High current switching supplies
 

I'm looking for a unit in the form of a power 'brick'.

Optimally, it is able to handle 20 amps at 12v. More would be nice if possible. I've found the large supplies with screened cases and such. I love those supplies for clean power but I don't want something that big. I don't need a 5v rail from the supply as the major load is from the CPU and video card which use the 12v rail primarily. I see cheap 10 amp ones on ebay but not 20+

Here is the application. I have a Dell GX270 (Small Form Factor measuring 4"x12"x13") desktop system. I love the small size of this thing but it limits my power supply options to... well, nothing really. I can only run the stock 160w power supply so this makes me need to either use an external supplemental power supply, gut the 160w supply and use a full size power supply for power tunneling through the stock supply, or open the stock supply and upgrade the regulators.

Using an external supply I feel is a bad idea. It makes for a big, heavy cable running into the back of the computer with a bunch of wires and it ruins the portability of the computer. The regulator upgrade MIGHT work if I can get heat sinking into that case but that is unlikely because of the size.

Those two reasons have led me to the supplemental supply. It will have a circuit that kicks it on and off with the system but can supply the processor load circuit and the video card with power.

This is the system specifications(pending power supply) *=maximum:
2.4GHz P4 HT 256k (3.2GHz P4 HT 512k* due to thermal limitations)
2x1GB PC3200*
200GB 7200rpm SATA HDD
128MB AGP Radeon x1050 (AGP 256MB Radeon HD 3450)
Creative Audigy SE 7.1

That's pretty much it. Not a blazing fast crazy computer but plenty for what I need. Light gaming with HD movies to come due to the new video card's ability to output component 1080i. This, however, will be my last upgrade as I've reached the thermal limitations of this system except for a blu-ray DVD for when it gets retired to HTPC-only use.

All the places I use you already read about in that supplier thread. 20A seems like a tough find. Could you use two of the cheapo 10A ones in parallel? Crack the cases and consolidate them perhaps?

Every time I use a power supply calculator online it gives me what seem like very inflated values. Anybody know a good one?

I might be able to do that. I could use two for the processor because it is a 4 pin setup that has two grounds and two 12v rails. Tonight I'm testing if a 60w 12v/5v combo supply I have is enough to run the video card. If it is I'll be able to use that for at least the card. If I could get my input connector to stay at around 9 wires or so it might not be that bad.

I'm going to have to do some testing. ATX specifications for the 'P4 connector' is 1 amp minimum and 13 amps max with a surge of 16.5 amps holding > 11volts for 10ms. I think a large capacitor could more than supply that and 13 amps is the max specification. I could use a single 12v/10a supply and see how loaded it gets. If I do that, I can get my wiring down to 1 sense pin, 3 ground, 2x12v rails, and a 5v rail. 7 pin connectors aren't hard to come by. I'm sure ratshack even carries that kinda stuff.

That would give me 180 watts of external, fanless power supply in a custom DIY box.

Do you have a UPS? If it uses a 12v battery maybe that could handle the surges for you. The capacitor idea definitely sounds reasonable if you've got a good supplier.

I've done some rough testing. The video card runs great off 2 amps 12v and 2 amps 5v from the switching regulator I have now so I believe the processor will run fine off of 10 amps.

Also turns out that I NEED a switch in line with the adapters so they turn off when the power supply turns off. With just the video card getting external power the dvd drive stays in an active error state unless the second supply is unplugged.

Can you use a relay hooked to the power supply to control the adapters? I don't know about you, but having to remember to throw another switch every time would drive me crazy.

A solution that requires less effort than relays would be to use the switch on a power strip, unless the computer automatically powers itself off (a la ATX). For auto-poweroff, it should be possible to pull the signal from the wire that tells the power supply to shut off and send it to the other power supply too.

I actually do the powerstrip thing myself. I've arranged my devices by function on a few of them so I can turn off anything I don't need. For example, if I'm running a simulation at night or doing some other computational task, I can kill everything but the computer itself. If I'm downloading something big overnight, I can just kill the speakers monitor, and my skype phone. Simple and a little crude, but it does the trick.

I'd actually prefer to not have the supplies come online via main power switching as I don't know what kind of start-up surges they may have.

The way I was thinking of doing it was to run a 12v line from the existing power supply to a couple of double pole relays in the main box that activate when the system is turned on. There will be a slight delay but not enough to make a difference I don't think.

This is also the preferred method if one day I am away from home and the power blinks and the computer itself doesn't come back on but the power brick does. Right now I have the computer set to automatically come back on in the event of a power loss but you never know.

There's plenty of pinouts and documentation on ATX power. You should be able to just tap the wires that signal the power supply to power up/power down and use them for the second power supply.

https://skylab.org/~chugga/mpegbox/pi...conatxmain.gif

I keep meaning to make one of those power strips that tun off when your USB port goes into sleep mode. I have a friend who could use it as he sleeps his mac mini most of the time, but the ready made ones are out of his price range. Make a nice gift for him.

I thought about that, but the ATX supply is turned on by shorting 'power on' with a ground so there is no real signaling there. I'm not sure how computers are built now but I'm sure an extra 90ma from 3 relays grounding through the atx on circuit will probably cause too much current to be flowing.

Toss a potentiometer on there and see what level it comes on at.

Usually the motherboard has the power on off signal from the front panel and it feeds through the motherboard to the power connector controlling the power in the supply from a connection pin 14 which you should be able to wire into the seond power supply and have them both turn on and off together. I don't see a need to have that many amps from the 12 volt supply as the hard drive takes less than an amp and the extra power to the video card is usually handled by the extra 4 pin connector from the power supply that would plug into the motherboard near the video card slot to provide extra current to it. Usually the processor takes 25-30 amps at 3.3 volts and regulates it down to the 1.6 volt or whatever a Intel chip needs with the switch mode regulators on the motherboard. The other thing you might be able to do is to take a bigger supply apart and stuff it into the case without an enclosure since the supply case is usually full of air anyway. Make sure you have a supply with the 24 pin power cable if the motherboard requires it so that the power connection pins are not overloaded.

I wish it was as easy as wiring into the motherboard connector but it really isn't. The green wire (pin14) needs grounded for the power supply to go active. I don't know what kind of current grounding that pin takes so I don't know if 2-3x relays with 30ma each will overload the switching circuit or not. I can only assume it would because we are talking about small SMD transistors.

As for the processor power. I see that the 3.3v rails have a lot of power through them and it makes 100% sense that the 3.3v would be regulated down but was wondering if you knew why the 'P4' connector is used because of higher processor power demand now and it is 12 volts instead of 3.3v?

https://www.playtool.com/pages/psucon...onnectors.html
https://www.formfactors.org/developer...12V_PS_1_1.pdf

This is where I'm getting my info from.

Well, having looked that up, I needed a standalone power supply this morning for a hard drive hooked up via an IDE->USB adapter whose power brick had gone missing. I looked for the nearest wire-like item and found some solder wick, stuck it in the green pin, and sat the power supply body down on it. Success! :)

I did learn something. I wasn't sure if the wire takes a pulse and the PS latches that, or if the wire has to have a constant ground. It does need to stay grounded, when you unground it, it powers off again.

Anyone know how to de-pin the ATX connector so I don't have to cut the wire to make a semi-permanent connection?

If you look into the connector each pin has two tabs on it that oppose each other. They are easily bent inwards with a tiny screwdriver or paperclip and it pulls right out.

I have read that some ATX supplies don't like to run without a motherboard connected to them because the PWR_OK line (Grey, pin 8) stays silent and the supply starts to swing above and below normal on the rails trying to make power that will give it the OK.

Yeah you should have loads on each of the major outputs since they rely on the 5 volt load to keep the 12 volt output from climbing too high since they come from the same switching transformer on separate windings they tend to interact a little. If you get into the actual motherboard specs you should be able to find out how much power is going into the various supply voltages. The newest power supplies provide a 4 and 6 pin connector for powering up the video card slot depends upon what the motherboard needs plus there are some extra power connectors for the processor - intel needs some extra power usually and not AMD.
The Cooler Master 460 watt supply I have here has 3.3@22a 5@25a 12@18a 12@18a YEAH two 12's -12@0.5a 5Vsb@2.5a
This supply is rated an Intel Form Factor ATX12V V2.3 and over 70% efficient at typical load.

what you could do about having to use 2 switches just use a STDP switch

Single Throw (one lever)

Double Pole (2 separate switch contacts that turn on and off at the same time)

Ever get this functioning? I'm curious how the capacitor for peak loads worked out.

Worked great!

The two bricks feed 3 capacitors directly, a 5v and two 12v. All of the caps are 16v 8,200uF caps. The voltage out of the bricks are .3v over what the normal rail is so it ended up being perfect (the motherboard runs off the bricks until it pulls their voltage equal to the power supply voltage where current demand swings quickly to the normal PS again). Across the caps I have 10k ohm resistors to apply a constant load and discharge the caps when I unplug the unit.

I have a sensor pin from the stock 12v supply voltage that gives me my on/off switch. The relay is after the capacitor so the regulators can keep the caps charged while the computer is off and then start-up current isn't having to be fed while also charging the caps(defeating the purpose of them).

Since they are switching supplies they don't go up in voltage beyond their rating as demand decreases like a lot of linear supplies with transformers do. I honestly don't think that I needed caps that large but it's better to be safe than sorry!

That's great. Pretty big caps! I noticed that all electronics has some 1/3 F supercapacitors for $1.15 which is what reminded me of your project. Only 5.5V though but I guess one could use several.

Unfortunately, you can't use several to increase voltage handling, that is a fixed limit. You can run them in parallel to increase capacitance though.

Yeah, they are big, I don't think the bricks like charging them either as they sound funny when they are first started driving those caps versus when they aren't on there. I've considered limiting the current from the bricks to the caps but I don't expect the regulators to be plugged in and unplugged a lot anyways.

Those super caps are sweet! They are meant for slow-discharge like in a solar powered LED torch or something similar... Project time!

Placing the capacitors in series should allow higher voltages, although you may need parallel large resistances if the leakage current isn't high enough. Of course you'll be sacrificing capacitance with the series setup. And it's a pain.

What project have you got in mind?

The voltage rating of a capacitor is the voltage at which the capacitor can safely run before the dielectric begins to break down. Running them in series will still cause you to go over that because on the negative side of the most positive cap will have 0 volts and the positive side will be over. The dielectric will break down and then the next cap will be over and it'll continue til every one of them is dead.

Not sure on the project, I'll come up with something!

Check it out, useful trick:

https://books.google.com/books?id=nMB...um=7#PPA339,M1

I see what they did there, they made a voltage divider using the resistors across the caps. That will work, for small headphone amps I do that to create a virtual ground.

Just know that circuit won't work without the resistors. Without the resistors they'll die like I mentioned above. :)

This is one of those tricks I learned in college that I've never had an opportunity to try out. Most of my stuff is digital.

You could actually use varying ratings if you put some thought behind it. If you have a 100v(DC) line and you put a 5k and a 10k resistor across it you'll have 33.3v and 66.6v respectively. Almost perfect for a 35v cap and a 63v cap.

What sort of longevity hit do you get if you go right up to the maximum limit? Or is there a safety factor in there already?

I'm sure there is a safety factor in there, almost everything has one. Usually electronics are de-rated by 10%. The exception is stuff like large motors but they usually have a specified amount of time you can run them over.

I forgot to go into additional details on what happened. The power situation is fine but the cooling turned into a problem as the stock heatsink is capable of cooling the 2.4GHz at 66.2 watts TDP but falls behind grossly on the 3.2 with 82 watts TDP. The funny thing is that I didn't buy the Pentium 4 3.2 Extreme with 1mb cache because it dissipates 89 watts and I thought the cooler couldn't handle that. Maybe I'll pick one up now... hehe

The 'hot rod' heatsink I got is 41mm tall instead of the stock 32mm causing me to have clearance issues in the small form factor case. The stock heatsink/fan is 64.26mm from the top of the fan to the processor core so I'm modifying the case to take a 120x120x25mm side fan that will just BARELY clear the inside of the case. But, that 120mm fan will more than keep my computer cool.

I'll finally be done modifying this thing as I'll have maxed it out.

Holy crap!

Pentium 4 Extreme Edition 2MB

I spent a fair bit less than that on all the new components for my current system.

I was thinking that if a larger number of caps were used and the voltage limit approached then resistor tolerances might start to matter. I imagine you'd need to use quite a few for it to really add up, though. Using better ones or just measuring them should get around it.

I don't know that I would do it to make up for voltage tolerances. Honestly, caps are so cheap now there really is no reason to skimp out on it.

When I make a virtual ground using two caps and two resistors I always make the caps meet full rail to rail voltage. If the virtual ground goes unstable for some reason and you get full input voltage across the lines you are still safe.

For example, I'll use two 16v caps in the supply even though the maximum input is 12v. I COULD use two 6.3v since each side of the virtual ground should only be 6v but if it happens to not be I don't want to be replacing caps because i was being lazy or cheap.

Sorry, I was just having fun speculating. I get the conceptual equivalent of a song stuck in my head a lot. Even bothered to run some numbers of the tolerances...it's a sickness. :) I can't really see where I'd use it, either. More of a neat gimmick than something practical.

You said you are into digital so it is a neat gimmick when it comes to that. I deal with analog and vacuum tubes most of the time and it can be a money/time saver when you only need two resistors and a couple caps to create a ground, +V, and -V instead of using a rail splitter with a cap in front of it and two behind it. It isn't as efficient or as stable as the rail splitter but it is easier.

Yep, I said it, I work with tubes. I love my tubes!

What do you use them for...audio?

I prefer my tubes tied, personally. :p

I do use them for audio but also power supplies, they are the ultimate in soft-start. I've made a few tube headphone amps and am in the process of building another one.

I'm going to be building a tube power amplifier for my new surround sound but I have to get my moving situation figured out first. They'll be interesting, I'm doing a point to point build using nothing but the schematic. It uses 11 tubes for 2 channels of audio(excluding power supply). The mono amp for the sub woofer uses 8 if you count the ballast tubes in the power supply