a cheap power supply?

[Alexontherocks]

Hi everyone. I have already posted in this section to seek some help with an ambitious project envolving a high powere TEC. Unfortunately I have decided to concentrate on the watercooling which is on hold due to exams.

Thinking about TEC I asked myself the everlasting question of what powersupply to use and rather than spend on a dedicated one, I tried to figure out how I could use those ceap 500/600w atx one can find for less than 70 dollars (45 euros) for a pair of them.

I have found some info on hooking them in series or parallel thus doubling either the volts or the available amps. I have also checked ou the possibilities of connectinf 5v and 12v to get 17v thus powering any TEC close to its efficiency peak (please correct me if I am wrong).

However I have several questions.

Let us assume that I have 2 atx power supplies with the following DC output:

+3.3v @22a // +5v @16A // +12v1 @15A // +12v2 @16A // -12v @0.8A // +5vsb @2.5A

• first question: given these are allied max 500w models is it safe to assume the 2 12v lines are actually 1 and are simply physicall separated outside the PSU? As far as I know few models actually have independent 12v lines.
  
• As a consequence would it be prudent to either join the 12v lines to have a max absorption of 30A @ +12v? and what about hooking up a single +12v with a +5v?
  
• Assuming that the answer to both questions is no then would it be possibile to get hold of 4 identical power supplies and achieve the following?

1. hook them in 2 sets (of twos) each group providing in parallel 32 amps on the +12v line? Then, (with appropriate modifications) connect these "two" "single" powersupplies in series to achieve 32 amps on a +24v line?
   
2. Alternatively hook 2 psus to achieve 32amps on the 12v lines and subsequently joining the +5v line (formed by linking the psus in parallel along the 5v lines) from the other psus to achieve a theroretical dc output of 17v @32a?

If point 1 is viable then would it be possibile to gather the dc output and somehow use a powerful potentiometer to modify TEC voltage?
Let us assume this is possible thus using 4 ATX power supplies to provide 768 watts max. The single cables providing current in parallel would be able to sustain the amperage. The cables connecting the "2" sets in series probably would not. More resistant cables would be required, however AFAIK the project could succeed in terms of current distribution safety-wise. Overall absorption @ 24v would be 32 amps (amps are on the low side but I would not be using this voltage anyways with any TEC) consuming 768w out of the 2000w overall possible wattage of 4 500w psus.

Last question: I have devised this ingenious contraption (which in all probability will cause severe injuries to me and/or my family members) assuming that it is better to have a layout envolving: [ joining 12v lines and 5v lines (or 12v lines) in parallel and then hooking the resulting lines in series ] but could it be done the other way around? [ hook 12v and 5v (or 12v) in series to achieve 17v and 16A and then hook the "2" 17v lines in series to parallel to achieve 17v @ 30/32A ]

I hope this will help anyone if the basis is sound that is. I intend to see the project through if viable but I thought of getting some feedback before hand. I must admit I have modest knowledge of PSU. I have learnt a lot in the past months and I am fairly confident of the work envolved in hooking 2 psus in parallel or in series. The combination of both sounds good but I cannot assess its viability on my own.

Thanks in advance

P.s. I ask forgiveness if I may have been unclear in my description. English is not my native language. I may post a bitmap picture if necessary and you all promise not to laugh..

ah what the hell...here ya go:


Very simple diagram which is not correct in terms of symbols or anything. Just couple of details: 2ohm reistance of all 5v lines to achieve correct functioning. Separation of ground wire from case in 2psus. (one for each "group" of parallel psus).

From here, if this is viable I would like to start investigating 5v+12v and such.

If joining the 2 "independent" 12v lines from each psu is possibile (and safe) then I would have a max ouput of 24v@ 64A from 4 psus. over 1500w!!

[NotSoCoolJ]

You could use a large lighting transformer. You would have to rectify and filter the line but it's much cleaner and safer than using a number of PS units in series/parallel.
http://www.usalight.com/indoor_600_w...1_pr_1225.html
That one should be good for around 40 amps at 12 volts. Your rectifiers amperage should be rated for at least 2x the intended draw and more wont hurt. Filter capacitors should be no less than 36v for 12v or 60v for 24v. Again more is better.

What is your intended application? Do you intend on using direct contact with the cpu die or are you planning on making a chiller? ATX PSU's are not the way to go with big power TECs. Most people forget that the molex cable is only rated for about 6 or 7 amps. If you try to draw 30amps on a molex you will surely have a fire.

[Alexontherocks]

You could use a large lighting transformer. You would have to rectify and filter the line but it's much cleaner and safer than using a number of PS units in series/parallel.
http://www.usalight.com/indoor_600_w...1_pr_1225.html
That one should be good for around 40 amps at 12 volts. Your rectifiers amperage should be rated for at least 2x the intended draw and more wont hurt. Filter capacitors should be no less than 36v for 12v or 60v for 24v. Again more is better.

What is your intended application? Do you intend on using direct contact with the cpu die or are you planning on making a chiller? ATX PSU's are not the way to go with big power TECs. Most people forget that the molex cable is only rated for about 6 or 7 amps. If you try to draw 30amps on a molex you will surely have a fire.

There are several easier ways to get the amount of power to get a TEC to work even at its max ratings. I was simply investigating alternative ways to use ATX supplies. A chance to get rid in a creative and, I hope, safe way of 2 PSUS I have at home.

Regarding wiring you are right. Infact I would personally stick to under 6 amps a-cable to avoid melting or worse. The point is that given the number of PSUs I could very well use all the +12v lines and +5.0 red lines up until I make necessary connections. Current would be equally spread across all wires. For instance I have 7 wires pumping all the +12v lines. Let us assume it is safe to have no more than 6 amps running through each I would still manage to get 36 amps across them. If you add the fact that either layout I plan on using, half of the absorbed amps will be provided by each one of the 2 "groups" (comprised of 2 psus each), I would conservatively not see more than 1 or 2 amps on each line stemming from each individual PSU. The problem arises, I believe, when making connections. Every step in the circuit decreases the number of cables until I hook the TEC. I would need to keep a fairly high number of cables conservatively no less 6 under a theoretical absorption of 35 amps of a full powered Tpeltier.

In regards to the use I plan for the TEC as I stated in my previous message I prefer to first get my watercooling done and only if I am sattisfied with it I will venture into the realm of peltiers.

I was just hoping for someone to check up on my work and provide some insight on the viability of using multiple identical psus in such a manner. Setting TECS aside I would definetely like a stable DC output with a high amperage rating for several other applications. A sort of DIY lab unit but with several times the continous current rating of most lab PSUs.

I was thinking (in my best of hopes) of stacking the PSUs in a black wooden box and have analog meters giving amp and voltage ratings plus a potentiometer (maybe a resistive dimmer for peltiers?)

[Alexontherocks]

small edit: Regarding my simple picture....one psu is lacking a cable to home powersupply, my bad. Another detail lacking is of course the necessary modification to one PSU in each "group". one PSU AFAIk must have the ground disconnected from the chassis to ensure correct functioning.

I also found one of the articles which gave me this idea: http://www.procooling.com/index.php?...s&disp=52&pg=1

[Uncle Jimbo]

Alex -
I have done something like this - it's not too hard assuming you can identify all the ground points to float the supply. That includes earth ground which is usually tied to the supply ground at multiple points. Easy to measure though - you just have to be thorough.

Most of the split rail supplies actually have separate regulation on each line. All of the big voltages come from a pre-supply which develops the primary voltage used by all of the regulating elements. It's not usually a problem to tie 12V rails together.

You don't really need 2 ohm on the 5V line - a 10 ohm is plenty of draw. Most modern supplies work fine with no load on the 5V line.

You really don't want to run at anything over 12V on a 16V TEC, and even 24V TECs are much more efficient at 12V if you can move the heat. The best efficiiency is at 25% of max amps, or below. Efficiency is very poor near the upper end of the power curve.

You may need to move over 200W. With good water cooling, you can do that with a 19933 (the '437' from FrozenCPU), as long as you are not looking for huge low temps. At 24V. that TEC burns 800W in addition to the 400W heat moved. At 12V, you only draw 15A, and instead of wasting 800W, you burn less than 200W, but you move about 200W of heat across 20C. Whatever operating point you are at, 20C will give you a lot more clock. If you run all the numbers, you get about 20% lower temps at full power versus 50% power, and you use 4 times the power to get there.

[Alexontherocks]

Thanks Uncle Jimbo! I mentioned a 24v TEC as an example. I am curious to know if infact I could hook all the 4 TAX together. I can indentify the floating points with a multimeter and patience. The Connections to the chassis of the atx can be easily removed or severed.

You mentioned having done something similar. Did you join two or more psus in parallel or series? Do you believe my ridiculously simple diagram is correct? Should I add diodes to prevent back current? Would you modify anything else? I assume that there is no problem in joining all the 12v lines (yellow cables) together since each PSU would provide infact very little current Not more than maybe 28 amps each with a massive, theretical, load of 64 amps (which I will never ever think of using anywhere near my pc).

Houston do I have a go for launch?

Thanks!

[Alexontherocks]

lucky me....searching for cheap power supplies I have found 10 keytech atx 500w monorail providing 21 amps on the 12v line
If the project is sound is there any real problem in creating something like this?



10 powersupplies providing theoretically 100 amps peak @ 24!.

The price is good and I would encase everything inside wood and closed with all the fans facing outside the box. 2 piles of 5 atx psus with analog voltage and current readouts. If viable a device absorbing 32 amps will load each psu (that is if load is shared equally) with 3.2 amps. Very high efficiency overall!

Please someone tell me this is viable I'm getting very excited at the thought of playing around with powersupplies...

[littleowl]

Keep in mind the each of your power supplies are only rated for 80% load in short sprits. not 100% all the time.

[Uncle Jimbo]

lucky me....searching for cheap power supplies I have found 10 keytech atx 500w monorail providing 21 amps on the 12v line
If the project is sound is there any real problem in creating something like this?



10 powersupplies providing theoretically 100 amps peak @ 24!.


The price is good and I would encase everything inside wood and closed with all the fans facing outside the box. 2 piles of 5 atx psus with analog voltage and current readouts. If viable a device absorbing 32 amps will load each psu (that is if load is shared equally) with 3.2 amps. Very high efficiency overall!

Please someone tell me this is viable I'm getting very excited at the thought of playing around with powersupplies...

Alex - Your drawing is correct, and with a little care, this is actually easy to do - well easy if you are careful. Using identical supplies for all of the parallel elements is the best way to assure good sharing.

The sense circuits in ATX supplies are tied at the internal circuit board. That's actually a good thing for this kind of design, because the internal wiring provides an effective buffer resistance for current matching.

A typical newer 500W single rail supply puts out 30A at 3.3V, 35A at 5V, and 20A on the 12V line. Given your spec, this is probably what the Keytech 500W has. Older supplies had more on the 12V lines, something like 25A at 3.3V, 30A at 5V and 32A at 12V. Exact numbers don't matter except in calculating the final amperage capacity.

You can also combine all of the supplies in parallel very high current. I would use power relays to do the switching, that way you can do a switch to select the desired output voltages. You can always draw any available output from the 'bottom' supplies (the ones tied to earth ground) no matter what the switch position. . With the right setup, you can select from any combination of those values. So you can get the following voltage/ amp combinations, given 10 supplies with 30A at 3.3V, 35A at 5V, and 20A at 12V:

Always available in any configuration:
3.3A at 150A
5V at 175A
12V at 100A

10 in parallel:
3.3A at 300A
5V at 350A
12V at 200A

Series sets are limited by the lowest amp leg:
3.3 + 3.3 = 6.6V at 150A
5 + 3.3 = 8.3V at 150A
5+5 = 10V at 175A
12+3.3 = 15.3V at 100A
12+5 = 17V at 100A
12 + 12 = 24V at 100A

So that setup can deliver a wide range of voltages at high current. In your layout, supplies 6 -10 are the 'base' supplies - they require no change except combining the wiring, and they are tied to earth ground.

The first step is to tie common wiring together. The existing wiring is adequate if combined - no need to open the supply and do any internal changes. I combined like wires in groups of 4 wires into a spade lug, and then used screw terminals to tie those together, one terminal block for each voltage on each supply. That makes for a nice neat installation and it is easy to change out a supply. I left about 8 inches of wire, just to keep routing options open. There will be more than twice as many black lines as lines for the other voltages. Use them all.

I put a 50A circuit breaker in the ground line for each supply just to avoid any fire hazard.

Next, combine the orange lines from the line to the main motherboard supply, using the same technique.

Finally, do the yellow and red lines. I used all the red and yellow lines coming out of the supply, as they were all common internally on the one I used. Check your supply - don't use any lines which tie to a different point inside than the lines on the MOLEX connectors.

Now you should have a block of voltages for each supply on a common connector, and a ground block with overcurrent protection.

Next, tie the 'banks' together. You don't need any diode isolation or anything like that - it will only create a voltage drop and an additional potential failure point. You should now have two sets of outputs, one for each voltage on each bank.

If you want to use the switching arrangement above, you need to get some big high current relays. You can often get those cheaply. I used 24V relays from an air conditioning control relay, which were rated at 50A per contact. I got them for about $15 each at a local supply house. I just added a standard 24V home heating transformer to run them, they use 24VAC.

That would be an awesome supply and power just about any load you could want. You could even use thermostat type control to pick the output voltage and achieve a wide range of power selections.

<Edit> one final point - to turn the supplies on, you need to connect the green wire from the supply to ground. A small relay for each bank can do that job, and turn the supplies on and off with the main power to your load.

Also, just to be sure the supplies come up, you should put a small resistor between 5V and ground on each supply. Unless for some reason the supply needs more current to start, I use a 33 ohm 5W. I have never had to use more. That resistor burns less than a watt so it doesn't get hot or waste power.

[Uncle Jimbo]

Keep in mind the each of your power supplies are only rated for 80% load in short sprits. not 100% all the time.

That's good advice, especially for less expensive supplies. You have an advantage if you are only using one rail and not all the voltages - the limiting factor in those supplies is usually the size of the internal heat sinks, and the 'big power' components are often on a single heat sink, which is why the combined load is less than that for a single rail.

[Alexontherocks]

Alex - Your drawing is correct, and with a little care, this is actually easy to do - well easy if you are careful. Using identical supplies for all of the parallel elements is the best way to assure good sharing.

The sense circuits in ATX supplies are tied at the internal circuit board. That's actually a good thing for this kind of design, because the internal wiring provides an effective buffer resistance for current matching.

A typical newer 500W single rail supply puts out 30A at 3.3V, 35A at 5V, and 20A on the 12V line. Given your spec, this is probably what the Keytech 500W has. Older supplies had more on the 12V lines, something like 25A at 3.3V, 30A at 5V and 32A at 12V. Exact numbers don't matter except in calculating the final amperage capacity.

You can also combine all of the supplies in parallel very high current. I would use power relays to do the switching, that way you can do a switch to select the desired output voltages. You can always draw any available output from the 'bottom' supplies (the ones tied to earth ground) no matter what the switch position. . With the right setup, you can select from any combination of those values. So you can get the following voltage/ amp combinations, given 10 supplies with 30A at 3.3V, 35A at 5V, and 20A at 12V:

Always available in any configuration:
3.3A at 150A
5V at 175A
12V at 100A

10 in parallel:
3.3A at 300A
5V at 350A
12V at 200A

Series sets are limited by the lowest amp leg:
3.3 + 3.3 = 6.6V at 150A
5 + 3.3 = 8.3V at 150A
5+5 = 10V at 175A
12+3.3 = 15.3V at 100A
12+5 = 17V at 100A
12 + 12 = 24V at 100A

So that setup can deliver a wide range of voltages at high current. In your layout, supplies 6 -10 are the 'base' supplies - they require no change except combining the wiring, and they are tied to earth ground.

The first step is to tie common wiring together. The existing wiring is adequate if combined - no need to open the supply and do any internal changes. I combined like wires in groups of 4 wires into a spade lug, and then used screw terminals to tie those together, one terminal block for each voltage on each supply. That makes for a nice neat installation and it is easy to change out a supply. I left about 8 inches of wire, just to keep routing options open. There will be more than twice as many black lines as lines for the other voltages. Use them all.

I put a 50A circuit breaker in the ground line for each supply just to avoid any fire hazard.

Next, combine the orange lines from the line to the main motherboard supply, using the same technique.

Finally, do the yellow and red lines. I used all the red and yellow lines coming out of the supply, as they were all common internally on the one I used. Check your supply - don't use any lines which tie to a different point inside than the lines on the MOLEX connectors.

Now you should have a block of voltages for each supply on a common connector, and a ground block with overcurrent protection.

Next, tie the 'banks' together. You don't need any diode isolation or anything like that - it will only create a voltage drop and an additional potential failure point. You should now have two sets of outputs, one for each voltage on each bank.

If you want to use the switching arrangement above, you need to get some big high current relays. You can often get those cheaply. I used 24V relays from an air conditioning control relay, which were rated at 50A per contact. I got them for about $15 each at a local supply house. I just added a standard 24V home heating transformer to run them, they use 24VAC.

That would be an awesome supply and power just about any load you could want. You could even use thermostat type control to pick the output voltage and achieve a wide range of power selections.

Wow...great post! I have ordered the psus (10 @35 euros ). I would like to choose a series type supply therefore having the benefit of intermediate voltages.

Just a couple of questions:

How should I connect the relays? could I be a total noob (not much acting required) and ask you for a simple diagram I could keep as a reference during the build?

To avoid having 10 schuko plugs could I just get the plugs and connect the cabling (phase, neutral and ground) and merge them all into 1 large (high current) cable? I will not use anything above 40 amps but just in case...

thanks again. I will submit the specs of the atx supplies

[Alexontherocks]

to avoid having half of the forum kick me out...I meant of course merging phase neutral and ground into a single cable comprised of 3 wires: phase neutral and ground.

[Alexontherocks]

just wanted to post a few pics: http://i3.ebayimg.com/01/i/000/f6/91/5f5b_1.JPG

http://i14.ebayimg.com/01/i/000/f6/91/7663_1.JPG

other questions: you mention common 12v lines. I am unsure what you mean by that. Should I use all the 12 v lines or just the ones along the molex rail? (thus excluding PCI-E SATA and motherboard 12v lines?)

[Uncle Jimbo]

Wow...great post! I have ordered the psus (10 @35 euros ). I would like to choose a series type supply therefore having the benefit of intermediate voltages.

Just a couple of questions:

How should I connect the relays? could I be a total noob (not much acting required) and ask you for a simple diagram I could keep as a reference during the build?

To avoid having 10 schuko plugs could I just get the plugs and connect the cabling (phase, neutral and ground) and merge them all into 1 large (high current) cable? I will not use anything above 40 amps but just in case...

thanks again. I will submit the specs of the atx supplies

I'll try and do a relay diagram for you, something simple like your layout. On power, the easiest thing to do is just use a power strip inside your box and plug into that, makes it easy to replace a supply if needed, and can be made to look quite tidy.

[littleowl]

just wanted to post a few pics: http://i3.ebayimg.com/01/i/000/f6/91/5f5b_1.JPG

http://i14.ebayimg.com/01/i/000/f6/91/7663_1.JPG

other questions: you mention common 12v lines. I am unsure what you mean by that. Should I use all the 12 v lines or just the ones along the molex rail? (thus excluding PCI-E SATA and motherboard 12v lines?)

That psu will probably only run 18a on the 12v line. I say this because the max is 22a and a good 90% of pc pcu's rather they are 20% or 80% efficient never ever make the full amp load amount.

edit: I may be wrong on the 18a but you can do the math when you get it to see what exactly it gets.

[Alexontherocks]

That psu will probably only run 18a on the 12v line. I say this because the max is 22a and a good 90% of pc pcu's rather they are 20% or 80% efficient never ever make the full amp load amount.

edit: I may be wrong on the 18a but you can do the math when you get it to see what exactly it gets.

From what I understood by looking into series and parallel layouts I believe I will never actually manage to have the setup supply more than 11 amps on each 12v rail. Let us say I take only 4 psus and hook them in two banks (parallel) and then connect them in series to achieve 24v. The amount of current I can have on each bank is indeed 22x2= 44 amps. A theoretical load of 44 amps would max out the current value on both atx supplies, each trying to give 22 amps on its own 12v rail.Now If I take the 2 banks and hook them up in series I get 24v but I do not double the current. If Load is shared equally (please correct me if this assumption is wrong) I still will be able to power a 44 amp device but nothing more and since I have 2 banks current load will be shared equally among available banks and from there equally among the single atx supplies. That is of course if all my assumptions are correct.

By quoting UncleJimbo and his chart of available voltages I gather that maxing out at 100amps total @ 24v I am technically loading each supply with only 10amps on the 12v line. Conversely hooking the psus in parallel (10 of them) would indeed let me tap into all the available current and theoretically have a device run at 12v@200A (loading each supply equally with 20A of current)

[Uncle Jimbo]

That psu will probably only run 18a on the 12v line. I say this because the max is 22a and a good 90% of pc pcu's rather they are 20% or 80% efficient never ever make the full amp load amount.

edit: I may be wrong on the 18a but you can do the math when you get it to see what exactly it gets.

The way to tell what a PSU can deliver steady state is to measure the heatsink temp and put a resistive load on which drives the max load. That needs to be done with the supply fully assembled so that airflow is as it would be in normal use.

If the voltage drops out of spec (11.5V on 12V line, 4.75V on the 5V line) then the load needs to be reduced. If the heat sink temp is under 85C after 10 minutes and the voltage is still in spec, the supply is well designed and you can use it at that rating. If it goes above 85C, then you need to drop the load until it stays at or under that temp.

In my experience, PSUs on the lower end of the price scale use about the same circuit design as more expensive supplies, but use smaller caps and smaller heat sinks. They can deliver 'rated power' for short periods - a minute or so - but after that, the heat gets out of control and they eventually shut down - or fail. Running the test above on several supplies listed as 500W, only a few delivered continuous power - the Ultra 500, Antec True power supplies, and other well designed units. Supplies from PowMax, Logisys, sunbeam, and the like managed around 380W continuous.

The safe derating on those supplies is 75% - so if the rating is 20A, figure it can deliver 15A continuous in normal use.

Not every inexpensive supply is cheaply made - for example, I I bought a 600W rated supply of a brand I had never heard of and never saw again - A-Span or somethng like that - for $29. It was rated for 40A on 12V and 24A on 5V. I had had low expectations, but loaded it to that level and it delivered 12.1V and 5.2V, and after 10 minutes, the sink was at 70C. I increased the loading to 30A on the 5V rail, and 50A on the 12V rail. At that level, voltage was 4.9 on the 5V rail and 11.8 on the 12V rail, and after 10 minutes the sink was at 78C. I didn't press my luck, since I was getting 750W from a 600W rated supply. Just goes to show, you can't always judge a book by it's cover.

Lower price supplies have a variety of different weak spots. Some can handle the amps but fall below regulation voltage long before they hit rated max. Some can deliver all day at rated power on one rail, but fall apart when multiple rails are loaded, even well below the max power.

One advantage that Alex has is that he will only be pulling significant power from one rail at a time. That is actually the best case for a lower priced PSU, since they almost always have all of the regulating MOSFETs on a single heat sink. The spec that Alex gave, if you add up all the power for all the rails, comes in at about 525W, not bad for a 500W rated supply. Even if the de-rating is 75%, we get 394W. The combined total of the 12V rails is 372W. So there is a decent possibility that he can draw rated power from the 12V lines without overheating, if there is no significant draw anywhere else. Running just the 5V or 3.3V lines is not even close to that power, so without any load on the 12V lines, I would expect that supply to run all day at rated amps.

Probably not a bad idea to do some testing on the supplies when they come in, just to get a feel for the safety margin and if any derating is needed.

[Alexontherocks]

Very interesting....I will conduct some testing as soon as possibile.

Couple of questions more:

1)to activate startup of all psus I was thinking of purchasing a small relay switch 12v-activated to close the connection between the green wires and ground. The switches on the back of the atx will be open thus the entire setup will be activated at startup by sensing the 12v outputed by a fan or molex connection in my pc case.

2) I still don't get the detail about common yellow wires. In another PSU I noticed that all 12v lines spawn from the same area (a small portion of the pcb marked by a black line). Does this mean the 12v lines are common and therefore can be used?

[Alexontherocks]

I thought of a cabinet sort of layout...




ventilation should be as good as can be.

here is the front:




plenty of space once the "door" is opened. Simple hinges will do. I was thinking of positioning the readouts and plugs on the top face of the box. Hinges will be used here as well to have easy access to the inside. Fans blowing opposite (lowest one will push air inside) and highest will push air outsides. Cables and switches should stay cool.




Suggestions are welcome...

[leuler]

That's what I would call a " tower of power "

[leuler]

From what I understood by looking into series and parallel layouts I believe I will never actually manage to have the setup supply more than 11 amps on each 12v rail. Let us say I take only 4 psus and hook them in two banks (parallel) and then connect them in series to achieve 24v. The amount of current I can have on each bank is indeed 22x2= 44 amps. A theoretical load of 44 amps would max out the current value on both atx supplies, each trying to give 22 amps on its own 12v rail.Now If I take the 2 banks and hook them up in series I get 24v but I do not double the current. If Load is shared equally (please correct me if this assumption is wrong) I still will be able to power a 44 amp device but nothing more and since I have 2 banks current load will be shared equally among available banks and from there equally among the single atx supplies. That is of course if all my assumptions are correct.

By quoting UncleJimbo and his chart of available voltages I gather that maxing out at 100amps total @ 24v I am technically loading each supply with only 10amps on the 12v line. Conversely hooking the psus in parallel (10 of them) would indeed let me tap into all the available current and theoretically have a device run at 12v@200A (loading each supply equally with 20A of current)

Actually, in the 100amp @ 24v arrangement, each PSU supplies 20 amps.
Each bank of 5 PSUs provide 100 amps ( 5 x 20 amps ) at 12v. Connecting
the two banks in series gives 100 amps at 24v ( 12 + 12 ).

The discussion keeps switching between rails and PSUs. Does each PSU have
two +12 rails or one? From one of the pictures you posted it seems that there
is only one +12 rail.

[Alexontherocks]

Ok I get it...the load is not shared equally if the two banks are set in series. Fair enough.

Regarding the rails, I initially looked into a PSU that had 2 rails v1 and v2. I was asking myself wether the two lines could be joined.

Now I have found 10 supplies which have a single 12v rail. As a reference take the pics I posted the link to. Those are the psus I will be using.

Anyway 100A on any rail is probably 3 times what I could ever want to use both in my pc or outside. The assembled unit would become sort of a massive (and somewhat portable) DC high current source. I could use it for various tasks, the most probable being a TEC.

[Uncle Jimbo]

Very interesting....I will conduct some testing as soon as possibile.

Couple of questions more:

1)to activate startup of all psus I was thinking of purchasing a small relay switch 12v-activated to close the connection between the green wires and ground. The switches on the back of the atx will be open thus the entire setup will be activated at startup by sensing the 12v outputed by a fan or molex connection in my pc case.

2) I still don't get the detail about common yellow wires. In another PSU I noticed that all 12v lines spawn from the same area (a small portion of the pcb marked by a black line). Does this mean the 12v lines are common and therefore can be used?

Alex -
That method of turning on the supplies is clean and easy. If you use a 5V relay, you can have an alternate way to turn it on - use a switch from the 5vsb on your stack of supplies, which is always on. That way you can turn it on from the switch, or from the 5V line from an external source like a fan in your CPU.

Whether the yellow wires go to a common area or to separate regulation MOSFETs (as in some split rail supplies), you are always safe in tying them together - you should tie all the yellow wires from a given supply together, before the first connector.

The same goes for the red and orange lines, and of course the black. The only caution is on the mainboard connector - some PSUs also use red for the 5vsb, and some use orange for the power good sense line. Look at a mainboard connector diagram and only use lines marked as regular power.

Finally, I think you realize that in series, the amps available is limited by the lowest amps in the serial line. My remark about 'always available' refers to the fact that the bank of supplies tied to earth ground (the 'bottom' bank) will always have all voltages available since they are all at the common ground point. The only caution is that based on testing, you may need to de-rate power if you use all the rails, per littleowl's comments.

[Alexontherocks]

thanks for the advice. I am aware that "always available" must be interpreted correctly. What you meant was just that those configrations (3.3 5 and 12) are always available in terms of nominal voltage no matter what layout I choose (serial or parallel).

I would like to ask what is the best way to make connections.....some people have advised me against using screw type connectors. Others use them regularly even for high current low voltage applications. What should I use?

[Uncle Jimbo]

thanks for the advice. I am aware that "always available" must be interpreted correctly. What you meant was just that those configrations (3.3 5 and 12) are always available in terms of nominal voltage no matter what layout I choose (serial or parallel).

I would like to ask what is the best way to make connections.....some people have advised me against using screw type connectors. Others use them regularly even for high current low voltage applications. What should I use?

Flat screw terminals with spade lugs can move 30A to 50A depending on screw size, limited by the capacity of wire on the lug. European-style with a tube and screw are good for 30A. Either can be made to work.

Large mainframe systems use screw terminals for many of the high current connections. I think people who have problems were trying to just put the wire under a flat screw, that doesn't work reliably.

I have flat screw terminals with 50A straps which can be used to bridge the terminals. You just need to use enough lugs to manage the maximum current. This is technology that has been in everyday use for 50 years, there are no mysteries and it is very reliable.

For example, 10 Position 50 Amp Barrier Block 603-10 for $4 at http://www.salvagetech.com/shopdispl...Barrier+Blocks



For the big power distribution, you might want to consider parts designed for that purpose. You can get 4 6GA to 1 2GA units, rated to 400A. Those have capture blocks which securely hold large cable. I think 2GA would be more than adequate...