TPS61200 Boost Converter Releasing Smoke :-(

[mribble]

I'm using this boost converter (TPS61200): http://www.mouser.com/ProductDetail/Texas-Instruments/TPS61200DRCR/?qs=sGAEpiMZZMtY9G8Xaw%2fcnvufCtuHXcIe

I plug in a single sell lithium batter and everything works fine. I turn it on and off a few times and everything is still working fine. After awhile I turn it on and smoke comes out of the TPS61200, it is always very hot after that, and it no longer powers my circuit. If I replace the TPS61200 with a new one things work for awhile and then the same think happens while turning it on. I have seen this happen with two out of two PCBs. It seems like turning it on/off quickly might have something to do with it, but I'm not sure about that since I ran out of TPS61200 chips to test with. The circuit is only drawing 37 mA so I'm not even close to overloading the boost converter. If I feel the boost converter while it's running normally it is barely warm to the touch.

Does anyone have any ideas what might be going wrong here?

Here is the schematic for my circuit: http://www.glacialwanderer.com/_blog/blog2011/07_July/circuit_debug.png

Here is the board layout around the TPS61200: http://www.glacialwanderer.com/_blog/blog2011/07_July/circuit_debug2.png

I'm not that experienced with boost converters so if anyone has any ideas what might be going wrong here, I'd really appreciate any help you can give.

[Psi]

Looking at your circuit/PCB and a quick skim of the datasheet i noticed you don't have GND pin 9 of the IC connected to anything.

With some chips you can leave some GND pins unconnected but they normally mention it in the datasheet. This chip looks to need it connected.

From the datasheet..

To avoid ground shift problems due to the high currents in the switches, two separate ground pins GND and PGND are used. The reference for all control functions is the GND pin. The power switches are connected to PGND. Both grounds must be connected on the PCB at only one point ideally close to the GND pin.

pin 9 looks to be the only GND pin for the control logic, so without that connected the control logic is running from a floating ground that will be all over the place.

When you had it working it may have been due to very small internal coupling between GND and PGND, and since GND would be very prone to noise, oscillation and spikes it wouldn't take much to make the IC destroy itself.

[mribble]

I don't know how I missed that.  I guess I just needed another set of eyes.  Thank you so much for taking a look at this.  Once I get some more of the boost converters in stock I'll add a jumper wire and see if that fixes my problem.  I'll report back with my findings. 

Thanks!!!

[mribble]

I've got one question about this data sheet.  It says:

"Use a common ground node for power ground and a different one for control ground to minimize the effects of ground noise. Connect these ground nodes at any place close to one of the ground pins of the IC."

The ground I use for the rest of my circuit should be the control ground or the power ground?  It's not clear to me if the control ground is only a control ground for this chip or if all the other grounds in my system (ICs, battery, LEDs, ...) should go to the power ground.  I think I'm supposed to use the control ground for all of these, but since they use a fair amount of power and I don't think the data sheet is completely clear I want to make sure by asking you experts.

Thanks!

[gregariz]

As a tip, look at the sparkfun module using this chip and look at the ground seperations.. thats what I did.

[Psi]

I've got one question about this data sheet.  It says:

"Use a common ground node for power ground and a different one for control ground to minimize the effects of ground noise. Connect these ground nodes at any place close to one of the ground pins of the IC."

The ground I use for the rest of my circuit should be the control ground or the power ground?  It's not clear to me if the control ground is only a control ground for this chip or if all the other grounds in my system (ICs, battery, LEDs, ...) should go to the power ground.  I think I'm supposed to use the control ground for all of these, but since they use a fair amount of power and I don't think the data sheet is completely clear I want to make sure by asking you experts.

Thanks!

If you look at the diagram on the first page of the datasheet it uses two ground symbols to show how to connect grounds and which ground to use for what.
You can also see both grounds are connected together only at the GND pin.

The key bit is really that the output capacitor ground is a separate ground track to the GND pin, separate to the feedback dividers ground track to the GND pin.

Since a copper pcb track has resistance you get voltage across it under load. The more current through the track the more voltage across it.  You don't want changes in current causing fluctuation in the feedback circuit, which you would get if the ground current for both power and feedback traveled down the same piece of copper track.

[Neilm]

"Use a common ground node for power ground and a different one for control ground to minimize the effects of ground noise. Connect these ground nodes at any place close to one of the ground pins of the IC."
I don't think the data sheet is completely clear I want to make sure by asking you experts.

Thanks!

Generally, if you have a multi-layer PCB with proper power and ground planes having all the pins attached to the ground plane by as short a length of track as possible will give the best results. Pay attention to the places that the currents will flow and keep these as short as possible.

Neil

[mribble]

I've tried to take all the suggestions here and rework this part of the board.  Here is an image of that part of my new pcb: http://www.glacialwanderer.com/_blog/blog2011/07_July/ca5_power.png

If anyone sees anything I could further improve let me know.  I'll probably place an order in the next day or so unless people have more suggestions.

Thanks for all your help!