Output at very low currents and efficiency

Flashlight Modding and DIY Parts
1

Hi ! I am not sure this is the right place. My question is related to dynamo bike light, so…

Well, I would like to know exactly how a led will behave at very low current. Texas_ace and djozz output tests are really useful, but they focus on flashlight use, which doesn’t really deal with low currents. In fact, I want to build a bike headlight, with the best possible output given the power given by the dynamo (and supercapacitor based standlight). It would be based on pilom’s circuit, and I know cri, tints etc. are important, but they are for later. Furthermore, those are independent considerations.
Thus, I am wondering if I would better just use XP-L HI, or “under-charge” some XHP’s in order to get the best possible efficiency.

According to texas_ace graphs, it looks like the lumen output tends to be proportional to the current at less than 0.1A. Plus, the efficiency is the highest for the minimum measured current (0.15A). So, the lowest the current, the better the efficiency, right ? Even if you “push” at 0.1, 0.05A and so on ?
Hence, for a given electrical power input, the lumen output will be higher if you increase the number of leds ? (if you leave every other parameters aside, including common sense)

Thanks !

2

No at some point the efficacy decreases, I measured the relative efficacy of the LH351D and the peak is at 180mA, under that it decreases : Testing LED efficiency decrease at very low current
So to optimally use this LED one would use PWM dimming under 180mA and analog dimming above.
I haven’t checked other LEDs but it seems that the LH351D reaches its peak at a fairly high current compared to other LEDs which don’t show the decrease at the min current in tests, they should all have this drop though even if the peak is at a lower current.

3

Nice, those are precious data ! Thank you !
I will read that tomorrow.
Any idea where I could find similar data for other leds ?

4

i am not totally sure, but i think the low currents you are talking about [where the lumens per watt is a bit higher], would be so low as to be useless on a bike light

also a bike light usually doesn;t worry about efficiency, usually they have 2 or 4 18650 or bigger, and it runs for a week or so of normal riding

i ride and use headlights, have been since the dark/incandescent/nicad/nimh days

li ion is WAY better in every way

just saying

if you start worrying about what you are worrying about, the cost will triple and the run time will be 5% more :slight_smile:

5

I don’t think you want to just look at the efficiency of the LED you’d probably want to look at the efficiency of the driver as well because even though the LED might be more efficient at low currents the driver might be the inverse. Hard to do because there’s no driver datasheet but there’s some good measurements posted above

6

Yes guys, you have a point, but I fear it is not relevant right there.

My light is powered by a dynamo hub. So no battery and no driver.

I am after efficiency because the power source is my legs :slight_smile:
There are other things to worry about ,like the circuit efficiency, but this is independent from how I should use the leds. (in this case, because it is pilom’s circuit and not a regular flashlight build).

One of the reason of my question is related to the standlight, which does not requires tremendous lumen output. Runtime on the other hand… That’s why I am concerned about efficiency.
Running a XP-L at 0.2A, or a XHP35 at 0.1, or maybe both in series, at 0.07A (random number, idk) might not be the same in term of lumen, runtime and overall efficiency.

7

Sometimes the information you need has to be obtained by experiment. It can be fun to play with parts and find out things the datasheet of the manufacturer doesn’t cover.

8

For constant standlight lasting for minutes I recommend a simple LDO in parallel to the main LEDs, which charges a supercap to 3.3V or 5V. From the energy stored in that supercap a small and efficient buck-boost converter can power an extra standlight LED (<50mW should suffice). Since the standlight only consumes a few mA, it does not need have leading edge efficiency… cheap parts will be okay and the standlight can stay turned on at all times.

9

Oh sorry I missed that part. Don’t know much about those.

You can look at the luminous flux vs forward voltage parts of the led datasheet and get an idea of where maximum efficiency is. Maybe also think about what kind of beam shape you want. You can direct a lot of light into a small area with a reflector. Reflectors take up space. Using a ton of LEDs might be a more efficient but could make it harder to focus the light where you want it and just make kind of send a little light in all directions.

As an extreme example have a not flashlight that is 600+ Samsung lm301b emitters stuck to a 2’x1’ slab of aluminum, no reflectors, with ~330mah to each one, and it’s really good at lighting up a few square feet from a few away but wouldn’t light illuminate that far up the road. Probably not that helpful but that’s all I got

10

Can you?

relative_output
relative_output849×937 79.8 KB

1 Thank
11

You can now

Relative luminous flux vs Forward Voltage (3)
Relative luminous flux vs Forward Voltage (3)700×400 15.3 KB

12

Well, this topic is going in several directions (and I must admit that this is good news), so I will try to keep things organised.

  1. Considering standlight.

For now, I have a supercapacitor that charges through a diode, from the voltage of 2 leds over 3, wired in series. Can be seen here (D8, D9, D10, C4, R4) Another Dynamo bike light (front + rear + standlight)
The supercapacitor discharges in one led through a resistor. It gives a very long standlight, but wastes some power and almost half of the energy stored in the superC (as it won’t give anything once it goes below Vf)

Thats why I though, for the next build, about a similar thing as you said. To discharges the superC in one Led, using a buck-boost. But I struggle to find one (using ti webench) that will admit 0.5 to 9V input, with correct efficiency.
A little arrangement of one tiny C in // with big R could sense when the dynamo is running. Putting a relay across this could trigger the stanglight circuit, so that it does not disturb the rest of the circuit in normal use.

  1. Beam

Reflectors are great (well, I guess as I never used one), but they have the tendency to be round, which is not ideal for road use :slight_smile:
That is why I consider using various carclo lens (10203, 10049 or 10759). They give an elliptical beam, which are better for road use, and only some tests will tell me which is better (obviously the bigger the narrower the beam, but I also want this build to stay rather compact)
One guy did similar things one reddit, I will ask him about his optics and how he got this cut-off.
https://www.reddit.com/r/flashlight/comments/gdgetx/my_front_bike_lights/
https://imgur.com/a/dVrScNz

  1. Efficiency

Actually that is what I did. If you consider (Lumen_output (O)) / E_power (P), you get :
O / P = (O/i) / (P/i)
Considering O/i, the second curves tells you it is a decreasing function of i. This is why two leds ridden at a fixed current will give more light than one led at twice this current. Also why running a led with pwm gives less lumen output than a constant but lower current.
Considering P/i = U*i/i = U : It is an increasing function of i.
In the end, a decreasing function divided by an increasing function equals a descreasing function. Hence, the better efficiency at low currents.

Otherwise you can just plot the data given by texas_ace in his tests, you get something like this. (y = lumen/W ; x = current) for a nichia 219c.
image

This “problem” here is that he never goes below 0.15A, with is not that low when you consider standlights. Plus nothing says that the efficiency keeps on increasing when you decrease the current (which is in fact what thefreeman measured.

13

Shouldn’t the analysis include the power your legs are producing? At low currents, I assume you’re cycling at a low speed and your legs are low power as well? So if the efficiency is lower in that case, relatively speaking the power loss maybe high, but in the absolute sense you’re not loosing that much?

14

I can’t.

15

It might not need to go that low. For my light I use a boost-regulator to power two standlight LEDs in series. It’s a super cheap AP3019, which discharges a 1F supercap from 5V down to below 3V. In theory thats ~half of supercap power not used, but keep in mind that supercap will not immediately discharge to 0V once the regulator shuts down. After a loooooong stop light, for example, the supercap will only need to be charged up by a fraction if its voltage again.

Of course it would be cool to double the standlight output by using a better regulator (or use a more compact supercap), but that’s something that can be done when more important things, such as beam-pattern are figured out.

Also keep in mind that this might double the total cost of standlight electronics. The LT1932, for example, can go down to 1V but costs ~900% more than the one above for ~30% longer/brighter standlight.

BTW do u mind sharing couple pictures of your current build? :smiley:

16

Thank you for these reference, I will check them.

You are right , the superC will stay charged and be useful very quick the next time, it is just frustrating to chose a “high” voltage high capacity capacitor, just to use half of it in the end. But well, that’s the game apparently :slight_smile:

Considering cost, it would be a real problem for mass production, but for a one shot “passion” build, 5€ is more like a detail. Actually it would be among the cheapest parts of the build !

17

If there are no money and space constraints you could use the same arrangement of regulators, which @thefreeman uses in some of his fancy flashlight drivers: a cascaded boost-buck

Then u can choose a 5V boost (e.g. TPS61202) which can deplete the supercap down to 0.3V. At the output of that 5V regulator u can then place a buck to drive a single LED. This way u can use up to 95% of supercap charge.

While u are at it, u should also implement a switching regulator to charge the supercap to 5V (instead of an inefficient LDO).

With these improvements you could use a small supercap and charge up your standlight within seconds.

18

The amount of light necessary to bomb downhill on a trail at night is represented in many bike lights.

The amount of light needed to see holes in the road at a low speed is less than people think.

Riding in the wilderness, I regularly switch between 20 and 1000 lumens, to conserve energy.

19

Did you see the awesome graph I made tho?

20

Sorry for the delay, I’m still here !

Well, I search a bit about this solution (boost-buck in series), but I fear that everything I would win by discharging the superC a bit more, will be lost 10 times because of the design constraints for the boost being able to pump until 1V…
What I found (well, what TI webench found for me), is that a boost regulator with 1V-5V input, and 5V output, giving approx 0.5W will have at best 80% efficiency, often 70% and even sometimes 60%…
And that’s only for the boost, there’s the buck after that.

This, added to the fact that most of the energy is at high voltages (you get back twice more energy from 4 to 2V than from 2 to 0V), it makes me consider forgetting the complete superC depletion, to focus on just 3V-8V operation at higher efficiency (easy above 90% from what I found). The last volts are expensive to gather, for almost no energy…

About your idea of charging the superC with a regulator, is there a particular interest in doing this ? I mean, my current design (that I planned to copy for the most part) wires the superC in // with 2/3 leds. (well, there is a diode to prevent it from discharging through them as soon as I stop).
Imagine 3 leds at 0.5A (typical dynamo use). for Xp-L, that’s roughly 8V total. Minus the schotky drop, that’s 7.5V left to feed in a regulator that would discharge in one led only. (there should be a switch to disconnect the regulator when the dynamo gives power) . This is a simple straight-forward way to charge the superC. The leds in // with the superC won’t light until its almost completely charged, so the design must take care of this point. (like superC in // with 2 leds, + one led in series with all the rest)

Considering the AP3019, you’re talking about, it seams like a great simple idea, but I fear that it wouldn’t suit my needs here. Being a current regulator, it requires to wire some resistors in series with the led you want to keep for the standlight. My design wires all leds in series, and there is ground at the bottom. And I don’t want to have a led dedicated only to standlight. It could be done with complex switching pattern, but that may overcomplicate the thing when simple (yet imperfect) voltage regulation can be done.