Need help with Buck Converter

[vace117]

Hi,

I am very new to electronics, and I am having a lot of trouble getting http://diodes.com/datasheets/AP3431.pdf buck converter to work reliably. The converter works properly for about 30 seconds, and then oscillations begin, with brief periods of normal function.

Here is how it is hooked up:


I am testing this converter with a custom made constant current sink:


And here's what the physical construction looks like:



For the first 30 seconds, everything is good. Vin = 5V, Vout=2.5V, FB = 0.79V:


Then the oscillations begin, and they are accompanied by a high-pitched buzz either from the converter, or the inductor:


Here's a short video that demonstrates the whole malfunction cycle and the high-pitched buzz:



A few further observations:

• Using a different load leads to the same oscillations eventually.
• Using no load at all, also eventually malfunctions! Here's an example with no load. You can see that the converter operates properly for brief periods of a few seconds, but mostly Vout is at Vcc:
  
• After the oscillations begin, restarting the converter does not fix the problem. It seems that unplugging and waiting for about a minute is what returns the converter to a working state for a short while.
• Nothing gets hot, or even warm.
• I've been having very similar problems with other converters as well. This is not my first attempt.

I have changed quite a few things since my initial attempts:

• I selected a better inductor, which has it's specs tested at 1MHz (switching frequency of the buck converter).
• I mounted everything, except the FB voltage divider resistors on the breakout PCB
• I have used a different breadboard, just to rule out a board defect
• I've used a different converter
• I was much more rigorous with calculating the inductor value
• I tried different power sources

Despite all these steps, I am still having the same stability issues.

What could I possibly be missing?

Any advise would be appreciated!

[HackedFridgeMagnet]

I think 1Mhz and breadboards dont work well together. 
That track for the inductor is unnecessarily long, your high current paths that are switching should be as short as possible. Remember any long track is an inductor.

Your Vcc decoupling capacitor should be a lot closer.

Hows the converter work with just a resistive load?

Someone is going to jump on me for this but: 10nF between FB and Gnd right near the pins. (Some value will affect stability but obviously not the best way to do it.)

Retouch those solder joints just in case.

[T3sl4co1l]

Eww... the controller that is.  Only a paragraph discussing ESR, and most of that in regards to output ripple.  As for the "transient response" part, suck it yourself and see.  Gee, thanks.

I never trust those "internal compensated" things.

Offhand, why 3.9uH instead of the 2.2uH shown in the datasheet, and what are the full spec on those caps?  Have you tried a tantalum for output?

A pot for feedback is a Bad Idea.  A pot with a resistor to GND is a slightly less bad idea, since it can't be adjusted to infinite output voltage (not that a buck can go above the supply, but trying may confuse or, who knows, even damage the controller).  The controller may well be expecting a known impedance on that pin, which a pot wouldn't handle very well.  (Input leakage is completely unspecified, which TI also does from time to time, sadly.)

My first picks for monolithic converters are TI, AD and LT.  Yeah, it's a royal pain sorting through the thousands on Digikey or whatever.

Tim

[Rigby]

ESR on input and output smoothing caps is too low?  Try electrolytic with same values.

[eecook]

Hi vace117,

Where is the PGND ground connection? I don't see it connected anywhere, or there's a trace below the ic package?

[kony]

Try loading it with purely resistive load first. It might be as well your CC load oscillating after some transient on input.
As others already said - what are full specs of ceramic caps used for filtration, namely material and voltage rating ?

[rob77]

as for a very beginner i would suggest a converter running on lower frequencies - you could save yourself a lot of time and frustration.
for example the MC34063 is a good choice for beginners - works like a charm on breadboard. it's not a fancy 1MHz and high efficiency device but ideal for beginners.
designing a PCB for SMT 1MHz converter is quite out of scope for a very beginner - at that frequency even the shape of the trace  matters (inductance).

[Rigby]

1 MHz isn't very high.  My understanding is that a breadboard should be able to support 1MHz circuitry without much trouble...

[TiN]

Rotate APW 90° CCW (as if looking on photo) and have switch node (SW) connect to inductor with wide nice beefy shape.
Locate MLCCs (caps) closer to APW, without thermal barrier (you have small single-layer board, it does not make a difference for soldering)

Route feedback (FB) by wire directly to your trimpot, lift that leg off the board a little, so you can run big solid uninterrupted GND plane.
Also sometimes not bad idea to decouple feedback with 47-330pF NP0 capacitor to ground, but that should not be necessary.

[marshallh]

Take your breadboard and throw it out. Then redo the supply on perfboard or make a pcb for it.

[rob77]

1 MHz isn't very high.  My understanding is that a breadboard should be able to support 1MHz circuitry without much trouble...

apparently is otherwise this very thread wouldn't be here

[vace117]

As others already said - what are full specs of ceramic caps used for filtration, namely material and voltage rating ?

Here are the full specs for caps and inductor:
http://www.digikey.ca/product-detail/en/CDMC8D28NP-3R9MC/308-1619-1-ND/3232394
http://www.digikey.ca/product-detail/en/LMK316AB7226ML-TR/587-2782-1-ND/2573976
http://www.digikey.ca/product-detail/en/CL31B106KOHNNNE/1276-1082-1-ND/3889168

Did I select these correctly?

[vace117]

Rotate APW 90° CCW (as if looking on photo) and have switch node (SW) connect to inductor with wide nice beefy shape.
Locate MLCCs (caps) closer to APW, without thermal barrier (you have small single-layer board, it does not make a difference for soldering)

Route feedback (FB) by wire directly to your trimpot, lift that leg off the board a little, so you can run big solid uninterrupted GND plane.

I'll make another board with your suggestions in mind. A lot of people told me that using a trim pot is a bad idea, so I think I'll just put fixed resistors on the board, with a very short trace from FB. That way the ground plane will not be interrupted either.

I'll post here once I have the board. Thank you for all the great suggestions!

[mariush]

Like everyone said.. the layout is bad.  Capacitors must be much closer to the chip, thicken the traces, keep the inductor traces very short.
Have a look at datasheets for other high frequency converters and see what they recommend, for example Linear Tech has nice datasheets with nice pcb layout recommendations.. see LT1308, LT1613, LT1949... they explain why and show how to arrange everything.
Capacitors you chose are relatively OK... just don't forget about voltage derating.. a 10uF 16v capacitor won't be 10uF when there's 5v on it, the capacitance goes down. Read http://www.maximintegrated.com/en/app-notes/index.mvp/id/5527

In your case, it's probably fine, but the layout is a mess still.

[T3sl4co1l]

Caps probably aren't bad, 16V rating used at 5V should still be over 50%, hopefully more like 80% of rated.  Question remains, why 3.9uH inductor when the datasheet says 2.2?

I'm not so offended by the layout.  Things could be closer, but that obviously doesn't matter on the inductor side, only the input side.  You can try tacking a small (1 or 0.1uF) cap across the VCC/GND pins, directly on the pins, to see if it affects anything; I'm guessing not.

A solid ground plane would be nice (e.g., via stitched to the bottom layer, rather than single sided), but as long as it's connected underneath the chip, it can't be terrifically bad.

I don't know why it takes time to go nuts, but the problem is definitely compensation.  Ripple is clearly visible during the oscillation, it's not shutting down or anything.  A prime suspect is using component values too large or too small for the internal compensation.  The 3.9uH inductor seems the most likely culprit.

And again, the datasheet says very little about ESR, so a tantalum (not electrolytic -- ESR too high) would be good to try, or a small chip resistor (0.1-2 ohm) in series with the output filter cap, to see if that helps.  Again, L, C and ESR all factor into the loop stability, and if they aren't exactly what the chip is designed for, you get this.

Tim

[vace117]

Question remains, why 3.9uH inductor when the datasheet says 2.2?

I followed the inductor value calculation formula given in the datasheet. The actual calculation is shown on my schematic diagram in blue.

[vace117]

How does this design look? Any further suggestions before I try this board?



Any further recommendations would be appreciated!

[TiN]

I would move 11k above 34k resisor to have ground voltage more close to output ground voltage. This helps with feedback.

Also where your bulk caps? Plant some 270uF 16V caps or similar near Vin and 100-470uF 6.3V near ur 22uF.

[vace117]

I would move 11k above 34k resisor to have ground voltage more close to output ground voltage. This helps with feedback.

When you say "output ground voltage", do you mean the power switch ground pin (PGND)? So, route the 11k resistor up to the PGND pin? Is that what you meant?

Thanks!

[vace117]

Also where your bulk caps? Plant some 270uF 16V caps or similar near Vin and 100-470uF 6.3V near ur 22uF.

The datasheet does not say anything about adding a cap in parallel with the 22uF. Is it safe to add one there? It will not effect the loop stability?

[vace117]

Gave this another go with the new PCB layout tonight:



I am happy to report that the oscillation problem was gone!

The new problem is - fire . The converter worked flawlessly for about 6 minutes, with varying loads. I was keeping my finger on it to see if it was heating up, and I thought that it was getting quite a bit warmer than it should have been. Than this happened:



If you look at the video of the burning process, you can see the conformal coating boiling around the chip, which suggests a lot of heat. It is strange that most of this action is going on around the EN pin, which is supposed to have ~2uA going through it.




Any idea what might be going on here?

[vace117]

And..... success!   

I decided to build this circuit again - exactly the same layout and parts.

As soon as I turned it on (with 800mA load), I heard quiet crackling, like something was repeatedly sparking (I heard the same sound when the EN pin burned). This time I could see the sparks and the smoke coming from the input capacitor, not the EN pin!

I turned the circuit off, doused the board in isopropyl alcohol and then blew it out with a can of compressed air. I saw quite a few chunks of dried flux (I  am assuming it was flux) fly off the board. Turned the converter again, and now it's been running cool and quiet for 20 minutes.

So what is that all about? I was sure flux was non-conductive... So what did I remove with compressed air that finally made the circuit work? I am using Acid Paste Flux SP-30.

[T3sl4co1l]

ACID PASTE FLUX?!

[IanB]

So what is that all about? I was sure flux was non-conductive... So what did I remove with compressed air that finally made the circuit work? I am using Acid Paste Flux SP-30.

WTF? 

From the data sheet: http://datasheet.octopart.com/SP30-Kester-Solder-datasheet-8865129.pdf

Like other inorganic acid fluxes, Kester SP-30 is too corrosive for electrical or electronics soldering applications.

[vace117]

Like other inorganic acid fluxes, Kester SP-30 is too corrosive for electrical or electronics soldering applications.

Oops. That's slightly embarrassing .

When I started to play with electronics a couple of months ago, I simply grabbed an old can of flux I had sitting on a shelf for years, w/o looking into it too carefully. I'll get something less crazy for future projects .

Thanks for pointing this out!

[Rigby]

Embarrassment is the tuition we all must sometimes pay in order to learn something valuable. 

The Moral of the Story: Remember to verify your assumptions once in a while.

[nsayer]

When I was about 8 I asked my dad to get me a little home finger galvanometer ("lie detector") kit from Radio Shack. He did, but he insisted on helping with the soldering. He used a giant gun iron that I'm guessing was >100W and solder that was left over from the last copper pipe sweating operation (undoubtedly acid core flux).

It lasted for an hour or so before it stopped working. We opened it up and all of the traces on the PCB (there was no solder mask) were all eaten away and/or corroded.

He had read the admonishment to use rosin core solder and pooh-poohed it.