Forums

Some progress in understanding. This burst-mode operation seems to be a startup mode. And with output voltage set to 5.5 V, the IC refuses to leave sturtup. If I set output voltage to 5.0 V by changing feedback divider, the problem goes away. I will explore this further since I need slightly more than 5 V to be able to smooth out the supply with an LDO. I should point out that in the datasheet, 5.5 V is included into recommended output voltage range (although it is exactly the maximum recommended). And even with 5.0 V output, the problem is still not over. The IC does not leave this burst-mode until the output voltage comes close to 5 V (the datasheet specifies the threshold of 1.8 V, which at 3.7 Vin is definitely overcome thanks to built-in diode). So it looks like the datasheet is wrong twice. It is actually more likely I am wrong...I feel I am right however. If someone could confirm my findings...

This is indeed odd.  The part can output 5.5V at much higher currents.  Could you attach your schematic and board layout?

Please follow the link: https://plus.google.com/photos/105822372665226114142/albums/5739898425366521057 I was absolutely sure I had provided this link in the first post.

Thanks for sharing this.  You might want to order the EVM, which has already been built and tested, to evaluate your circuit.

The layout states that there are two ceramic input and output caps but I only see one in the photo.  Did you stack the caps or is there just 1 there?  The datasheet recommends ceramic caps for both input and output.  The 6.8uF would work but just the 0.1uF may not.  I would also recommend removing that giant tantalum on the output and seeing if that helps things.

You also might have noise getting into your FB resistors or have switching currents running through them.  This would explain things.  You might try reducing their values to 10k and 1k and seeing if this helps.  Or take the 10k resistor (lower resistor) and stand it up (tombstone it).  Then, run a wire from its ground right back to the IC's ground pin.

Chris Glaser

Did you stack the caps or is there just 1 there?

I did stack them. Initially there were only 0.1uF ones, and I soldered 6.8uF on top in attempt to fix the problem. Gained 1 mA - not much.

Chris Glaser

I would also recommend removing that giant tantalum on the output and seeing if that helps things.

Chris Glaser

You might try reducing their values to 10k and 1k and seeing if this helps.  Or take the 10k resistor (lower resistor) and stand it up (tombstone it).  Then, run a wire from its ground right back to the IC's ground pin.

I'll try. Thanks!

Results.

"Removing giant tantalum" - does not help (causes extreme output ripple 1.5Vp-p). Placing 47uF aluminium electrolytic instead - simply does not help either.

"take the 10k resistor (lower resistor) and stand it up (tombstone it). Then, run a wire from its ground right back to the IC's ground pin." - No difference.

Reducing resistors - well, I didn't do that. Instead, I have placed 1K-potentiometer in parallel to the divider. It did not change the behavior at 5.5Vout, but it allowed extra experimentation.

What I have found is that as I turn the voltage up (under 68 Ohm load), everything is good until I reach a value of about 5.15V. After that, the output voltage snaps to 4.8-5V and remains there, more or less independent of load. Next, as I turn the pot backwards (to lower the output voltage), at some point the output jumps up to c.a. 5.0 V and starts to respond to the rotation. So, it appears I was wrong about the startup mode, it looks more like overvoltage protection mode.

The snap voltage (5.15V) depends on the temperature of the IC (or something else?). E.g. if I keep the converter under load for some time, the snap voltage increases to 5.35V. If I further heat the IC with hot air gun to 50°C, the snap voltage reaches 5.9 V. That is, at 50°C the converter works perfectly well for me.

Without load, the output voltage can be increased up to 7.2V.

I have also captured the startup of the converter under load, and noted interesting behavior. It starts up correctly, gaining 5.5V successfully, and begins to regulate. But soon afterwards, the output voltage drops to that damned 4.8-5V. The duration of good regulation is somewhat random, attaining times from about 10 ms to more than 0.4 s.
I have added the captured plot to the album (please follow the link in the post number 4).
Again, this looks like overvoltage protection comes into play.

PS. For now, I am satisfied with the converter at 5.1V output. I will change the divider accordingly and move on to more essential parts of my project. Although the original problem remains unsolved, I have to move on. So, should I mark the question as answered?

Thank you very much for your help!

It does sound like noise is getting into the FB pin and we would need to see more waveforms to be sure.  But if you have a solution you are happy with, then go with it.

I am sorry for emotions, but what else can be done to reduce that noise? The layout is almost as compact as I can make it, and FB is quite far from the noisy L-pin net. It shouldn't be that sensitive! 

By the way, I have also tried to short FB to ground right at the IC (the pins are adjacent, I had sticked something metallic in between). Doing so under load does not change the output voltage (or makes it jump to 4.8-5V if the one set is lower). So I really do not think that noise being fed into FB is the cause of the problem.

Can you tell me the list of the waveforms you'd like to see?

 PS. Does anyone know, how tolerant is TPS61220 to soldering by hand? I haven't found anything in the datasheet on this. I have soldered it with a fine tip at 250°C, and for some pins it took me two or even three attempts to get it right.  

Let's leave the profanity out of these public forums.

The device is able to be soldered by hand.  We make EVMs that come pre assembled with a tested working circuit.  You can simply change one of the feedback resistors to make it output 5.5V.

You would need to look at Vin, Vout, L, and the inductor current to know what the IC is doing.  After all these experiments and soldering, I would suggest replacing the IC since it may be damaged.  I would also confirm the components values you have on the board.  Measure your ceramic caps and resistors before you install them to make sure they are what you think they are.

Sorry again. Profanity removed.

I have captured some of the waveforms, and uploaded those ready to the album. I will make more later, but it's going to be a lot of pictures. I will probably have to organize them somehow. I also have to note that from now on I am not running the converter alone. I have assembled a part of my circuit and in its current state it should be drawing 2.5 mA (on waveforms, I call it "light load", while "heavy load" is when a 68-ohm resistor is connected to Vout). 

I also have done some experimentation on heating, and now I am convinced that it is the IC's temperature that matters, not ceramic capacitors' or something else's.

By the way, I am waiting for new chip to be delivered. I am going to replace it, along with ceramic capacitors. I hope it will help. 

 Thank you! 

Hello, again.

The new board with slightly different layout. New parts. New chip from a different source. Each pin soldered with one touch of the iron.... Soft-starting Vin to prevent large surge through built-in diode... Same results. Unfortunately. 

I'm sorry that you're still having issues.  Do you have part numbers for the caps you used?

No, I don't know the part numbers. For ceramic caps, I know they are Samsung, 10 uF 16 V  X5R (measured capacitance turned out to be 8.7 uF). Most probably the part number is CL21A106KOFNNN. Input side tantalum cap (now 100 uF 6.3 V) was bought looong time ago, I can't find the part number. The output side tantalum (330 uF 6.3 V) was desoldered from a dead notebook motherboard.

Thank you very much for the help.