Just a few comments, based on the many interesting replies. There are too many to make it practical to comment on each individually, so here is a summary.
1) Yes I had considered using Droitwich @ 198 kHz, and might well do that, although according to Wikipedia, it use a rubidium frequency reference, which makes it pretty useless. I don't know if that information is correct though - I would have expected a cesium source.
2) I had not considered using DCF77, but I will do so.
3) I measured the impedance of my loop (1.5 x 1.0 m, 19 turns) using an HP 4284A precision LCR meter. At 20 Hz, it has a magnitude of 0.7 Ohms at a phase angle of 17 degrees (0.65 +j 0.17). This is pretty close to what I would expect based on the DC resistance of the wire. But at 100 kHz, it is very different (14.2 + j 1000 Ohms), which is 1.6 mH in series with 14.2 Ohms. (All these readings were read from the meter, with no averaging, as I changed the display format, so don't expect them to convert exactly.)
The loop has a self-resonate frequency somewhere between 300 and 400 kHz, but I don't know exactly where as my LCR meter only has discrete frequencies it works at - it is not fully tuneable.
4) The common base amplifier was a circuit posted by someone on the time-nuts mailing list, who wrote
"I use a home-made untuned loop antenna with 4 windings of 2.5 mm2 insulated wire on a 80 x 80 cm wooden frame, and with a grounded base pre-amplifier mounted on the antenna frame. A schematic is enclosed for you to copy. The pre-amplifier is powered through the cable, and loads the FS700 input as required. I live about 290 km from the island of Sylt, and get nice noise margin figures from the FS700, normally about 40 dB, often up to 46 dB."
The circuit is attached
If someone has something that works, it is tempting to follow it. I think the use of common based is because of the low input impedance of the amp (estimated to be 3 Ohms), although my loop impedance seems far greater than 3 Ohms.
I changed the transistor for a MAT12, as the original is impossible to obtain from reliable sources. There are plenty on eBay from China, but I suspect they are all fakes.
5) The bandwidth of Loran-C is quite wide, and therefore I am tempted to not tune the loop. Again I go back to the comment from someone who is getting very good S/N ratios with an untuned loop and the common base amplifier.
6) I'd not considered that voltages of 10's of volts could be inducted by lightning. For that reason I will probably put some high-speed back-to-back diodes, to keep the voltages in sensible limits.
7) Someone else mentioned the use of Litz wire, and given how far the real part of the input impedance of the loop (around 14 Ohms) is different from the DC resistance (estimated about 0.7 Ohms), clearly the skin effect is having more effect than I had given it credit for.
8) Yes, I had considered a whip and FET amplifier. In fact the FS700 was supplied with an active whip antenna, but I don't know what the amplifier was.
9) I believe a loop will give better S/N than a ferrite rod antenna, and expect this will lead to better stability.
10) I have a GPS disciplined crystal oscillator (HP 58503A), which should be more accurate short-term than the long-wave systems, but I want to compare them. Long term they should all be the same.
Dave