Description

This is an hysteretic buck LED driver in a pill form factor. It is built around the Infineon ILD4001 hysteretic buck controller.
The advantage of this driver is the building simplicity and the ANALOG control capability, without forgetting the higher power capability.

Can be driven directly from any 2.5V, 3.3V, 5V or higher logic level, with analog and PWM signals, allowing precise light adjustment.

Due to the high input voltage range up to 40V, with full current up to 1000mA, can be used to drive up to 40W LED based luminaires and also being powered with battery packs without intermediate voltage conversion, keeping high efficiency. By default, is set at maxium of 605 mA, but analog and digitally dimmable.

Again, it is failsafe, providing over-heating, short and open circuit protection. It is made with the same safety philosophy of Glighter-S. But is still cheap, smaller and purple, like a tiny amethyst crystal.

Details

This project is a simple, yet efficient buck LED driver, or capable to provide basic constant current capabilities. Its existence is on the tail and for the same reasons of the Glighter-S module, but this was made to satisfy a compromise on another perspective: cost reduction and analog control vs. fixed frequency operation. This gives the need to pay more attention on filtering the output and is addressed in the project logs herein.

But has the enormous advantage to be so simple to build and being cheap with the same reliability of Glighter-S. I can now experiment more freely and use them as a jelly bean part for my LED projects.

In the picture is quite clear that the module is smaller, since requires less parts, due to its intrisinc different, yet simple, technology of an hysteretic buck converter.

Project summary

Technology: hysteretic buck regulator
Input voltage: 2.5V+V_LED to 40V (V_LED the forward voltage of the LEDs)
Output voltage: V_LED to 39V
Sourced current: up to 900 mA (set by default to 640mA)
Size: 13 mm x 21 mm (0.51 in x 0.83 in)
Efficiency: up to 95%
Operating Temperature: -40°C to 60°C
Safety features:
- Short circuit immune
- Open circuit immune
- Safe state (off) when no control is applied
- Overtemperature protected (T < 120 °C)
Control features:
- PWM control: voltage swing from <0.4V to >2.5V (safe up to input voltage supply), max PWM frequency 20 kHz
- Analog control: voltage swing from 0.4V to 2.5V for linear output current adjustment
- OFF: voltage < 0.4V
- Fully ON, max current set with sense resistor: voltage > 2.5V up to supply
Other infos:
- Input impedace EN/REG pin: 39kΩ
- Max contrast ratio with PWM: 100:1
- Max contrast ratio with analog dimming: 6:1
- Max combined contrast ratio: 600:1

Source files

Any external link is shown in the project's links.

How to use it

The driver has 3 input pins

VIN: the input voltage
GND: ground
EN/REG: the control pin (analog+PWM) and enable pin

Any board handling any type of digital signal can be used to drive the EN/REG pin. Any signal lower that 0.4V result in a full off driver; any signal higher that 2.5V fully turns on it. Analog signal can be safely used.

The output is made by 2 pins

LED anode
LED cathode

Just connect the LED respecting the polarity, or the LED can fail.

Setting the current

Assuming that the input voltage is respected, the driver can light any LED which is designed to take around 640mA, or lower, if proper analog dimming is provided. By changing the sense resistor, the maximum current can be increased up to 1A. Can also be reduced to increse the analog resolution of the output current.

I made this resistor fixed (but interchangeable) for a security reason: avoid wrong configurations. The driver uses a sense voltage Vsense = 0.121 V. The maximum output current shall be sey by I = 0.121V / R.

Project Logs

Collapse

LED current estimation
Enrico • 08/23/2018 at 19:52 • 0 comments

So far so good? Maybe! But how can you practically measure the current in the LED to demonstrate the ripple reduction?

I don't have a current probe for my scope and neither I can afford it. Secondly, measuring in a conventional way required a sense resistor, which would alter the results, and I should manipulate too much the already delicate measurements to compensate this effect and get the correct result.

KISS my LED

I opted for Keeping It Super Simple, you Stupid (this acronym is always different each time a read it). Here the setup, which I was staring at with sun-glasses, hopeless, trying to find a solution.

Then I realized that I had an OPT101 somewhere. I still can measure the light. But this is a somehow slow sensor. From the previous log, with my setup with a single LuxeonRebel LED, I had an operating frequency somewhere 300 kHz. Different supply and analog dimming will affect this, since it does not have any oscillation: is just the speed of the inductor to reach the thresholds!

The OPT101 was used with external programmed gain:

With a useful table shown here:

Since I don't have any capacitor with me and 1M of gain bring a 23 kHz of bandwidth, a gain 500k could set the game of a bandwidth between 44 kHz and 23 kHz, without making stability issues. Therefore the only output component was a 500 kΩ resistor and the crude schematic look like this:

Here I don't care about the quantitative ripple reduction in absolute way, but more in a relative manner with respect to the solution with no ripple attenuation.

One last thing was to be sure that the LED could provide sensible light variation w.r.t. current variation. In my case, the operating point of the LED was somewhere near 400mA. The LED was a LXW8-PW40, with the following characteristics:

Still pretty linear I would say.

Finally some measurements

The results are showing an output from the photosensor of 35 mV_p-p which is amplified by 500k times.

This is my reference point. Here with 300 Ω of sense resistor, reference voltage of 116 mV, results in a I_set = 386 mA. If the regulator provides +/-15% of ripple current, then 35 mV_p-p will corresponds to the 32% of ripple measured in the previous log, where in current is 150 mA_p-p. Applying 2.2 uF reduces this to 25 mV_p-p.

With 6.6 uF this reduces down to a value between 11 mV_p-p and 13 mV_p-p.

Now I have a comparison value: mathematically speaking, with 6.6 uF the ripple shall be 27 mA_p-p. From the original 150 mA_p-p this is 5.5 times lower. Here, 11 mV_p-p of ripple corresponds to a reduction of 3.2 times from the original 35 mV_p-p of the first reference acquisition.

I would say that considering the non-linearities of the flux/current characteristics of the LED, parasitics of capacitors, their tolerances, imperfections in the LED dynamic resistance and scope acquisitions far from perfect, this is quite a good result.

I would say that what was theorized in the previous log seems correct: a ripple of +/-2.5% can be achieved with a ceramic of 7.2 uF.
LED current ripple reduction/optimization
Enrico • 08/21/2018 at 22:08 • 0 comments

The board carry a space for 2 output capacitors. This is done to leave space to put as many caps as possible, to experiment with EMC problem solving and accurate lumen output, where needed.

In the first assembly I put an output cap of 1uF to filter out some mid-frequency disturbance and reduce a bit the output ripple. An additional 100nF may be good as well.

Initial measurements

The controller is an Infineon ILD4001, and states a ripple of +/-15% in the current. With a setup at an I_set = 386 mA, I have measured the ripple, according to a sense resistor of 300 Ω. The upper current peak is 493 mA and the lower of 343 mA, with a delta of 32%.

Here how appears the current measurement the current ripple, measured across the sense resistor, in purple:

This is the overall current, which according to the following schematic, is the one flowing through R_sense:

To filter out the ripple, is needed to "steer" the AC component out of the LEDs, which translates in putting a parallel capacitor across them. If the LED has a given dynamic resistance according to its bias point and the capacitors a certain ESR, the (small signal) schematic looks like this:

From which the current divider created let me find the ΔIf, and subsequently, from the equivalent capacitor impedance, finding an equation which provides the output ripple in function of the capacitor's impedance:

If the ΔIf reduced is, for example, from +/-15% of the initial to +/-2.5%, from the above formula:

Where Rd = 0.35Ω and the ΔIf−reduced in this use case is +/-25 mA. So, with an original 30% of ripple (+/-15%), to go down to 5% I needed at least 7.2uF.

Two low cost MLCC 0805 4.7 uF, 50V capacitors in the two available slots will do the job in all the LED configurations. Again, I did not put them (for now is the standard BOM) since it may never be an issue. But the user has the freedom to tweak it and here I just gave some suggestions.