I want something that can boost 2.4V, 2.0V in view of drop-down, to approx 5V, and can flow 1A or more, or 2A with a margin or more.
High efficiency is also important because it affects runtime.

I thought about making my own using the NJM2360AD or 8A71 chip, but I was not sure if I could get the required amperage and space was limited, so I chose a commercially available module.
I'm not sure if I can trust the specs on this one, which is a very cheap Chinese product, about 800 yen for 10 pieces.

Adjust the output voltage with a potentiometer. Tests with a measurement device connected showed that from 2.0V or 2.4V, the output drops down when trying to boost the voltage to about 5V, and the heat generated by the chip becomes severe and exceeds 60 celsius. When the output is suppressed to about 4.5V, the temperature stabilizes at about 30 celsius and the device operates normally with good current value and efficiency.
If the potential difference between input and output or the ratio of the two is too large, it will not be stable, and there is not much margin for error. For the time being, it seems to be suitable for the purpose, so I decided to make it with this. If not, I will replace it with something else.


postscript:The dcdc is a well-known switching boost module that uses the MT3608 chip. It is sold at a ridiculously low price by several Chinese suppliers, and the chip is probably interchangeable and not always at the recommended program values. The ripple is so bad that modification methods to stabilize it can be found on some enthusiasts' pages.
And there are not many chips that can boost voltage from 2.0V input to around 4.0V and output 2.0A or more. In the use of this light, we were able to use up to 800lm (2 lights x 400lm), which is close to the limit.



Many commercial drivers for Power LEDs cannot be used because they require an input voltage of about 12V. In addition, many of them are inefficient and generate a lot of heat, although they are supposed to be efficient.
Therefore, I decided to build a two-transistors constant-current circuit, which is a stable circuit that emits light reasonably well as long as the Vf is above Vf. Since two Power LEDs are used, two constant-current circuits are also built in, and the output of dcdc is run in parallel.

It's a simple circuit, and can be set up to 800 lumen class just by changing the resistor values of R2 and R4.
I did not show in the figure, but I put an electrolytic capacitor on the output side of the dcdc. It is OK without it, but it will be more stable.

postscript: The constant-current circuit shown in "Circuit Experiment" below is better in many respects; a constant-current circuit using an OP amplifier and MOSFET would facilitate loss suppression and high output power.
I have not verified this yet, and it may depend on the selection of transistors, but it is possible that the constant current circuit of the feedback method of these two transostors is limited to around 0.75A (250lm) per circuit (one light). For high illumination, we should use MOSFET.



We have calculated the current control and illuminance here.

Since Q1 and Q3 are 2SC1815 and the installed potential difference of R2 and R4 (Vbe) is 0.6V, the current values flowing through Q2 and Q4 can be determined by their resistance values (A = Vbe / R).
Power LEDs are easy to drive with high illumination: 433lm/1.5A, 325lm/1.0A, 243lm/700mA, 130lm/350mA, Vf is 3.25V when 1.5A is applied.

To obtain 400lm illuminance, R2 and R4 should be set to 1ohm to obtain 208lm per lamp and 416lm with two lamps.

To obtain an illuminance of 600 lm, 303 lm per lamp and 606 lm for two lamps can be obtained by adding a 2.2ohm resistor in parallel to the 1ohm of R2 and R4.
If 1ohm is added in parallel instead of 2.2ohm, 1.2A will flow and the two lights will exceed 800lm



I used two 5W Power LEDs from Cree. This is a standard, with relatively low heat generation due to its high efficiency.
XPGWHT-L1-STAR-G51. Efficiency is over 100lm/W. It is easy to handle high intensity Power LED with relatively low Vf.

One of these is sufficient to provide 400 lm of illumination, but due to personal preference, two lights were placed side by side.
There are advantages. Heat removal is easier, and efficiency is improved by reducing the heat and illumination per light. It can be easily modified later by adding a switch to make it possible to switch only one light to 200 lm, or to make it an 800 lm class light.

Cree's Power LEDs also have a dedicated collimator lens. They are very compact and highly efficient, but after comparison, a more efficient commercial lens was used. Both have a half angle of reduction of 10 degrees.



We have already mentioned that the old light was diverted. Unlike the recent angular design, it is rounded.
The dcdc is embedded in the head section without the original reflector and lens, and the constant current circuit, Power LED, and heat sink are mounted on a circularly cut substrate to cover it.

The head part that can be rotated about 90 degrees is convenient and good. It can be turned downward by pushing it from behind.



At first, 400lumen



A red LED pilot lamp was installed, but it was superfluous because its lighting status is obvious from the rear.





The belt was being replaced with a plastic belt that could be bolted on, but when it was installed, it seemed to be too far away from the bar, so I went back to the original fiber belt.

It is rare that there are no streetlights or that the streetlights are so sparse that it is too dark, such as in some parks or on rural roads. On such roads, bright lights are certainly useful and perhaps even essential.

In the promotion of this type of high-intensity light, images taken at night on a street without streetlights are always used. The brightness looks more impressive that way, but on the other hand, when photographed on a street with streetlights, it is clear that they are not very effective in illuminating the street.

If street lights are bright enough, then a 400 lm light is useless when it comes to visibility.
On streets with a lot of pedestrian traffic, such as shopping streets, lights that are too highly illuminated are only light pollution. On such streets, this light should be turned off and only sub-lights should be used, or turned extremely downward.



Even a moped bike, for example, has about 1000 lm?�Even so, they do not cause light pollution because they can be set to low beam or louvered lighting.
Some products incorporate this into bicycle lights, but it is rare. Even in cars and motorcycles, if the optical axis adjustment is not good, modern LED lights are too bright and light pollution level.



What I built it for was for driving on backroads.
On backroads, most cyclists ignore the stop sign. I can only wonder if they lack imagination, but I'll leave the pros and cons of that here.

I mentioned that it is not so effective in illuminating the road surface on streets with streetlights, but it is effective for joggers and bicyclists on intersecting streets.
By illuminating the area around intersections, they can sense that a vehicle is approaching and many will begin to slow down. Of course, there are always exceptions.
Sometimes they will even stop, even if we have paused and the bicyclists on the intersecting road have priority. Of course, in such cases, we let the other person go first.



However, 400lm is not so powerful for this purpose, although it has a certain effect. In this respect, I think it would be better to have about 800 lm. Even a moped, for example, has more than that level of illumination to clearly see approaching vehicles at intersections.



In the calculation, if we assume that the battery power is 9.6 Wh without considering the loss of circuit efficiency, the run time is about 2 hours and 20 minutes assuming 100 lm/W. If we assume the circuit efficiency is 90%, the run time is about a little over 2 hours.

Of course, I actually tested it. The result was 2 hours and 30 minutes. I'm not sure if Cree LEDs keep this long even though they are more efficient, maybe the effect of having two lights is a factor.

What I found useful is that the illumination begins to drop slightly before the NiMH rechargeable battery drops to the proper end voltage due to a certain high load, and this lets me know when it's time to replace the battery properly.



postscript : My assumption of 100lm/W was wrong. Many Power LEDs are like that, and even Cree LEDs are like that when emitting 400lm with one light, but when emitting 200-300lm with this LED, the Vf decreases and the efficiency increases to 120lm/W, which is very high and unexpectedly long runtime.
This is exactly the effect of having two lights.



I will change the resistor in the constant current circuit to make it 600lm class. Runtime would be about 1 hour and 30 minutes, though.

As already mentioned in the circuit diagram explanation, the current can be increased or decreased by simply changing the values of Q2 and Q4. Since it is troublesome to remove them, we reduced their value by adding a resistor in parallel.
I have already mentioned that the theoretical illuminance is 606 lm.

I didn't notice a big difference from 400lm to see it in a room with the lights off. I guess that's about it.




In addition, the pilot lamp, which seemed unnecessary, was not removed but replaced with a resistor to reduce the illuminance. 1k resistor was used, so the theoretical value is only 0.4mA, and it is not really bright, but it is enough for a pilot lamp.



If you look at it in the absence of streetlights, you can see that it is brighter than 400lm.
I would think 400lm would be enough for such a road, unless you are going very fast.

On streets with streetlights, I did not notice much of a difference. I suspect it was due to light distribution.

Although the collimator has a half angle of 10 degrees, the boundary between light and dark is unclear and widely illuminated, making it difficult to compare the two. It may be more suitable for our purpose to find a collimator with a clear light distribution boundary than to shorten the run time by increasing the illuminance.



Maybe I should have put a switch on the resistor I added in parallel. That way I could switch between 400/600lm.







In the image, you can see that 600lm is brighter than 400lm because you can compare



It seems that 600lm would be more effective than 400lm in getting vehicles on intersecting roads to notice you on back roads. However, 600lm is still not enough to ensure this. It would have to be more than 1000lm, but it would be a little too much with two NiMH batteries.



Tested. The result was 2 hours and 10 minutes. At the point where the illumination seems to have decreased slightly, 2 hours and 10 minutes. It kept too long. Maybe it has not reached 600lm illuminance. I'm guessing around 500lm based on runtime.

Although it might damage the battery somewhat, I was prepared to over-discharge the battery and kept the light on, and at 2 hours and 25 minutes it began to blink slightly. It is an interesting feature of the constant-current circuit that it blinks when there is a shortage of power. I removed the batteries and put them through a discharger. One of the batteries had just dropped to the proper termination voltage, but the other still had room to spare. I knew it was holding too much.

It is possible that after the 400 lm test, the battery has been charged and discharged several times, which is a familiar phenomenon to RC hobbyists, and that the battery performance has started to reach its full potential. That could explain why the runtime has not changed much even though the illumination has been increased to 600lm. Even in that case, I think it would be about 500lm.

The fact that the illumination did not reach the calculated level may be a limitation of the constant-current circuit. I have not verified this yet, and I suppose it depends on the choice of transistors, but for high illumination, a MOSFET constant-current circuit should be used.


postscript : My assumption of 100lm/W was wrong. Many Power LEDs are like that, and even Cree LEDs are like that when emitting 400lm with one light, but when emitting 200-300lm with this LED, the Vf decreases and the efficiency increases to 120lm/W, which is very high and unexpectedly long runtime.
I don't think it's possible for 600lm to last a little over 2 hours, but 500lm or so is reasonable.

I want to make the boundaries of the light distribution clearer. The angle of illumination can be narrowed down a little more.
I attempted to change the light distribution by placing an aluminum pipe over the collimator and installing a convex lens at the end of the pipe. The focus can be adjusted by adjusting the height of the aluminum pipe. However, it is difficult to align the optical axis, and even if the optical axis is aligned, the irradiated light is spotty.
The convex lens is probably to be used in conjunction with a reflector. It is unlikely to work well with a collimator.

postscript: I mentioned that the angle of illumination could be narrowed down a bit more, but that is only for roads with streetlights. On roads without streetlights, it is scary to drive with a light with a narrower light distribution (angle of illumination). The same is true for clear light distribution boundaries. If we cannot see not only the end of the road but also both sides of the road, we will not feel safe.
I think a light suitable for bicycles should have a wide light distribution, and since the two LEDs are placed side by side, it might have been better if they were angled a little to the left and right.


Eaves were added (right image). An aluminum pipe, cut in two lengthwise and rounded at the tip, was placed over the collimator. The light emitted horizontally is weakened by simply turning it slightly downward.
The light distribution is not adjusted, but only the excess light distribution is blocked, so there is a loss of light intensity, but this is to avoid glaring light to oncoming cars and pedestrians.
This is installed and illuminated slightly downward from the height of the handle, it would block the strong direct light to the person.


Needless to say, I'm not trying to completely block out horizontal light.
Ideally, the light distribution would be adjusted by using a back-reflective system, but simply adding eaves can be quite effective.
The two images on the left show the left image at lighting without eaves and the right image at with eaves. Taken at the same location from a distance of about 20 meters. I've reduced the image too much, but the lighting without the eaves is too bright and inhibit spatial perception of the surrounding area.

From what I hear, glaring headlights are prohibited under German traffic laws. That might be light pollution.


I would like to use LEDs with light bulb color.
Nowadays, all streetlights are white LEDs, and they are always installed at intersections, even on back roads. Even if I shine a white LED light here, it will not stand out unless the light is very bright.
Therefore, if an orange light is irradiated by a bulb-colored LED, the irradiated light itself will be highly visible.
Likely, however, bulb-colored LEDs would be less efficient, and the runtime might be shorter.
Unfortunately, no bulb-colored LEDs are manufactured by Cree.


I want the light to be in the 1000 lumen class.
To do so, it is necessary to reconsider the selection of LEDs. The current single dcdc is not sufficient in terms of output. The runtime will be about one hour. For actual use, 3-4 cells or lithium-ion rechargeable batteries should be used.


I guess I shouldn't even think about updating it, except the eaves.



For example, 1200 lumens; it would be better to build a new one if it were to exceed 1000 lumens.
As for LEDs, there are commercially available ones in the 1000-2000 lumen class that are suitable for the purpose. The question is what to do with the battery. The choice is between 4-cell Ni-MH or 1-cell Li-Ion. In either case, the maximum power supply will be in the 1200-1600 lumen range.
The reason why I assume one cell in the case of lithium-ion rechargeable batteries is because it's easier to control and less risky.


As for dcdc, with that much voltage and power, it is possible to make a high efficiency and very stable step-up chopper using LMR62421 chips, etc. The output can easily be about 12V.
The power supplied would be about 10-15W, depending on the battery.


As for LEDs, I think Cree's XHP series is the best.
Illuminance ranges from 1000 to over 2000 lumens, and Vf ranges from 3V class to 12V class.
Most of the ultra-high-intensity lights on the market use this series (although the specifications include impossible figures such as thousands or tens of thousands of lumens).


Even at 1000 lumens, the power supply will be about 10W, so the parts must be carefully selected when building the constant-current circuit.
Exhaust heat will also be greater, so it would be better to use two LEDs.


Given the runtime, mode switching would be necessary.
If four AA Ni-MH rechargeable batteries are used, it can only hold 1000 lumens for about one hour. If four C batteries are used, it can last for two hours, but they are large in size and a little heavy. Even when using a single cell of 18650 lithium-ion rechargeable battery, a 3000-3500 mAh battery will last only about 1 hour and 30 minutes.
A Lo mode with about 500 lumens will last for 2 to 3 hours.


Considering the casing, I wonder if it would be better to use 4 AA type cells and switch between Hi mode with 1000 lumens and Lo mode with 400 lumens.



I took this photo on a street with sparse street lighting. I think the illumination is sufficient as far as visibility of the road surface is concerned.

The two left images and the two right images are taken at different points, with 400 lm on the left side of each and 800 lm on the right side (800 lm will be updated in the next chapter).
Unfortunately, 400lm and 800lm are not much different. Runtime is exactly halved, though.

The whiteness of the irradiated light depends largely on the quality of the road surface.

Ended up updating it, it will be 800lm/400lm light.
While experimenting with a constant-current circuit made with an OPamplifier and MOSFET, shown in the next chapter, I wanted to rebuild it. The purpose was twofold as follows:
-- First, to replace it with a slightly more efficient circuit.
-- Second, to allow switching between 400lm and 800lm.
Also, the transistors used in the constant-current circuits of the two transistors have already been disconstructed, and the stock of these transistors is running low.

The fact that both the voltage and resistance, which determine the current value, can be set relatively freely and that MOSFET can be used, offer advantages such as suppression of losses and ease of achieving higher illumination levels.

The update does not mean that only the circuitry has been rebuilt, but the entire head section, including the constant current circuit and Power LEDs, has been newly created without changing the power supply section, so it is easy to replace and revert back to the previous version.

It looks the same as before. The circuit itself has been revamped. The only difference from the diagram in the next chapter is that there is no monitor lamp in the circuit that sets Vref.
Since the illuminance is different, the LEDs used are different and the values of each parameter (program) are also different.

Incidentally, dcdc is close to its limit. If you try to increase the output of dcdc to increase illumination, the output will drop and oscillation will occur, causing the chip to heat up. I adjusted the output to just before it does not happen.


Here is a circuit of the update, including the battery monitor described later.



The increased illumination is evident to compared to past images.
The intensity of light is also impressive in the absence of streetlights.

The light would have to be directed downward a bit more to be annoying. As you can see on the wall in the left image, the upward light distribution is cut off by the eaves.


The irradiation light is quite noticeable where streetlights are not so strong (right images).













The irradiation light is not so noticeable in brightly lit areas (left images).
It would have to be over 1000lm to do that.



With 800lm it works 1 hour and 15 minutes, and with 400lm it works 2 hours and 15 minutes. Basically, I would use 400lm and switch to 800lm as needed.

I was hoping for a longer runtime at 400lm, but I think this is probably the result of adjusting the dcdc output to somewhere near the limit for this configuration. If I tried to increase the output beyond this, it would start oscillating even with sufficient battery power, and the MT3608 chip would also overheat, which would result in more losses.

I think 800lm is the limit with two C NiMH rechargeable batteries (2.4V, 4000mAh)


The batteries are two 4000mAh C NiMH rechargeable batteries.
When the power runs out and the voltage drops sharply, the light begins to oscillate. The oscillation is intermittent at first, but if it continues to be used as it is, the oscillation cycle will shorten and the light will become an extremely annoying light pollution light. Stopping the light when it starts to oscillate is very convenient, because the light can be taken out just before it reaches the appropriate termination voltage.

This means it's easy to avoid the risk of using NiMH rechargeable batteries until they are damaged by over-discharge.



Since the head part can be easily replaced, I made the head part as an option.
The cable connections were also made easier to replace using pin connectors, so all I have to do is remove two bolts.

The first head part I made is only backward-compatible, so I don't use it anymore; I disassembled it partly because I missed some parts such as LEDs and collimators, but the circuit parts assembled on the board can only be used for the same purpose.
Those parts were diverted to make a head part with different characteristics as follows:

-- Uses light bulb-colored LEDs
-- Low illumination level of 400lm/150lm for longer runtime for long rides.

High-intensity bicycle lights always use cool white LEDs. They do so because they are highly efficient, but that's no fun either.
On brightly lit streets, 150lm is sufficient as long as visibility is maintained. Even on roads without streetlights, 400lm is enough to see the road surface. This might be more practical for street riding.


Due to the compatibility of the LEDs and collimator, the light distribution is narrower than the previous products. However, compared to cool white LEDs, there is no eye-piercing glare.
Mode switching is done for 2 lights individually, so effectively 3 modes of 400lm/275lm/150lm


Runtime is 2 hours 20 minutes with 400lm and 5 hours 30 minutes with 150lm. I'm estimating that with 275lm it would be about 4 hours.

Unfortunately, this circuit does not oscillate when the time comes to change the NiMH rechargeable battery. This is not to say that the circuit will not oscillate, but by the time it does, it will be too late and the NiMH rechargeable battery will have been over-discharged.
Therefore, we incorporated a circuit that turns off the monitor lamp at low voltage (see below).



Even without streetlights, there doesn't seem to be any problem with visibility of the road surface.

Of course, its brightness is inferior to 800lm, so if the course has many streets without street lights, it is better to use 800lm head part.



Although irradiation light can be seen under streetlights, it is not very strong.
I guess it's easier to see because the color tone is different from the white LEDs in the streetlights.












Irradiated light is not very noticeable in bright streetlights.



I think it will be positioned as a light for visibility in town driving.



NiMH rechargeable batteries must stop discharging at the appropriate termination voltage. Any further discharge will damage the battery, resulting in loss of capacity and making charging/discharging itself impossible.

When using an 800 lm head part, it would start oscillating as it approached the termination voltage, at which point the battery could be replaced. It was a very convenient phenomenon indeed. The drop down occurs rapidly as it approaches the capacity limit with a certain load.
However, when the optional head part is used, it does not oscillate even when the termination voltage is below the termination voltage. Continued use until oscillation occurs would certainly damage the battery. The difference may be due to the difference in load, but the difference in circuitry seems to be the cause of this difference.

Therefore, a system to monitor the battery was incorporated. If the voltage falls below the specified voltage, the monitor lamp will turn off. The circuit diagram is as follows:

One transistor and two resistors. Simple circuit but fine-tunable and reliable.

The light turns off when the transistor base voltage falls below 0.6V. VR is essential to eliminate the influence of subtle individual differences. Adjust using batteries those have been lowered to terminal voltage with a discharger.

While ordinary battery checkers measure the voltage dropdown with a load of different capacities, this battery monitor is incorporated as the power supply unit of this light, so the load is the light and the battery is monitored when the light is turned on.


There is a multi-cell battery monitor to monitor individual cells, but it is an overkill system to use for such a device.
If you are concerned about the two cells being out of balance, just increase the detected voltage a bit to "take the battery out earlier". Then run them through the discharger individually.


Ideally, three or four LEDs would be arranged in a row and turn off sequentially as the battery is depleted, or use a full-color LED to change hue. A monitor that shows the degree of depletion like that would be good. It would be very good if the remaining usage time could be estimated.
This is because, as a security component, lights cannot be discontinued while riding at night. If I knew I was going on a long ride, I would carry a spare battery, but this is not always possible.

In the circuit diagram on the right, the LEDss turn off in order from LED1 as the battery is depleted, and LED3 also turns off when the termination voltage is reached. Assuming two NiMH cells, R2 and R3 are relatively small resistors, R1 is relatively large, and the VR should be 50 to 60% of R1 when combined with R2 and R3. The timing of turning off LED1 and LED2 is a matter of what level of remaining battery power to close up, and is focused by the values of R2 and R3. It would be possible to estimate the remaining usable time of each of LED1 and LED2 when they are turned off.
It's too much of a hassle to go this far, though (lol).

Personally, I use a small but 100lm sub-light in conjunction with it, so even if I take care of these NiMH rechargeable batteries, no problem with it for city riding.
This small sub-light and the tail light share a battery. I was concerned about the separate batteries for the lights I built this time, and even considered the possibility of having a single, powerful battery to share.
But I thought there was an advantage to having distributed and redundant batteries, since one could be discontinued.

I took the opportunity to add a monitor circuit to the sub-light battery as well.

Depending on the efficiency, it will provide 7-10W of power and 2-3A of current, and if two lights are used, 1-1.5A of current will be drawn per circuit.
For dcdc, the LMR62421 chip is used so that a stable voltage can be supplied. It could be OKL-T/6-W12N-C, which has higher output and is capable of not only up-conversion but also down-conversion.
As for the constant-current circuit, although commercially available products can be used, we decided to build a constant-current circuit using an OP amplifier and MOSFETs for the reasons already mentioned.
If the question is about efficiency, it is how to suppress the MOSFETs from generating heat for the voltage that exceeds the Vf of the PowerLED. The loss of R3 is not negligible. The heat generation must be suppressed while still providing sufficient illumination. Trial and error with a variable voltage regulated power supply and a breadboard is the only way to solve this problem.

The red LED is also used as a monitor lamp in the actual device, but I considered using this Vf as it is for Vref.
Normally, the current value would be variable by dividing the voltage with a potentiometer and making Vref variable, but I thought that if Vref is fixed to make various adjustments simple, it would be easy to adjust the dcdc output according to the Vf of the PowerLED since Vref does not change even if the dcdc output is changed.
However, if the Vf of the red LED is 2.1V and R3 is 2.2Ħ, a current of about 1A will flow, resulting in a large loss of 2W.
The output of dcdc to obtain sufficient illumination was about Vf + 0.5V of PowerLED, so Vref was set to 0.5V by voltage divider of resistors based on this voltage. 500mA if R3 is 1Ħ, and 1A current will flow if 1Ħ is added in parallel. Adding a SW to the additional resistor will switch the mode.

The constant current circuit shown in the figure is one circuit for one light. dcdc and Vref can be shared when using two lights and two circuits. it would be better to put a capacitor at the output of dcdc.


The illuminance should be calculated here. The PowerLED used is the Cree XHP50.x.
This series has three classes of Vf 3, 6, and 12V, all of which have a maximum standard of 18W and 2320 lumen. In other words, even the maximum illumination is 128.9 lumen/W, which is a high efficiency. If you drive at low power and low illumination, we can expect around 140-150 lumen/W.
To obtain 1000 lumen illumination with two lamps, 3.5W should be applied to each lamp. 1.1A is sufficient if Vf is 3.25V, and 0.6A is sufficient if Vf is 5.75V.
Since we will connect two 1Ħ resistors in parallel to R3, adjust Vref to 0.55 to 0.60V and let 1.1 to 1.2A flow through R3.

It may seem like a waste to use two PowerLEDs with 1000 lumens when a single PowerLED has over 2000 lumens, but if this leads to higher efficiency and longer runtime, then so be it.



I have all the materials.
But now I am beginning to wonder if NiMH is really the way to go. Wouldn't it be better to make it in a lithium-ion rechargeble buttery?

I had two 18650 rechargeable batteries and I found a battery box that looked good, so I bought it on impulse along with the control circuit. A good battery box is one that does not use dovetails and is unlikely to cause contact defects. This is an important point.
For control safety, the system is powered by a single cell, but if two cells are set and switched, they can be used in case of sudden battery failure.

I own 18650 rechargeable batteries and chargers from a reliable manufacturer. My personal concern is whether the control module is reliable when discharging.
It is not used for charging, and single-cell discharge control is also simple. And most ignition and explosion accidents are caused by the battery itself or the balancing circuit of the multi-cell control.

Lithium-ion rechargeable batteries not only have superior power to capacity ratio but are also easier to handle, so there is no reason not to use them if safety is guaranteed.


Therefore, each part should be modularized as needed. The battery part should be easily interchangeable between different sized cases. The case can be used with either 4 NiMH rechargeable batteries or 2 Li-ion rechargeable batteries.
Use of lithium-ion rechargeable battery with one cell, switchable between two cells in a case

The old light I diverted to 800 lumen light was so useful that I want to carry on the concept and make my own. The sketch looks like this;
Red line is a case of Li-ion, and blue line is a case of NiMH.

The power supply part and the light-emitting head part are morphologically separated, and the angle of the head can be easily adjusted.
The joint is a little too large. I would like to make it a little more compact.

The LEDs will be Cree XHP50.2 with a Vf of 5.75V, and the dcdc will be boosted using a TPS61088, so whether the input is 4.8V of NiMH or 3.7V of Li-Ion, the output will be stable.
In other words, the two types of battery cases and the voltage booster parts can be detached and replaced with pin connectors, etc.

The combination of a constant-current circuit and LEDs can be handled even if the LEDs are replaced by adjusting the dcdc output and Vref with a potentiometer.
However, it is too cumbersome to minimize losses with MOSFETs and R3, and I don't think there are any LEDs that would want to remove the Cree XHP50.2 and install it instead, so I'll fix this combination.

Install a voltmeter with LED display so that the voltage of the battery and the voltage after boosting can be switched and monitored. The monitor lamp will no longer be necessary.



The LED is a Cree XHP50.2 with Vf of 5.75V. The maximum power is 2546lm/19W. The maximum power is 3000mA, so the Vf would be about 6.3V. 1400mA would increase the efficiency to 1120lm/8.05W.
There are not many collimators that fit the 5mm square tip. Either the collimator or the base needs to be modified.

dcdc is a boost module using the TPS61088 chip. The output is very stable even with a slightly larger load and can output up to 10A. Put a capacitor on the output side just in case.

The control circuit for the lithium-ion rechargeable battery is a well-known Chinese module using TP4056 and DW01A.
In the December 2018 issue of Transistor Technology, this module is positively evaluated for its multifunctionality and safety. It is used exclusively for discharge, but the overcurrent control is 3A, so an output of 7.5 to 11.1W can be expected, which is suitable for the purpose of 1000lm.
For an AA NiMH 4-cell, an output of 8.0-9.6W can be expected.

This also means that with this power supply section, whether it is a NiMH 4-cell or Li-Ion, the limit is about 1200 lm. dcdc has much more leeway.

The maximum illumination by two LEDs has a potential of 5000lm, so the module with constant current circuit and LEDs fixed should be designed to bring out that potential. Although this power supply unit cannot be used and 5000lm is too high illumination, a mode that can be set up to 5000lm in stages by connecting to a high output power supply should be provided.
For 1200lm or so, a boost module using LMR62421 chip for dcdc would have been sufficient, but the use of TPS61088, which can output up to 10A, is to leave the possibility of extracting this potential.

Prior to fabrication, combinations of Vref settings and limiting resistors were tested in a spreadsheet to examine mode settings.
As a basic, it switches between 436lm Lo mode and 1076lm Hi mode with a single switch.
As a case of further exploiting the potential of LEDs, a setting of 2000lm/3000lm/4000lm/5000lm was calculated. However, adding 3-4 switches for this purpose is too much, so I can add 2 switches to switch between 2036lm/2996lm/3956lm. 2 switches can add 3 modes.

But this is only possible if I have a battery that can output 30W for 4000lm. For example, for 5000lm, 36W output is required.
Some lithium-ion rechargeable batteries have a large capacity and 2C discharge capability, and are capable of 30-40W output. Discharge control will also use dedicated high amperage ones, but it is questionable to what extent these can be trusted. Also, the runtime will be about 30 minutes, although the illuminance is not always used. A typical control circuit using 18650 batteries would require about four batteries. Multi-cell lithium-ion control is risky and a bit too portable.
The only possibility is to use it on an electric bicycle.



Here is the actual circuit diagram, not an experiment. The constant current circuit is the same as the experiment and the 800lm circuit, but with different parts and program values. For mode switches, one switch is shared by two light circuits. The resistors associated with it are individual in the two light circuits.


First, the head and joints part, which would be the most difficult hurdle to hand-building.

This turned out unexpectedly beautiful.
I tried on some replaceable head parts. I dare it to use the same size and the same fixation method.

There are two rotation points, and the head can rotate up and down nearly 180 degrees. Since it only needs to be able to rotate slightly downward, the one that joints to the power supply unit can be fixed. If this would make it more compact, that would be better.

The rest is just a matter of time.



Waterproofing measures should be carefully implemented.
However, the only thing that is a concern is the mechanical switch. Semiconductors have no problem even if they get wet, and if the entire board is potted, there is no problem at all. The board can be covered with epoxy adhesive. Of course, the surface of the chip that generates heat is excluded.
Potting is also an anti-vibration measure. Although it is highly unlikely to be damaged by vibration, potting the light's circuit board is indeed an appropriate measure.

I have long wondered why common commercial products break due to flooding, but I thought it was simply not possible to waterproof them in order to keep the chips that generate heat exposed. Or they simply skimped on the cost a bit. If it is hand-building, it could be an easy task to waterproof it, except for the surface of the heat-generating chip only.

As a rule, the pending mechanical switches are installed at the bottom of the structure. This alone has some effect against flooding, but it does not prevent moisture.
Therefore, we should apply contact grease. If the switch can be disassembled and injected with grease, it will be quite waterproof, but just applying it to the exposed parts will have some waterproof effect.
Contact grease should also be applied to mechanical contacts such as battery terminals and connect pins. Prevent oxidation, etc., and keep contacts permanently energized.

While contact cleaners and contact resurrectors are designed to eliminate defective contacts, contact grease is designed to prevent contact deterioration and is highly effective in preventing moisture and waterproofing. Upwardly compatible, so to speak.

As an aside, I would like to note a few things that concern me about commercial products.

Spec fraud is very common indeed. They are fraudulently specifying either illumination or runtime, or often both.
That is often the case with lights made in China, but it is not limited to them. Even now-famous manufactures such as GENTOS are overstating the illuminance and runtime of their products. On the other hand, there are Chinese manufactures such as MagicShine, which have been highly evaluated for their truthful specifications and high cost performance.
Of course, brands like Gaciron and CatEye do not lie in their specifications (in guessing). However, they are very expensive.

Some examples of what kind of fraud is common.

There are too many lights that can use AA NiMH rechargeable batteries or dry cell batteries that fraudulently claim both very excessive runtime and illuminance.
For example, there is a product manufactured by GENTOS that tout a runtime of 4.5 hours with two AA types in Hi mode at 400lm. Two AA-type lamps have a power consumption of only around 5.0Wh at most. Even with the high efficiency of recent LEDs, a drive with 400lm illumination will only last about one hour.
There is a product from the same manufacture GENTOS that claims a runtime of 12 hours with two AA batteries in Hi mode at 210lm. This is a big lie, not to mention a calculation. A review from an online retailer states that two 1900mAh NiMH rechargeable batteries only lasted 4 hours and 90 minutes when dry batteries were used. In this case, the illuminance is only about 100 lm when calculated backwards from the fact that it lasted for 4 hours.

There are quite a few products in the several hundred lumen class that use lithium-ion rechargeable batteries and do not indicate capacity, illuminance, or runtime. It is advisable to exclude them from the list.
One Chinese product claims to be able to maintain 800 lm for 4.5 hours on a 2600 mAh built-in battery, but if 800 lm is correct, it should only last for an hour or so. Moreover, it proudly states that it uses CREE XPG2 S3 LED, but the maximum illumination of this LED is only 500lm.
Many buyers will have difficulty evaluating lumen counts. If you can spoof them on the lumen count, it will be easier for them to balance the books at run time.

In the 1000lm and 2000lm class, a single cell of lithium-ion rechargeable battery is the limit of the power supply. Even at 2000 lm, a special high-capacity battery would be required. All products that state more than 3000 lm are considered to be false unless they have special power-supply specifications.
As you can see from the table in the next section, this class has a runtime of 1-2 hours even with a very high-capacity battery.
Many in this class also note specifications that are impossible. Many of them proudly state the model number of the LEDs they use, and then state specifications that exceed the maximum illuminance of the LEDs. They seem to think that they can sell their products if they write "military grade" on the label.
What seems troubling is that it's hard for buyers, who can't compare and evaluate, to see the lies, and it's hard for them to be reflected in reviews, etc.



The capacity (type), number of cells, and illumination (lumens) of the battery are used to calculate the possible runtime. AA and C are limited to nickel hydride rechargeable batteries. Dry cell batteries are excluded because their capacity varies due to load, and the higher the load, the more severe the drop in capacity. Also, cases that require a discharge of 1C or more are excluded and left blank.

The conditions were set as very high efficiency: 90% power efficiency with boost and constant current circuits, and 125 lm/W LED efficiency. If a specification is written that exceeds these results by more than 10%, it can be considered a spec fraud.




How to estimate the specs of the lights you have?
In most cases, the calculation will be based on battery capacity. If you are using nickel-metal hydride rechargeable batteries, of course, but the same applies to lithium-ion rechargeable batteries, as the battery capacity is probably the most reliable base. If you own a charger that can measure the capacity, that would be ideal.

Then, actually measure the runtime and calculate the illuminance.

The power efficiency of the boost and constant-current circuits and the luminous efficiency of the LEDs in such cases can only be guessed at on a hunch based on the grade of the light in question. The efficiency of each of the boost and current control circuits should be at least 90%, and if these two circuits are multiplied together, the efficiency will be at least 80%. Usually, we can assume that the efficiency of all circuits is about 90% through optimization.
If you're using Cree LEDs, you can set it around 120lm/W. With other LEDs, unfortunately, it's around 100lm/W.

This will reveal runtime and illumination. I doubt many of the products will meet the catalog specs.

The old light that was diverted was the HL-300 manufactured by Cateye and sold in the late 80s. The image on the right shows the light converted to a bullet-type LED instead of a bean bulb.

It was a popular product that sold well worldwide, and is still being traded as a vintage on eBay and elsewhere. The price is several tens of euros, so if I buy it from Japan, it will cost me about 10,000 yen including shipping costs. A French bicycle accessory manufacturer has even ripped-off of it (see next section). In response to a request from the American market, a product with a red lens was also made so that it could be used as a taillight.

Although halogen lamps sound good, they are so-called "bean-bulb" lights with very low illumination and are not suitable for visibility on the road surface. The large reflectors may be used for visibility.

The attractive feature is the morphological separation of the shaped and rounded power supply part and the light-emitting head part. The head can be moved 90 degrees up and down. It can also be removed and installed in the opposite direction. The belt can be rotated 90 degrees and set in four different directions, and can be attached to either the handle or the stay.
These, very high degree of freedom.

Well built internal structure.

The battery box is not simply divided into sections for inserting and removing batteries, but rather is intricately constructed so that bolts and switches do not obstruct the space. This makes the interior of the power supply unit compact, eliminating excess space. The distance between the two cells is only one thin plastic plate.

There is a metal terminal on the rotating part to conduct electricity to the head, but the structure is designed to keep this terminal from being exposed to the outside air. The head is not likely to be flooded even when used in the rain, but there is a hole at the bottom that may be used to drain water.

Such detailed construction may be unique to Japanese products of the 1980s.
The HL-700 (right image), the successor to the HL-300, was marketed, but it was simple and uninteresting in terms of structure. After the 1990s, it was the era of marketing and feasibility study, and although it would have been better if business feasibility had been emphasized, the emphasis became more on cost reduction, and industrial product design became obsolete.



The Soubitez AV-828, AR-829 is a rip-off of the Cateye HL-300. They had a similar lineup to Cateye, and were world famous as a bicycle light manufacturer.

The mounting position of the joint part is different from that of the HL-300, which reduces the degree of freedom. For example, it is not possible to locate upward the head part even if it is suspended under the handle.

The internal structure and switches are cheap to look at, and there is no sign of ingenuity at all. There is also a large amount of wasted space. The quality is on the level of made in China at one-coin stores.

They should have made a perfect copy of the HL-300 anyway, but I guess they didn't have the technology to do that.
French products are sometimes interesting, but they are basically cheap. Sales promotions are sometimes very clever, though.

I got those because those was on sale at a low price at an auction. Of course, it's intended for diversion and modification.



I have written extensively about the Cateye HL-300 diversion; the HL-300 diversion was the standard.

I was thinking of diverting the Soubitez AV-828 in much the same way. I also thought that if I changed the location of the head joint so that it was transplanted, I could mount it on the bike in the same way as the HL-300. That's why I bought it. I'm not a collector.

However, it would be smarter to make the head and joint parts by myself. Compatible with the HL-300 head and comparable in shape and quality(left images).
MT3608 dcdc can also fit inside, though just barely.

Then I began to worry about the poor quality of the power supply unit and began to think that it would be better to choose a commercially available battery box that looked good.
I could use a belt for mounting, but something like that could be diverted from one-coin shop product and would be sufficient.

That's why Soubitez AV-828, AR-829 are in storage. No, maybe throw it away.

To be diverted, if the product is not attractive enough I'm not sure I'd be inclined to do so.



If the modern lights had been attractive, there would have been no need to divert it.

How did this happen? Lights these days are all square-like tubular or cylindrical designs, with no internal part-module functions appearing in the shape.
Since industrial design is also the conceptualization of objects, I think it is attractive to the owner if the conceptual function of the object is manifested in its form.
Is there no such morphing of function in the design of the electronic age?
At least, new functions associated with high illumination are beginning to be created, and it is not impossible for these functions to become apparent in the form and appearance of the product.

Unfortunately, lights manufactured by Japanese companies seem to be capable of nothing more than increasing illumination intensity unnecessarily. The design and concept work of the lights seem to have become rigid.
Cateye is no exception, going so far as to install a TIR (collimator) in its ultra-high-intensity lights (right image), and continues to produce lights that are offensive to people and oncoming vehicles with impunity. They do not seem to have thought about the concept of a light for a vehicle that is appropriate for high illumination.
GENTOS, for example, has been promoting its own fraudulent specs to the point of excessiveness, which even Chinese distributors do not do.
Both are just eager to create an alibi for a price.

Some Chinese products are already superior in terms of ideas and systems for vehicle lighting, such as StVZO-compliant products, and there are even products that can switch between Hi/Lo.
Not all Chinese products are good, to be sure, but some good manufacturers such as Gaciron, MagicShine, and Lumintop are beginning to emerge. Perhaps Chinese manufacturers were able to be adventurous in their design and manufacturing during the period of a steadily rising economy. Even if not all of them have been successful.
This period coincided with the time when lithium-ion rechargeable batteries began to be used as a matter of course and bicycle lights became more highly illuminated. It may be fair to say that some Chinese manufacturers were chosen by the times because they had the opportunity to design products that responded to the need for higher illumination.

American companies such as LEZYNE are also a hot topic. One product complies with the StVZO but does not compromise on light distribution for vehicles. Visibility is high for the illumination level. LEZYNE is probably the most advanced manufacturer when it comes to light distribution. Some of their designs are unique and interesting.

The Lumintop B01 is a popular product with excellent value for money and the designer's commitment.
The light distribution would be StVZO compliant with the back-reflective system if not for flashing mode, and boasts a high illumination intensity of 850 lm in Hi mode. Its colour tone is yellow, more like the fog lights of the past than warm white, and it is said to be highly visible even in the rain. An attractive light with a high degree of perfection.

The B01 has a chronic condition. The switch becomes inoperative. As far as I know, I hear multiple instances of the same symptom.

Personally, I do not like the B01 very much. First of all, its design is not interesting.
The StVZO-like light distribution is much admired, but it is questionable whether it is really suitable as a front light. In town, it may not be a problem, and it is very desirable that it does not confuse oncoming traffic, but on dark country roads with no street lights, it will certainly shorten the visibility distance.
I like it if it is warm white, but I think yellow may be too particular. The catalogue specs say 850lm, but in reality the maximum illumination is only about 550lm.

I obtained such a B01 because it developed a pre-existing conditions and the owner decided he didn't need it. It's a rare LED, and might be interesting for a parts bin.



If we can correctly estimate how they were assembled, it is easy to correctly disassemble them.

It was surprising to find all the circuits on a single circular board 23 mm in diameter. The voltage booster circuit, constant current circuit, mode control, as well as charge control and discharge control of the lithium-ion rechargeable battery, are all integrated on this small board.

The switch did not work at all, but when the board was taken out and operated directly, it turned out to be perfectly normal.

If it can be dismantled to this extent, it can be diverted as we like.
It occurred to me to use NiMH batteries. The runtime will be shorter, but I want to eliminate as much risk as possible.
Nevertheless, let's repair it back to normal first.



Preexisting conditions will recur even if they are repaired to their original state. If you're going to repair it, you better be able to remove the cause of it.

During the disassembly process, it was observed that the board moves inwards when the switch is pressed. This is because the board has moved out of position, but the silicon waterproof cover on the outside of the switch makes it somewhat difficult to fix this board in place.
I don't feel like doing anything difficult, partly because I don't want to use this light on my bike.

A simple remedy would be to attach a tact switch, but doing so would impede waterproofing. Flooding in equipment using lithium-ion rechargeable batteries is to be avoided. A waterproof tact switch might be a good idea.
Therefore, I decided to discard the silicon cover and replace it with a new one made of conductive rubber. I tried to take out it from a calculator sold for one coin, but it was the right size, including the buttons, so I used it as it was. The feeling of operation is also very good.
Take steps to ensure that the main unit will not be flooded in the event of flooding through the cover.

It is not difficult to apply preventive measures for this chronic condition if they are taken before it develops.