I had originally designed the circuit so a switch could select between flashing and solid modes, but it turned out that the Zener diode that needs to shunt the dynamo current during the off-cycle of the flash wasn't up to the task. I'm quite convinced of the value of flashing lights in urban settings, but in this case I've settled for solid lights while moving and a flashing standlight. The circuit has had a reasonable amount of bench testing on my dynamo testing jig, but it's now time to bundle it all into the lamp housing, put it on the bike and start riding with it. Here are a few pictures of the board with switch and copper heat sink mounted LED (described elsewhere):
The orignal selector switch on the top of the housing proved unreliable, so I enlarged an existing hole in the bottom of the housing to hold the switch, which has a nice rubber boot. Mounting the PCB was a little fiddly. A tiny L-bracket with one #4-40 tapped hole is used to mount the circuit board to the lamp housing:
It all fits together nicely, with the retrofitted LED/heat sink/optics and switch looking fairly inconspicuous:
The taillight LED is ready to go, although I need to get the brown plastic taillight chromed (a gift from a generous reader!). I intend to save my really nice NOS SA lampset for some other project. Once the taillight is chromed (technically, vacuum metallized), I just need to hook up the taillight and dynamo wires and mount the lamps on the bike.
Future improvements could include a taillight flashing system, where the headlight is never disconnected from the dynamo, alleviating the problem of surging dynamo voltage when flashing both LEDs. I implemented this on a breadboard using a low-side N-FET to short the taillight LED, under the control of an Attiny10. This also has the advantage of smaller and fewer components than a 555-based flasher.
Another improvement I'd like to eventually make is to upgrade the optics. The headlight uses a Cree XM-L, which can take up to 3A of current. Unfortunately, small format narrow beam optics aren't yet available for the XM-L, but they are for its cousin the XP-G, whose 1A current maximum can easily handle the dynamo's output. The XM-L optic certainly improves the beam shape, but it still doesn't have as much throw as I would like. The best solution would be to fabricate a new heat sink for the XP-G and uses a narrower optic.