Help! Killed three converters already

[goldserve]

I need someone check this board layout for me to see why it is not working.

First, I am using the converter LCT3454

https://www.linear.com/pc/downloadDocument.do?navId=H0,C1,C1003,C1042,C1116,P15693,D10838

**broken link removed**

The boad layout is as follows:

**broken link removed**

**broken link removed**

The circuit is basically a copy of the reference but i've added some other support circuitry. I have prototyped the board to to just include the converter circuit.

Input and Output capacitors: 10uf
L1: 4.7uh
Vc: 0.1uf
Iset1: 3.6kohm
En1: 4.2V
En2: GND
Vout: LED +
LED: LED -
Vin = 4.2V

Initial observations before powering circuit: LED is connected to gnd internally in the chip

When I power on the chip, I get a blink and then the converter chip gets hot. After that, I check the chip and find GND and SW2 are shorted together. That means the chip must be fried or busted now right? Can anyone figure out why that happens?

 

[hjames]

And here I thought *I* did confusing layouts. Between the irregularity of the silkscreen and the oddball 'x' scattered though the screenshots, I can barely follow the circuit. Either clean those layouts (make the silkscreens more consistent, show where all the actual pads are, something...), or just take a couple pictures of your prototype or else you'll just get a lot of wild guesses...

Well as for the wild guesses:
1) routing on the inside of a LLP package is risky, but doable - as long as you have a soldermask...

2) have you tested for shorts on everything? Getting a solder short *under* a part is one of the joys of the new package styles. Ditto for the inductor.

3) Use current limited power supplies when trying out new circuits - stick a 10 ohm resistor in series with the battery and temporarily set the "iset" resistor to something like a 100K resistor just to set the limit current to something very low.

 

[goldserve]

Thanks for the words of wisdom.

I have included a better screen shot. The X's are GND pads that were flooded over with.

**broken link removed**

1. I have checked for shorts between the pins and GND. The only pin that seems to be connected to GND is LED which is confirmed on a unsoldered part. I do have solder mask on the board so should be no fears.

2. Good idea about setting the isense resistor to a very low current for now.

Two boards failed in the same way so i'm thinking something else may be wrong here...

 

[hjames]

Incidentally, how are you soldering the part down? A good ground contact is critical with these kinds of parts.

Other than random shorts under the chip or inductor, I can't see anything wrong with the layout. If the boards are consistently failing, I'd pull off the 3454 chip and start doing continuity checks on everything until I'm satisfied that everything that is supposed to be connected is, and nothing beeps out in odd places and all the resistances to ground/vcc are close.

The only other possibility I can think of is that the inductor is sufficiently underspec'd that it actually causes the chip to die - or you have the LED or a capacitor in backwards or something - but those are complete "WAGs".

 

[goldserve]

Alrighty...I'll get some more parts and check every pin more carefully. I solder the parts by using a hot air blower to blow the part on. The ground contact should be good becuase i've put a little solder and flux on the back of the part first.

I don't get something in the schematic though. The LED + goes to pin 7 and LED - goes to pin 5 right?

 

[hjames]

Alrighty...I'll get some more parts and check every pin more carefully. I solder the parts by using a hot air blower to blow the part on. The ground contact should be good becuase i've put a little solder and flux on the back of the part first.

I don't get something in the schematic though. The LED + goes to pin 7 and LED - goes to pin 5 right?

Okay that sounds reasonable - it does take the bottom pad a lot longer to reflow than the pins on the outside. Usually I have a pair of tweezers and poke the part while it is being heated to make sure that it has melted.

pin 5 is essentially a .1Ohm resistor to ground, and pin 7 is the output of the switching regulator, so that is correct.

 

[Electrode]

I've played with the Luxeon- Lumiled units before. Are you using any heat sinking? I know the LED portion usually needs it with the current(s) used, but wondering if your circuit does as well. Just something to consider if not already.

 

[goldserve]

Yeah, I can tell you all about the heat these things generate. The circuit only lit up dimly for around 200ms before it turned off and converter died. Think i'm doing something wrong...

 

[goldserve]

Argg..I built another one very very carefully but it something is still wrong.

The led turns on after 1 second, after a few seconds, the led starts blinking. I'm thinking the chip gets too hot and goes into thermal protection so it goes on and off. I'm thinking i need to measure the current and temperature of the chip.

What do you guys think?

 

[Nigel Goodwin]

What do you guys think?

I think switchmode designs are VERY critical, you make a slight mistake and everything dies!. From what I recall about similar problems a while back, the 0.1 ohm current sensing resistor is absolutely critical, as is the layout around it?.

 

[goldserve]

Is there something obviously wrong with my layout I still think it is the overtemperature of the chip because I felt the chip and it gets really really hot. My next pcb will have vias to draw the heat to the bottom layer of copper as well.

 

[Oznog]

1. What is the part number of the inductor you used? It would be easy to have a tiny inductor which will saturate or melt under such currents.
2. If this is a Luxeon, did you electrically isolate the thermal slug on the back? For example, it's easy to accidentally to sink the slug to an aluminum flashlight body which is also the battery ground. This will cause the device to malfunction.

 

[goldserve]

There are two things I haven't done yet. I need to get a resistor that is 1ohm and can handle at least 1W so I can do some current measuremeants to see if the current sense is proper. I supposedly have set it for 150ma but it seems much brighter.

I am using the 4.7uh inductor from this series:

https://www.coilcraft.com/lps4018.cfm

Improved layout:
**broken link removed**

**broken link removed**

 

[philba]

The schematic you showed isn't the EXACT schematic you are using, correct? (since it is straight from the LTC datasheet) I hate to be suspicuous but I have seen that before - "it's just like the in the book (or what ever)" and when you eliminate everything else you look at the actual schematic and there's a difference that accounts for the problem.

You appear to be programming the current for .85A which is within the chip's spec. What I don't see are the 2 10uF caps from the schematic (and really couldn't tell from the layout). both need to be ceramic according to the datasheet. I assume that's for a low esr. I know that SMPSs are very finicky about that.

edit: you posted a better layout as I was composing so i see 4 caps now. still don't know if they are 10uF ceramic, though.

too bad ltspice doesn't have that part in the library.

 

[goldserve]

C3 and C4 are the caps. First I thought you had to use LOW ESR tantulum caps. I've installed a 10uf ceramic on C3 which is the input and a tantulum on the output. I think i'll change it to 10uf ceramic.

The schematic should not differ too much from the one given. All i've got are pins connected to the EN1 and EN2 of the chip for use once I make sure the boost converter is working perfectly. R3 and R4 are pull down resistors to turn off the converter by default. Currently not populated.

 

[philba]

you said you have it set for 150 mA but from the datasheet and your current programming resistor value of 3600, I get 850 mA. 3850*(.8/3600)

I've always thought low ESR tantalum was an oxymoron.

 

[goldserve]

I used the 20.5K in the position of R1 and 3.65K in R2 but I have only tied EN1 to 1 and EN2 to 0.

 

[philba]

I used the 20.5K in the position of R1 and 3.65K in R2 but I have only tied EN1 to 1 and EN2 to 0.

sigh... that's not what you said in your first post. this is why you should post the real schematic you used to produce your board, not some clip from the datasheet. I wonder what else lurks...

 

[hjames]

Low ESR tantalums are definitely out there - you can get something that has ~20mOhms in a B sized package (versus the typical 2Ohm of a normal tant.) Of course you end up paying up the nose for them. Another benefit is that usually these are the "polymer" variety which is less likely to go up in flames...

If you still haven't found the issue, you really need to use an oscilloscope to figure out what's happening - debugging "blind" like this is just really difficult. With the proper tools, debugging this should be a 15 minute exercise.


BTW, where are you located? If you're reasonably close by, I'd offer to debug one of your dead boards - After the last thread where I suggested this part, I snagged a couple chips from Linear in case I felt like building my own flashlight.

 

[goldserve]

Hi James,

Thank you for your offer. I do have some issues with the via underneath the part which is causing a short to the GND underneath the part when the solder flows so i've redesigned that part of it. I've also added better heatsinking to the part by drawing the heat with vias to the bottom plane.

I will re-run the proto boards and try again. In the meanwhile, I'll have to go to Linear in Milpitas to pick up some more parts. As you've guessed, I am currently located in the southern bay area and I've only moved here a few weeks ago and will be staying for 4 months. I don't have all the equipment that i'd like to work with as well.

Once I get the new boards and try things out, I'll let you guys all know. Thanks!