Lightning Interface

I purchased the LSU-2002 from Stormwise (they have a pretty wacky web site). The LSU-2002 is no longer available, but the patent that describes it (4,823,115) is online, and various people on the weather mailing list have considered building one. They now sell the LA-101, but is rather more expensive ($50) and I have no idea whether it would work in the same way.

The simplest form of interfacing is just to wire it (via an optocoupler) into a DS2423 and count the pulses. However, this does not (apparently) adequately distinguish between manmade interference and natural lightning. The difference can be discerned by looking at the duration of the pulses.

According to patent 5,977,762, the detection is of natural lightning if either the number of strokes is high, or the total duration is large. Accordingly, I present the circuit below. Note that I don't think that this circuit violates the patent as the patent requires that counting the number of strokes and their duration is performed by starting a timer when the first stroke is seen. The standard way of running 1-wire devices is to poll them at regular intervals. Hence there is no timer that is started by the first stroke.

The Circuit

The actual circuit is just as DS2423 which uses one channel to count strokes, and the other to count the total number of milliseconds that the pulse is active. A 1kHz waveform is generated by a 555 (which is held in Reset by the absence of a lightning pulse).

The power for the board either comes from a hub, or directly via a jack and regulator.

There is a fair amount of protection in the form of TVS and diodes on the inputs. This is all to try and prevent things from blowing up badly!

The circuit is available in a PDF file.

I have prototyped the circuit using my experimenter's board plus a bit of breadboard. The original experimenter's board only counted the number of strokes, and I felt that it was a bit of a waste of the second DS2423 channel!

Actually I now realize that there is a potentially serious problem - very short pulses will not get counted and will be eliminated by the debounce circuitry within the DS2423. In particular, a pulse shorter than the debounce time (400usecs) will be ignored. This is a side effect of making the pulse 'active high'. The reason for that was to directly connect it to the reset line of the 555. Maybe I can just claim this as a feature!

The PCB

The PCB is small (1.3 inches by 2 inches). I'll post the top and bottom views sometime when I'm happy with them.

The Software

None as yet. Currently I'm doing all my testing with TMEX! I don't really have any good ideas on how to reduce the data. I feel that the following things would be interesting:

Total number of strokes in a 5 minute period
Total number of (say) 2 second periods with strokes in a 5 minute period. This (maybe) gives a better feel for the number of flashes.
Peak number of strokes in a 2 second period.
Peak number of milliseconds of strokes in a 2 second period
Use the millisecond counter to reject very short strokes
Dunno what else.

I feel that it may become more obvious once lightning season returns. Actually, here in New England, we get lightning storms in the winter (during snowstorms), so I may be able to do real testing.

Back to the main sensors page.

Comments to Philip Gladstone