Killing LED driver

[piguy101]

Hi, I created a PCB from the following schematic I made that's supposed to be a flashing light (at rate of 2.4 Hz) that is powered off of 24 VDC. It is a basic switchmode current driver using the AL8805 to drive a string of LEDs. This circuit works for a few minutes, but then it dies randomly upon powering it up again. The LED driver IC U2 gets very hot. I pulled off and replaced U2 with a new IC and the circuit works again for a few times, but the whole thing breaks again after plugging it in a few times. And this process repeats when I replace the chip again.

Any idea what is going on?

Thanks

[phil from seattle]

You calculations look right (100 mA from a 1 Ohm) resistor.  What is your layout? The datasheet has a section on layout/thermal considerations and shows thermal vias.  Though, I'm not sure that it's just a thermal layout problem as you are "only" driving 100 mA LEDs. Still worth looking at.

A minor note - you are using a 2.2 uF cap but they mention going to 4.7 when approaching 36V. Probably not consequential, though.

[hli]

You are missing the input capacitor. It should be at least 2.2µF ceramic - the data sheet shows 150µF+100nF in the example schematic at the end. You circuit might be oscillating, or the the inductance of the input wires might create voltage spikes at the input.

[danadak]

What is the Isat rating of L1 ?

http://cds.linear.com/docs/en/design-note/dn008f.pdf

Regards, Dana.

[piguy101]

What is your layout? The datasheet has a section on layout/thermal considerations and shows thermal vias.  Though, I'm not sure that it's just a thermal layout problem as you are "only" driving 100 mA LEDs. Still worth looking at.

Attached is my layout. Please let me know if there is anything I can do to make it more amenable to viewing. I am not using any thermal vias because I didn't think it was a big deal. I could be wrong.

You are missing the input capacitor. It should be at least 2.2µF ceramic - the data sheet shows 150µF+100nF in the example schematic at the end. You circuit might be oscillating, or the the inductance of the input wires might create voltage spikes at the input.

You are right. I have C2 on the input, but it is only 1 uF, not a parallel combination of capacitors. How important is this?

What is the Isat rating of L1 ?

I was worried about this. The inductor I am using is CB2518T330K https://www.digikey.com/product-detail/en/taiyo-yuden/CB2518T330K/587-3038-1-ND/2763418. The saturation current is listed as 130 mA. Since my LED current should be 100 mA, does this make me okay?

[tpowell1830]

R8=100 ohms

[piguy101]

R8=100 ohms

Why do you say that? The AL8805 datasheet says Iled=Vthd/Rset and gives an example for a 660 mA load, a 150 mOhm resistor is used. I am using a 1 ohm resistor for a 100 mA load. Is this correct?

[phil from seattle]

Here's their recommended layout. Notice what I believe to be thermal vias. Also, in the example board there are a couple of vias to the ground plane and the other side (top in this case) has a relatively large ground plane area.

[piguy101]

Here's their recommended layout. Notice what I believe to be thermal vias. Also, in the example board there are a couple of vias to the ground plane and the other side (top in this case) has a relatively large ground plane area.

Okay, here is my layout in a closeup with the GND plane showing. I did try to match the datasheet somewhat closely when I made it, but I see I may have pretty been off the mark. I have no thermal vias and the ground trace for U2 is pretty skinny. Should these issues actually lead to total failure and damage to the IC?

[hli]

Without the thermal vias your switcher might get too hot (since this is the only way of dissipating the heat)..
Your C2 on the input is placed poorly - its connection to ground is way too long, creating additional inductance. Just rotate it by 90 degrees and connect it to the big ground plane.

[tpowell1830]

R8=100 ohms

Why do you say that? The AL8805 datasheet says Iled=Vthd/Rset and gives an example for a 660 mA load, a 150 mOhm resistor is used. I am using a 1 ohm resistor for a 100 mA load. Is this correct?

Ok, after looking up the datasheet, the resistor should be a 500 mOhm, since CTRL is ~ 5 volts. In the future a link to the datasheet would be helpful.

I am confused what you are trying to accomplish with the inverting Schmitt trigger on CTRL? This is the dimming pin. Are you trying to switch the LEDs on and off at 2.4 Hz.?

EDIT: Corrected Hertz typo.

[phil from seattle]

Okay, here is my layout in a closeup with the GND plane showing. I did try to match the datasheet somewhat closely when I made it, but I see I may have pretty been off the mark. I have no thermal vias and the ground trace for U2 is pretty skinny. Should these issues actually lead to total failure and damage to the IC?

I'm not 100% sure. You are running the chip well under the limits but not much copper area for heat sinking. Could be in conjunction with the lack of an input cap. Do you have an oscope? worth at least poking around to see if there's something amiss.

No matter what, I'd add an input cap, re-layout the switcher part to have more ground copper, add 10-20 thermal vias to as large a ground plane as you can get on the bottom. Give the heat a big fat channel to get away from the IC. You can cram the vias together fairly tight. [edit]also, move L1 closer to the SW pin. The DS talks about minimizing ringing[/edit]

Also, I would double check that you haven't gotten any of the other switcher components misinstalled. Check see that all your caps and resistors aren't damaged. Inspect and cont. check the PCB itself.

[piguy101]

R8=100 ohms

Why do you say that? The AL8805 datasheet says Iled=Vthd/Rset and gives an example for a 660 mA load, a 150 mOhm resistor is used. I am using a 1 ohm resistor for a 100 mA load. Is this correct?

Ok, after looking up the datasheet, the resistor should be a 500 mOhm, since CTRL is ~ 5 volts. In the future a link to the datasheet would be helpful.

I am confused what you are trying to accomplish with the inverting Schmitt trigger on CTRL? This is the dimming pin. Are you trying to switch the LEDs on and off at 2.4 Hz.?

EDIT: Corrected Hertz typo.

The whole circuit flashes the LED string to make it more noticeable. The RC network and inverter on the CTRL pin is used as an oscillator. The CTRL pin is weird. According to the datasheet, it can be used with an analog voltage for dimming (0.5 V to 2.5 V), or it can be used as an enable, if the voltage on it is greater than 2.5 V, which is what I'm doing (5 V).

[piguy101]

Okay, here is my layout in a closeup with the GND plane showing. I did try to match the datasheet somewhat closely when I made it, but I see I may have pretty been off the mark. I have no thermal vias and the ground trace for U2 is pretty skinny. Should these issues actually lead to total failure and damage to the IC?

I'm not 100% sure. You are running the chip well under the limits but not much copper area for heat sinking. Could be in conjunction with the lack of an input cap. Do you have an oscope? worth at least poking around to see if there's something amiss.

No matter what, I'd add an input cap, re-layout the switcher part to have more ground copper, add 10-20 thermal vias to as large a ground plane as you can get on the bottom. Give the heat a big fat channel to get away from the IC. You can cram the vias together fairly tight. [edit]also, move L1 closer to the SW pin. The DS talks about minimizing ringing[/edit]

Also, I would double check that you haven't gotten any of the other switcher components misinstalled. Check see that all your caps and resistors aren't damaged. Inspect and cont. check the PCB itself.

Well, I soldered on another input capacitor on top of C2 (1 uF + 2.2 uF now) and put in a new regulator and the circuit seems to be working okay for now. I will definitely add those vias in a future rev of the board.

I poked around with my scope, but I'm not really sure where to look or what to look for, frankly. Attached is the voltage across the inductor, but I'm not sure what a "good" waveform should look like from the switchmode regulator. It looks like to me that the regulator is switching the inductor a number of times and then stopping briefly. This is not the constant PWM I had expected.

[james_s]

When working with switchmode regulators the PCB layout tends to be very critical. You might try prototyping the circuit deadbug or island style on a scrap of blank copper PCB, I find that tends to work well. You need very wide low impedance grounds and connections between the switching device, inductor and capacitors need to be as short as possible. The placement of parts in the feedback loop can be critical too, you don't want it picking up noise from the switching.

[phil from seattle]

I'd look at the output and input power rails.  They should be relatively stable within the bounds of PWM.

[james_s]

That switching waveform looks ugly, it appears there is a great deal of oscillation there, which is exactly the sort of symptom you get from inadequate grounds and stray inductance/capacitance.

[mikerj]

Ok, after looking up the datasheet, the resistor should be a 500 mOhm, since CTRL is ~ 5 volts. In the future a link to the datasheet would be helpful.

A PWM signal (low level <= 0.4V and high level > 2.6; transition times less than 1us) allows the output current to be adjusted below the level set by the resistor connected to SET input pin.
...
Note that 100% brightness setting corresponds to VCTRL = VREF, nominally 2.5V. For any voltage applied on the CTRL pin that is higher than VREF, the device will not overdrive the LED current and will still set the current according to the equation VCTRL = VREF.

Seems to me like the OPs 1 ohm resistor value is already correct for a maximum LED current of 100mA.