PSU conversion - help needed please!

[seven7seven]

Hi all,

I am busying converting an ATX PSU to be used as a normal standalone 12V DC power supply for someone (like what this guy did here).

I have bought an Aopen 350W ATX V2.2 PSU (link) to use.

Unfortunately I cannot find any specs for it regarding the minimum load requirements (so that I can workout what resistors to use as dummy load.).

I did find Intel's ATX 2.2 specifications, and they say the min. loads are (for a 350W):

12V1 - 1A
12V2 - 1A
5V - 0.3A
3.3V - 0.5A

The specs for a similar PSU, a Sparkle 350W V2.2 (link), say:

12V1 - 1A
12V2 - 1A
5V - 1A <--- the only difference to the official ATX specs above.
3.3V - 0.5A

The various guides that I've read on how to do this PSU conversion only mention a min. load requirement on the 5V line, and don't say anything about needing a min. load on the 12V and 3.3v.

So, my two questions are:

1. Do you I need to put a min. load on 12V and 3.3V in addition to the 5V?

2. What dummy loads must I use (incl. for 5V)?

Thanks very much, I really appreciate the help!!

God bless & kind regards,

Richard

 

[Yotch]

I would first try to get the power supply to turn on by connecting the green wire to any one of the black wires with a short piece of wire just gently pushed into the 20 or 24 pin connector. See these links for more details.

http://www.dansdata.com/images/digidoc/runatx360.jpg
http://leoricksimon.blogspot.com/2006/08/atx-psu-diagnose.html
http://www.aitechsolutions.net/pchwtrblsht.html

Once you have the supply running you should be able to decide if you need a load on the 5V supply or not depending on the voltage that you read on the 12V terminals or if the supply switches off after a few seconds. It seems that a 10 Ohm 20Watt wirewound resistor should be an adequate load across the 5V supply if necessary. I don't think that you will need to have any other loads. What are you planning to power off the 12V rails? If you need to add a resistor you will need to carefully check that it is not going to overheat after several hours of use - if it gets very hot you will need to increase the rating to 30 or 40W. I wouldn't recommend buying a new supply for this kind of application as you can buy very cheap secondhand ones on Bidorbuy for <R50.

More relevant links.
http://www.wikihow.com/Convert-a-Computer-ATX-Power-Supply-to-a-Lab-Power-Supply
http://forums.techpowerup.com/archive/index.php/t-25.html
http://www.hardforum.com/archive/index.php/t-710299.html

 

[seven7seven]

I would first try to get the power supply to turn on by connecting the green wire to any one of the black wires with a short piece of wire just gently pushed into the 20 or 24 pin connector. See these links for more details.

http://www.dansdata.com/images/digidoc/runatx360.jpg
http://leoricksimon.blogspot.com/2006/08/atx-psu-diagnose.html
http://www.aitechsolutions.net/pchwtrblsht.html

Once you have the supply running you should be able to decide if you need a load on the 5V supply or not depending on the voltage that you read on the 12V terminals or if the supply switches off after a few seconds. It seems that a 10 Ohm 20Watt wirewound resistor should be an adequate load across the 5V supply if necessary. I don't think that you will need to have any other loads. What are you planning to power off the 12V rails? If you need to add a resistor you will need to carefully check that it is not going to overheat after several hours of use - if it gets very hot you will need to increase the rating to 30 or 40W. I wouldn't recommend buying a new supply for this kind of application as you can buy very cheap secondhand ones on Bidorbuy for <R50.

More relevant links.
http://www.wikihow.com/Convert-a-Computer-ATX-Power-Supply-to-a-Lab-Power-Supply
http://forums.techpowerup.com/archive/index.php/t-25.html
http://www.hardforum.com/archive/index.php/t-710299.html

Thanks for the reply!

I think I will try with a 4.7Ohm (~1A load) on the 5V to be safe. Will a 10watt be fine if the PSU is used continuously for a few hours? From my calculations it will only dissipate ~5W, but do you recommended I rather use a higher wattage such as 17W/20W?. I will mount the resistor on the PSU chassis (will try to put it in the airflow too).

Another q: what percentage of the rated max. current can I safely, and continuously, put on the 12V rail? This PSU's max. output is 8A on 12V1 and 13A on 12V2 (does that mean there is total 21A on the 12v rails available??). I might, however, return this PSU and end up using a used PSU which is rated at max 10A on the 12V rail.

I guess it does make more sense to use a 2nd hand one rather than a new one, since I'm gonna void the warranty anyway (mounting the resistor, removing unneed cables), thanks.


Thanks so much!!

 

[Yotch]

In some of those links that I posted the recommendation was for a 10 Ohm >10W resistor. If you want to double the current then you also need to double the wattage i.e >20W. You can buy power resistors that have an aluminium body which are better at dissipating heat via a heatsink - there is a picture in one of the links. You don't want the resistor to get too hot otherwise it could malfunction over time or even start a fire. If it gets too hot to hold then I would say that you need to either increase the wattage or to improve the cooling. Remember that a small soldering iron generates about 30W of heat. In the article that you posted the resistor has a ceramic body which would be less effective at conducting heat. You must decide whether the benefits of a 1A load actually offset the complications of generating more heat in the resistive load. One person commented about using a fan on the 5V supply instead of a resistor which I think is quite a good idea. Alternatively you could even use a suitable light bulb as a load and then not have to worry about it overheating.

It may even be simpler if you buy an old AT type supply - these can be bought online for a few rand and don't require any extra control signals to turn on. You really need to try out the supply on the intended load to see if the extra 5V resistive load is required. What are you planning to power that needs so much current? Does the intended load need a very accurate 12V supply? Most electronic equipment is quite flexible and will run on say 11V instead of 12V.

Regarding the maximum power output for the 12V supply, I would be reluctant to draw more than 50% of the maximum figures continuously. It may well be possible to use nearer to the maximum figures particularly if no power is being drawn from the other rails but just remember that if a switching power supply blows under load that it can destroy any load that is connected to it. Lots of people have destroyed motherboards/CPUs and graphic cards like this. It may be wise to be a bit cautious. Look around on the Internet to get feedback from people who have actually drawn these type of currents successfully. I would e-mail the manufacturer of your particular supply to make very sure about the 12V1 and 12V2 power outputs because if there is any problem you could end up with a catastrophic failure. If you calculate the total power of the supply by summing up the power contributions from each rail you can see that the total exceeds 350W which makes me think that all of these figures are very suspect. I would think that continuous currents of 4A on 12V1 and 6.5A on 12V2 should be safe. How are you planning to combine the outputs from the two 12V supplies to supply a single load?

If you are in the Johannesburg area this company often sells old AT or secondhand ATX supplies. Phone first to confirm. Bidorbuy is usually cheaper though - search for power supply.

http://www.starsun.co.za/starsunimp/index.php

 

[seven7seven]

Hi Yotch,

Thanks a lot for the help!

Sorry I forgot to mention in my previous reply that I am going to use the PSU to power a bunch of CCFL kits. They each draw about .75A, and I would need to power 8 or 10 of them (so would need to draw 6A - 7.5A in total) . Each CCFL kit has its own separate power connector so I plan to just plug some into the one rail, and some into the other, would that be ok?

I have no idea how sensitive the CCFL's inverters are, hopefully not too sensitive though! My reasoning for wanting to put a larger load on the 5V is to prevent the PSU from being damaged if the load is not large enough. But I see the ATX v2.2 specs require that the PSU must not be damaged when switched on with no load, so it is safe to say that the only danger would be the PSU shutting itself off? In that case, I will first try the PSU with no load.

I realised that the load won't actually be continuous, it will only be used sometimes (maybe once a week or something) and when it is used, it will only be on for a few hours (max 24 hours at a time).

If I used one of the aluminum body resistors, could I just mount it directly onto the PSU's chassis?

 

[Yotch]

What you are planning to do should work fine as you have several loads which you can just power off either rail. Seeing as you will only be using the supply to power a group of CCFL tubes I would look for a cheap secondhand 300W ATX one since this application really doesn't require anything fancy. I recently bought a 250W ATX supply for R30 off Bidorbuy. You have to check regularly though as they sell quickly. You could even power those tubes using several 12V 2A DC power packs or 12V battery chargers although the ATX supply will probably work out cheaper.

Regarding the resistor, you should first see what you can get and how much it is going to cost you. It's unlikely that you will be able to find the exact value that you want and you may have to settle for what is available. It seems that the aluminium body resistors are quite expensive ~R30 in which case a much cheaper ceramic one ~R4 with a higher rating would probably be adequate. You may find that the CCFL tubes are producing too much heat and decide to change the resistor value at a later stage to lower the 12V supply voltage in any case so the cheaper resistor may be a better option. In that article that I posted the author just bolted the aluminium resistor directly to the chassis. Alternatively you could just use several cable ties to tightly bind either the aluminium or the ceramic resistor in place through the air vents in the chassis to avoid drilling the chassis. Try to get some suitable heatshrink tubing to insulate the solder joints to the resistor etc. as electrical insulation tape is not suitable and a short circuit across the 5V rail could be quite serious. You need to choose a diameter that will shrink down to insulate the joints tightly i.e. neither too big nor too small.

Computer switching supplies are normally quite robust and they shouldn't blow when no load is present but it's not impossible. You may need to fiddle around a bit with a resistor or even some of the other ATX signals like power good etc. to get the supply to turn on. Just take a look at that wiki that I posted. Hopefully the supply turns on at the first attempt without any problems.

This is copied from the wiki - I'm not sure if this includes the 12V supply - just experiment yourself.

"ATX power supplies are "switch-mode supplies"; they must always have some load to operate properly. The power resistor is there to "waste" energy, which will give off heat; therefore it should be mounted on the metal wall for proper cooling. If you will always have something connected to the supply when it is on, you may leave out the power resistor."

 

[seven7seven]

Thanks for all the advice. I got the PSU to turn on (and stay on) by doing the green wire trick thanks

When there is no load on the PSU the 12v rail measures at about 11.99 - 12.03. However, when I start loading the 12v rail the voltage starts to drop. With a ~4A load the voltage drops to 11.68V. When I put the 10ohm on the 5v line the voltage improves to 11.80 though.

I'm gonna test the voltage when I put ~8A load on. I don't know how close to 12v these CCFL inverters need though, because maybe ~11.5 is fine which would mean I don't need to use any resistors. What do you think ??

BTW I'm using a 10hm 17W cement resistor - after a few minutes it reaches ~60c! When I put it in the airflow of a fan it stays at ~40c. So if I do have to use the resistor I'm gonna try get an aluminum clad one and mount it on a heatsink with a fan.

Thanks for all your help!

 

[The_Unbeliever]

Good morning

Whilst on this subject - is there any schematic diagrams for building a PSU with only 12V as input, and which you can use to power a standard PC off any 12V battery?

My reasoning is this - it is more efficient this way than to reduce 240V to 12V, charge a battery, and step it up to 240V again. (UPS)

Regards

Libs

 

[seven7seven]

Good morning

Whilst on this subject - is there any schematic diagrams for building a PSU with only 12V as input, and which you can use to power a standard PC off any 12V battery?

My reasoning is this - it is more efficient this way than to reduce 240V to 12V, charge a battery, and step it up to 240V again. (UPS)

Regards

Libs

Sorry I don't know about that:/ maybe try checking some car PC forums as I'd think they might have a solution?

To add about the CCFLS: Unfortunately I don't have a datasheet for these inverters, but having a look on the web at the datasheets of a few other 12vDC CCFL inverters, they seem to say 12V +- 10% which would make 11.5v fine. What do you guys think?

 

[seven7seven]

The CCFLs seem to be working perfectly at ~ 11.5V.
I am, however, a bit concerned of the possibility that running the PSU without any load on the 5V rail (but with the 12v rails loaded) could be dangerous It would be a lot easier to leave out the 5v rail resistor but if I need it then I need it...

 

[Gnome]

The CCFLs seem to be working perfectly at ~ 11.5V.
I am, however, a bit concerned of the possibility that running the PSU without any load on the 5V rail (but with the 12v rails loaded) could be dangerous It would be a lot easier to leave out the 5v rail resistor but if I need it then I need it...

Is this for a engineering course or something? (just interested)

Anyway I built a lab PSU that seems to work so far, I basically bought two 10W wirewound 6.8Ω resistors for the +5v rail and put them in parallel, no need to cool them the resistors are designed to dissipate 10W of power and the air blowing from the fan alone will be more than sufficient.

Here's the math:

Ωparallel = 1/6.8Ω + 1/6.8Ω
Ωp = 2/6.8Ω
Ωp = 6.8Ω/2
Ωp = 3.4Ω

V=IR
5v = I(3.4Ω)
5v/3.4Ω = I
1.47A = I

P = IV
P = (1.47A)(5v)
P = 7.35w

As you can see the resistors have plenty of headroom and you'll dissipate around 7w of current (the voltage is according to the ATX standard +-10% in the range of the reported voltage). You can bump that to 10w if you want although I doubt you'd need to. I did the same for the 12v rail where, it's around 8w or so, basically I aimed for 10W but got those final values because of the E12 resistors values that were available to me.

EDIT: I might seem a bit unclear from the way I typed but they are connected in Parallel, so they're combined resistance was 3.4Ω therefore they dissipate less than half their rated current...

 

[Gnome]

Good morning

Whilst on this subject - is there any schematic diagrams for building a PSU with only 12V as input, and which you can use to power a standard PC off any 12V battery?

My reasoning is this - it is more efficient this way than to reduce 240V to 12V, charge a battery, and step it up to 240V again. (UPS)

Regards

Libs

I am quite interested on the subject of power conversion and did some research on the subject and the easiest answer to your question, well at least as far as I could see that unless you buy a powersupply that can give you a +12v DC regulated output from +12v DC you might as well give up while you're ahead, building one is very difficult!
The +5v, +3.3v, -5v, -12v rails are far easier tho...

 

[The_Unbeliever]

I am quite interested on the subject of power conversion and did some research on the subject and the easiest answer to your question, well at least as far as I could see that unless you buy a powersupply that can give you a +12v DC regulated output from +12v DC you might as well give up while you're ahead, building one is very difficult!
The +5v, +3.3v, -5v, -12v rails are far easier tho...

And what should there be two batteries, giving 24VDC - so you can regulate it to 12VDC? Should be better - so if the voltage of the batteries dip, then it will still deliver 12VDC up to a point.

Or am I mistaken?

I'm thinking of using solar power to charge the batteries during the day, and eksdom power (if available) at night...

 

[Gnome]

And what should there be two batteries, giving 24VDC - so you can regulate it to 12VDC? Should be better - so if the voltage of the batteries dip, then it will still deliver 12VDC up to a point.

Or am I mistaken?

I'm thinking of using solar power to charge the batteries during the day, and eksdom power (if available) at night...

Yes... Although I have to apologize because I think I have misled you a bit. Creating a DC->DC regulated computer PSU could *possible* be more efficient if you disregard the +12v rail. If it is included tho, it might be more efficient to use the off the shelf Inverter and standard off the shelf PSU combo even if you have +24V DC, the DC->DC route will probably be less efficient still

This is a bit hard to explain without writing a novel but in short you'd need a switchmode powersupply to produce a efficient regulated DC->DC powersupply. And switchmode powersupplies work internally with AC, so even if you were to create a switchmode powersupply that uses 12vDC as input and gives you all the ATX 2.0 spec'd outputs it would still convert the 12vDC to some form of AC.

The inverter is basically a extra step, so if you were to use a linear regulator it *could* be more efficient in certain circumstance (IE. low current, say <10-15amps, such as the +5v rail for example)...

I don't know too much about the subject tho, I got interested in electronics after doing 2 engineering courses @ varsity for extra credit but we only did digital circuits and microcontrollers (Like PIC's, 4000 series IC's and so forth), so the stuff I know about PSU and so forth I researched myself and I might be completely wrong.

Anyway the reason I said it's best to give up is even some of my friends in their final year of BIng(EE) consider the switchmode powersupplies quite complex circuits, building one yourself will not only be complex it'll also take lots of time, so unless you can buy it off the shelf it's best to use the Inverter -> PSU configuration.

Those setups can still be quite efficient tho, the inverter can usually be up to 90% efficiency and the PSU about 80%-85% efficient so let's say you have a computer that uses 400W (the real amount of power it uses) then your overall power used would be: 530W(for a inverter with 90% and a PSU of 80% efficiency) which isn't too bad! That's a overall efficiency of 75%

EDIT: As a sidenote most PSU's are more efficient when running with 110V AC @ 60hz than 220-240V AC @ 50hz so if you have a universal PSU you can squeeze that extra bit of efficiency out of them by using a DC->110V AC inverter, although that would mean if you wanted to run anything other than the PSU from it, it would also have to support 110V AC.

 

[Yotch]

@Richard7: Good to hear that you managed to get it working. I don't think that you will have any problems running your CCFL tubes at even 11V - I think I read somewhere of people doing this to stop them from heating up too much. Regarding the 5V resistor - I'm not sure whether it could cause long term damage to the supply if you leave it out. Maybe you could e-mail the manufacturer first to get their opinion. You could try to monitor the temperatures of the components in the supply like the toroids and switching transistors/FETs etc. both with and without the 5V resistor to see if there is any obvious temperature change. Maybe you could even use a digital or video camera that is infrared sensitive in a darkened room to do some tests - apparently many digital cameras can resolve the IR from a TV remote - so they might also be able to find the hotspots. Be really careful though that you don't touch anything while the supply is on as you could get a really nasty high voltage/high frequency shock. If you switch the supply on for say ten minutes and then switch it off and then try to measure the temperatures on the different components you may be able to get a good idea.

You might also consider trying one of those high power/high brightness Lumileds luxeons instead of a resistor. Maybe you could somehow incorporate it together with the CCFL tubes. I think that Mantech and Electrocomp keep them but the prices vary quite a bit.
http://www.mantech.co.za/Stock.aspx?Query=luxeonand See Page 2

@The_Librarian: Just search for the thread - A boer maak 'n plan to see how one South African is using two car batteries to power his home office. If you want to switch from mains to battery I think it's going to be a lot easier in the long run if you rather look at a more energy efficient computer to start with. Desktop PCs use quite a bit of power on their own and usually have an external screen etc. Even after spending a considerable amount on car batteries and inverters etc. you are at best going to get 2-4 hours use betwen charges. I would seriously recommend that you rather look into a cheap secondhand laptop that you can power off its own or a car battery or one of those $200 Asus laptops that are supposedly going to be available soon or possibly a Mac Mini. If you want to use solar power effectively you need to have energy efficient appliances. Consumers will have to get used to the idea of using much less energy if they want to switch to solar power with the current technology.
http://crave.cnet.com/8301-1_105-9760930-1.html

 

[seven7seven]

Thanks guys!

Gnome: Nope this is just for a customer (he bought the CCFLs from me).

The PSU needs to be ready by tomorrow so unfortunately I don't have time now to email manufacturer or do heat testing, so I think I'm going to just use the one 10hm 17W resistor on the 5v rail.

I'll mount the resistor with cable ties on the outside of the PSU at the back, on the exhaust fan grill/vent. Doing so will block some of the PSU's exhaust airflow but not too much, do you guys think that will be safe?

 

[Gnome]

@The_Librarian: Just search for the thread - A boer maak 'n plan to see how one South African is using two car batteries to power his home office. If you want to switch from mains to battery I think it's going to be a lot easier in the long run if you rather look at a more energy efficient computer to start with. Desktop PCs use quite a bit of power on their own and usually have an external screen etc. Even after spending a considerable amount on car batteries and inverters etc. you are at best going to get 2-4 hours use betwen charges. I would seriously recommend that you rather look into a cheap secondhand laptop that you can power off its own or a car battery or one of those $200 Asus laptops that are supposedly going to be available soon or possibly a Mac Mini. If you want to use solar power effectively you need to have energy efficient appliances. Consumers will have to get used to the idea of using much less energy if they want to switch to solar power with the current technology.
http://crave.cnet.com/8301-1_105-9760930-1.html

It depends entirely on the computer, the only way to know for sure is to buy one of those plugs that measure the watt/hour rating of the appliance, then check the computer at idle and at full load, from what I've seen most decent PC's use about 250-300w on load including power lost to inefficiency (If you bough a energy efficient setup it'll be far less). If it's 300w with inverter efficiency included you get +-330w.

If you're running that from a quite cheap 45Ah battery you'll get almost 2 hours on just one battery. Make it 4 batteries and laptops can't possible compete, most laptops have 2 hours on full load anyway....

 

[Yotch]

I'm sure that what you are doing with the resistor should be fine. Hope everything works as planned. If the resistor starts turning brown in colour after a few hours of use you may need to think about using either a higher wattage ceramic type or an aluminium one otherwise it should be fine. These resistors can usually withstand quite high temperatures for hundreds of hours.