
{ 62, 30 },
{ 70, 40 },
{ 78, 48 }, /* severe dsp quantization error */
{ 101, 65 }, /* at high end of scale */
} }
(Literary event: I just defined "S meter" in the Wikipedia! See my article.)
The Orion has a couple of peculiarities when measuring power. The minimum reading (on my unit) is about S 1.5 on all bands when looking at a dummy load. What does that mean? If S9 = 50 µV, then S1.5 is 6 dB x 7.5 S-units or 46 dB down in power = 2.5 x 10-5. The voltage ratio is the square root of power, or about 5 x 10-3. So the equivalent RF voltage level is 50 x 5 x 10-3 = 0.25 µV. The Orion's sensitivity spec (main receiver) is "< 0.18 µV" says that the receiver's internal noise will be about this level, which is consistent with the meter reading. In other words, Ten-tec seems to have this right.
(On the other hand, the tried and true RST system says that S1 means "Faint signals barely perceptible". That's an argument for using 5 dB per S-unit, as some have defined it, at least with our current generation of receivers.)
The other feature of note on the S-meter is the behavior at very high signal levels. As you can see from the calibration table above, the raw S-meter units are spread out roughly 8 units per 10 dB. However, the meter reading is highly quantized. You only see values of S9+48 and S9+65, nothing in between. This indicates that the software computation of received power has a quantization problem. Of course, it's hard to complain too much, because such power levels are almost never seen in real life. (If they are, better switch in some attenuation!)