...
It is a circuit for flashing LEDs using a bicycle dynamo hub as a power source.
...

Wouldn't it be better not to shunt the power to ground?

Bicycle generators make drag when cycling, so the best option would be to make the whole circuit go open circuit when the LED is not on, then maybe use an energy efficienct system (like a buck SMPS) to drive the LED when it is on.

 

I had considered suggesting that kind of approach, but had second thoughts because of the output characteristics of a typical bicycle "dynamo" This is usually an alternator with a large leakage inductance to restrain the rise in output current at high speeds. As a result, at higher speeds a quasi constant-current output is obtained, and the rise of voltage on open-circuiting it may be quite large.

A linear series regulator is probably best avoided, as the series drop might become inconveniently big if the load current is less than the available input. I agree that a buck switcher looks like the best option in principle, but there are some snags which the student would have to deal with.

The input of a regulated buck converter has a negative incremental resistance, and fed from this type of dynamo its input voltage could become unstable. Care would be necessary to avoid difficulty with starting from low speeds (lock-down at high duty cycle) and the converter could have to cope with a relatively large input voltage range. There could be oscillation at the input unless the control characteristic was carefully managed.

It must be possible to make a system like this work, but I wonder if it is appropriate to the level of this project?

 

Wouldn't it be better not to shunt the power to ground?

Bicycle generators make drag when cycling, so the best option would be to make the whole circuit go open circuit when the LED is not on, then maybe use an energy efficienct system (like a buck SMPS) to drive the LED when it is on.

It would be better not to have to lose the energy during the off cycle as heat, but I can't conceive of a way to do it other than adding an additional LED, which complicates the physical design.

I've actually prototyped a buck converter that works - I posted the details in the Projects forum. The point of it is to extract more current from the hub at higher speeds, which it does reasonable well.

 

I had considered suggesting that kind of approach, but had second thoughts because of the output characteristics of a typical bicycle "dynamo" This is usually an alternator with a large leakage inductance to restrain the rise in output current at high speeds. As a result, at higher speeds a quasi constant-current output is obtained, and the rise of voltage on open-circuiting it may be quite large.

A linear series regulator is probably best avoided, as the series drop might become inconveniently big if the load current is less than the available input. I agree that a buck switcher looks like the best option in principle, but there are some snags which the student would have to deal with.

The input of a regulated buck converter has a negative incremental resistance, and fed from this type of dynamo its input voltage could become unstable. Care would be necessary to avoid difficulty with starting from low speeds (lock-down at high duty cycle) and the converter could have to cope with a relatively large input voltage range. There could be oscillation at the input unless the control characteristic was carefully managed.

It must be possible to make a system like this work, but I wonder if it is appropriate to the level of this project?

So you don't suggest your large reservoir capacitor solution?

I'm quite determined to get it working, even if the off cycle energy is lost as heat. That's a compromise I can live with with as the drag would appear the same as if the lights were on all the time. Perhaps the best solution is your power transistor solution - shunting the current through a package more capable of dissipating the heat.

It would be better, of course, to save that energy and use it during the on pulse, but as everyone has suggested here, it may be overly complicated. Ideally, I don't want to have to regulate the current going into the LEDs, as the dynamo's saturation current is below the maximum current of the LEDs.

The other thing I've considered is something TRIAC-based that would disconnect the AC dynamo input before the rectifier. Not sure if this is a viable solution yet - need to read more about it.

 

This is a separate argument to trying a larger reservoir capacitor, but note my earlier comment that trying to store 0.25 seconds' worth of charge may require an impracticably large capacitor value.

I do not know whether a rear red light is customary or even permitted on bicycles in you region, but if so would it be worth considering as a means of utilising some of the surplus current?

 

What you could do is use a large 63v cap, and maybe a 60v zener across it, just in case it pumps up to that voltage (but it probably won't from a bicycle generator).

Then you can use a very simple 2-transistor buck regulator like this;
http://www.romanblack.com/smps/smps.htm

which unlike most IC type buck regulators is tolerant of a wide input voltage range and has no "max input voltage" provided you use a transistor suitable for the voltage and tweak some resistor values. The result would be the minimum drag, and a stable enough +5v (or whatever) DC output.

 

It would be interesting to look at the starting characteristic of this circuit when fed by a current-limited supply. I suspect that the starting lock-down issue still requires attention.

In a formal set-up with a PWM controller, a maximum limit on the duty cycle is among the measures used to help starting with a current limited input. Other methods may include imposing rules dependent on the input voltage.

I am not saying that this won't work - I would be interested to see how it goes, but having had some experience with current-fed systems in submerged equipment I'm aware that there are pitfalls. (A similar situation exists with solar batteries, addressed by MPPT controllers.)

 

This is a separate argument to trying a larger reservoir capacitor, but note my earlier comment that trying to store 0.25 seconds' worth of charge may require an impracticably large capacitor value.

50-100 mF capacitors seem to be available from Digikey for $5 or less in the voltage range I need. They are supercapacitors rather than standard electrolytic types. I wonder if I could trouble you to comment on my description of how a large reservoir capacitor would act to absorb the dynamo energy during the off cycle. Knowing whether my description is way off the mark or not would go a long way to improving my understanding. Here it is from page 2 of this thread:

Let's say I go with a conventional diode rectifier and then put a large reservoir capacitor on the output - 75mF as you suggest. With the LEDs connected the dynamo voltage would be Vf of the LEDs (about 5V in my case), so the capacitor would charge to 5V, right? Then, during the off cycle the capacitor starts charging above 5V until it is clamped by the Zener at 6.8V. The size of the capacitor (or the timing of the pulse) is calculated so that it doesn't have time to charge to the clamping voltage of 6.8V, thus preventing the Zener from actually shunting any (or very little) current. Then during the next on pulse, the LEDs are on, dropping the capacitor voltage to 5V. That extra reservoir of energy in the capacitor is realized as extra current for the LEDs? Then during the next off cycle the capacitor is back at 5V and has 'room' to absorb more energy to prevent the Zener from clamping?

I do not know whether a rear red light is customary or even permitted on bicycles in you region, but if so would it be worth considering as a means of utilising some of the surplus current?

My circuit is driving 2 LEDs in series, one for the headlight and one for the taillight. I'm starting to consider some amber side flashers just to deal with this dynamo voltage spike during the off cycle, but it is doesn't really work well with the actual lamps I have in mind.

Thank you all for you comments and suggestions. It is a big help!

 

It would be interesting to look at the starting characteristic of this circuit when fed by a current-limited supply. I suspect that the starting lock-down issue still requires attention.
...

The simple 2-tran buck regulator I linked to will usually run in linear mode at very low currents BUT will still regulate the output voltage correctly.

As any significant current is drawn or any current pulses (like from a flasher) it will start oscillating very well due to the large amount of positive feedback.

It is a really forgiving SMPS for a system like this dynamo + flasher where lots of factors (Vin, Iout) will change over time.

 

This is probably best settled by actually trying it. A simulation might give a rough idea, but it would be necessary to get a reasonably good model for the dynamo to get a meaningful result. Simulations of oscillating devices can also be long to run, and starting is not always well modelled.

I probably have a bit of a "thing" about lock-down, but as they say, once bitten... I think though that many would agree that it can be very awkward when something just will not come up at the appropriate time. The issue I refer to is not so much reluctance to start oscillating, but depression of the input voltage, possibly followed by an abrupt start if the input current gets big enough to give full output with the switch stuck full on.

With a high-impedance supply, this may happen at a higher input level than the minimum which could give full output with a narrower duty cycle. This effect can delay or even prevent proper starting - not a good idea for a dynamo which will stop and start a lot.

This might be helped by moderating the load, perhaps reducing the load current in response to low input voltages.