Buck boost converter help

[ampennes]

Hey guys do any of you have experience working with the LTC 3534 buck boost converter?  I am having an odd problem with it and can't seem to figure out why.  I'm using the example listed on the first page of the datasheet which should result in a 5V output.  However when I measure the output I only get 2.3V and with no load the current from my supply is 183mA which is many orders of magnitude over what it should be.  I drive the Vin and on/off pin to 5V while keeping the PWM/Boost pin low.  If I drive the PWM pin high then the chip seems to short to ground and hit the current limit on my supply which is set to 800mA right now.  I have inspected the chip for shorts to ground repeatedly and found none and I have tried 3 different chips all of which have the same response.  I'm running this on a pcb but I can try and mock it up on a bread board and see if anything changes but most of my parts are smt so it will be a bit difficult.

I have checked the pins for bridges repeatedly using a magnifying glass and a digital microscope

I have checked for shorts to gnd on the input, output, and feedback pins and havent found any

I do have an oscilloscope

excerpts of board and schematic attached

link to 3534 datasheet http://cds.linear.com/docs/en/datasheet/3534fb.pdf

Anyone have an idea about what is happening?  I'm stumped.
Thanks!

[Ice-Tea]

I suppose there's supposed to be a GND plane on top? Not seeing any GND connections?

[ampennes]

Yea there is a gnd plane I just think it is easier to look at the board without the plane cluttering everything.  here is a pic with the ground plane

[suicidaleggroll]

Your decoupling is not great.  Remember current flows in a loop, every milliamp you pull out of your input bypass cap must be returned through your ICs ground back to the cap's ground.  Your traces between the IC and the pos terminal of the decoupling caps (both input and output) are decent, but the return paths are quite long and convoluted.  Page 12 in the datasheet has the recommended layout, I highly recommend sticking to it.  This may not fix the problem, but it does need some work.

[Ice-Tea]

Checked for shorts between 12-13 and its neighbours?

[ampennes]

Yup, the resistance between 12/13 and 5V or Feedback are both in the tens of mega ohms so no problems there.
I agree that I should address the gnd loop in the next rev.  C16 should be flipped and placed to the left of the buck boost and I'll figure something out for C17 but that shouldn't be the cause of the issue here as far as I know.

[amitchell]

If I was you I would be looking at the schematic of the demo board.

http://media.digikey.com/Photos/Linear%20Tech%20Photos/DC1227A.jpg

http://www.digikey.com/catalog/en/partgroup/ltc3534edhc-demo-board/13200

From my limited boost experience, layout and component selection is mission critical.

[ampennes]

Put the scope on single shot mode and looked at the input.  There is some high frequency noise about 0.5V ptp riding on the 5V input but no collapsing.

Thanks for the link to the demo board I will order one soon and check it out if I can't find any higher res pictures.

[chris_leyson]

Never used the LTC3534. From the outline on your PCB it looks like you have used a small 5uH axial inductor and I bet it's saturating and then the chip goes into current limit when you turn on PWM. A small axial inductor won't have enough ferrite core area to store the energy, 1/2LI², the current will reach the right value but L is tiny because the core is saturated. If you can find an app note for the LTC3534 eval board then it will have a part number for the inductor, well at least I hope so rather than a list of phone numbers like on the data sheet. I had to look twice, a list of phone numbers what the... that's about as useful as shoes for snakes. Thanks amitchell you posted before me  bigger inductor. You could probably use a surface mount inductor mounted sideways with a pair of wires on the SMT pads. Easy mistake to make if your're not clued up on magnetics design and I've probably made the same mistakes in the dim and distant past, so don't worry.

[ampennes]

Originally yes I used a small axial inductor which I assumed to be the cause of the problem but now I have a larger can type indutor linked here http://www.mouser.sg/ProductDetail/Coilcraft/RFS1113-682ME/?qs=W38ilblRkRnrYqeuPGGArw%3d%3d
Roughly the same footprint so swapping it over was no big deal.  This one is 6.8 uH though instead of the suggested 5 but I wouldn't think that would have much of an affect.  This one is rated for over 5 amps so it definitely won't saturate.

[chris_leyson]

Hi ampeennes, sorry missed your post I tweaked my reply a little bit, does it work now  hopefully.

[ampennes]

nope, same behavior as before sadly.

[ampennes]

Actually something I did seems to have fried the chip as I'm now getting 5mV output with very little current draw and both sides of my inductor are at 0V and are basically shorted to gnd

[amitchell]

http://cds.linear.com/docs/en/demo-board-manual/dc1227A.pdf

It's a 3.3uH Toko inductor and from looking at the size of the inductor and the current rating of the demo, I am guessing it's saturation current is rated at 1-1.5Amps.

[ampennes]

is it much of a problem to be using a larger inductor?  I'm not sure if this is one of the cases where you just have to be over a specified value or if the value has to be exact.

[Ice-Tea]

Larger uH wise tends to reduce the ripple and little more, larger in current rating matters even less.

[ampennes]

mocked it up on a breadboard to remove the possibility of errors in the PCB being the root cause and it still doesn't function properly so I'm not really sure what is going on.  I noticed in the datasheet that it says the PWM pin should be pulled high on startup.  Perhaps not pulling it high is bad for the chip?

On a separate note does anyone have a relatively wide input band buck boost converter that they would recommend and is potentially less temperamental than this chip?  I have an input voltage range of 3.2 ish to 6V output of 5V and efficiency while "off" is important to me

[amitchell]

Breadboards don't seem to be SMPS friendly.

I would check out TI's WEBENCH, parametric design and selection tool that will spit out a schematic and BOM. Neat thing is that you can find your sweet spot between BOM count/cost and efficiency.

[amitchell]

I should add that TI's stuff is well documented with examples and the support forum on their website is well staffed and visited. Small fish like me can get answers from staff!