A new dynamo LED driver to get more power from the dynamo, as needed

Experiment 1: Birth of the Frankenlamp2: Philips LED bike light modified for use with dynamo

On 1 December 2010 I modified a Philips LBL for use with dynamo. This is simple when one makes use of the dynamo's 'feature' that it can only supply about 500mA (note: clawpole generators eletrically limit current output, this is not so for the Sunup eco 3-phase generator but that has its own regulation system which electronically limits the outputcurrent also to 0.5 A). I used some old stuff I had lying around, standard rectifier (no Schottky diodes), and an old 4700 µF capacitor:

LBL for dynamo, 1: Direct
Click on the picture for full size Click on the picture for full size Click on the picture for full size

This experiment with the LBL running direct from dynamo was a bit disappointing: The beam is the same great beam but power was in between that of the LBL's low and high mode, did not give me a 'wow' feeling which the LBL on high gives, and it was dimmer (but of course the beam is a lot larger) than that of regular dynamo lamps such as the Edelux.

Experiment 2: More power to the Philips LBL: Using an analog circuit with tuning capacitors

As you can read on candlepowerforums (and forum.mtbr.com), it's possible to extract a lot of power from a standard dynamo using simple resonance circuits. So on 6 December, after the previous straight dynamo powered setup, I used tuning capacitors. The capacitors for resonance I used were also old and possibly no longer the right capacity as electrolytic capacitors dry out and have less capacity after a long time. This is probably why the results were not very good, because I didn't see a difference with the previous experiment:

LBL for dynamo, 2: With tuning capacitors
Click on the picture for full size

What we would want from a dynamo LED driver

A problem with analog circuits using tuning capacitors is that they are not flexible. You need to choose a tuning (or resonance) capacitor such that you extract the maximum amount of power from the dynamo, but the maximum power extracted differs at what speed you want to ride with most power, i.e. the top of the power curve is at different speeds with different capacitors. What you would really want is to have a variable capacitor such that the power can be optimized to the maximum possible (if you can use that much for the LEDs) at each speed. Such variable capacitors are not available, but but there's no need... The circuit simply needs to give the dynamo something to aim for as it were. I've seen this mentioned in a few places before, but I have not really seen it used effectively to give maximum power. Next, you need to control the current. Whatever the LED wants, is not what the dynamo gives so you need a voltage converter.

Commercially available alternatives or other solutions previously described on the web

These are interesting from an experimenting point of view, but not especially suited for the task at hand, so now we come to the really interesting section:

Microprocessor controlled LED driver

The failure to get the 'wow' factor really was the impetus for me to start diving into LED drivers but I also wanted to run the LBL with a b2flex or maxflex. The LEDs in the LBL are driven at 0.689A according to Olaf Schultz (www.enhydralutris.de) and I have done some measurements with the limited equipment I have, which correspond very well. From runtime calculations I would estimate current through the LEDs to be about 0.80A (see also my posts on candlepowerforums), but when measuring I found the voltage drop over a small resistor of 0.05Ω to be 35.3mV which gives a current of 0.706 A (measurement error at worst 5% due to the resistor but as I averaged resistors I think it's much better). So the driver in the LBL is not that efficient, but also, there is room for a lot more light output when running the LEDs at 1.0A (maximum the Rebels are rated for).

I contacted George from Taskled early December with my ideas for a dynamo driver which should be a cool product and I think could sell very well, but he wasn't interested.

So I was diving into microcontroller programming (which really means more the programming environment, as I've programmed in loads of languages, I even made a programming language a few years ago), and I was looking for which microcontrollers would be useful, when I got an email early December from someone who had already implemented a system that does exactly what I wanted to do...

Due to various circumstances, here, and with him, there was a long delay before I got my hands on it. I couldn't do any developing myself in the mean time either.

I got my driver on 9 June...

What it can do

It can drive LEDs at 10W, it can supply power for USB devices. It should be changed so that the lamp has priority, as now, light will dim if too much power is drawn from the USB output. There's provision for standlight via a LiIon battery. There is a taillamp driver for 5mm red LEDs driven at 20mA (at least 3 LEDs, using a boost circuit from 5V).

Experiences, using it with a Philips LBL

Test in progress

Pictures of the provisional setup

1. Headlamp and taillamp running off of the 18650 battery which is used for standlight, the electronics is in the saddle bag hanging on the handlebar.
2. on-off-on switch to switch modes with a LED attached as a mode indicator.
3. Taillamp consisting of 9 red LEDs wired in series (20 mA).

In the standard LBL the LEDs are driven at 0.689A, and this driver runs the LEDs at about 0.90A (I've done measurements with a bunch of dynamo hubs and I got to about 0.85 A using a triple XM-L, I will later measure a double XM-L and/or the LBL to measure power output with 2 LEDs). This should give about 330 lumen out-the-front. So, will that power increase noticeably improve the lamp? Is the standard LBL power already good enough? I suspect it will not improve what you see by much even if you were to run the LEDs at the maximum the Rebels are rated for, which is 1.00A. This is because of the logarithmic nature of human vision, i.e. you need a lot more intensity to experience a noticeable increase in perceived brightness, this is probably also the reason why the LBL was so impressive compared to the Edelux when I first saw it in action in August 2010: It has the same intensity with which is lights up the road, but it lights up much more (wider and longer beam), so the extra light output is mainly used to make the beam larger and this is quite noticeable. I'll soon see if this estimate is correct!

The driver has a battery for standlight, which is a 18650. This is charged while riding...

2011-6-25: 5 km ride: First ride on wet road 1 was interesting in that I felt that the range of the LBL-dynamo was about 70m despite the road being wet (it wasn't raining any more, so no layer of water on the road, and some later sections of that road were drying up). I encountered a car and the driver went as far to the left hand side of the road as possible, while slowing down almost to a stop. This was interesting, it hadn't happened with any other dynamo lamp on that road before, it also didn't happen with the LBL on battery power when I first tested it in August 2010. A kilometre or so further I encountered another car and exactly the same thing happened. I did have the LBL's cutoff below the horizon in case you're wondering...

I didn't get the impression of the lamp being brighter than the battery powered LBL but the road was wet, so perhaps it was a lot brighter but I didn't notice it because of the wet road (yes, I should have another LBL to check at the same time).

There are a few settings for the driver. The operating mode is set with a momentary on-off-on switch and there's a status LED. Oh yes, and it comes with a bunch of LEDs for a taillamp... When setting the lamp in auto mode (auto switches between daylight and night modes) it will give maximum brightness. There are 3 other fixed modes: low, medium, high (0.90A). In high I noticed the brightness increased to a maximum when I got to about 24 km/h.

I didnt' notice any vibrations from the dynamo hub (this time I used a Shimano HB-NX70) but road 1 isn't perfectly smooth asphalt and the bike I tested on has less issue with dynamo hub vibrations anyway, so for more on this I need to check on road 4.

2011-6-26: 20 km ride: Testing with the same bike on road 4 I did notice vibrations, but there isn't a sudden peak as much as with a standard headlamp such as the Edelux, it's smoother over a larger speed range. That was my impression. To be done: More tests with this bike, then tests on the other main testing bike with DH-3N80 and SP PD-7 (= new name for the HB015) dynamo hubs.

2011-6-27: 10 km ride: Tested more and made some videos: video page (direct links:      road 3 (reddish asphalt)      Vriezekoop even (good asphalt) )

The LBL with dynamo driver puts light on the road that is at least as strong as that of streetlamps, its light doesn't get drowned out by them as with most other dynamo lamps.

To come: more rides and beamshots...

Experiences, using it with a 3x XM-L

Triple XM-L, bin T5, colour 1A, LEDs running at 0.90A (well, measured it seems 0.85A, but that's easily offset with a higher bin LED :)), which should give about 800 lumen out-the-front, none of that Supernova fairy tale stuff.

First testride 2011-7-3: Optic = Cute-3-M-XM (medium, +/- 12.5°). Throw is no more than the QL-269, the setup is provisional of course, not even proper LED housing for the XM-Ls. I needed to put my hand over the side of the optic as it manages to put a lot of light directly into my eyes, well, you can see in the video.

     Triple XM-L with dynamo driver vs. Q-lite QL 269

To be tested: Cute-3-W-XM although the medium optic is already very wide with little throw...

3rd testride 2011-7-4: Optic = Cute-3-SS-XM (narrow, +/- 9.5°. I compared it to the E3 triple 2009. The E3 triple is completely blown away.

Compare the following 2 and see how much better the triple XM-L with special dynamo driver is:
     2011-07-04: Route with trees and under a road: triple XM-L, bin T5, colour 1A, Cute-3-SS-XM narrow triple optic, with special dynamo driver (mp4) (ca. 17 MB)
     2011-07-04: Route with trees and under a road: E3 triple 2009 supposedly 550 lumen (mp4) (ca. 17 MB)

The following ride compares the E3 triple and 3x XM-L in one long video:
     2011-07-04: 2nd piece of the same road: with trees and past a few houses: triple XM-L, bin T5 colour 1A, Cute-3-SS-XM narrow triple optic, with special dynamo driver vs. E3 triple 2009 (mp4) (ca. 73 MB)

3x XM-L directly on dynamo (compared to E3-triple-2009)

Here I use only a bridge recitfier + 4700µF smoothing capacitor. No tuning capacitors!

The result is predictable: a lot of light, but not as bright as when using the special driver. The 3 x XML gives much more light than the E3-triple-2009. This is mainly due to not having a crap LED driver, as the XM-L isn't that much better at low currents than older LEDs, it's mostly a lot better at high currents. See also the E3 triple page for estimates of how much light the E3 triple could produce if Supernova had made/used a decent driver.

     2011-07-11--EX1-3XML-vs-E3triple2009.mp4 (23 MB)

This video is not great as the E3 triple was first pointed too low which is why it looks very bright at first, then I adjust it while riding (aim it a bit higher) to give the best view of the road, so for more throw, which the camera doesn't really pick up. I felt that the 3x XML direct on the dynamo gave me the same amount of light as I remembered from the Ktronik 3x XP-G (unfortunately I didn't have time in August 2010 to make beamshots of the Ktronik lamp).

After the ride, I did some calculations using measured XM-L T5 and XP-G R5 output (not datasheet values): XP-G R5 measurements, XM-L T5 measurements:
This gives, using junction temp = 40 C, optical efficiency = 85%:

 3x XP-G R5 at 0.5A (direct on dynamo): 3x ca. 154 lm x 0.85 = 394 lm
 3x XP-G R5 at 0.7A (tuned resonance circuit): 3x 204 lm x 0.85 = 520 lm

 3x XM-L T5 at 0.5A (direct on dynamo): 3x ca. 188 lm x 0.85 = 479 lm
 3x XM-L T5 at 0.7A (tuned resonance circuit): 3x 261 lm x 0.85 = 666 lm
 3x XM-L T5 at 0.9A (special driver): 3x (ca. 0.9 x 360 lm) x 0.85 = 826 lm

A T6 would produce about the same according to later measurements at these currents. Strangely the measured T6 was less bright than the earlier (but pre-large-production) T5 at high currents, so there could be variations in the LEDs despite selection? The difference at higher current suggest not the LED core but the thermal interface to be worse in that T6 sample that was tested): XM-L T6 measurements

The Ktronik triple XP-G will be somewhere between the values 394 and 520 lm (depending on speed and tuning capacitors), so it can indeed be close to the triple XM-L direct on dynamo. The triple XM-L running on the special driver feels as if it's much brighter than what I remember of the Ktronik triple XP-G. I hope I remember correctly, but, seeing the calculations it all fits. And yes, I did the calculations after the experiment with the triple XM-L direct on dynamo, at which time I thought "hey, now the triple XM-L looks more like what the Ktronik triple XP-G produced", so I was NOT adjusting my experiences in my mind to fit the calculations!

This was the last experiment with the triple XM-L (at least for a while) as I went an experiment too far and fried the LEDs :(

Power measurements

I tested the SP PD-8, SP SD-8, SP PD-7/HB015, Shimano DH-3N80, Sanyo NH-H27. Still to do: Shimano HB-NX70.

Using the special dynamo driver with which I power 3 XM-Ls I've measured the power supplied to the LEDs. The results show that the Sanyo NH-H27 indeed supplies less power than the other dynamos. Further it turned out the power supplied by the SD-8 (dynamo meant for small wheels) in a 622mm wheel is a lot lower, in line what I had expected at first considering the specified wheel size; a true 2.4W dynamo should provide a bit more power. A test as done according to StVZO/TA with a resistor of 12 Ω is yet to be done.

2011-12-10: Here are the results in a graph:
Dynamo power putput measurement

The experimental dynamo driver is optimized for Shimano dynamo hubs, so perhaps with some tweaking power output can be improved with the others (Power output for the DH-3N80 is highest of the dynamos I've tested so far).

Note that the current through the LEDs is limited to ca. 0.85A in the driver (as it was originally meant for the LBL; the Rebel LEDs are rated to 1.00A maximum) and when using the Shimano DH-3N80 you get to this current at below a speed of 30 km/h, which causes the curve to go down. So it's possible to get more power from that dynamo at 30km/h.

The measurements of the PD-7/HB015 are not linear while all other dynamos's have a nearly linear relationship of power/speed. This could be the result of a few problems I had at one point with measurements due to connectors not making good contact from vibrations. I need to check that by redoing the measurement.

Raw data:

All measurements have done with an old multimeter but it seems to be giving fairly accurate results:

R=0.10 Ohm, driver with 18650:

SP PD-8, 
          Voltage over R, V total (R+3 LEDs)  Ptotal (R+3 LEDs)  Taillamp
 15 km/h: 36 mV            8.40 V              3.0W               steady (ca. 0.4W)
 30 km/h: 80 mV            8.70 V              7.0W               steady (ca. 0.4W)

R=0.05 Ohm, driver with 18650:

SP PD-8:
          Voltage over R  Voltage  Ptotal (3 XM-L+R)  P+taillamp      Taillamp
 15 km/h:  20   mV         8.40 V   3.36 W             ca. 3.76W       steady (ca. 0.4W)
 25 kn/h:  36   mV         8.70 V   6.26 W             ca. 6.56W       steady (ca. 0.4W)
 30 km/h:  42.5 mV         8.80 V   7.48 W             ca. 7.88W       steady (ca. 0.4W)

         V (resistor R)   V total  Power (3 XM-L+R)  Taillamp
 15 km/h  20-22mV           8.45V    3.38 - 3.71 W    steady (ca. 0.4W)
 20 km/h  26mV              8.60V    4.47 W           steady (ca. 0.4W)
 25 km/h  36mV              8.70V    6.62 W           steady (ca. 0.4W)
 30 km/h  42mV              8.80V    7.39 W           steady (ca. 0.4W)



R=0.05 Ohm, No 18650:

          V (resistor R)  V total Power (3 XM-L+R)  Taillamp
  5 km/h : 4.0   mV        8.08 V   0.65 W           flashing (? W)
 10 km/h : 13-14 mV        8.29 V   2.25 W           steady (ca. 0.4W)
 15 km/h : 20    mV        8.47 V   3.39 W           steady (ca. 0.4W)
 20 km/h : 27-28 mV        8.62 V   4.74 W           steady (ca. 0.4W)
 25 km/h : 33-35 mV        8.76 V   5.96 W           steady (ca. 0.4W)
 30 km/h : 40-42 mV        8.80 V   7.22 W           steady (ca. 0.4W)

Shimano DH-3N80:
          V (through R)  I       V (3 XM-L+R, est.)  Power   + taillamp
  5 km/h : 5    mV       0.10 A   8.08                0.81 W  ?
 10 km/h : 13   mV       0.26 A   8.29                2.16 W  ca. 2.56 W
 15 km/h : 23   mV       0.46 A   8.47                3.90 W  ca. 4.30 W
 20 km/h : 31   mV       0.62 A   8.62                5.34 W  ca. 5.74 W
 25 km/h : 39   mV       0.78 A   8.76                6.83 W  ca. 7.23 W
 30 km/h : 43.5 mV       0.87 A   8.80                7.66 W  ca. 8.06 W

Again fairly linear curve between speed and power at 10 km/h and more, except
at 30 km/h.

Sanyo NH-H27
          V (through R)  I       V (LEDs+R) Power   + taillamp
  5 km/h : 3.5 mV        0.07 A   8.10       0.57 W  ?
 10 km/h : 10  mV        0.20 A   8.35       1.67 W  ca. 2.07 W
 15 km/h : 17  mV        0.34 A   8.52       2.90 W  ca. 3.30 W
 20 km/h : 23  mV        0.46 A   8.62       3.97 W  ca. 4.37 W
 25 km/h : 29  mV        0.58 A   8.70       5.05 W  ca. 5.45 W
 30 km/h : 32  mV        0.64 A   8.75       5.60 W  ca. 6.00 W

So the Sanyo produces less power than other 3W hubs, and it's probably not StVZO
compliant as the Sunup people found out.

SP SD-8:

          V (through R)  I       V (LEDs+R) Power   + taillamp
 5 km/h : 2.0  mV       0.04 A   7.95       0.32 W  ? (flicker)
10 km/h : 7.5  mV       0.15 A   8.22       1.23 W  ca. 1.63 W ?
          (here the taillamp is on the edge of flickering)

15 km/h : 11.5 mV       0.23 A   8.36       1.92 W  ca. 2.32 W
20 km/h : 16.5 mV       0.33 A   8.50       2.81 W  ca. 3.21 W
25 km/h : 21.5 mV       0.43 A   8.61       3.70 W  ca. 4.10 W
30 km/h : 26.5 mV       0.53 A   8.69       4.61 W  ca. 5.01 W

A lot less power... To get to the almost 8W that the PD-8 reaches at 30 km/h I
would need to cycle ca. 47 km/h.


This driver can extract ca. 8 W at 30 km/h from many dynamo hubs (such as the SP PD-8/PV-8 and PD-7/PV-7 (=new designations for HB015/HB011) hubs, more from the Shimano hubs). I suppose some tweaking could be done to extract the most power from the dynamo for specific hubs. As it is the driver was originally tested for and thus made to extract the most power from a few Shimano hubs. Not all hubs act the same and it shows here as it gets a bit more power from the DH-3N80 than the SP PD-8 and SP PD-7/HB015... The most impressive feature is however the power output at low speed which is higher than any other system out there, and even with 3 LEDs after a short while, the power output is such that the LEDs light up flickerfree and they are plenty bright at 5 km/h. Also nice is the fact that it doesn't need choices to select the best power output for a given speed (with a system using resonance capacitors, the capacitor used must be chosen so the power output gets highest at a certain speed, with this system power output is optimal at all speeds).

Power output of the SD-8 was disappointingly low confirming again that my view of using a small wheel dynamo in a big wheel, is not a good idea if you want to be ready for future developments.


If you're interested in this system, well, you will have to wait... I can't say yet when or how it will be made available.

2011-11-18: I'm going to design a new system soon, similar in features to the tested driver but I want to drive USB output not from dynamo output but from the 18650 so that USB output never influences the headlamp. Hopefully I can get something working soon. I will put the PCB/battery in a aluminium IP66 watertight housing, I'm still researching plugs and supple cables.

To email me go to the email page