LM3410 led boost driver

[justDIY]

Has anyone worked with the LM3410 (Dan?)

I've built a design to power a 4x5 4x7 array of yellow + red leds, at approximately 3 watts (2.2v * 7) * (0.05a * 4).

I've destroyed three chips so far, and made two pcb revisions, but haven't found the cause of my problem.

My schematic is laid out as pictured in National's datasheet, the chip is so simple, it just needs a coil, a resistor and a diode for external parts.

The third time, I thought I'd test the circuit with a resistor on the power supply first ... so I powered it up using 4.5v from AA's through a 220 ohm resistor, it worked fine.  Next I used a 100 ohm, followed by a 10 ohm.  All these tests went well.  I finished by connecting the power directly.  I get a brief flash from the led array, and then the chip starts smoking.  This has me confused, since National claims the chip has both an internal current limit protecting the switch, and a thermal limit protecting the chip itself.   If I were over-taxing the chip in some manner, I'd expect one of these limiters to protect it?

I'll post my schematic and board layout tomorrow, they're on another computer at the moment.

edit: corrected a typo in the array size statement

[cadstarsucks]

Has anyone worked with the LM3410 (Dan?)

What makes you think I have used it?    Seriously though, nice part and I have not.

I've built a design to power a 4x5 array of yellow + red leds, at approximately 3 watts (2.2v * 7) * (0.05a * 4).

I've destroyed three chips so far, and made two pcb revisions, but haven't found the cause of my problem.

My schematic is laid out as pictured in National's datasheet, the chip is so simple, it just needs a coil, a resistor and a diode for external parts.

The third time, I thought I'd test the circuit with a resistor on the power supply first ... so I powered it up using 4.5v from AA's through a 220 ohm resistor, it worked fine.  Next I used a 100 ohm, followed by a 10 ohm.  All these tests went well.  I finished by connecting the power directly.  I get a brief flash from the led array, and then the chip starts smoking.  This has me confused, since National claims the chip has both an internal current limit protecting the switch, and a thermal limit protecting the chip itself.   If I were over-taxing the chip in some manner, I'd expect one of these limiters to protect it?

I'll post my schematic and board layout tomorrow, they're on another computer at the moment.

I can think of a few things not having seen your circuit and values. 

At those frequencies wiring parasitics are a big factor, luckily the switch current sense is internal.

At 50mA I am assuming you are hitting the LEDs hard.  Parasitics again.  The LED current sense might not be getting to the chip in time.

Inductor saturation.  This is a biggie when it comes to the chip blowing.  The protection circuit tries to shut down the chip but there is a delay of ...well that is annoying- it is unspecified, they are relying on the inductance to be relatively high with respect to the frequency.  Somewhat reasonable at 525KHz or 1.6MHz. 

Oh well... inductor saturation... for 2.2V*5*50mA*4/2.7V/80% out you need a continuous inductor current of over 1A, more if you are compensating for current sharing resistors.

Output capacitor...you need one unless you do not care if the LEDs see 250mA peak.  More if you need to compensate for current sharing resistors.

Input capacitor...again you need one.  Your batteries have a finite source impedance that the regulator does not like, and batteries do not like to deliver high currents.

National's webench is showing 2.7A in the inductor and 500mA peak in the array at 4.1V.  The call out is for a 1.7ADC inductor, but what you reaaly need to look for is at least a 3A saturation.  The array peak current goes up with increased supply voltage since the duty cycle is lower and the 10uF capacitor can not absorbed the energy from the inductor.

I hope this makes some sense, it seems I am rambling a bit.

Dan

[justDIY]

I can think of a few things not having seen your circuit and values. 

At those frequencies wiring parasitics are a big factor, luckily the switch current sense is internal.

At 50mA I am assuming you are hitting the LEDs hard.  Parasitics again.  The LED current sense might not be getting to the chip in time.

Yes, I'm driving them at their max recommended current.  The specs claim a pulse current of up to 100mA.  They're superflux type automotive leds.

Inductor saturation.  This is a biggie when it comes to the chip blowing.  The protection circuit tries to shut down the chip but there is a delay of ...well that is annoying- it is unspecified, they are relying on the inductance to be relatively high with respect to the frequency.  Somewhat reasonable at 525KHz or 1.6MHz. 

Oh well... inductor saturation... for 2.2V*5*50mA*4/2.7V/80% out you need a continuous inductor current of over 1A, more if you are compensating for current sharing resistors.

After I had run the simulation, I figured I might need a bigger inductor.  I originally was using a 10uH, 1A inductor and switched to a 22uH, 4A inductor.

Output capacitor...you need one unless you do not care if the LEDs see 250mA peak.  More if you need to compensate for current sharing resistors.

I don't have any caps with a safe voltage rating right now... I have some coming in next week.  So there is no output cap at this point.

The LED array is wired as four parallel strings of seven series diodes, no current sharing resistors.

Input capacitor...again you need one.  Your batteries have a finite source impedance that the regulator does not like, and batteries do not like to deliver high currents.

I have a 10uF, 10v ceramic cap on the input side of the switcher

National's webench is showing 2.7A in the inductor and 500mA peak in the array at 4.1V.  The call out is for a 1.7ADC inductor, but what you reaaly need to look for is at least a 3A saturation.  The array peak current goes up with increased supply voltage since the duty cycle is lower and the 10uF capacitor can not absorbed the energy from the inductor.

Looks like I'll need to order some proper parts

Here are the links to my layout and schematic.  The design has room for a microcontroller and status led, which are not populated currently.   Also, there is both a pull-up and pull-down resistor on the enable line for the switcher.  Currently the pull-up resistor is populated.

layout - http://projects.dimension-x.net/pictures/bikelite/boost_board2.PNG
schematic - http://projects.dimension-x.net/pictures/bikelite/boost_sch2.PNG
here is a very shitty picture of the actual pcb... I really hate this canon camera for macro work! - http://projects.dimension-x.net/pictures/bikelite/busted_switcher.jpg

[cadstarsucks]

I don't have any caps with a safe voltage rating right now... I have some coming in next week.  So there is no output cap at this point.

The LED array is wired as four parallel strings of seven series diodes, no current sharing resistors.

I ran the simulation assuming 5 LEDs since you originally said it was 4x5, though I now see your calculation was for 7.  You can force current sharing by putting a BE junction in one string and using that to feed BE junctions on the other strings.  It is not perfect but it makes sure one string does not hog the current and blow.

Did the new inductor prevent the chip from smoking? 

[justDIY]

Did the new inductor prevent the chip from smoking? 

nope.  as soon as I connected it without a resistor to the batteries, it started smoking.  The chip is instantly destroyed ... it won't power up if I let it cool and try again, with a resistor to limit the current.

[cadstarsucks]

that really sounds like the inductor saturating.

[justDIY]

I agree, but I wonder why it's running away like that.  The inductor recommended by webbench is a coiltronics dr74 33uH, with an Isat of 1.73a

I don't have a datasheet for that 22uH coil I'm using, but the notes I have written down in the bin which stores them claimed the current rating was 4a.  I don't know if that is at saturation or working current, but either way, it should be enough.  Not having enough inductance (22 vs 33 uh) is just going to change the duty cycle if I understand correctly?

It feels like I've overlooked something simple in my design.

[cadstarsucks]

are you using a really fast schottky?  are you using the 500khz or 1.65mhz version of the chip?

[justDIY]

500k version, LM3410Y.  I am using a schottky and not a normal silicon rectifier, but I will have to look up the part # and datasheet for it - I'm not sure how fast it is. 

[cadstarsucks]

500k version, LM3410Y.  I am using a schottky and not a normal silicon rectifier, but I will have to look up the part # and datasheet for it - I'm not sure how fast it is. 

An under rated schottky might do it.  1/(500khz*100*2)=10nS.  While 0.5V*1A*10%=0.05W does not sound like a lot it depends on the package and does not include switching losses at 500KHz.

Dan

[justDIY]

A little more than a year later, i've made progress on this project.

Here is the working prototype, still wet from an alcohol wash:


and a quick video of the different flash patterns:
http://www.youtube.com/watch?v=34TDFJGER-M


I'm using a 33uH coilcraft dr74 inductor rated at 2a or there abouts, and a vishay schottky diode rated at 40v, 3a.  The current sense resistor is 1% 1ohm.  Input cap is a 22uF 16v tant, output caps is a pair of 4.7uF 35v tants. Microcontroller is a 12f683 running at 4mhz.  This was the last LM3410 in my cut tape stock, so I had as many fingers and toes crossed as possible when I applied power.  There's still some inefficiencies I have to deal with, at full power the switcher is drawing about 1.2A RMS from the batteries, which is higher than it should be for 3w output ... vBat is 3.4v under load, which is about 4.08 watts into the system.  Using the microcontroller, I PWM the chip to about 50% output, and things run a lot smoother.  I'm going to rework the board with a 1.6 ohm current sense resistor, to cut power into the array down to 120mA (from 200).

After testing and programming, I put in a fresh set of batteries (3 MAHA PowerEX 2700 AA's, NiMH), and let it run, starting at 4pm.  By 11pm it was still bright as hell, batteries were down to about 3v under load.  It was really pulling a lot of current from the batteries at that point, the inductor was warm and had started singing, batteries were warm, so I shut it down.

[cadstarsucks]

The national webench simulation shows 33uH crashes at full load.   you want a 15uH indutor.  it allows you to simulate different input conditions and component values. 

at 15uH and 2.7V in you have 1.5A draw on the battery at steady state, but the 33uH tanked.  at 3V in the 33uH was marginal.

Dan