2
\$\begingroup\$

schematic

simulate this circuit – Schematic created using CircuitLab

I am working on something very similar to this, Adaptive LED Headlights – Model 8790 Adaptive 2 as a hobby project. I have controller part done, VID_20180728_195351.mp4, and this was the simpler part of the job.

Now I am trying to build a circuit which would light up up to five powerful LEDs depending on the lean of the motorcycle.

LEDs are Cree XPG2, they need 3.5 V and will run at ~1.5 A. The controller is ESP32 which gives 8  mA and 3.3 V. Motorcycle gives ~13.5 V and I'd like to avoid using voltage converters and keep it as simple as possible dropping 13 V to 3.5 V on transistor, power waste is not a problem since these LEDs light up only for a short periods of time.

I've tried BD135 as a switch but it cannot handle this load. Then TIP120 it's ok but in order to run at required current it requires ~4.5 V on gate.

ESP32 obviously doesn't have 4.5 V, so in order to stop burning further transistors I am looking for a way to select the right transistor for this project. I've run through many transistors datasheets but I am not sure what characteristics are important here.

On similar topics I saw that MOSFET transistors like IRF520 require 1-2 V on the base to be fully open, but I am afraid that voltage drop may not be as significant as on darlington TIP120 and it won't be possible to select an appropriate resistor between ESP32 pin and transistor's gate to have required 3.5 V at emitter.

CLARIFICATION: the system will have 10 lights in total, 5 items from each side. 1 to 5 leds may be fired at a moment depending on the lean angle.

The whole system runs on a massive aluminium heat sink(about 1 kilo) with active air flow.

\$\endgroup\$
  • 1
    \$\begingroup\$ Max sink/source current of a GPIO of ESP32 is 8 mA, iirc. \$\endgroup\$ – Long Pham Oct 29 '18 at 15:26
  • 1
    \$\begingroup\$ TIP120 is an NPN Epitaxial Darlington Transistor, so it needs a base current to turn on. I think you should have a read about basic electronic first. \$\endgroup\$ – Long Pham Oct 29 '18 at 15:30
  • 1
    \$\begingroup\$ MOSFETs have a gate and not a base. \$\endgroup\$ – Andy aka Oct 29 '18 at 15:37
  • \$\begingroup\$ Sounds like you are using the transistor in an emitter-follower configuration - LED between emitter and ground. It is more common to ground the emitter, and put the LED between the collector and positive supply, controlling the current with a resistor in series with the LED. \$\endgroup\$ – Peter Bennett Oct 29 '18 at 15:38
  • \$\begingroup\$ Basically, look for logic level mosfets or mosfet drivers. But trying to use the transistor or fet to drop 9.5V at 1.5A or 14 Watts doesn't seem wise. \$\endgroup\$ – Passerby Oct 29 '18 at 15:41
1
\$\begingroup\$

This is a typical automotive application.
The way this is normally done is the LEDs are wired in series and you use a boost LED driver.

It is unlikely you can drive these beyond 1 amp as they run very hot. At 1A you will need significant thermal management.

XP Gen 3 (XP-G3) are now available which are more efficient.
XP-G2 142 lm/W, Vf = 2.9V @ 350 mA
XP-G3 158 lm/W, Vf = 2.73V @ 350 mA With the High Efficiecy XP-G2 or XP=G3 the Vf ≈ 2.9V @ 1A.
5 x 2.9 = 14.5V

150 lm @ 350 mA = 143 lm/W
If you could increase the current and maintain 85°C Tj then
At 1000 mA we get 2.5x more lm: 375 lm @ 1000 mA = 125 lm/W
At 1500 mA we get 3.4x more lm: 510 lm @ 1500 mA = 113 lm/W
At 2000 mA we get 4.2x more lm: 630 lm @ 2000 mA = 105 lm/W

Example:

This MIC2282 boost LED driver takes an input up to 15V.
It is very simple with only 5 external parts.
The SNS resistor sets the current.
It can drive between 5 and 10 LEDs at 13.5 Vin

enter image description here


Suitable Transistor

If you need to turn and off each individual LED then I would use a automotive grade N-channel MOSFET with low RDS(on)

The Rolm RUL035N02FRA 20V, 3.5A, 43 mΩ RDS(on)
This MOSFET is classified for Class III life sustaining medical devices
and AEC Q101 qualified.


Running single LEDs below 3V from a 15.5V supply requires either a switching regulator to drop the supply voltage or big (i.e. 15+ watts) power resistors for each LED.

Heat is going to be a problem.

At 1500 mA that's 4.5W per LED and over 16W per resistor.
A massive heatsink is not going to solve this 200W (10 LEDs) issue alone. It is not easy to design a PCB that can spread the heat from each LED to the heatsink fast enough.

Because I would not want my headlights going out on my bike at 100+ mph, reliability is important.

Heat is a major factor in reliability of semiconductors. Step down switching regulators are required (IMHO). High efficiency equates to low heat dissipation and higher reliability.

The LMS3635 is a 5.5A 92% efficient buck step down regulator.
1 or 2 of these in parallel would do.
Assuming no more than 5 LEDs are on at the same time.
To get 2000 lm from five XP-G3s you would a little over 4 amps.

\$\endgroup\$
  • \$\begingroup\$ not sure what is your exact idea \$\endgroup\$ – Serg Fillipenko Oct 29 '18 at 16:38
  • \$\begingroup\$ use 5 XP Gen 3 per light module, gives 2x5x5=50 leds? or use a converter module per light module what gives 10 converters and 10 leds? \$\endgroup\$ – Serg Fillipenko Oct 29 '18 at 16:39
  • \$\begingroup\$ 1 to 5 leds may be fired at a moment depending on the lean angle? So you need to control each LED individually? What I proposed was a single driver that drive 5-10 LEDs. No on-off, just on when powered with 13.5V. If they all are not going to be on continuously, how are you going to control them? What does 2x5x5=50 leds mean? What is the second 5? \$\endgroup\$ – Misunderstood Oct 29 '18 at 16:51
  • \$\begingroup\$ if you take a look at the video in question you'll get the idea. yes, 10 leds or groups of leds will be controlled separately, this is where all the story with transistors starts. 2x5x5 means 2 sides, 5 leds in the group, 5 groups on each side. please see the link to the original light I am trying to copy. I am working on those side light elements which light up when a motorcycle leans. \$\endgroup\$ – Serg Fillipenko Oct 29 '18 at 18:11
  • \$\begingroup\$ Yesterday the video would not play. Tried it again today and now I get it. So this is supplementary lighting in addition to the factory headlights? Now I'm not sure how serious the heat issue will be if the LED will only be on intermittently. I was thinking they would be on continuously. If only on during a turn then that's just 1 second at 30 mph. \$\endgroup\$ – Misunderstood Oct 30 '18 at 18:47
0
\$\begingroup\$

Ok one problem I'm immediately seeing here is that you're dissipating all the excess voltage in your transistors. This is why they're burning up.

You can easily calculate the power dissipation in the transistor by multiplying the voltage drop(~10V) by the current(1.5A). This gives 15W, which to put it simply, is too much.

A simple solution is to use a power resistor in series with the LED; and switch the transistor fully on. This will dissipate most of the power in the resistor, instead of in the transistor.

A more efficient solution would be to use a smaller resistor, and then use PWM to limit the current to an appropriate level.

Another solution would be to use a switching regulator to step the voltage down to 5V, then branch that off to the various leds. This will be much more efficient than using 13.5v directly, and it will keep you from burning up your transistors. The supply will need to handle 5leds x 1.5A = 7.5A. Search switching regulator on amazon, you'll find something.

Side notes about mosfets:

Mosfets are great. They're easier to use than transistors in switching applications. The trick is to look in the datasheet for a chart of Rds(on) vs Vgs. Most mosfets will have this, but some don't. Find the resistance of the mosfet at 3.3v, from that you can easily calculate the power it will dissipate at 1.5A.

There are some tricks you can use to boost the gate drive voltage and get a lower Rds(on), for example using a gate driver or an additional transistor. But that does add complexity.

One more thing, just because a mosfet has an Vgs(on) of 2V does not mean it will have a low resistance at 2V. It starts conducting at that point, but you might have to get it up to 5V or more before the resistance becomes acceptably low. Always check the chart.

\$\endgroup\$
  • \$\begingroup\$ the most explaining comment so far. is there such a thing as MOSFET array or integral MOSFET? so instead of soldering 10 transistors, 10 power resistors, probably 10 usual resistors, I can solder 1-2 integrals and the rest? \$\endgroup\$ – Serg Fillipenko Oct 30 '18 at 20:16
  • \$\begingroup\$ people write that PMV16XNR can be a good choice for 3.3v logic \$\endgroup\$ – Serg Fillipenko Oct 30 '18 at 20:20
  • \$\begingroup\$ You probably will have to use individual components. There are arrays of mosfets and resistors, but they're specialized parts and not usually designed for high power applications like this. \$\endgroup\$ – Drew Oct 31 '18 at 0:40
  • \$\begingroup\$ The PMV16XNR looks pretty good. The Rds(on) is very low at 3.3V, and the current rating is high enough. The only worrying thing is that the max Vds is only 20V. The 13.5V from the battery might have spikes in it that go over that. A 40V mosfet would probably be better. \$\endgroup\$ – Drew Oct 31 '18 at 0:47
  • \$\begingroup\$ PMV16XNR are super tiny. almost impossible to solder by hand \$\endgroup\$ – Serg Fillipenko Nov 4 '18 at 21:08

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.