ATX PSU as AC Adapter Replacement

[blake7984]

Hello,

I've got a network closet running about 10 various devices, each of which has their own AC adapter. After having a total of 3 of these adapters go out in the past year, I'd like to quit using them. About half are 5V, and the other half 12V. Current ranges from 750mA to 2.5A. Does anyone see any potential issues with getting an average $50 ATX PSU, adding a 10ohm 10W resistor across 5V line, a few binding posts outside, and replacing each AC adapter with power from the PSU?

It seems to me like a great solution to prevent anymore $20-$30 AC adapters from going out, be a little more efficient, and also to get rid of power strips and big adapters everywhere. Main thing I want to verify is that I'm not looking at some potential fire risk or any other serious issues from running this 24/7 in a non-normal way (outside of a computer, using a resistor and a couple wiring bypasses to fake a motherboard).

Thanks for the input,
-Blake

 

[MikeMl]

You need to check that all the devices you plan to power can share a common ground. Wall-warts are usually totally isolated; you are forcing all of your downstream devices to share a common ground, which might induce some nasty ground-loops between the various devices...

 

[Sceadwian]

If it's network equipment it shouldn't matter, the output of all 10/100 connections are transformer isolated because of the 802.X physical layer specifications, however as Mike said it depends on all the equipment you're using and what for. I think some gig ethernet adapters are DC coupled but I'm not sure and it might not even matter. If you really have that many adapters I'd say it's worth a shot. You can easily solder molex connectors to the devices so you don't have to use exposed binding posts, then all that's required is the switch and load resistor on the 5 volt line. Which may not be required because you're starting the supply up under load if you have 5V devices on it.

 

[blake7984]

Any good way to check if any of my hardware will get upset about sharing a ground? Here's a brief rundown of the devices:

2 Netgear Gigabit switches
1 Wi-Fi Router
1 Cable modem
1 Linksys VoIP Adapter
1 Silicondust HDHomeRun Network Tuner
2 ATI USB Tuners
1 USB Hub
1 Western Digital USB HDD

What issues would the ground loop cause here? I know what it would do with audio equipment, but with the stuff listed above, I really don't know.

As I said, my main concern here is safety. I'm usually not a big fan of hacking hardware that uses 120vac and runs 24/7 without my supervision, so I like doing my checking around on such a project.

 

[blake7984]

Bump?

$15 MIR on the PSU I plan to order ends today, so I was hoping to get some sort of confirmation on good idea / bad idea before days end. Anyone?

 

[Sceadwian]

I think worrying about issues with a common ground shouldn't really be issues at all, you'll have to try it and see.

 

[MrAl]

Hello there,


You'll have to have a way to power all the devices to use this method, so you'll probably have to have all the wall warts up and
running before you can start this procedure.

With everything up and running, you can try measuring the AC and DC voltages between grounds. If you dont measure any difference
you can probably get away with tying the grounds together. If you do measure a difference, you might try connecting a 1Megohm resistor
between the grounds to see if the voltage goes down a lot. If it goes down a lot, connect a 100k resistor and see if it goes down a lot
again, and if it does then connect a 10k resistor and again see if it went down, and if so then a 1k resistor, then 100 ohms.
At some point the resistor should bring the voltage down to zero or very close, at which time you can assume that you can connect
them together provided everything still works as usual.
It's difficult to say exactly what voltages you might see because there are a bunch of factors that would govern that. In general,
you dont want to see too much current flow between grounds and probably 100ua should be tops but again that is a bit difficult
to nail down too.
If it turns out that you can in fact connect them together, then you can try powering them with the ATX supply one at a time
until they all get powered.

Another interesting thing to think about is that when you connect an AC wall wart to a computer to power a device like a hub
or something, you are immediately connecting the ground of the wall wart to the ground of the computer (most cases).
When you power another device, you connect still yet another wall wart to the ground of the computer so both wall warts
are connected to the ground of the computer. This suggests that most grounds can be tied together, but the tests above
will help to ensure they really can be.

There are a few other issues that can come up too though. Some wall warts are stamped 12v but they really put out an
unregulated 12vdc that can peak much higher. If the device to be powered depends on that higher peak then a constant
regulated 12v might not work. Also, the device to be powered may rely partly on the internal impedance of the wall wart
that came with it (battery chargers are a good example of this) and so you would have to connect a series resistor in
series with the ATX supply for that device. The value of the resistor might be determined by measuring the current into
the device to be powered and then instead of connecting it directly to the ATX connect a small series resistor that
provides the same current as the wall wart did. This means you would probably have to start with say 20 ohms and work
it down to 1 ohm, and finally if the current doesnt increase above the wall wart measurement connect it directly.
Some resistance may be desirable anyway though, like 5 ohms, provided the device still works well.
I realize that this procedure can be a bit involved, but if you want to make sure you get it right it's a good idea.
Keep in mind that you only have to do this one time only.

Lastly, i can not guarantee that nothing can go wrong unfortunately because i am not there to see what you have exactly.

Another idea entirely is as follows...
The wall warts you are using are probably switchers, as these things seem to fail much faster than the simpler 60Hz
transformer type. You can consider building your own switching regulators for each wall wart that burns out, one
at a time as they go out. The basic design is the same for the different voltages, and because you can use a better
more reliable design you can expect to get many years of service life out of each one. That will keep everything
isolated and still provide long life.

 

[Sceadwian]

Well lets start with the list.

The gigabit router should be fine, I've never read anything that suggests gig ethernet isn't isolated like 10 and 100 Ethernet.

The wifi router is fine, as it produce RF.

The cable modem might possibly have issues as it may be internally grounded to the coax ground of the incoming cable connector. No idea what would happen there, prolly nothing.

No idea about the VOIP adapter, though I'm guessing it as well has isolated outputs for it's ethernet and the other connections should have any issues unless it's being fed through other audio eqpiment, but even then it won't harm anything. I can't think of any reason any of the other devices would have grounds tied in an unusual manner.

 

[blake7984]

Well, I just measured DC and AC voltage between the first two random devices I picked (HDHomeRun Network Tuner and WiFi Router) and it measured DC voltage starting low and the measurement kept increasing until around 300mV or so. AC voltage difference was about 43V, which to me seemed unusually high, right? Those are the only devices measured so far (its not an easy process in that closet), and no resistor tests done yet. With a voltage difference like that, is this something I should continue looking into or shoot it down now?

Sceadwian:
Good point with the cable modem and the coax. A similar issue could easily be possible with the 2 ATI tuners or the HDHomeRun network tuner too I would assume.


There probably are some interesting potential loops there after you mention that, since there are basically 5 devices (4 tuners + modem) with coax inputs, and theres also a coax cable amp plugged in shortly before the coax splits for all those tuners. Between the AC powered coax cable amp, the coax cable input itself, all those tuners, and the networking...I can definitely see some potential loops.

 

[blake7984]

Couple more quick tests I just did:

Between WiFi & Gigabit switch: same as before, 3-400mVDC, 43VAC

Between Gigabit Switch & HDHomeRun network tuner: 50mVDC, .5VAC

Seems like the WiFi is what was causing the high readings before, but no idea why.

 

[MrAl]

Hi again,


Well it would be better if you already had a spare ATX to start with so you didnt have to buy one first.

43vac could just mean that there is some residual ac because the two are already isolated and the slight capacitance is allowing
some ac voltage across the two grounds. You need to connect a resistor next, as per the procedure i outlined in a previous post.
This will tell you if the voltage is real or just a by product of the small capacitance. You have to realize that the meter has a
high impedance so it can see these voltages even when they are not of sufficient power to maintain that voltage across a
relatively large value resistor. In other words, a 100k resistor would probably knock that voltage down quite a bit. If not,
then that means there could be something else causing the voltage and you would want to be very careful going to 10k.
If 10k doesnt do it, i'd probably not continue.
Sometimes even a finger across the grounds will knock it down, but you have to be careful about getting a shock.
One finger only The resistors provide more positive results though.
Do you have any resistors around?

0.5v is almost nothing, so they are already almost equal.
For the 43 volter, you can also try rotating the wall wart so that the prongs go into different holes, if that is possible.

 

[blake7984]

Unfortunately most of my parts are elsewhere still until I can get them moved to this house. I did find a few resistors lying around though, and two of the ones still having long leads are 22k and 220k. Tried 220k which knocked it down to about 7VAC. 22k knocked it down to around .7VAC, so thats encouraging. Reversing the wall wart had no effect, but as you noted before, most of these adapters are switching.

I just looked around to make sure, but I'm out of extra ATX PSU's now. I used to have a few extras, but I guess they've all been used up or thrown away by now.

Guess I'll go back and measure a few more voltages and see how things go. I assume if the 22k keeps voltages under 1V or so, then I should be alright?

Out of curiosity, say I go through with all this and 2 or 3 of the devices get pissed about ground loops...what happens? Do they turn off, do they fry, do they wait 18 hours until I'm out of the house and then start a fire? (Hate being overly paranoid - but a coworker just lost his house to a fire, which has me on increased awareness)

 

[Sceadwian]

1 volt at 22k is only half a milliamp of current, hardly anything I'd worry about. The main problem you'll have is current will flow where it's not wanted, this will do any number of random things from affect the devices function to start the house on fire, mainly depending on the amount of stray current. From a homeowners perspective I'd be a little leary, in the off chance something could be traced back to the setup and it's determined that it caused the fire it could void your insurance claim, this is a bit paranoid but yeah it's possible, even if your setup isn't at fault and it's traced to an electrical fire in the same room, but things like that pretty much only happen in movies =P

 

[MrAl]

Hi again,


Oh that sounds great about the 22k resistor tests. It sounds like you are good to go with those devices.

As far as the effect of the ground loops, if the wires are short it should keep this problem to a minimum,
and the most likely result if any would be a lowered noise immunity with the signals that travel from
the device to the computer. You'll have to test everything and see if any devices start to slow down
or start noting errors. I dont believe that this would cause a fire however, that is usually caused by
high currents or arc overs from bad connections.
If you are that paranoid about fire then you might want to look into an enclosure and of course fire
and smoke detectors that can notify someone close by.
You can never really guarantee that a fire will not start no matter what you do, because something
somewhere, somehow can always go wrong. You can only provide for the detection of said fire
or smoke and make sure there are ways to subdue it immediately.

 

[blake7984]

Finished testing a good sample of the devices, other than the USB HDD and USB hub, neither of which I would suspect to be an issue (since as you noted before, many of those are designed to share grounds anyway). Never got more than a few volts measured, and always reduced to under 1V with the 22k.

Overall this sounds like a safe bet that should work out fine, but the paranoid side is kicking in and making me wonder if I should just stick to the normal safe approved (mostly working) method. If everything had a dedicated grounding connector on the outside of it (like the 2 gigabit switches do) it would probably make me feel safer. I'm sure after a day of it all running, it would pretty much prove it to be safe, but I'm thinking I would always be scared when going on a 4 or 5 day trip out of town.

With all this green talk everyone is doing lately, can't someone come up with a solid safe system for people with 5+ 5/12V devices to start sharing a single PSU? Surely I'm not the first person to want this....


Guess now its time to decide if I wanna put the work into this thing and have a hard time trusting it, or just accept having 10 ugly wall warts all over the place.

 

[Sceadwian]

Do consider the energy you're saving via efficiency from all those wall warts. They're honestly is no reason to have every input be isolated. If you've gotten low voltage readings with a 22k resistor, just use it. Heck, switch to a 10k resistor and see what the voltages drop to, if it's below a diode drop then you're probably just being paranoid. Then try 1k see what you get...

 

[MrAl]

Hi again,


Yeah i know what you mean here. There's always that nagging feeling that something might go wrong and then you would be kicking yourself
for the next 40 years. That's one of the things about life i guess, you can never know the future for sure.
You can however take steps to ensure a safe setup, after all, i think i would be paranoid leaving 10 wall warts running for 4 or 5 days too, not
only a computer power supply. Of course you have logged lots of run time hours with the current system, but that just lulled you into that
false security
Perhaps you can find a large metal cabinet and install everything in there with good detection systems and possibly even some sort of
fire extinguishing scheme (think sprinkler system) just in case anything does go wrong when you are not there. I think that's what i
would do if i had to leave my system running that long without some attention. I dont even let my computer run for five minutes
when i go out, i always cut the power completely.

 

[blake7984]

Power savings were definitely factored in, even though I honestly expect it to be minimal. Current usage of this setup is around 50W. I figure the *most* I could get that down to is 40W (most ATX PSU's aren't extraordinarily efficient either). If I actually saved 10W, that would be around $11/yr savings, not huge, but not bad.

I wish I had seen documentation of someone else doing this for an extended period of time with no issues, but surprisingly I couldn't find where anyone else has tried. Don't other people have networks like this these days??

I wish I didn't have so much equipment that needed to stay on 24/7, but guess thats just how it is. Having a large metal cabinet would probably make me feel better, but theres no room and that starts to bring the expense even higher .

If I don't go through with this, I hope I can at least find some more of these nice 12V 4A adapter packs, since they pretty easily handle the load from those power sucking switches (which is what fried 2 out of 3 of my adapters).

Decisions decisions....

 

[Sceadwian]

With that many devices? Not that many people really, not even that many people use a router unless it's with build in wi-fi.

 

[mneary]

I built a multiple adapter for a box of wireless mic receivers... ten male coaxial connectors like this 2.5MM DC POWER CABLES, 24 AWG 6' CABLE | AllElectronics.com or this 2.1mm DC Power | AllElectronics.com depending on the inner diameters required.

[edit] Your electricity rates must be low; my next kWh will cost $0.23, and if I exceed 307kWh, it jumps to $0.26.