Merde with SMPS

Your schematic is mirrored. It says the minimum switching current is 200mA So try loading the output? (but I don't know this IC.) Have you tried running it first from a bench supply? (rather than a battery.)

George H.

Dropbox crashed me ,....again! so can't see schematic you are using. TI's website gave me the chip spec.

Usually an upconverter shorts the inductor to GND building up a flow of current through it, THEN opens the switch to GND causing the outpuit end of the inductor to pop up to almost any voltage it wants to go to. At least the inductor pops up to 5V plus Vbe drop, or so [looks like TI put a low drop switch in there to improve efficiency]. When all the current is exhausted from the inductor, its output leg goes to the same voltage as the input, usually with a lot of high frequency [scarey looking] ringing. So the voltage at the inductor's output should go to GND, pop high for a bit, then ring down to the battery voltage coming in. Then cycle repeats.

*if* something doesn't go right, all this action gets fouled up. for example, the switch to GND never lets go of the inductor, not good. It is very important to match the inductor, the capacitor [the FIRST capacitor], and the load so that the chip is happily running in its 'sweet' spot.

Or, there's an open in the resisttor feedback causing the chip to think the output is ALWAYS too much and thus voltage IN goes straight thru and no switching occurs at all.

I agree put a nice minimum load on the output, switchers historical abhor no load.

When you get it running we can all help addressing how to make the output as quiet as a battery.

Fixed the mirrored schematic. Sorry about that. There is some load already - about 5 mA in theory, but I haven't measured it.

I don't think the minimum load is anywhere near 200 mA. Figure 6 in the datasheet shows happy looking efficiency figures for Io from 0.1 mA all the way up to 100mA, with 5 V out, same as mine should be.

One thing that bugs me: I used a copper pour all around the chip and the passives (dotted line on the board) that is tied to ground. Was that a good idea?

--
Douglas Beeson

Everything looks kosher to me, but the fact that the output voltage goes up with the other converters disabled indicates that you may be overloading the battery. What's the supply voltage with the one converter running? You may be sucking the battery dry right off the bat.

How much current are you asking from the thing?

Barring weak batteries, and given that I get the idea that you're new to surface mount, I suspect wiring errors. Go over the thing with a strong magnification or an assembly microscope if you have one, and look for plain ol' solder bridges, bad joints, and whatnot. Going over it with a continuity checker is probably a good idea, too.

If you do as suggested and put it on a good stiff bench supply, then one or more wiring errors may cause a smoke signal to be sent up, or at least cause some informative heating of parts (just lay your finger down on the circuit -- if you snatch it back in pain, the shiny spots on your finger will indicate the hot components).

--

Tim Wescott 
Wescott Design Services 
http://www.wescottdesign.com

Thank you for the great info. I did a quick capture of the output (near the output capacitor) and graphed it in Excel:

The output looks regular to me but seems to have a lot of ripple. Is 100mV on 2.5V OK?

--
Douglas Beeson

The batteries were fresh the other day when I started debugging and gave a clean 3 V. Now they're down to about 2.75V.

I don't have a bench supply but I can wire up a LM317-based thing pretty quickly and give it a try.

In theory, should be about 5mA (a PIC and an LT1720 plus a few things that want nano-amps. But I can't figure out how to actually measure it!

Ha! I did notice that the TLV61220 gets a little warm (actually got very hot before I desoldered the other two SMPS), but nothing else seems overly hot.

But I will triple check the wiring.

Thanks, Tim.

--
Douglas Beeson

You are welcome. Graphed in Excel?! wow, you should get a copy of free octave [Matlab clone]. I first started using it to replace Excel plotting and calculations now use it to even analyze DSP before pouring the program into concrete, to do complex communication systems analyses, to 3D plot a myriad of data sets, to make Color Images to also plot 3D, make video plots, and basically just use as one whiz bang calculator.

Remember dropbox crashes my systems so cannot retrieve images.

sounds like your FB path is not right. Did you 'measure' the resistors before using? I've had some badly mismarked ones find their way into the stock room. from memory most regulators use 1.2 Volts based upon what is called the Widlar reference. About as constant as the energy gap in silicon, or so. so the center tap of the resistor string is supposed to be

1.2 volts and then across the top resistor is supposed to 3.8 in your case.

From memory, linear supplies routinely yield 0.1% ripple while switchers generally had 1% ripple which means at least 50mVpp at 5 Vdc.

You didn't say how much current the supply is requiring just to do 2.7 Vout.

Doug, Yeah sorry about that. I replied in haste. The 200 mA is the smallest maximum current. Well try a fixed supply. (And then flip that polarized tantalum cap around so it's in the right way. :^)

That shouldn't be a problem. (I had a ground plane spoil the (high) Q of an inductor, but I don't think that matters here.)

George H.

dl=0

2.75 under load, or open-circuit? If it's open-circuit, then the cells are pretty much toast.

Put a resistor in the power lead, enough to drop less than 100mV at your predicted current (so, 10 ohms). Then measure the voltage drop across the resistor.

"got very hot before I desoldered..."

Do you have an oscilloscope? Can you check the battery leads for oscillation?

--

Tim Wescott 
Wescott Design Services 
http://www.wescottdesign.com

What are the time units?

Did you mention tantalum caps somewhere? The data sheet recommends ceramics.

Once you do get it working, using the +5 to drive the other switchers may be a problem.

--

John Larkin         Highland Technology, Inc 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

e is

l=0

=0

m

two

ful?

TI's

f

nd

ut a

is

s

ng.

a

ts.

is

or],

k

nd

hor

put

the output capacitor) and graphed it in Excel:

0

mV on 2.5V OK?

No I mentioned tantalum caps. It's just (for me) the most common mistake t hat get's by testing and then fails in the field... the d@mn things can las t for hours/ days with a reverse voltage... I had one that was on a 5V supp ly and lasted for years. (35V I think) My only solution is a visual inspe ction.

George H.

dl=0

Is he using the output of the one switcher to drive the input of others? I had missed that. Yes, it can be an issue.

I recently helped a local colleague here in PDX with a problem he was having with driving a switcher from a switcher -- he was having intractable oscillation problems. Once I pointed out that an efficient switching supply that's supplying constant power out has, BY NECESSITY, a negative input impedance, his problem became much clearer to him.

He ended up just scraping the second switcher off the board, although I still think that he could have tuned the first switcher into stability with the second switcher as a load. He was uncomfortable with playing games with the first switcher's compensation, though.

I've driven switchers from switchers with good success - however, in my case it was a bulk supply switcher that was powering EVERYTHING, including a bias supply switcher that presented a very light load. Presumably the positive impedance of the "everything" swamped out the negative impedance presented by my bias supply.

--

Tim Wescott 
Wescott Design Services 
http://www.wescottdesign.com

Soft-starting the secondary switchers can help a lot. Otherwise, as you note, the negative Zin of the secondaries can bog down the primary. Adding "real" resistance across the intermediate bus doesn't help if the primary switcher is current limited trying to start up the downstream ones.

I've had the same concern when powering a product from a switchmode wall-wart; it works fine once it's up, but may never get there. Some of the warts fold back, even worse.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Units are 40 ns. Total sample length is 2.5 ms. Eyeballing it, frequency is about 125kHz which if I recall (not at the bench right now) is about what my oscilloscope measured.

Yes, I use 10uF 16V ceramics for both VBAT and VOUT. I am also using the exact same 4.7 uH Murata inductor mentioned in the spec.

That was my intention, yes. I see more people have added their thoughts to this so I better read on!

Thanks, John.

--
Douglas Beeson

No tantalum caps here ;-). By the way, you said you were having problems with Dropbox so I used some IT magic to extract the URLs for the raw files. These should load the PDFs straight away with no Dropbox Javascript cruft:

Schematic:

Board:

Switch output:

doug

--
Douglas Beeson

Yipes. So you're saying that even once I get this primary 3-5V switcher to actually work, I might not easily be able to use its output to drive the 5-15V and 15->-15V converters that I need? How the hell...we just landed on a comet!

I presume that by "soft start" you mean holding off on the drive "enable" pin until some delay has passed, enough for the upstream switcher to stabilize? Is a simple RC network good enough for that?

--
Douglas Beeson

20schematic.pdf?

dl=0

dl=0

Actually he means having the secondary converters ramp their voltages up slowly. Normal switching converters can ask for a big surge of current on start-up if they're not made to be current-limiting or otherwise soft- starting.

If you need so much charge to start up those secondary converters that your primary power rail sags, then you're in a world of hurt.

(This wasn't the problem my friend was experiencing -- he was getting outright oscillation. So that's another thing to worry about!)

--
www.wescottdesign.com

Here, I have extracted the actual file URLs so that no extra Dropbox stuff loads. These should be safe!

Schematic:

Board:

Switch output:

Funny you should mention the FB path. On the TLV61220 the FB pin is supposed to sit at about 500 mV, according to the datasheet. Mine is at 350 mV or so. I have double-checked the resistors - 1M and 110K - just as the datasheet suggests, which should be the right values for 5 V out. But that's clearly not happpening.

This is what my oscilloscope shows:

After unsoldering and resoldering most of the parts, I am now getting about 3.3V, but this varies under load!

3.42 V at 10 mA load 3.31 V at 19 mA load (LED)

Nothing gets hot anymore, though. Ideas? Thanks, Robert.

--
Douglas Beeson

they seem to be temporary URLs.

it looks like a 404 to me

--
umop apisdn