Boost converter producing exactly 1/2 expected voltage

[RussJK]

Hi all,

I'm an experienced software engineer designing my first commercial hardware device (we're a startup with no money to hire a real EE). It's a battery-powered sensor using the TI TPS61025 synchronous boost converter to provide a fixed 3.3V output. I designed and built it according to TI's spec sheet (

http://www.ti.com/lit/ds/symlink/tps61020.pdf).

But it's producing 1.6V, exactly 1/2 of what I expected. I cannot guess why. Can anyone give me a clue about how to go about "debugging" this? The QFN package was of course particularly challenging to solder, so I'm wondering if a screwup in soldering could produce this rather curious result.

-Russ

P.S. If there's an equivalently efficient device to do what the TPS61025 does in a non-QFN package, I'd love to know about it, because I've come to hate QFN.

 

[MacGyverS2000]

What are your feedback resistor values?

What voltage do you measure at the feedback pin?

Dan - Owner

http://www.Hi-TecDesigns.com

 

[RussJK]

I measure less than 10 mA, but I don't have the feedback pin wired at all. I got the impression from TI's doc that that was only used with the adjustable voltage components (61020, 61028, and 61029). Figures 28 and 29 show FB unwired. Did I misunderstand?

-Russ

 

[Comcokid]

This is a moderately high frequency switching device. Make sure you're using ceramic capacitors with short leads (probably < 1/4" each) for the input and output. Don't use only electrolytics - you need ceramic capacitors.

I assume from your indication that you don't like QFN devices that you may be trying to use the chip using crude prototyping techniques.

Do not try to prototype a device like this with wire leads soldered to the chip and to a solderless plug-in type prototype board. Even at a few hundered kilohertz a couple inches of wire on a several uF capacitor will be enough inductance to make it like the capacitor isn't even there. Electrolytics have enough internal inductance (usually expressed partially in their ESR rating) to make them useless even with short leads.

If you are not using low lead inductance capacitors (i.e. ceramics) or have long leads, you could get the chip into a funky start/stop mode where it's not putting out the proper voltage. If you are prototyping using solder and a breadboard, I would recommend you use surface mount caps with very short leads soldered to them.

 

[LiteYear]

You are correct, the FB pins are not required in the fixed voltage versions.

The first step is to ensure what you built is what you've intended to build. Use a multimeter in audible continuity mode to test that every point your schematic shows should be connected is in fact connected. If you want us to check your schematic, post it.

Then, to debug, you need to understand how a boost converter works, and you need an oscilloscope. The half voltage output is probably just coincidence. What is your input voltage? If it's a single battery cell, maybe you're actually seeing the input voltage?

Start by applying a fixed resistance for the load - this is the easiest load for the converter to supply. Something like a 10 Ohm resistor will give you a nice 330mA load current.

Then measure from GND to VOUT, from GND to SW and from VBAT to SW with a CRO. What you're looking for is evidence of the boost converter sawtooth wave. It's presence or absence should narrow down the problem.

 

[RussJK]

Comcokid, thanks so much for your response. This component is on a PCB I designed based on TI's requirements as I understood them, keeping everything as close as I could. I've attached a photo. C1 (C2 on my PCB) is a 10 uF ceramic. C2 (C4 on my PCB) is a 2.2 uF ceramic. C3 is a Kemet 100 uF tantalum capacitor with a 9.0 mOhm ESR. (I don't recall exactly why I chose that one in particular.) R1 and R2 (R1 and R5 on my PCB) are each 390K. As you can see, everything is surface mounted and soldered directly to the PCB, but I can't entirely vouch for the soldering. I have a fair amount of standard soldering experience, but the 61025 was done with a hot air solderer with something I hope was within the appropriate reflow profile. I don't like QFN because it's so difficult to verify soldering quality.

Is the exactly 1/2 voltage significant, or just a random artifact?

What I'm especially hoping for is some approach for debugging this problem. I know how to deal with software, but hardware debugging is outside of my skill set.

-Russ

 

[RussJK]

LiteYear, thanks for your response. I'm using an external variable power supply that I've ranged between 2 and 5V; I get the same result across the range. I have not applied any load other than the Extech voltmeter. Might that be a problem?

I can certainly post my schematic if my previous post is not sufficient (I use DipTrace; what output format is appropriate?).

I don't own an oscilloscope, only a logic analyzer. Is something like the DSO Nano oscilloscope V2 sufficient (and essential)? Will that help me detect soldering issues?

Thanks especially for the debugging process. I'll get right on that.

-Russ

 

[IRstuff]

If you post a schematic, it's better to post a screenshot or PDF.

TTFN
faq731-376

 

[MacGyverS2000]

The soldering looks clean... the board layout isn't as tight as I would like for production, though likely okay for test purposes.

Dan - Owner

http://www.Hi-TecDesigns.com

 

[IRstuff]

Can't really tell, but I'd be concerned about your ground traces. Additionally, I couldn't tell of EN was actually tied to Vbatt.

TTFN
faq731-376

 

[RussJK]

IRstuff, what aspect of the ground traces are you concerned about? As soon as I get this power supply working correctly, I'll be designing V2 of this board, so any suggestions are greatly appreciated.

Dan, what would you tighten for production layout? Would you move the capacitors even closer? What about the inductor?

As LightYear suggested, I verified all the connections are correct using my multimeter in audible continuity mode. And in doing so I noticed something that's definitely wrong: I have the tantalum capacitor backwards, with the stripe to ground instead of VOUT. D'oh! Could this be causing my problem? Can I just unsolder it (assuming I even can) and put it right, or might it have been damaged by being in backwards?

I've attached the relevant portion of my schematic.

-Russ

 

[RussJK]

According to Wikipedia, "Most tantalum capacitors are polarized devices. When subjected to the wrong polarity, the capacitor depolarizes and the dielectric oxide layer breaks down, causing it to fail." So it seems that I should assume my capacitor is toast. Now this particular one costs $3.50 in single units. TI's spec sheet for the boost converter output capacitor says "For economical reasons, this is usually a tantalum capacitor. Therefore, the control loop has been optimized for using output capacitors with an ESR of above 30 mΩ." So is this 100 uF tantalum with the 9.0 mOhm ESR really a good choice for this application?

-Russ

 

[LiteYear]

Yes, the tantalum cap has probably popped. They can only tolerate a fraction of their rated voltage in the reverse direction. I like to associate the band on the capacitor with the straight line in the schematic symbol - they both denote the +ve terminal.

And yes, it's probably not suitable. Those Kemet caps have very low ESR. A standard "low ESR" tantalum cap will have about 10 times as much ESR. Also check the frequency V capacitance of your chosen cap - if it rolls off significantly by 600kHz, it might not be suitable.

And yes, that cap is critical to operation - anything could happen if it's not working.

Since you're using a power supply to drive the input, I recommend a 0.1uF cap directly across the VBAT/GND pins (in parallel with your 10uF input cap). It probably wont be necessary if you're powering the device with a battery and you're in a low noise environment, but then again it certainly wont hurt.

The typical switching frequency is 600kHz, so the DSO Nano V2 wont give you much resolution on the switching waveform, but it will be better than nothing.

The circuit should still work with no load, but the device goes into its power saving mode so the output is just occasionally "pumped" to keep it at the required output voltage. That's going to make for a pretty nasty waveform that many multimeters will not measure very accurate. Adding a load resistor will put it in its normal operating mode and make a much flatter output waveform possible. On second thoughts, use a 33ohm 0.5W resistor to keep the output power down.

 

[RussJK]

Bingo! The backwards cap was the problem! Only two components to position correctly and I screwed up one of them. What a n00b! :-} Thanks so much for all the help and great advice.

LightYear, thanks especially for the advice on the tantalum cap. 70 mOhm ESR caps are 1/3 the price!

-Russ