LED CC booster design – feedback needed

[3dgeo]

Hi,

designed 9x15 3528 UV LED matrix driver circuit.
Analog circuits is not my strong side, so I would like to get some feedback on inductor and diode (efficiency is not extremely important as this is not battery powered project).
Inductor: YNR8040-470M or SLD10D40S-330MTT?
Diode: BAT46W or BAS100CS-AU_R1_000A1?
Booster IC: XL6008E1

At the moment I don't know how high I can push LEDs and still maintain reasonably low temps (45x75mm matrix size, passive cooling, 100*100mm aluminum PCB). To limit and balance current I will be using 64R 1% resistor for 20mA(180mA total) and two in parallel for 40mA(360mA total) – I assume setup like this will balance current between 9 lanes well enough? Matrix needs to be turn on/off, so to save extra BOM I designed matrix to be off with 12-24V input (way below 37.5V(2.5*15) forward voltage). Input and output will have few 10uF caps (double them up for output to use 50V rated caps).


Also 3.3v <100mA is needed from the same 12-24V input voltage (different double sided PCB). I assume I can just reuse same diode even tho it will be way ovekill, but how about inductor? What value should I use to get stable output for MCU and other logic ICs?
Regulator IC: ME3116AM6G


P.S. this will end up as an open source project...

[Benta]

A very general comment:
you seem to presume that all LEDs have the same forward voltage. Are they selected for this?
If not, I'd recommend active current control per LED string.
Like this:

[mariush]

I'd encourage you design it as a step-down / buck driver or as a linear driver.

Your leds have forward voltage of 3.3v typical, up to 3.6v .. so let's say you design this to use a regular laptop adapter (18v...20v - and 24v would also work but less efficient) then you could have chains of 5 leds to form chains giving you 3.3v x 5v = 16.5v leaving 1.5v+ available for the led driver.

So each of your columns could be either 3 channels on a led driver, or you could parallel the 3 chains of 5 leds to a channel.

with chains of 5 leds, and 20-60 mA per led, you're looking at 27 groups of 5 leds so 27x 20 - 60 = 540mA - 1620 mA  (a 18..20v 3A lcd monitor/ laptop charger would be perfectly fine for this)

You have led drivers like AL8843 which can do up to 3A of current and allow you to set the max current through one resistor, and then use either a potentiometer to adjust brightness, or a pwm signal from your microcontroller
AL8843 :  https://www.digikey.com/en/products/detail/diodes-incorporated/AL8843QSP-13/10668306

another similar driver : https://www.digikey.com/en/products/detail/diodes-incorporated/PAM2863ECR/4033265

[3dgeo]

A very general comment:
you seem to presume that all LEDs have the same forward voltage. Are they selected for this?
If not, I'd recommend active current control per LED string.
Like this:

Well, as LEDs will be from a same reel I assume that differences between 9 lanes with 15LEDs in series shouldn't be big issue... ?
My design is a constant current as booster IC vref is 1.25V and 62R will give 20mA@1.25V per lane. Tho I'm not sure about other lanes...
Your suggestion seems a bit too BOM/PNP intensive, if lane balancing will be an issue I think I'll go with CC LED driver on a low side.

I'd encourage you design it as a step-down / buck driver or as a linear driver.

Your leds have forward voltage of 3.3v typical, up to 3.6v .. so let's say you design this to use a regular laptop adapter (18v...20v - and 24v would also work but less efficient)...

You probably missed that input is 12V to 24V. Even if input range would be 18-20V I'm also having hard time understanding – I assume you suggesting paralleling  27 lanes of 5 LEDs, but in that case there is still a lane balancing issue, and in your case this may be *3 more problematic as there would be *3 more parallel lanes and *3 current.

I was considering 16lane LED drivers, but due to dynamic input voltage they are not best choice in this case in my opinion.



Any feedback on inductor or diode?

[viperidae]

As long as an led doesn't fail short circuit in one of the unmonitored strings. That would cause excessive current in the remainder of the string.
An open circuit failure in the current monitored string would cause excessive voltage output of the boost regulator, since the top of its feedback circuit will be open.

[mariush]

I have a bit more time so let me say my thoughts.

Your last sentence in the first post is "P.S. this will end up as an open source project..." 
Considering this, you should put some thought into the design and try not to use some parts that may be hard to obtain and may have specific pinouts and requirements.

For example, rather than using a no-name part available only on LCSC like that Microne ME3116, maybe pick a part that's 20-30 cents more expensive but it's mass produced, available at multiple stores and maybe requires less parts around it and uses a more standard footprint.
You could pick a part that's doing synchronous rectification, so you won't need a diode and boostrap capacitor and compensation network and all that. After all, it's a plain 12-24v to 3.3v buck regulator chip, they're not hard to find.
I'd even use a jellybean 34063 before using such IC - only the inductor would be slightly more expensive.

As for the other ... if it's gonna be an open source project, I'd be reluctant to do something in my project that runs higher than 24v. That's just me... it's also not necessary to go to higher voltages, since it's just a matrix of leds, they can be arranged in various ways.

Boosting 12..24v to 51v 360mA is kinda wishful thinking, with that chip. 

The chip claims up to 92% efficiency, but that's doing 12v to 18v boosting - just 1.5x   Boosting 12v to 51v is 4.25x times the input voltage and if it's even possible, it won't be very efficient, I could be wrong but I think you'll get maybe 80-85% efficiency at best.  In addition to that, you'd have to use a diode rated for at least 75-100v, same for electrolytic capacitors on the output (at least 1.5x the output voltage recommended which makes it more expensive - you can use 35v electrolytic or even polymer capacitors on the output of a 24v boost regulator)

So if that's the case .. with 51v x 0.36A = 18.36  watts,  the chip would probably pull 100 x 18.36 / 85 = 21.6 watts from input  and at 12v that means your power supply must do at least 2A of current.
That rules out basic wallwart adapters doing 12v 1-1.5A ... and moves you into the laptop adapter style bricks where you might as well go for 16.5v / 18v / 20v adapters which are more typical.

It would make more sense to design for 12v..20v input and boost to 24v and then configure your led driver chains to be just the right length, and parallel 2-3 of these at a time.

Your leds have a typical forward voltage of 3.3v but they go to max 3.6v and you have 9 x 15 leds or 135 leds ...
Assuming the leds are all close to 3.3v forward voltage, you could have 7 leds x 3.3v = 23.1v, leaving around 0.9v for the current limiting resistor or for the internal transistor of a led driver that limits the current, whatever you use.
But, it would make more sense to have groups/chains of 6 leds, because that gives you  6 x 3.3v .. 3.6v = 19.8v .. 21.6v

With 6 led chains, you could use 18 x 8 instead of 9 x 15  = you'd have 144 leds instead of 135 leds but with just 8 columns you could use a basic 8 channel led driver instead of just resistors, and give you the ability to set a maximum current for all channels using a single resistor, and then adjust output current using i2c/spi commands to the led driver (or analog / pwm signals)

Alternatively ... take 12v...24v in, and use a buck regulator to output 11..12v and use groups of 3 leds (3.3v x 3 = 10v , 3.6v x 3 = 10.8v).. you'll have 45 groups of 3 leds or 48 groups of 3 leds if you go with 18 x 8 arrangement.

This way if some led blows up or dies shorted, worst case scenario just one group of 3 leds would be out of circulation or dying due to extra current, and the voltage regulator won't boost to higher voltage than the one set.

[3dgeo]

As long as an led doesn't fail short circuit in one of the unmonitored strings. That would cause excessive current in the remainder of the string.

Issue like this is very unlikely in my opinion, and even if it will occur at that point entire unit will be unusable anyway.

An open circuit failure in the current monitored string would cause excessive voltage output of the boost regulator, since the top of its feedback circuit will be open.

Good point, that could be an issue. I guess MCU could monitor output voltage to prevent that... Or may be a ~53V zener diode to feedback pin?

I have a bit more time so let me say my thoughts.

Your last sentence in the first post is "P.S. this will end up as an open source project..." 
Considering this, you should put some thought into the design and try not to use some parts that may be hard to obtain and may have specific pinouts and requirements.

I did, most parts these days are hard to obtain.

The chip claims up to 92% efficiency, but that's doing 12v to 18v boosting - just 1.5x   Boosting 12v to 51v is 4.25x times the input voltage and if it's even possible, it won't be very efficient, I could be wrong but I think you'll get maybe 80-85% efficiency at best.  In addition to that, you'd have to use a diode rated for at least 75-100v, same for electrolytic capacitors on the output (at least 1.5x the output voltage recommended which makes it more expensive - you can use 35v electrolytic or even polymer capacitors on the output of a 24v boost regulator)

I'm fine with 80% with reasonable temps. In main post I did provide 2 links to 100V rated diodes and I've addressed caps too (only ceramic caps will be used).

Your leds have a typical forward voltage of 3.3v but they go to max 3.6v and you have 9 x 15 leds or 135 leds ...
Assuming the leds are all close to 3.3v forward voltage, you could have 7 leds x 3.3v = 23.1v, leaving around 0.9v for the current limiting resistor or for the internal transistor of a led driver that limits the current, whatever you use.
But, it would make more sense to have groups/chains of 6 leds, because that gives you  6 x 3.3v .. 3.6v = 19.8v .. 21.6v

With 6 led chains, you could use 18 x 8 instead of 9 x 15  = you'd have 144 leds instead of 135 leds but with just 8 columns you could use a basic 8 channel led driver instead of just resistors, and give you the ability to set a maximum current for all channels using a single resistor, and then adjust output current using i2c/spi commands to the led driver (or analog / pwm signals)

Alternatively ... take 12v...24v in, and use a buck regulator to output 11..12v and use groups of 3 leds (3.3v x 3 = 10v , 3.6v x 3 = 10.8v).. you'll have 45 groups of 3 leds or 48 groups of 3 leds if you go with 18 x 8 arrangement.

I fail to see benefits in your suggestions in terms of BOM and efficiency – multi lane LED driver would most likely be linear + regulation circuit is still needed, I appreciate your post tho.

[Benta]

Well, as LEDs will be from a same reel I assume that differences between 9 lanes with 15LEDs in series shouldn't be big issue... ?

That's called wishful thinking. You have zero knowledge of manucfaturing, packaging, number of batches, number of wafers etc.

[mariush]

I did, most parts these days are hard to obtain.

*34063* : 600k in stock on Digikey, millions at avnet, farnell, mouser , RS Components , tme.eu : https://www.findchips.com/search/*34063*
*33063* (pin compatible with 34063) : same story

For 3.3v at 100mA , you don't necessarily need 450kHz switching frequency and some obscure manufacturer.

I fail to see benefits in your suggestions in terms of BOM and efficiency – multi lane LED driver would most likely be linear + regulation circuit is still needed, I appreciate your post tho.

You don't need to use drivers but they can open the doors to adding features at some point.
If the output voltage is close enough to the sum of the voltage drops on the series of leds, then a simple resistor would be enough as a current limiter.

BUT, the forward voltage of leds will drift with temperature so as the leds get warm the current going through the leds will increase because usually as the leds get warm the forward voltage will decrease.

At 25c ambient, the forward voltage may be 3.3v, at 60c the forward voltage may be 3.2v for the same current.

If a led dies shorted, then you suddenly have less voltage drop so the current on the remaining leds goes up, to the point where you could damage the whole series of leds.

Ex.  15 leds, 3.3v forward voltage = 49.5v and your input voltage is 51v ... resistor to limit to 40mA  : V- Vforward = current x R = > R = (51-49.5) / 0.04 = 37.5 ohm 

One led dies shorted and now your forward voltage is 14 x 3.3 = 46.2v  so  (51v - 46.2) = ?  x 37.5 ohm  -> current = 0.12 A or 120mA - congratulations, you'll have a whole column of 14 leds much brighter than the other columns, until the leds burn out because they can't handle the much higher current amount.

Leds can also fail open (bonding wire breaks or gets desoldered) and then your whole column of 15 leds fails to light up... One could live with 3 non-working leds, or even 6, as it's just a lower amount of UV in that local area, but a whole column of 15 leds would make your product less usable.

It's just a matter of arranging the traces on the PCB to account for making groups of 3 or 6 leds or as many leds as you decide and  then optionally parallelling them 2-3 such groups to get a

Led drivers were suggested as a way to add features, like adjusting intensity, or turning off individual columns. It's not a must. You could easily have one resistor for each group of leds in series, resistors are cheap and you can even squeeze them between leds if you want.
I'd do 1 per 3 leds and supply 11v-12v to the leds but you can do whatever you want. If one led dies shorted, then just the 3 leds in series will be damaged and the other remain functional. If one dies open, again just the 3 leds will stop working.

Benefit of led drivers is that they account for the forward voltage changing and they maintain the same amount of current through the leds.. each channel independently controls the current even in a 8/16 linear led driver chip. You set the maximum current and then some drivers can reduce current per channel through pwm or analog input (a potentiometer)
But if you want something plain, then you can just use resistors and set the maximum current fixed.

For example, maybe you'd want to make it more fool proof and add over temperature protection ... add a temperature sensor on the pcb with the leds - if / when the temperature goes above 70-80c, start to pwm the leds to reduce power and/or start a fan to push some air over the aluminum of the pcb. In such scenario, having a driver instead of plain resistors could help you.

[3dgeo]

I probably wasn't clear enough – main goal of this thread was to get feedback on inductors specs.
I'm open to design suggestions but all aspects in the first post must be considered.

[Marco]

The second diode will run ridiculously hot, the first a little hot. Just get a couple amp schottky.

[Benta]

I probably wasn't clear enough – main goal of this thread was to get feedback on inductors specs.

If this was just for yourself, I'd say do what you want. But you threatened to open-source it, and bad open-source designs don't help anyone, rather the opposite.