Feeler: Driver Circuit for max LED Power from Hub Dynamo

Eddy

Newly Enlightened
Joined
Oct 15, 2007
Messages
13
Location
Rockwall, Texas
Today I visited my local bike store for a 1.125" steerer tube as a physical sample, to make sure my electronics will really fit inside. I was lucky, right at the top of the scrap metal container there was a 1.125" rigid steel fork that I could carry off for free.
Hammered the star nut all the way thru the tube and out the bottom, ....
....Why not go for the 3rd version right away ?
Because a) Given the time I plan to spend on this every week, it would take very long to complete and b) Feedback from the 1st version improves a later one.

Trying to find some way to contribute to this remarkable driver circuit effort... I looked at bikes and discovered headlight nirvana on a dirt jump bike... (edit: from an electronics vibration perspective it may be better to orient the PWB in the steerer so that one sees planform when looking from the right or the left. That way there is less deflection (but perhaps more shock?) when one lands a jump... or rolls over a piece of gravel.) Marzocchi makes this wonderfull fender that bolts right over the hole in the bottom of the steerer tube. Here are three pictures that show why this is good:
208950189-M-1.jpg
208950523-M-1.jpg
208950221-M-1.jpg


The reason this is on topic is: If the single unit headlight assembly people like (because it reduces interconnect) consists of a fender with the electronics configured as a candle that fits into the steerer from the fork crown side of the steerer... The inner diameter of the aluminum steerer is a little bit smaller for strength reasons where the fork crown attaches. This is a bit different than the ID on the cromoly steerer on the single speed (procured the same way as the benchmark):
208950278-M.jpg


Should have data tomorrow on IDs. I am not so enthusiastic about scoring the inside of the steerer by driving the star nut all the way through or setting a new one, as they rarely go in straight without proper tools/fixtures. Especially if it is a star nut for steel steerers in an aluminum steerer! Suprisingly the difference in ID is enough to warrant two different star nuts!
 
Last edited:

ktronik

Enlightened
Joined
Oct 12, 2003
Messages
779
Location
Australia
you could always run it in a small tube strapped to the 'down tube', behind the fork... then the whole system will be removable... maybe some PVC pipe or something... I will build both to show, when we get the PCB up & running...but mounting not really a problem external or internal...with the PCB shape that is being used.

K
 
Last edited:

Martin

Enlightened
Joined
Apr 5, 2006
Messages
584
Location
Germany
Eddy, the MOSFET rectifier is not that simple. Once the MOSFETs are conductive, they are conductive throughout the half-cycle (or most of it). If the smoothing capacitor is below the dynamo voltage this is OK, but a large part of the half cycle the smoothing capacitor would be above, so I must quickly open the MOSFETs then, needs way more sophisticated control logic. The opposing dyno terminal is of unknown phase because of the capacitors I put between, so forget that. It really needs a dedicated controller like they use it for high power MOSFET rectifiers.
Now consider the benefit: Half-bridge replaced by MOSFETs = 0.3 V saving out of 10V LED string voltage = 3%. You will never ever notice the extra amount of light and your money is better spent on higher bin LEDs. How you arrive at 15% ?

Stand light ? Many ways to do it. The easiest way is to connect a standard tail light across the string of 3 LEDs. All German tail lights must meanwhile have the stand light feature. A later version of my PCB will include stand light for the head and any tail you may want to connect.

You can increase the number of LEDs up to the max voltage rating of the IC when you adapt the passive components (different bias, different switching point). I will come up with multiple BOMs:
Bottle dynamo, 3 LED
Hub dynamo, 3 LED
Hub dynamo, 4 LED
Hub dynamo, 6 LED

Some forks' steerer tubes DI is very small on the bottom end. My RS Jett XC was like 25% narrower. No hope to insert a PCB that way. Only the top is a safe entry, because this is always designed to accept a standard star nut.

Calina, the crowbar short-circuits the power if the voltage exceeds a certain maximum as it happens when the LED string disconnects. Without the crowbar, the voltage then climbs so high that the circuit is damaged and the LEDs too, if they reconnect to that high voltage. Once the crowbar attacks, the voltage across it drops below 1 V so that the power dissipation in the circuit will be small and nothing will ever happen to the LEDs when they reconnect. The dynamo will start to hum somewhat.
 

PhxCycler

Newly Enlightened
Joined
Sep 20, 2007
Messages
26
the crowbar short-circuits the power if the voltage exceeds a certain maximum as it happens when the LED string disconnects. Without the crowbar, the voltage then climbs so high that the circuit is damaged and the LEDs too, if they reconnect to that high voltage. Once the crowbar attacks, the voltage across it drops below 1 V so that the power dissipation in the circuit will be small and nothing will ever happen to the LEDs when they reconnect. The dynamo will start to hum somewhat.
I just toasted my LEDs as a result of this. I had a 3 LED light working great and wanted to see what a 6 LED light would look like. I built up another 3 LED fixture and added some switches to select A, B or A+B. I was running the 6 LEDs in series (A+B) when I switched to the A light. I saw a brief flash, then nothing.:mecry:At the time, I wasn't sure what happened. When I switch to the B light, poof! :banghead: Now I have two cool fixtures and no working LEDs. DX, here I come.

Martin, would your crowbar circuit work with the original passive circuit shown here? This is the circuit I'm using. https://www.candlepowerforums.com/posts/2156522&postcount=49

If so, could you give me the values of C11, R15, and D5, or tell me how you computed them? I obviously need to add some protection to the output so I don't do this again.

Thanks,
-Mike
 

Eddy

Newly Enlightened
Joined
Oct 15, 2007
Messages
13
Location
Rockwall, Texas
Thank you all for humoring my naive/sophomoric thoughts (that is earnst).

Ktronic, thanks for suggesting alternatives before I hit the wall measuring steerer diameters.

Martin, yes, I measured numbers as low as 19.95mm for the crown end and 25mm for the top in steel and an MX comp alloy steerer was 22 at the crown and 24.29 at the top. There were also things going on that I could not measure above the race in some forks. The bottom 4cm of length looked untrustworthy.

15% came from something like this rough calculation:

At .5 amps the 1n5818 schottky diode shows 0.6V drop. In the full bridge two are always on so that is .6 W. If they are FETs with a .03 RDSon the power dissipated is i^2R or .01. If two are on that is .02 Watts. So you have about .58 extra Watts.

In the three LED configuration at the beginning of the thread I see 5 W of LED light in the full bridge range. So, .58/5= 12%

I think my original thoughts on the doubler overlooked duty cycle...

Yes, the caps make the diode voltage ambiguous, but it does seem like D1 and D3 could be driven in a complimentary way off the capacitor side of the dynamo without a control chip... For a 6% improvement. Or even just D3 for 3% improvement. The suggested circuit was a 100 ohm from the Cap side of the dynamo to the N-Channel gate with a pair of opposite polarity schottkys (one in series with a zener) taking that gate to the negative rail...

I have no idea how much these FETs cost. [$0.90]

These better diodes cost $0.25 http://www.vishay.com/docs/94324/94324stp.pdf and the SMT package is easy to solder...?

Also, what happens if D2 and D4 are compimentary Headlight LEDs?
 
Last edited:

Martin

Enlightened
Joined
Apr 5, 2006
Messages
584
Location
Germany
Eddy, the drop across my 1N5818 diodes is around 0.37V at 0.5 A. The dynamo is pretty much a current source, so there's really not much benefit to expect. But for a definite result, I need to measure the MOSFET rectifier. And I will try shunting two of my rectifier diodes with MOSFETS, too. The part that you linked (or an alternativ low drop one) I will try to accommodate in the next layout which will probably be mixed SMD and thru-hole.
D2 & D4 complementary headlight LEDs ? This would violate the LED's reverse voltage rating, so you'd have to connect a rectifier diode in series.

PhxCycler, sad to hear what happened to you.
I'd recommend to put any smoothing capacitor into the headlights so it's impossible to remove from the LEDs.
The overvoltage crowbar is calculated like this: SCR's trigger voltage / resistor defines the zener current. This is picked for best temperature stability (curves in datasheet). Zener voltage (at given zener current) + trigger voltage is attack voltage of the circuit. Leave out the cap unless you find that the circuit fires sporadically without that you exceed the set limit. Verify your design by putting it through PSPICE (best free version on the linear website). Then you can also make sure that your peak current is within the SCR's capability.
I still suggest you put the smoothing caps into the headlights, this is better, cheaper, easier (if it fits).

Now working on the PCB for the auto-switching circuit. It looks like it could fit onto a 25x100mm strip (inclusive the crowbar and a strain relieve). It's very dense, tall components can't be close to the edge because it has to go into the (round) tube. I need to spend a couple of evenings and optimize.
 

PhxCycler

Newly Enlightened
Joined
Sep 20, 2007
Messages
26
PhxCycler, sad to hear what happened to you.
I'd recommend to put any smoothing capacitor into the headlights so it's impossible to remove from the LEDs.
The overvoltage crowbar is calculated like this: SCR's trigger voltage / resistor defines the zener current. This is picked for best temperature stability (curves in datasheet). Zener voltage (at given zener current) + trigger voltage is attack voltage of the circuit. Leave out the cap unless you find that the circuit fires sporadically without that you exceed the set limit. Verify your design by putting it through PSPICE (best free version on the linear website). Then you can also make sure that your peak current is within the SCR's capability.
I still suggest you put the smoothing caps into the headlights, this is better, cheaper, easier (if it fits).
Thank you Martin. That's a great idea to put the caps with the LEDs. Do you think this would be sufficient to keep them from being blown again without the crowbar circuit? That would seem to solve the problem quite simply and elegantly.

The reason I'd like to be able to switch between 3/6 LEDs is for power generation (i.e me) minimization. I frequently ride distances of 1200 km and I don't always need uber brightness. If I'm cruising along at 25-35 kph on flat roads, I'd like to minimize the power I need to generate. The 3 LEDs are more than sufficient in this case. However, it's also not uncommon for me to encounter a 60-80 kph descents. In these cases, more light is definitely better.
 

Eddy

Newly Enlightened
Joined
Oct 15, 2007
Messages
13
Location
Rockwall, Texas
Eddy, ...
D2 & D4 complementary headlight LEDs ? This would violate the LED's reverse voltage rating, so you'd have to connect a rectifier diode in series.

Thanks for all the information. The proposed circuit is becomming more understandable with every bit of dialog.

... Boost converters used to put a bunch of diodes in series to support the reverse voltage. With a 5V reverse voltage could D2 and D4 be replaced by 4 LEDs (2 per side) on a 3 LED primary headlight?

If I understand correctly in doubler mode they are not used and in full bridge mode you are going faster and they might be useful with a focused beam. I do not have a good understanding of the voltages yet. Is the linear PSPICE model able to help that understanding?

Eddy
 

Martin

Enlightened
Joined
Apr 5, 2006
Messages
584
Location
Germany
That's a great idea to put the caps with the LEDs. Do you think this would be sufficient to keep them from being blown again without the crowbar circuit?
Absolutely. Better than any crowbar (which will allow a certain overvoltage before acting). If you cannot get the size of cap into the headlight that you wish, then use multiple smaller ones. Only have the tuning caps outside.[/quote]
The reason I'd like to be able to switch between 3/6 LEDs is for power generation (i.e me) minimization. I frequently ride distances of 1200 km and I don't always need uber brightness. If I'm cruising along at 25-35 kph on flat roads, I'd like to minimize the power I need to generate. The 3 LEDs are more than sufficient in this case. However, it's also not uncommon for me to encounter a 60-80 kph descents. In these cases, more light is definitely better.
So do I get this right, you are using the full-wave / doubler circuit and you still want to switch b/w 3 and 6 LEDs ? So you get 3 levels ? If so, you might want to check out the basic concept of a quadroupler circuit on my website. By the way, asking for less light power might not save you much pedaling effort, it might actually cost you a bit more. But try it, depends on how the dynamo is made.

Eddy, D2 & D4 carry only half the current each. Putting LEDs in these positions would defeat the purpose of the driver circuit, it would consume lots of LEDs and underutilize them, plus the main LEDs would lose a lot of light.

Pretty sure now that I can fit the components on a 25x100mm strip. Probably shorter. Working on the PCB layout.
 

Eddy

Newly Enlightened
Joined
Oct 15, 2007
Messages
13
Location
Rockwall, Texas
Martin said:
....
Pretty sure now that I can fit the components on a 25x100mm strip. Probably shorter. Working on the PCB layout.

I'm sold! ....How much shorter?

209747651-S.jpg


Having no idea how bleary eyed folks are, on this fork the right leg is hollow, with plenty of room for a 25mm x ... 85mm pwb. ...How much shorter? :)


A fitting and a drip loop will make 25mm x 100mm work fine. As you can see, this is my primary night riding vehicle in that the dynamo is mounted. Electricity exits right beside that empty fork leg.

When will the "exactly what I need" pwb be ready?

...This is exactly what I need.

Thanks again to all the contributors.

Eddy
 
Last edited:

Martin

Enlightened
Joined
Apr 5, 2006
Messages
584
Location
Germany
I'm sold! ....How much shorter?
Possibly 85mm, if you decide to cut away the overvoltage crowbar and strain relieve could be 70mm. All prelim figures.
Be aware that 3 to 4 larger electrolytics are on wires off the board.
So you will take up the task of vibration-testing the board in your fork leg :D
 

Eddy

Newly Enlightened
Joined
Oct 15, 2007
Messages
13
Location
Rockwall, Texas
Possibly 85mm, if you decide to cut away the overvoltage crowbar and strain relieve could be 70mm. All prelim figures.
Be aware that 3 to 4 larger electrolytics are on wires off the board.
So you will take up the task of vibration-testing the board in your fork leg :D

Yes, I will do that. Mind, I will pot it in a wax filled microballoon epoxy mix first! Hope it does not catch on fire.
 

PhxCycler

Newly Enlightened
Joined
Sep 20, 2007
Messages
26
So do I get this right, you are using the full-wave / doubler circuit and you still want to switch b/w 3 and 6 LEDs ? So you get 3 levels ? If so, you might want to check out the basic concept of a quadroupler circuit on my website. By the way, asking for less light power might not save you much pedaling effort, it might actually cost you a bit more. But try it, depends on how the dynamo is made.
Martin, Thank you for the confirmation about the caps across the LEDs.

As for the 3/6 LED question. If I look at your power vs speed graph and specifically look at 30 kph, there is either 5 or 11 watts power of power in the 3/6 comparison. I can ride at this speed for hours on end. When I ride a 1200 km event, I may be riding with lights from 8-12 hours/night for 3 nights straight. In these events, conservation of one's energy is crutial. I have a SON generator and I now have two 3 watt incan lights which may be switched on or off in the same arrangement that I want for the LEDs. I can definitely feel the 3 vs 6 watt difference from the hub.

Also, I ran my 3 LED setup last weekend at a 200 mile event. I rode in the dark a total of about 3 hours. I could feel the difference with the LEDs on and off. So, yes, I can definitely feel the power difference in the hub and the longer the ride, the bigger the issue becomes. The 3 LED setup is awesome for all but big descents. In that case, gravity will be helping me generate the power and I want all the light I can get.
 

Martin

Enlightened
Joined
Apr 5, 2006
Messages
584
Location
Germany
I have a SON generator and I now have two 3 watt incan lights which may be switched on or off in the same arrangement that I want for the LEDs. I can definitely feel the 3 vs 6 watt difference from the hub.

Also, I ran my 3 LED setup last weekend at a 200 mile event. I rode in the dark a total of about 3 hours. I could feel the difference with the LEDs on and off. So, yes, I can definitely feel the power difference in the hub and the longer the ride, the bigger the issue becomes. The 3 LED setup is awesome for all but big descents. In that case, gravity will be helping me generate the power and I want all the light I can get.
This made me curious, so I tested right now the difference in load with 3 LEDs and 6 LEDs at both 30 km/h and 50km/h on my Shimano DH-3D71. I saw no difference in the current intake of the driving motor (of my test setup). However, there is a noticeable difference b/w LEDs connected and open circuit. What is interesting, short circuit of the hub also reduces the drag almost to the same level as open circuit.
It might be the SON behaves different from the Shimano, though.
By the way, powering 6 LEDs from a voltage doubler equals 3 LEDs on a full bridge rectifier. I did above experiment with 6 LEDs, switching b/w doubler and bridge rectifier and the result was the same, no difference in drag.
If you need to minimize the energy you spend, I suggest you charge a LiIon while you power the light at the same time. When charged, run the light off the LiIon and completely disconnect the hub until the battery needs charging again. Pretty much like a hybrid car.

Hey, what do you eat so you can ride that long ? I couldn't even drive that many hours...
 

PhxCycler

Newly Enlightened
Joined
Sep 20, 2007
Messages
26
Hey, what do you eat so you can ride that long ? I couldn't even drive that many hours...
Everything I can get my hands on! :grin2:

I just received a motor and controller that I ordered off eBay. So now I'm going to build a test setup to rotate my wheel at a constant speed. That'll make it a lot easier for me to make measurements and look at the signals on these circuits. I want to see what transients occur when I switch between the different modes. I'll also be able to easily monitor the voltage and current at the hub and the LEDs.

Your assistance thru this has been most helpful. Many thanks!
 

Martin

Enlightened
Joined
Apr 5, 2006
Messages
584
Location
Germany
Now here's something very close to Version 1.
Dimensions are 24 x 87 mm, with the overvoltage crowbar cut away it's 24 x 64. If the SCR is put on the back of the board, the crowbar can remain functional while the board can still be cut down to 72 mm length.
Anything obvious that I missed ?

dynauto20.gif
 

ktronik

Enlightened
Joined
Oct 12, 2003
Messages
779
Location
Australia
OOh man, do you sleep?? that looks great!! A lot of work just to get to here I bet... well done..

K
 

Eddy

Newly Enlightened
Joined
Oct 15, 2007
Messages
13
Location
Rockwall, Texas
Wow! That looks good. My focus now is obtaining parts in the right package to fit that board. If I put on my DFM hat... for single sided PCBs it is really easy to lift pads with the soldering iron. But big pads with generous clinch areas for lap type solder joints sometimes mess up really tight layouts.... Does the Eagle software allow for an oval piece of copper i.e. pad to be manually added at each throughhole? Manual placement of the pad will allow clinch direction to be in a direction that suits the local constraints, but still provides shock and vibe robust lap solder joints with lots of area for copper adhesion... Remember, for some of us this goes in a fork leg!

Guess I am kind of dating giving away my age when I speak about Bishop Graphics and hand placed component pads for every single solder termintion... If you do it that way you can aim the clinch however it needs to go - out, in or sideways (any angle).

Now, how do I get one of these?

...(Google Check Out lets you accept credit cards for free through 2007 and costs nothing to impliment. It is my first chioce for online transactions as either a buyer or seller.)
 

Martin

Enlightened
Joined
Apr 5, 2006
Messages
584
Location
Germany
The gerbers have gone out to the boardhouse. I should receive the first 4 boards within a week.

Eddy, that's a very valid point. I was also concerned about the tiny pads of a lot of the components in the library. The packages appear to be made for plated-through boards. I wasn't unhappy, it made things more compact, zero jumpers and single-sided.
To achieve the mechanical strength needed, my plan is to have the boards plated-through (but still single-sided). This should be stronger than the use of larger pads.

My first 4 boards are for bring-up and experimenting, and will not be plated-through. If I'm happy with these, I will order the real ones.

At that time I should probably start a new thread and ask around how many people would like one. If the total is below ten, I may just give them away for free.
 

PhxCycler

Newly Enlightened
Joined
Sep 20, 2007
Messages
26
This is awesome Martin. I'm interested in a couple of these boards. This will be a big time saver. Man, you've got a lot of work into this. :twothumbs
 
Top