Post by thread:[EE] 5V/5A (25W or so) SMPSU controller?

Hi guys, Does anyone have a favourite 5V/5A switch-mode PSU controller chip? I'm designing the PSU section of an FPGA-based device that needs 5V (for some pull-up resistors and LSTTL), 1.2V (FPGA Vcore), and 3.3V (everything else). The 3.3V and 1.2V supplies are generated by a Texas Instruments TPS75003 SMPSU controller. I was thinking about just using a 5V external, but then I started wondering about the possible effect of Sod's Law on my device: - If I use a standard connector (5.5x2.5mm coaxial DC power jack) then what happens if Joe "Dumb" User plugs in a 12V supply from something else? Obvious answer is "big frickin' boom". - It seems most DC plugpacks are either 4.5V, 6V or something higher, and typically top out at 12V. So I'm contemplating rigging the thing to take an input of 12V, then step that down to the 5V needed to run the TPS chip. But what I need is a chip to do that, and all the ones I've found are incredibly expensive (the TPS chip is £2+VAT; the 5V/5A chips I've seen are £6 and up). Does anyone have a favourite SMPSU chip that meets these criteria? - 12V (preferably up to 15V) input - 5V output, 5A (25W) - Low total cost, few external parts (I can live with an external power FET) - ~65% or better efficiency I've seen the NatSemi Simple Switcher chips (LM22677/8), but these seem to be "just a bit" on the expensive side for what they are. The other option I thought of was to use an MC34063 with an external FET or BJT.... but AIUI that's going to be about as efficient as a 7805. :-/ Thanks, -- Phil. spam_OUTpiclistTakeThisOuTphilpem.me.uk http://www.philpem.me.uk/

I'm sure I've used the LM2678 before -- it's an older chip, but worked well. (National's switchers work well with minimum of futzing). But I tend towards the newer chips as they run at higher clock speeds and passives come down in size, bringing footprint and cost down. Their webench tool gives you an estimated BOM cost so that's a good place to look for a net answer. Alternatively, isn't there some type of protection mechanism you could put in? Or perhaps change to a custom power connector? Cheers, -Neil. Quoting Philip Pemberton <piclistspam_OUTphilpem.me.uk>: {Quote hidden}> PICdude wrote: >> How about National Semi's LM22678-5.0? Simple and relatively low >> cost, but I'm sure they'll have others with a lower parts cost (though >> more components). >> >> http://www.national.com/pf/LM/LM22678.html > > That's the part I've been looking at -- £6 in singles and not much less > in quantity until you get to >1k-off... > > Its brothers, the LM2679 and LM2678, seem to be considerably cheaper. > They're still £6 in 1-off, but order ten of them and that's almost > halved. Total cost of about £8 for a pretty decent 5V SMPSU that'll take > anything from 8V to 40V on the input... > > And I doubt anyone's likely to have a >40V PSU with a 5.5x2.5mm DC power > plug on it :) > > -- > Phil. > @spam@piclistKILLspamphilpem.me.uk > http://www.philpem.me.uk/ > >

PICdude wrote: > I'm sure I've used the LM2678 before -- it's an older chip, but worked > well. (National's switchers work well with minimum of futzing). But I > tend towards the newer chips as they run at higher clock speeds and > passives come down in size, bringing footprint and cost down. Their > webench tool gives you an estimated BOM cost so that's a good place to > look for a net answer. I had a play with WEBENCH, and the stupid thing kept crashing out (usually taking Firefox with it). The parametric search thing isn't much better... At this point I couldn't give a flying f00 whether it's a synchronous or non-synchronous regulator, has an internal switch or wants an external FET, or whatever. All I want is a 5V / 5A switching regulator that can take 12V or so on the input (or more, more is good). At this point I seem to have a few options: - 5V 5A external PSU. Cost ~£25, comes with standard power plug. Problem scenario 1: someone plugs in a 12V PSU and blows the power sequencer / SMPSU chip and potentially the FPGA and PIC too (if the FET shorts out). Not good... Problem scenario 2: PSU plugged in has the wrong polarity. Same result as above. A nice big Schottky diode should stop this happening... Potential solution: custom power connector. I don't even want to think how much THAT would cost. - 12V 3A external PSU and LM2678. Cost ~£25 for the PSU brick, £12 ish for the National SMPSU chip and support components (and an RPP Schottky). Problem scenario 1: Someone plugs in a 12V PSU with reversed polarity. Add a Schottky diode across the power plug to push the PSU into short-circuit protect mode. Should work as long as the PSU is smaller than ~50W (i.e. not a Nemic-Lambda EWS600 or something equally brutish). In an ideal world I'd buy a box of Cisco ADP-20s and use them -- they're cheap on the surplus market, made by a fairly reputable PSU manufacturer (Delta), use a stock-standard Molex connector, and provide +12V, +5V and -12V. Problem is that I'm left up the creek if/when Cisco discontinue them (which would seem to be "soon"). What would be even better is a bunch of PSUs for external USB drive caddies -- but those seem to be extremely hard to find as spare parts. Why is it always the "simple" things that cause the most trouble? > Alternatively, isn't there some type of protection mechanism you could > put in? Or perhaps change to a custom power connector? See above -- a Schottky diode will deal with reverse-polarity, a Zener+SCR crowbar will *theoretically* deal with overvoltage (but IME they're slow to react -- slow enough that the SMPSU chip will be quite dead by the time the SCR triggers). To put it simply: I don't want to have units coming back from the field that have been damaged (needing a new crowbar SCR, Zener, etc counts as "damaged") just because some complete fool plugged the wrong PSU in. 8-40V input covers just about all the PSUs I've got within easy reach. I think I might be able to Rube Goldberg my Farnell L30B into giving me 60V, but that's pushing it. My two standard bench PSUs (Instek PSP2010s) are 20V/10A. -- Phil. KILLspampiclistKILLspamphilpem.me.uk http://www.philpem.me.uk/

Quoting Philip Pemberton <RemoveMEpiclistTakeThisOuTphilpem.me.uk>: > I had a play with WEBENCH, and the stupid thing kept crashing out > (usually taking Firefox with it). The parametric search thing isn't much > better... Webench is actually awesome, but I've noticed that the latest release of Firefox has been crashing quite often. I really like Firefox, but don't know how to stop the automatic updates, as I'm not an early adopter of any technology. But I digress. I'm on Webench right now, and it works fine. It tells me that the lowest-BOM-cost circuit is for a LM25085 circuit @ $2.69 ea. f/1K units. BTW, National's support is awesome. Even when I've told them in the past that I was building one unit for myself, they were very helpful in explaining things to me. > ... > At this point I seem to have a few options: > > - 5V 5A external PSU. Cost ~£25, comes with standard power plug. > Problem scenario 1: someone plugs in a 12V PSU and blows the power > sequencer / SMPSU chip and potentially the FPGA and PIC too (if > the FET shorts out). Not good... I really wouldn't do this -- I've learned that now matter how strong the warnings, if there's a chance that some user will wire up something incorrectly, they will. > Problem scenario 2: PSU plugged in has the wrong polarity. Same > result as above. A nice big Schottky diode should stop this > happening... Reverse-polarity protection is easy. Not your real issue here. > Potential solution: custom power connector. I don't even want to > think how much THAT would cost. Why not? I generic molex type connector with pins should be under US$1. The labour (<-- UK friendly spelling) to wire it up will be more. > - 12V 3A external PSU and LM2678. Cost ~£25 for the PSU brick, £12 ish > for the National SMPSU chip and support components (and an RPP > Schottky). > Problem scenario 1: Someone plugs in a 12V PSU with reversed > polarity. Add a Schottky diode across the power plug to push the > PSU into short-circuit protect mode. Should work as long as the > PSU is smaller than ~50W (i.e. not a Nemic-Lambda EWS600 or > something equally brutish). Why short-circuiting for reverse protection? Why not just inline? Option 4: How about a small circuit (PIC 10F? with zener regulation perhaps) that comes on when the power connector is plugged in. With a voltage-divider, it can monitor the power actually plugged in, and if correct, switch on the power to your main circuit. Haven't actually tried this, but it works awesome in my head right now. :) > In an ideal world I'd buy a box of Cisco ADP-20s and use them -- they're > ... Not familiar with those, but also look for PS'es from oil inkjet printers -- they put out quite a bit of power. >> Alternatively, isn't there some type of protection mechanism you could >> put in? Or perhaps change to a custom power connector? > > ... > To put it simply: I don't want to have units coming back from the field > that have been damaged (needing a new crowbar SCR, Zener, etc counts as > "damaged") just because some complete fool plugged the wrong PSU in. > ... But how about a simple fuse. Design for say 3.3V, make it 5V tolerant, but at 5V, there'd be enough current to pop the fuse. Again, I haven't drawn this up, but off the top of my head I can't see why it can't work. How many of these thingamajigs are you planning to make anyway? Cheers, -Neil.

PICdude wrote: > Webench is actually awesome, but I've noticed that the latest release > of Firefox has been crashing quite often. Yeah, I've noticed that. It seems to have issues with wheel-scrolling too -- if there's a tooltip visible (e.g. the ALT-tag for an image) and you scroll up or down, a grey box is left where the tooltip was... > I'm on Webench right now, > and it works fine. It tells me that the lowest-BOM-cost circuit is > for a LM25085 circuit @ $2.69 ea. f/1K units. Another part not available from Farnell... > BTW, National's support is awesome. Even when I've told them in the > past that I was building one unit for myself, they were very helpful > in explaining things to me. Beats ADI then. I reported an issue with a datasheet (ADV7125 video DAC) a few weeks ago. Aside from a quick email asking what my "application" and "expected production volume" was, I've heard nothing. Basically the datasheet specifies that the part has an internal voltage reference, but doesn't explain how to use it -- only how to use an external reference. >> - 5V 5A external PSU. Cost ~£25, comes with standard power plug. >> Problem scenario 1: someone plugs in a 12V PSU and blows the power >> sequencer / SMPSU chip and potentially the FPGA and PIC too (if >> the FET shorts out). Not good... > > I really wouldn't do this -- I've learned that now matter how strong > the warnings, if there's a chance that some user will wire up > something incorrectly, they will. Philpem's wording of Murphy's Law: if there's a way for someone to wire something up incorrectly, they will. Corollary to Philpem's wording of Murphy's Law: if it's possible for someone to wire multiple somethings up incorrectly in such a way that none of them work, they will. Corollary #2: if there are multiple ways to wire something up wrong, then the most damaging wiring method will be used. 100W soldering gun on a glass-to-metal seal on a vacuum tube anyone? >> Potential solution: custom power connector. I don't even want to >> think how much THAT would cost. > > Why not? I generic molex type connector with pins should be under > US$1. The labour (<-- UK friendly spelling) to wire it up will be more. 1) Changing the power connector makes it a "custom part". This means I have to pay for tooling costs, etc. if I want the manufacturer to do the work for me. 2) I hate custom power connectors. Manufacturers discontinue connectors with alarming regularity. 3) I don't want to cut the plugs off of 60-odd PSUs and crimp / solder on new ones. Too much effort, too mind-numbing and boring. I can just about live with SMD soldering (thank $DEITY for solder paste, stencils and Argos mini-ovens!) but when there's manual work involved, forget it. > Why short-circuiting for reverse protection? Why not just inline? You mean something like a forward-diode? Voltage drop over the diode, and the loss in forward voltage were the main issues I was considering. Though with a Schottky diode, that's not really a big issue... > Option 4: How about a small circuit (PIC 10F? with zener regulation > perhaps) that comes on when the power connector is plugged in. With a > voltage-divider, it can monitor the power actually plugged in, and if > correct, switch on the power to your main circuit. Haven't actually > tried this, but it works awesome in my head right now. :) Bit overkill maybe. You could do the same with a window comparator (dual/quad opamp chip and some resistors). You'd still need, say, a bridge rectifier to feed the Zener. Also, if it was MOSFET driven, the body diode would conduct if reverse-biased. So we're basically back to a forward-biased Schottky, a FET, and some other fluff... >> In an ideal world I'd buy a box of Cisco ADP-20s and use them -- they're >> ... > > Not familiar with those, but also look for PS'es from oil inkjet > printers -- they put out quite a bit of power. They're power supplies for the Cisco PIX and 1700 Series desktop routers. > But how about a simple fuse. Design for say 3.3V, make it 5V > tolerant, but at 5V, there'd be enough current to pop the fuse. > Again, I haven't drawn this up, but off the top of my head I can't see > why it can't work. "Yes, Mr. Jones, just send it back to us -- the fuse needs replacing." > How many of these thingamajigs are you planning to make anyway? At the moment, 30 or 40, though potentially a few more depending on demand. -- Phil. spamBeGonepiclistspamBeGonephilpem.me.uk http://www.philpem.me.uk/

Alan B. Pearce wrote: > 1N5819 type diodes are readily available, and seem to be reasonably robust. The 5819 is only good to an amp though. I've had a quick look and everything that's good to 5A seems to have a ~0.6V forward drop. It looks like a synchronous bridge rectifier (four MOSFETs) would cost about 80p, so it's not really worth whining about. It's basically two Vishay-Siliconix Si4599 Dual PN FETs (an enh-PMOSFET and an enh-NMOSFET in an SO8 package). Rds-on is 29 milliohms, so the voltage drop at full load would be (5 * (29e-3)) = 0.145V max. over each FET. Two FETs active at once (one PMOS, one NMOS) would give a voltage drop of a shade under 0.3V. 12V - 0.3V = 11.7V. The spec for the external power socket is 12V +/- 0.5V, so this should be within spec (though only within a fairly poor margin). > Also if using an inline diode, depending on the size of any filter caps and > total current draw, the unit should operate off an AC supply, if that is > what the operator plugs in. Hmm, would that apply to a 4-FET synchronous rectifier too? (common sense says "yes, as long as Vpk < Vgs")... For reference, I was thinking something along the lines of: <http://i1.tinypic.com/rvvhfn.jpg> (from here -- <http://www.eng-tips.com/viewthread.cfm?qid=149960>) -- Phil. TakeThisOuTpiclistEraseMEspam_OUTphilpem.me.uk http://www.philpem.me.uk/