dynamo light with single LED

[minisystem]

Just pondering...

Most of the designs I've seen posted for dynamo LED lights take the rectified, smoothed output of the dynamo and dump it into two or more LEDs. Current regulation is handled by the fact that the hub saturates at below the maximum current the LED can handle (ie. there is no active current limiting circuitry). With 2 3V white LEDs in series your combined Vf is 6V, right? That just happens to be the rated voltage fo the hub, which we all know is really dependent on the load.

So, let's say I want to make a light that just uses a single white LED with a Vf of 3.0V or so. Based on my understanding of the way the hub saturates and how its voltage output is load dependent, I can't see how a single power LED could come to any harm. Am I missing something?

[Steve K]

didn't someone here do this recently? I recall something about retrofitting a single white led into an old headlight being used with a S-A dyno-hub.

The only harm done is that you aren't getting much light from the dynamo. I keep thinking about throwing together a little buck converter that is set to 50% duty cycle. It ought to knock down the incoming voltage in half, which in principle (he said while crossing his fingers...) ought to draw 6v @ 0.5A from the dynamo and provide 3v @ 1A at the output of the buck converter.

Steve K.

[minisystem]

ah yes. good point re. the buck converter. I was ultimately disappointed with my buck converter for two LEDs in series with a Vf of 5V (white & red), but I only tested it with a Sanyo H27, which is a lower power hub dynamo. I should really go back and try it again with a Shimano or Schmidt.

I've been thinking about this again actually and was trying to come up with some kind of schematic for a buck powered front light that would be able to bleed a limited current to a rear LED, like 100-200mA as that is plenty for rear illumination. I expect this is how the Supernova system works.

I think one of the limitations of my buck converter was that I needed more than 5V in order to convert higher voltage into higher current. With a single 3V white LED, that 2V difference might translate into better low speed performance.

[Savvas]

I keep thinking about throwing together a little buck converter that is set to 50% duty cycle. It ought to knock down the incoming voltage in half, which in principle (he said while crossing his fingers...) ought to draw 6v @ 0.5A from the dynamo and provide 3v @ 1A at the output of the buck converter. Steve K.

Steve, this is something you should do. There aren't enough really bright single led dynamo light designs in this world, especially not ones that also have enough output for a rear red led as well... The world is waiting...

Savvas

[Steve K]

Steve, this is something you should do. There aren't enough really bright single led dynamo light designs in this world, especially not ones that also have enough output for a rear red led as well... The world is waiting...

Savvas

Well, there are a lot of other people that could do this too.... not exactly rocket science. Get a clock source (a LM555? or something faster?), a gate driver, a suitable mosfet, inductor, flyback diode, etc., and that's pretty much all you need. Well, you also need some spare time to work on it, which is where I am limited.

My bigger goal would be to get a little microcontroller to generate the buck converter pwm signal based on dynamo speed. This would get you the max power out of the dynamo at any given speed. For someone who has been writing code for microcontrollers, this shouldn't be a huge task. You'd probably want to throw in a standlight function too.

Steve K.

[jdp298]

What? How complex? Let's answer the question more simply.

This stuff:



In here:



Looks like this when you've done it twice:



And if you have to run a wire to the back of the bike, just do this instead of trying to make 1 unified circuit



2 front lights, each with a 1W 90 lumen LED. Plenty for any unlit road above 7 mph.

Each light, rectifier, 6.3v zener to protect the 1 or 2 supercaps (which are rated at 5.5v but the documents say safe a little over this) and then a 5V LDO with something like a 6ohm resistor then the LED. Does mean when you stop it get dimmer, but then it lasts 5 minutes. Simple, small, flickers bright enough at 4mph and the zeners mean I've touched 40 mph without a problem. What's not to like?

[minisystem]

My bigger goal would be to get a little microcontroller to generate the buck converter pwm signal based on dynamo speed. This would get you the max power out of the dynamo at any given speed. For someone who has been writing code for microcontrollers, this shouldn't be a huge task. You'd probably want to throw in a standlight function too.Steve K.

As posted previously, I tried this with mixed success. Using two LEDs in series at 50% duty cycle only resulted in a worthwhile increase in current at relatively high speeds, >25 kph. It wasn't clear to me that active load matching with pwm provided much benefit other than dropping from 100% to 50% at some threshold speed. I suspect this is because I'm not getting anywhere near optimal load matching even at 50%?

I was also discouraged by swhs's reports of his mysterious microcontroller-based dynamo driver, which claims remarkable power output for triple XM-Ls at low speeds. Then I stumbled across this paper: Electronic Power Management for Bicycles, which actively engages in load matching, both by switching a capacitor from series to parallel with the load at some threshold speed as well as by duty cycle. The prototype is bulky and complex but has some nifty bits like using a buck converter and negative-to-positive inverting converter sychronized to the AC output of the dynamo to eliminate the need for a bridge rectifier. Unfortunately, details are scarce.

Despite my discouragement, I thought I'd revisit my microcontroller-based buck converter with just a single Cree XM-L, this time using a Schmidt SON hub rather than the Sanyo H27 I'd used previously. The low speed results are definitely better:





Meaningful amounts of extra current are available at speeds just below 20kph and roaring down a hill at >35 kph gets you well over 1A. Efficiency on the messy breadboard is about 78% at 30% duty cycle by my calculations.

As nifty as this is, I was hoping for more current at lower speeds. In theory, the hub should be putting out its rated power at 12 mph, right? So with 100% efficiency, a 3V LED should be getting 1A at 12 mph/20 kph at 50% duty cycle. I get about 700 mA at 30% duty cycle. I guess I could chalk this up to both low efficiency and a lack of a load matching capacitor?

Anyway, I guess it works well enough for single LED designs, which I like/need for my retro upgrades. I will happily accept advice for improvements though!!!

[Steve K]

I didn't realize/remember that you had gotten this far on the project. Congrats!

If I understand the results correctly, it does show that adjusting the impedance seen by the dynamo does allow the circuit to draw more than 3 watts at higher speeds.

As your blog suggests, there is still room for improvement in your circuit. The input cap is certainly a place to start looking. I'm not sure what would be best, since part of the point of the buck converter is to make the input impedance look like a higher value. A larger cap would actually look like a lower impedance... but a matched load for the dynamo is a cap, so maybe a larger cap is better??? Might be educational to try something significantly larger, such as 100uF.
It would also be interesting to see a scope photo of the cap voltage. It shouldn't change very much while the buck converter is running, and the low switching frequency of your converter isn't helping. Thinking back to some of the circuits I see at work, the 100kHz switchers usually have quite a bit more than 2uF at the input.

As far as getting more power from the dynamo at low speeds, I think you're limited by the ability of the dynamo to generate enough voltage to forward bias two LEDs in series. IMO, the ideal would be to have this sort of circuit drive a single LED that can handle a few amps of current.

Other comment: obviously, a proto-board isn't ideal for switching converters. The best would be to build it on a circuit board with a proper ground plane and wide, low inductance traces. This will be essential to being able to run with higher switching frequencies.

Any plans to optimize the design, taking it from an Arduino development kit, putting the code into a small uC and building a small circuit that can fit into a modest headlight?

regards,
Steve K.

[minisystem]

I didn't realize/remember that you had gotten this far on the project. Congrats!

and I thought you were just ignoring me!

If I understand the results correctly, it does show that adjusting the impedance seen by the dynamo does allow the circuit to draw more than 3 watts at higher speeds. As your blog suggests, there is still room for improvement in your circuit. The input cap is certainly a place to start looking. I'm not sure what would be best, since part of the point of the buck converter is to make the input impedance look like a higher value. A larger cap would actually look like a lower impedance... but a matched load for the dynamo is a cap, so maybe a larger cap is better??? Might be educational to try something significantly larger, such as 100uF.

I seem to remember back in the old thread we did some math and came up with 2.2 µF. Or maybe I happened across that value from some switching converter tutorial where I inputed frequency, duty cycle, voltage, etc and some design guide calculation spat out 2.2µF. In any case, it worked but was woefully inadequate at dealing with the transients produced at lower duty cycles (dynamo voltage spikes during the off phase of the switching cycle). I bumped it up to 1000µF (what I happened to have in the parts box that had the right voltage rating) and that smoothed out the transients and allowed me to run it at duty cycles below 50% without frying MOSFETs, but it did not seem to have a significant impact on the power output/performance/efficiency. The paper I linked to suggests a 300µF input capacitor for optimization at 15 km/h, which can be switched between a parallel and series configuration depending on speed. This kind of sounds like using a tuning capacitor a la Martin's low-speed boost circuits. I intend to pick up a few different values in the 100-1000µF range and see if that helps. Right track?

It would also be interesting to see a scope photo of the cap voltage. It shouldn't change very much while the buck converter is running, and the low switching frequency of your converter isn't helping. Thinking back to some of the circuits I see at work, the 100kHz switchers usually have quite a bit more than 2uF at the input.

The voltage is nice and smooth, even at 30% duty cycle it remains constant, whereas with 2.2µF there were transients that were clipped by the input zener. I'm thinking of bumping the switching frequency up to around 100kHz. 32Khz was chosen as it was the easiest (and maybe highest frequency?) PWM to get out of the arduino. I think other atmel µCs support higher PWM frequencies.

As far as getting more power from the dynamo at low speeds, I think you're limited by the ability of the dynamo to generate enough voltage to forward bias two LEDs in series. IMO, the ideal would be to have this sort of circuit drive a single LED that can handle a few amps of current.

Yes, I'm much happier with the single LED performance, but I was hoping to see close to 1A at 12mph/20kph. Hopefully a bit of tweaking will get me close.

Other comment: obviously, a proto-board isn't ideal for switching converters. The best would be to build it on a circuit board with a proper ground plane and wide, low inductance traces. This will be essential to being able to run with higher switching frequencies.

Yep. Good point. Before I move up to 100 Khz I should maybe consider designing an optimized PCB for prototyping rather than dealing with the breadboard. Kind of a pain, but getting a run of boards is so cheap now I think it would be worthwhile.

Any plans to optimize the design, taking it from an Arduino development kit, putting the code into a small uC and building a small circuit that can fit into a modest headlight?

Yes, I'd like to. I'm getting reasonably adept at programming the Atmel family of µCs, both with Arduino and a bit in assembler. Coding isn't the problem, it's my lack of electrical engineering experience that's the barrier to optimization. That's where you fine folks come in!

[Bandgap]

Hi folks.

Glad to see you have done it MiniSystem. Nice work.
I admire your dedication - The bits I got to try something similar are still in a draw :-/

What was the switching frequency in your design?

For reasons to do with feel rather than calculation, I was going for a high frequency to keep the input capacitor as small as possible.

And to use a CLC (pi) filter to allow it to be even smaller.

I realise there was no rectifier in the way, but I got loads of power out of a bottle dynamo (7.2W) using series resoanant capacitors. And I wonder if an empty capacitor on the far side of a bridge can cause a high-amplitude half cycle to appear? - need to think about this some more.

The-Other-Steve

[Steve K]

hi gang,

regarding the whole scheme of adjusting the duty cycle of a buck regulator as a means of maximizing the power from a hub dynamo... has anyone tried running simulations as a way to evaluate some of the variables such as input capacitance?

One reason I ask is that it's a bit of a nuisance to build hardware to run some of these tests (especially if you tend to blow up mosfets in the process!). Another reason is that I recently downloaded a version of Spice that is available for free from Linear Technology:
http://www.linear.com/
The upper right of the web site has links to the download for the software (LT Spice) and a couple of pdf documents to introduce you to Spice and a more complete document that provides the details of what Spice can do and how to use the various models that are available.

I used to use Spice in the early 90's for simulating switching regulators and other circuits. The computers back then just didn't have enough processing power and a simulation might take 8 hours to run. There was also a bit of an art to get a closed loop control to actually get started (Spice calls it "converging").

I'm just getting familiar with LT Spice, so I don't know if converging is an issue or not. For the simulation of the fixed duty buck, there is no loop to close, so it's not an issue. The bigger issue would just be developing suitable models of the hub dynamo at various speeds. Has anyone done this already???

thanks,
Steve K.
a.k.a. the other-other Steve

[minisystem]

Hi folks.

Glad to see you have done it MiniSystem. Nice work.
I admire your dedication - The bits I got to try something similar are still in a draw :-/

What was the switching frequency in your design?

32 KHz, the fastest PWM I think I can get out of an Arduino. Thinking of going up to 100 KHz using a different Atmel controller. My rudimentary understanding of switching converters is that there are tradeoffs with switching frequency and efficiency. I should probably read this: http://www.eetimes.com/design/power-...C-DC-converter

[minisystem]

hi gang,

regarding the whole scheme of adjusting the duty cycle of a buck regulator as a means of maximizing the power from a hub dynamo... has anyone tried running simulations as a way to evaluate some of the variables such as input capacitance?

One reason I ask is that it's a bit of a nuisance to build hardware to run some of these tests (especially if you tend to blow up mosfets in the process!). Another reason is that I recently downloaded a version of Spice that is available for free from Linear Technology:
http://www.linear.com/
The upper right of the web site has links to the download for the software (LT Spice) and a couple of pdf documents to introduce you to Spice and a more complete document that provides the details of what Spice can do and how to use the various models that are available.

I used to use Spice in the early 90's for simulating switching regulators and other circuits. The computers back then just didn't have enough processing power and a simulation might take 8 hours to run. There was also a bit of an art to get a closed loop control to actually get started (Spice calls it "converging").

I'm just getting familiar with LT Spice, so I don't know if converging is an issue or not. For the simulation of the fixed duty buck, there is no loop to close, so it's not an issue. The bigger issue would just be developing suitable models of the hub dynamo at various speeds. Has anyone done this already???

thanks,
Steve K.
a.k.a. the other-other Steve

I played a bit with LT Spice last year, but never having done SPICE simulation before, I quickly got overwhelmed. I did not accurately model the dynamo output, instead just used a 6V p-p sine wave as the power source, which, of course, behaves nothing like an actual dynamo output.

When I have some time without kids pulling at my trouser legs I'll get some screen grabs from the scope to share so you guys can see what's going on.

Cheers,

Jeff.

[Steve K]

I played a bit with LT Spice last year, but never having done SPICE simulation before, I quickly got overwhelmed. I did not accurately model the dynamo output, instead just used a 6V p-p sine wave as the power source, which, of course, behaves nothing like an actual dynamo output.

Spice is probably easier to pick up if you've already had circuit analysis classes as a EE. When I was learning Spice, I used a book titled "Spice - A Guide to Circuit Simulation & Analysis Using PSpice" by Paul Tuinenga. ISBN 0-13-834607.

Simulations are good for situations where working with the real hardware is difficult or it's hard to change some of the parameters that will vary. For instance, I did some work with large solar panels delivering a few kilowatts. As much fun as it would be to play with actual large solar panels, it was much more practical to run simulations where I could easily adjust the panels' open-circuit output voltage, short-circuit output current, etc. I had voltage-current curves from the actual solar panels so I could generate a good model for them, though.

This leads me to the downside of simulations.... they are only simulations, and require good models to be useful. For instance, for a bike dynamo light, you'll need a decent model of the dynamo that has the correct internal impedances and voltage source. If you want to simulate the dynamo at a single speed, you'll need an AC voltage source equal to the dynamo's open-circuit output voltage. The model should include (at least) a series inductance and series resistance. To calculate the inductance and resistance values, you'll need some characterization data for the dynamo, where the output is measured as the load is varied. Nick Ray did this for a range of SON speeds back in the BikeCurrent days:

http://www.flickr.com/photos/kurtsj0...57617009273346

Fortunately, models for common semiconductors are easier to find, and even capacitor manufacturers provide some information on the internal resistance and inductance of their caps.

Ultimately, it might take more time and work to simulate a circuit than to just build the darned thing! Even when you've got the simulation running, you can't expect it to be 100% accurate, and some days it's just hard to get a simulation to run, especially if it is a closed loop control. That's never a problem with the actual circuit... although sometimes the circuit does blow up when you apply power, and you might not know why.... that's when a simulation can be handy.

So... in summary... simulations aren't the answer to every question, but they are a handy tool to have available!

Steve K.

[Savvas]

Hi jdp298,

Winter's here and with daylight saving, it's bike-light season again! Can you please answer a couple of questions for me:

- the LDOs, when placed in parallel, simply pass more current in an additive manner? Is that how they work? I can only get 100ma ones...

- do you just connect your 2 separate headlights and tail lights to the dynamo in parallel? No problems with different impedances?

thanks,

Savvas.

[jdp298]

1. Yes, three 100mA LDOs will safely pass 300mA. I can get higher current ones from Maplin which I used in 1 of the front headlights but I'm not happy with it; it's efficiency isn't nearly as good as 3 lower current versions in parallel. That headlight fades out far faster.

2. Each light has its own bridge rectifier and then the rest of the circuit, so essentially all 4 lights, front and rear, are in parallel. I'm not sure the impedances of each are too critical, but all come on and stay as bright as I would expect. I use a current limiting resistor for each so they're all run at the right power. The total power is also under the 3W rated for the dynamo so there's no need to tune some funny filter impedance reactance things to drive the power higher earlier. Keeps it simple.

[swhs]

I was also discouraged by swhs's reports of his mysterious microcontroller-based dynamo driver, which claims remarkable power output for triple XM-Ls at low speeds.

It's not mysterious, details are out there in various places on how to get maximum power. Low speed power is indeed really good. The key is using the dynamo itself as part of the circuit, but you still need a few converters after that. Note that I didn't make it, I was discouraged as well to make my own once I got the details of this one (I had just started before the complete system was presented to me), however, due to circumstances I will need to make my own as the developer is MIA. I will document how it works later (I know the tricks). I didn't have time to do much last few months, and last year I could do even less. That's how it goes

Edit: Switching in the microcontroller system that I tested is done at 25kHz...

[minisystem]

It's not mysterious, details are out there in various places on how to get maximum power. Low speed power is indeed really good. The key is using the dynamo itself as part of the circuit, but you still need a few converters after that. Note that I didn't make it, I was discouraged as well to make my own once I got the details of this one (I had just started before the complete system was presented to me), however, due to circumstances I will need to make my own as the developer is MIA. I will document how it works later (I know the tricks). I didn't have time to do much last few months, and last year I could do even less. That's how it goes

Edit: Switching in the microcontroller system that I tested is done at 25kHz...

Ha. Well, it's still mysterious to me. I look forward to your documentation of this special dynamo power converter. I'm fascinated by its ability to produce so many watts of power for 3 LEDs at such low speeds.

In the mean time, I'm pressing ahead with a development board to tweak my buck converter configuration. It will send real time speed and current measurements over a USB connection to a PC. I have a rough working version using an Arduino, but want to layout a proper circuit board to see how much increasing the efficiency improves performance.