More than a decade ago I built a battery-backup unit for my dynamo bike lights. The idea is that when you stop at traffic lights or wherever, or as you struggle very slowly up a steep hill, a battery cuts in to replace the dynamo. At other times the battery is trickle-charged by the dynamo.
My unit had some unusual features; most interestingly, it measured the frequency of the dynamo output to determine when to switch over to the battery. On the other hand it suffered some problems: I was never very happy with the mechanical aspects, especially how to mount the unit on the bike, and the charging circuit was troublesome.
I eventually gave up on the thing, for various reasons. First I got a much more reliable dynamo: one of the problems that had driven me to investigate battery backup was that my dynamo would slip when cycling in the wet. After a disasterous experiment with a bottom-bracket dynamo, I eventually fitted a Nordlicht bottle dynamo with a rubber ring, which works superbly. Then LED lights came along; with an LED rear light as well as the dynamo, I can stop at traffic lights and am still conspicuous from behind. And batteries for LED light seem to last almost forever!
I've put together this page because Chris Juden from the CTC published a letter asking for circuits for battery backup systems. So here's my contribution. I hope it's useful.
The obvious approach is to monitor the voltage provided by the dynamo, and to switch over to the battery when it falls below a certain threshold. Of course the problem with this is that because the dynamo has significant internal resistance, as soon as its load is removed its output voltage will immediately rise, causing a naive circuit to reconnect it. The standard solution to this is to add hysteresis to the circuit so that a higher threshold must be passed before re-enabling the dynamo. But I thought up what I think is a cleaner solution.
My solution relies on the fact that the so-called dynamo is actually a generator with an AC output. As the bike slows, the frequency of the output drops, and this can be used to switch over to the battery. Unlike voltage monitoring, frequency monitoring doesn't need any hyteresis.
My frequency-monitoring circuit is described on its own page.
How to switch between the two supplies? I spent some time trying to avoid using a relay, assuming that it would be big. But I don't know enough about triacs to design a solid-state circuit to do the job. Do triacs impose a diode drop that would be unacceptable with our meagre 6v supply? But in the end I was delighted to find a truely tiny relay in the Maplin catalog that does the job splendidly; it is an incredible 10x10x7mm.
I chose to use a 6v sealed lead-acid battery of the sort that you often see in burglar alarms and applications like that. The main reason for this was that lead-acid batteries have more straightforward charging characteristics than NiCads. At the time, people talked about the memory effect - the idea that NiCads needed to be thoroughly discharged before charging again in order to maintain optimum capacity. In the intervening ten years this has become less of an issue, either because the batteries have improved or because the memory effect was a myth. I think that today I'd use five AA or AAA NiCads.
You might think that charging a 6v battery from a 6v supply is impossible. But no: the dynamo output is 6v AC RMS, which means about 8.5v peak. You then need a rectifier diode, which will impose a drop of at least 0.7v. This should be enough to charge the battery, but only just. I never got it to work properly with my lead-acid battery.
Chris Juden suggests that cells can be reconfigured from a series to a parallel organisation for charging - a clever lateral idea, but it needs extra relay contacts: I reckon you can just about get away with a DPDT relay. With five cells you'll need to break them into groups of two and three, with different series resistors and hence different charging rates.
I'd strongly advise fitting a fuse in series with the battery to guard against accidental short circuits. It is all too easy to accidentally short out your battery, and rechargeables can have very high short-circuit currents. Your unit could easily catch fire!