100W LED Driver V1 – Patrick's Blog

Overview

The LED driver is made specifically for driving 100W LEDs. I couldn’t find another driver that fit my needs online. I needed something for an affordable price that was dependable. I would need to buy about 8 units.

I didn’t want to buy a really cheap 12V-34V boost converter because of the unknown quality of the components and circuit performance and it would be fun to build an entire board from scratch. I will design it to have dimming capabilities since I was going to use the driver for studio lights and for general lab lighting. The lighting in my lab is really bad, 1X 15W fluorescent light in the middle of the room.

Design Steps

I decided to design my own led driver in the summer of 2014. I started off researching how to drive high-output LEDs. The best way to drive them is with a constant current. As the LEDs heat up, the forward voltage of the led drops [1]. This leads to the current increasing through the led with a constant-voltage driver. This can lead to further heating of the LEDs and might result in a thermal runaway which eventually can melt internal bond wires. A constant current driver fixes the problem by varying the voltage to keep the current constant.

Loss Calculations

The second issue was whether to use a linear drive system or go for the more complicated switching driver. I did a few simplified equations to figure this out. My hypothesis was that the switching driver would be more efficient because the linear driver will need to dissipate the excess power. Using the formulas below and a white 100W test led [2][3], the power dissipation of a linear converter was calculated.

$P_{dis} =V \cdot I = (V_{in} - V_{out}) \cdot I_{out}$

$42\text{ watts} = (48\text{V}-34\text{V}) \cdot 3\text{A}$

$\eta = \frac{\text{Power Output}}{\text{Power Input}}\cdot 100%$ [4]

$\eta = 70.4% = \frac{100\text{W}}{142\text{W}}\cdot100%$

At this point, I didn’t have any components picked out so I must have skipped calculating the efficiency of the switching converter. Switching converters have switching efficiencies around 80-90%. I will include the calculation now since I have completed the design.

High-side switch conduction losses:

$P_{COND_HS} = I_{OUT}^2 \cdot R_{DS(ON)} \cdot \frac{V_{OUT}}{V{IN}}$ [5]

$0.234 \text{ Watts} = 3.0\text{A}^2\cdot 0.11\Omega \cdot \frac{34\text{V}}{48\text{V}}$

Switching losses:

$P_{SW_HS} = \left(\frac{V_{IN}\cdot I_{OUT}}{2}\right) \cdot (t_{RISE}+t_{FALL}\cdot f_{SW})$ [5]

$0.326\text{ Watts} = \left(\frac{48\text{V}\cdot 3.0\text{A}}{2}\right) \cdot (21.2\text{ns}+39.1\text{ns})\cdot 75\text{kHz}$

Power diode losses:

$P_{COND_D} = V_F\cdot I_{OUT}\cdot \left(1-\frac{V_{OUT}}{V_{IN}}\right)$ [5]

$0.585\text{ Watts} = 0.78\text{V}\cdot 3.0\text{A}\cdot \left(1-\frac{36\text{V}}{48\text{V}}\right)$

Inductor:

$P_{DC_L} = I_{OUT}^2 \cdot \text{DCR}$ [5]

$0.175\text{ Watts} = 3\text{A}^2 \cdot 0.0584\Omega$

Adding up all the component losses:

$\text{Total} = \text{Conduction } + \text{ Switching } + \text{ Diode } + \text{ Inductor}$

$1.32\text{W} = 0.234\text{W} + 0.326\text{W} + 0.585\text{W} + 0.175\text{W}$

$\eta = 98.7%= \frac{100\text{W}}{101.32\text{W}} \cdot 100%$

The figure of 98.7% seems kind of high, practical measurements are needed here.

Boost Vs. Buck

~~I decided on a buck configuration because of the increased efficiency over boost designs [6].~~ Efficiencies are similar for buck and boost converters if the output voltage from the boost converter is less than 3-4 times the input voltage. The wire coming from the power supply will carry less current resulting in fewer cable losses over a boost design. A 12V supply will need to output 100W/12V = 8.3A. A 48V power supply will only need to output 2.08A.

Driver Chip Selection

I did a search on Digikey for led driver chips. I limited the search to drivers that could survive up to 60V with a current capability of 3.5A. PWM also needed to be an option. I picked the LM3409HV IC because of the simplicity of the reference circuit while maintaining high current accuracy.

LM3409HV Calculations

The calculations for components is presented in the PDF document below.

Click to access calculations.pdf

Design Files

The files for this project are uploaded to github. I created the schematic and PCB with Eagle PCB Design software.

https://github.com/doublEElectronics/100W-LED-DRIVER

Results

The current on the 2.5A and 3.0A ranges are stable with a minimal current error. There is an issue with the 3.5A range, the system goes unstable and the current rises to about 4-5 amps.

The led driver does not get too hot when it is operational. I am using the PCB as a heat-sink. At 3.0A, the driver gets warm to the touch. I would recommend airflow over the PCB when operating at 3.0A.

Videos

Sources

[1] http://www.ledsmagazine.com/articles/print/volume-4/issue-8/features/driving-led-lamps-some-simple-design-guidelines.html

[2] https://www.sparkfun.com/tutorials/217

[3] http://www.wayjun.com/Datasheet/Led/100W%20White%20LED.pdf

[4] http://formulas.tutorvista.com/physics/efficiency-formula.html

[5] http://electronicdesign.com/power/fundamentals-buck-converter-efficiency

[6] http://electronics.stackexchange.com/questions/83817/which-has-better-efficiency-step-up-or-step-down-switching-voltage-regulator

4 Replies to “100W LED Driver V1”

Thierry says:

October 9, 2016 at 11:20 am

Hi,
Your driver is exactly what I’m searching for a dive lamp!
Do you still have some extra I may buy? If yes at what price?
Two will be great but if you have more I’m sure few friends will be also interrested 😉
Kind regards,
Thierry

1. patrick says:
  
  October 10, 2016 at 1:48 pm
  
  I don’t have any for sale right now, I have only made 10 of these. The project files are on Github if you want to assemble them. Keep in mind that the driver takes a 48V input. 48V batteries are really rare. A better option would be to use a step-up converter and a Lipo battery. Lipo batteries do not like to be discharged below a certain voltage so there has to be a voltage cutoff for safety.
  
Kim Le says:

May 19, 2017 at 4:03 pm

HI! I HAVE problem

I have sample design iled 1.4A fsw 500khz. But LM3409 and P_FET are hot. IF i have serial diode from Lm3409 to gate P-FET. So LM is not hot and FET verry HOT. I think forward current gate of FET to PGate of Lm3409

1. Patrick Rice says:
  
  May 19, 2017 at 11:06 pm
  
  Hi Kim,
  There should be a direct connection from the gate pin of the LM3409 to the gate of the PMOS. A diode in series (cathode at the PMOS gate) would prevent the charge from leaving the gate, leading to the mosfet always being on (overcurrent, overheating). The data sheet recommends that the mosfet gate capacitance be under 30nC when the switching frequency is over 300kHz. If the gate capacitance is too large, the LM3409 will overheat since it will have to supply more current per cycle (charge/discharge) than it can dissipate. Make sure that there is an adequate amount of copper on the PCB to dissipate the heat from the exposed pad.
  
  Hope that helps,
  Patrick