Experience with series capacitors and hub dynamos?

gilesa

Newly Enlightened
Joined
Dec 13, 2019
Messages
12
I have been looking at building a headlamp circuit for a hub dynamo that gives increasing output with speed.
A series-resonant circuit that peaks near the expected top speed looks a good way to do that, but capacitor
choice seems not straightforward. Electrolytics are compact and have adequate ripple current ratings, except
that all the data sheets I have looked at have a steep derating of maximum ripple with falling frequency.
The tables typically have a low value of 50 or 60Hz, so it is hard to know if such circuits will over-drive
the capacitors.

On the other hand, this article suggests such ratings are very conservative at expected temeratures:
https://www.newark.com/pdfs/techarticles/kemet/Ripple-Current-Capabilities-Technical-Update.pdf

Does anyone here have practical long-term experience?
Reports of loss of brightness, measured loss of capacity, or even popped components would be
interesting, but so would reports that everything is fine. And how aggressive is your circuit?
Capacitor bias method (if any) may also be relevant.
 

gilesa

Newly Enlightened
Joined
Dec 13, 2019
Messages
12
Thanks, but that probably does not apply to the question I am interested in.
The site says: "While there's not a big difference at very low and very high speed, the boost in the middle is real nice to have."
The recommendation for a single LED, as I will be using, is 1650uF series capacitance. With my measured values, that puts
the resonance at 13.5Hz, or 9.5kph. I want to put the resonance around 25kph, so about 250uF.

That means much reduced ripple current rating and much higher current, and I am realy interested in any experience with
that sort of circuit.
 

jdp298

Newly Enlightened
Joined
Mar 22, 2011
Messages
115
Location
With your sister, somewhere you wouldn't like
I will admit, even in this niche of the internet, it's not a specific application I've seen much (any) activity on.

There's a chap on here who uses voltages and frequency waveforms to switch in another pair of LEDs at speed = x, but the analogue version with a single LED, while theoretically possible, hasn't been popular. I too have a home-brew cricuit, but it's not with this aim in mind. Sorry we've not been more help old bean.
 

gilesa

Newly Enlightened
Joined
Dec 13, 2019
Messages
12
The reason for preferring a single LED is that I am converting an old headlamp and there is only room to mount a single emitter.
Multiple LEDs in one package like Cree XHP3.5 are available, but beyond my soldering skills. I already wrecked a couple of
single LED SMD devices trying to mount them. There is no room for the ready-mounted star board modules.

I have done some more searching and found that the Forumslader USB charger devices use series capacitors as low as 110uF
(2 X 220uF). Pictures show what look like polymer or hybrid SMD electolytics, which typically have steep derating with falling
frequency. There are no circuit diagrams for the current version, but older ones have no capacitor biasing.

My German is not great, but a search for "Forumslader kondensator ausfall" (capacitor failure) found no reports of trouble with these devices.
I also found a Vishay datasheet that includes a chart for ripple current vs. temperature vs. lifetime that shows ripple current can be tripled
while maintaining lifetime, so long as temperature remains below 40C. The worst quoted lifetimes are 1000 hours, which is a lot of night cycling.

Given that, my guess is that there is not much to worry about, but I will be using capacitors with the best ripple and derating figures I have seen:
Panasonic FM series.
 

znomit

Enlightened
Joined
Aug 1, 2007
Messages
979
Location
New Zealand
The reason for preferring a single LED is that I am converting an old headlamp and there is only room to mount a single emitter.
Multiple LEDs in one package like Cree XHP3.5 are available, but beyond my soldering skills. I already wrecked a couple of
single LED SMD devices trying to mount them. There is no room for the ready-mounted star board modules.

What are you retrofitting into? You'll need to properly mount it with a thermal path to the outside or you'll fry the LED.

You can get 6V or 12V Cree easywhite XLMs which are ideal for retrofitting to many cheap battery lights to convert to Dynamo.
https://www.cree.com/led-components/products/xlamp-leds-arrays/xlamp-xm-l-easywhite
 

gilesa

Newly Enlightened
Joined
Dec 13, 2019
Messages
12
Nice parts, except that SMD and "Electrically neutral thermal path" means the central pad and two 5x0.5mm strips need to be soldered.
Not for me thanks!

The headlamp is a chrome-plated Sturmey-Archer with nickel-plated metal mirror, made about 1952.
I made an aluminium bush to fit, threaded to take an M8 brass screw that will carry the emitter.
That allows for focus adjustment and should be a decent heat sink.
 
Last edited:

formdeformed

Newly Enlightened
Joined
Dec 25, 2017
Messages
7
A little late to the conversation, but hopefully not by too much.
I've been using ceramic capacitors. Specifically, these. Murata 22uF 50v
They're more expensive than electrolytics and will likely take up a bit more space, but I think they fit the bill pretty nicely.

It seems that space will be a limiting factor for you. I'd just build the circuit in a different enclosure, separate from the lamp. But I'm a rookie at electronics stuff and have only recently progressed from the dead bug assembly method to using through-hole breadboards.
You can fit a fair amount on an Adafruit board in an Altoids tin.

As for setting the capacitance with a single LED. Consider that at 250uF you'll see well over 1 amp through that one LED at your target speed, which is fine, expect it's the same current through the generator.
I measured the internal resistance of a SON 28 at 3.6 ohm and an SP PD-8 at 3.0 ohm. So over 3 Watts are being dissipated in the hub winding in that cruising range.
I think a better method is to use a larger capacitor for a lower resonant frequency. Say 600-700uF. Then use a step down converter for faster speeds: higher voltage at the rectifier limits current in the hub and increases it through the LED.
Same power from your legs, but more light and less waste.

I borrowed the methodology that a few others have brought up on here, mtbr, and other places. I use a 555 timer wired as a Schmidt trigger with the AC signal, and use its output to trigger a 556 timer wired as two monostables to act as a missing pulse detector.
It's a pretty reliable way to automatically switch things up at a certain speed and fairly easy to build in some hysteresis. I'm sure the frequency to voltage chip that others use is a good option as well. It's just that I had the 555 chips and only have enough freetime to learn so many new tricks.
Use that to change the duty cycle on the step down regulator. As a bonus, use it to switch the caps for a different resonant frequency.
 

gilesa

Newly Enlightened
Joined
Dec 13, 2019
Messages
12
Thanks, I think there are two good suggestion there, but perhaps slightly in conflict as adding a buck converter implies higher rectified voltage, less current, so less capacitor stress.

I had considered ceramics, but the cost/benefit seems poor. From the Murata data sheet, the ESR is slightly worse than electrolytics. But the heat tolerance is higher and the greater surface area of multiple units should certainly fix any heat problem. After seeing the Forumslader, I am inclined to think electrolytics will be OK.

Driving a single LED through my hub does have problems. The static resistance of the GH6 is 6ohms, but I get 9ohms from a probably shaky dynamic measurement. So at 1amp there will be a severe mismatch with very poor electrical efficiency. I started using SPICE to look at circuit designs and as expected an active replacement for the buck freewheel diode makes a significant difference. Early results suggest that a very simple buck controller may be good enough.
 
Top