degreedays.net
, over the last 36M (1121d) the number of days below a baseline of 5°C, ie at which the above device might hope to generate something of use, is 287 or about 95 days per year, and provides good (not complete) coverage over the Nov/Dec/Jan winter PV low. See images including of the tiny in-line DC-to-DC boost converter.
2014/06/29: I finally gave the unit a test today, with the bottom end under the running kitchen hot tap (~50°C) and the LED that Thermalforce.de attached lit up after a short delay: a mere 20kWh+ of hot water input for maybe 100mW of light!
(The unit had arrived on about the 19th, well packaged between substantial wood planks.)
The output at the (white) LED was at about 2.5V, presumably clamped by the LED; the input upstream of the in-line barrel-shaped converter was ~0.45V.
This may well get put into the loft until closer to winter when I may need some help to install it as it requires quite a deep narrow hole.
As a side-note, it looks like I may be replacing the SheevaPlug with a 'tuned' Raspberry Pi (Model B) which should run (or idle!) happily on well under 2W, including the battery-sensing and external control I/O, which means that I might only need 20 of these units to power it mid-winter!
The TEG was installed this morning as I expect an air temperature below 0°C overnight which may be just low enough to allow the TEG to function.
2015/01/16: the TEG was installed physically this morning, and this evening I connected up the DC-converter and LED (in a transparent freezer bag to keep water out while keeping the LED visible) and with an air temperature of ~5°C the LED did not light (as expected).
2015/01/17: 8am, ~1°C, and the LED is still off!
2015/01/23: 8am, ~-2°C, LED still off, about 330mV available across it.
2015/02/02: 8am, ~-2°C, LED possibly very faintly on.
2015/02/03: 8am, ~1°C, snow on top of TEG slushy suggesting heat pipe OK, LED just possibly very faintly on.
London is not cold enough in winter for this to work, as the suppliers warned me would be the case!