I'm going batty after spending 3 straight days trying to wrap my head around this.

I've never used a switching regulator before. My current project has users who prefer to use 3S Li-Poly batteries (it is remote control hobby related), which means a max Vin of 12.6V. My device runs on 5V and 3.3V (but let's not worry about the 3.3 for now). The max draw is probably around 250 to 300mA but I'm basing it on 500mA to be safe.

I believe that is too much dissipation for a linear. (12.6 - 5) * 0.5 = 3.8W. Without significant heat sinking, I haven't been able to find a linear reg IC that even comes close (2.3W was highest). Even with head sinking, that's ridiculously wasteful.

So that just begs for a switcher, but that has proven MUCH more complicated than I thought.

I finally settled on the Richtek RT7251A (data sheet) for a few reasons, but partly because it appeared to spell out for me what to use for externals! I mean, it straight up tells you in a chart that for 5V regulation you use a 22uH inductor, a 10uF Cin and 2x22pF Cout. Right on! But then, the devil is in the details, isn't it? I couldn't rectify those numbers with the equations they provide further down. I emailed them and finally learn that those figures assume 1.5A full-time current. Well, that's not very realistic, now is it? Plugging in the numbers, I determine I really should be using closer to a 100uH inductor so that chart's out the window.

Now I'm stuck on the capacitors and I'm pulling my hair out.

The datasheet seems to suggest that ceramic shouldn't be used due to ringing that might cause another Manhattan project if it gets out of control and gives vague-ish warnings that polymer caps can be used (try finding one that's 10uF!!) and how "critical" it is to select ceramic properly if you choose to. Well, that's clear as bloody mud!

Between capacitance, ESR, ESL, and temperature coefficients... after all this reading, I feel like a lost child. I'd like someone to hold my hand please!

As a start, I found one page that seemed like it would help me (link), but for the life of me I can't figure out what "D" is supposed to be in their equations!

Right now, D is for Dumb and that's how I'm feeling. This just shouldn't be this difficult.

 

OK, I took a breath and looked at the equations again.

I just now noticed the tiny "D=0.5" so I guess it is taking the worst case (highest RMS) duty cycle for that variable. If that's assumed, then why not just put "0.25" instead of "D(1-D)"? Anyway... if I do have that right, and IF "Fs" means switching frequency in Hz and assuming Vpp is "1% of input voltage" then I get a Cin of 2.92uF. That's a lot of "ifs" but if that's right, then the 10uF recommended in the datasheet should be fine.

I'm still no closer with the other specs, but after another breather and reading through it again, maybe I'll get it.

Sorry, I should know better than to post when frustrated.

 

If you want to design and build one from scratch, there's lotsa help available here. Consider the National Semiconductor (now TI) "Simple Switcher" line of chips. Excellent app notes, easy to use, etc. Also, pre-built switching regulator modules are available on ebay cheap. If you want to just spend $5 and get on with your life, go industrial. Searching for a DC/DC converter will kick up a lot of fully isolated devices that are way larger and more expensive than you need. Instead, look for POL - point of load - devices. These usually are small, non-isolated pc board assemblies. Digi-Key and Newark sell small switching regulators that are about the size and shape of an 7805 linear regulator and are intended as "drop-in" replacements (with extra input and output caps):

http://www.digikey.com/product-sear...e=0&rohs=0&quantity=&ptm=0&fid=0&pageSize=500
http://www.digikey.com/product-detail/en/OKI-78SR-5/1.5-W36-C/811-2196-5-ND/2259781
http://www.digikey.com/product-detail/en/OKR-T/1.5-W12-C/811-2782-ND/3674288
http://www.digikey.com/product-detail/en/BR300/BR300-ND/4454206

ak

 

Anything other than "from scratch" (assuming you mean starting with an IC, not *building* the IC's circuit using transistors, diodes, etc) is not up for consideration. The eventually is for a production device with a minimum build of 1000 pcs. I expect the entire BOM to be not much over $5, so no I can't go that route, unfortunately.

That's not to say I didn't already look at the board-mounts. I sorted by price and saw the cheapest one (which met our V and I requirements, of course) and took a hard look at it (LMZ21700SILT) because @ $2 - while being the single most expensive item on our board - was tempting given my aggravations (which at the point wasn't nearly what it was when I posted!). And it is reasonably small. But I became skeptical when I saw the value of the inductor strapped on its back... 2.2uH doesn't come close to what is necessary based on equations. The device utilizes discontinuous mode, and I suspect it does so more than not with that low-value L. I suspect it really would only work well for a device that is maxing out its rated current all the time. Ours is not.

Anyway, I was extremely frustrated when I posted. I would still appreciate some interpretive help on the capacitor selection equations to make sure I'm properly using the formulas (ie. my assumptions on D and Fs)... but otherwise, after a moment of lucidity at around 2am this morning (and I'm up at 6... nice), I think I got a handle on it.

 

Post your schematic and some ideas you've considered and discarded. Have you considered making a switching regulator using an existing linear regulator? All you should have to do is add an external pass transistor, inductor, and diode.

 

Not to muddy the waters, but...
Look at this. I've used it several times.
And the '34063 price is right, too: as low as 53 cents at Digikey

 

Post your schematic and some ideas you've considered and discarded. Have you considered making a switching regulator using an existing linear regulator? All you should have to do is add an external pass transistor, inductor, and diode.

I hate posting schematics lol. The software I use... I like it for a variety of reasons, but pretty schematics isn't one of them. But here's the v-reg portion:



Ideas I've considered and discarded? Mainly, linear-only. We're too space constrained for effective abundant heat sinking and it is so wasteful.

No, I haven't considered making a switching regulator using a linear. I didn't think that was possible

Not to muddy the waters, but...
Look at this. I've used it several times.
And the '34063 price is right, too: as low as 53 cents at Digikey

Handy calculator, but not very precise. At least as a generic tool for other devices (might be perfectly fine for the MC34063A it is intended for). For example, my Cout is 2x 22p but that calculator goes from 1uF to 0uF with nothing in between.

I'm not sure where you get 53 cents... the MC34063A actually goes for under 20c at the stated volume. But I never considered it due to its restrictive operating temperature rating. Our device is used outdoors in sub-zero climates.

 

No, I haven't considered making a switching regulator using a linear. I didn't think that was possible

The Nat Semi LM317 datasheet has 2 examples. You can use LM117 or LM217 if you need a wider temperature range than LM317.

 

Look at the data sheet for an LM2576 (simple switcher). It uses a lot less parts than the RT7251.

 

The Nat Semi LM317 datasheet has 2 examples. You can use LM117 or LM217 if you need a wider temperature range than LM317.

Thanks for that, I'll take a look for future reference!

Look at the data sheet for an LM2576 (simple switcher). It uses a lot less parts than the RT7251.

And costs precisely three arms and two legs more! And it is enormous The smallest one takes up some 140mm2 of board space compared to the RT7251's 4mm2.

Again, I have it sorted now but thanks.

I also want to give props to RichTek USA for their service. They were prompt to answer emails and provide assistance, and they are even getting their distributor to send me a development/evaluation board for free despite the fact I was honest and told them I'd be purchasing from an overseas distributor (all the same to Richtek, but not to the distributor!). So for that alone I'm going to see how this chip performs.

 

So I looked at the LM*17... that's neat! I had no idea you could do that. However, I'm not sure on the practicality. Unfortunately, it isn't really a cheap chip to begin with. There are switchers for less money, so I fail to see why one would take a linear and make it a switcher unless it was to save money. I was expecting to see a chip that was costing, in volume, in the 5 to 10 cent range.

But can the application be extended to any linear? Or is there something particular about the LM*17 that makes it suitable? Because there certainly are linears in the 5 to 10 cent range that might be an attractive option.

For another day I didn't even have time to look that up, but you got me curious. (back to work)

 

Those 22pF output caps should be 22uF, I'm pretty sure... unless C24 is doing all the work.
Also, the cap-symbols look upside-down.

 

Ha! I was about to write, "I can read a datasheet, but thanks" but then I figured I better quadruple check... and you're right! They are 22uF. Thanks for that catch! Now it makes sense why the manufacturer was stressing in his email not to skimp on them. I was thinking, "skimp? They are tiny and cheap... what's to skimp?". Lol.

As for my upside-down cap symbols... they are not polarized. The software I use just doesn't have a straight symbol and I'm too lazy to rotate them. Although, now I better go quadruple check that they aren't supposed to be polarized!

 

So I looked at the LM*17... that's neat! I had no idea you could do that. However, I'm not sure on the practicality. Unfortunately, it isn't really a cheap chip to begin with. There are switchers for less money, so I fail to see why one would take a linear and make it a switcher unless it was to save money. I was expecting to see a chip that was costing, in volume, in the 5 to 10 cent range.

You must be referring to the price for the military version; LM117. The Industrial version doesn't command much of a price premium if it operates in the temp range you want.

But can the application be extended to any linear? Or is there something particular about the LM*17 that makes it suitable? Because there certainly are linears in the 5 to 10 cent range that might be an attractive option.

It will work with any 3 terminal regulator, but most aren't spec'ed for operation at less than 0C and ripple voltage will probably be worse.