wiring psu's in series?!

[glmclell]

I'm trying to find a source for high amperage low-voltage PSU for some experiments with thermoelectric cooling.

I was reading an old article on a ham-radio site that talks about wiring the +5 rail of three AT power supples (switching) in series to increase the voltage to +15vdc @ XX amps (whatever 5v was rated at)

Does this really work? I can't believe its not dangerous or something, since I can't find any examples of anyone doing it recently.

500 watt ATX (should be the same technology, just with a 3.3v rail?) is like $20 ... so three of those is $60 ... 15 volts, 38 amps, less than $100 sounds like quite a bargain.

of course, 15v is a bit high for most TEC's, would a parallel wired 'bank' of 7812's with heatsinks regulate that down to 12v?

 

[ante]

glmclell,

The 7812 thing it’s not so good for this many amps. I would get some high Amp diodes (from a scraped welder or similar.) and just put them in series with the output. This will lower the voltage by around 0.5V for each diode.

Ante :roll:

 

[Nigel Goodwin]

glmclell,

The 7812 thing it’s not so good for this many amps. I would get some high Amp diodes (from a scraped welder or similar.) and just put them in series with the output. This will lower the voltage by around 0.5V for each diode.

Ante :roll:

Or use two 5V rails and one 3.3V rail, to give 13.3V, which is pretty close to the nominal 13.8V of a car battery.

Presuming the 3.3V rail can provide the same power as the 5V rails?.

 

[ante]

Nigel,

That’s even better if possible! I should work as long as the lowest Amp. value is not overdrawn. There is another issue; the 5V and 3.3V neg. is probably connected to the metal cover so they have to be galvanic separated when assembled. Apart from this I can’t see any problems. Can you?

Ante :roll:

 

[glmclell]

ok, so wiring +5 to +5 to +3.3 across psu's 1, 2 and 3

from the specs on the psu's I'm considering buying, they put out 10A less on 3.3 than on 5V ... so I think going with 15V (yielding more amps from the psu) and then dropping the voltage with diodes (4 to 6 diodes to get down to 12-13V range)

are these diodes going to get scorching hot?

I have a source for 600V/40A lug mount diodes... **broken link removed**, I assume the lug is for bolting it to a heat sink?

about the galvanic separation .... mount psu's on something non conductive so their cases do not touch?

is there a danger should someone touch two cases at the same time, do I need to build a box?

lastly ... the article indicates as "one last thing" that psu 2 and 3 need to have their earth ground separated from the DC ground but it doesn't give an explination why???

 

[Nigel Goodwin]

ok, so wiring +5 to +5 to +3.3 across psu's 1, 2 and 3

from the specs on the psu's I'm considering buying, they put out 10A less on 3.3 than on 5V ... so I think going with 15V (yielding more amps from the psu) and then dropping the voltage with diodes (4 to 6 diodes to get down to 12-13V range)

are these diodes going to get scorching hot?

Do the sums! - W=VxI - 4 diodes at 0.7V each will drop 2.8V, for every 10A of current that's 28W of heat!.

I have a source for 600V/40A lug mount diodes... **broken link removed**, I assume the lug is for bolting it to a heat sink?

Yes, preferably a large one :lol:

about the galvanic separation .... mount psu's on something non conductive so their cases do not touch?

Yes, the cases are common to 0V.

is there a danger should someone touch two cases at the same time, do I need to build a box?

No, no danger in touching them, there will only be 5V or 10V between them.

lastly ... the article indicates as "one last thing" that psu 2 and 3 need to have their earth ground separated from the DC ground but it doesn't give an explination why???

That's to prevent you having to insulate the cases, by removing the connection between the casing and 0V you can then bolt the cases together - you can leave the lower of the three connected if you wish, so the 0V point of the 15V has a ground connection - or you could disconnect all three, and have an external connection.

BTW, ATX PSU's have a small standby PSU as well, and you have an input to the main PSU in order to turn it on.

 

[Dean Huster]

Isolate the cases from each other all you want to, but as soon as you plug them into the power receptacle, all of them will be automatically connected together through the third-wire safety earth ground. You'll have to internally "float" the supplies and that may not be an easy task.

Current from any series lash-up will be limited to the current capability of the supply with the lowest current rating.

How hot will series diodes get? P = EI. With a typical 0.7 volt drop, 1 a will dissiplate 700mW; 2a will be 1.4 watts; 10a will be 7 watts. However, remember that a rectifier diode spec may rise to a 1 volt drop at full current, so 10a could mean as much as 10 watts of power dissipated. Yes, the diodes will get hotter than firecrackers. You'll have to mount the stud-mount varieties on heat sinks, and you'll have to be careful of electrical isolation.

Dean

 

[ante]

Dean,

There is now reason to use the earth wire on more than one of the units. Preferably the one on which the negative output is used should be grounded. Just omit the yellow/green wire in the second and third unit that’s what galvanic separated means. It is reasonable to assume that the lowest current rating of the units will be the maximum current. I suggested the diode voltage drop solution as an alternative to paralleled 7812:s becurse I though it was a little more manageable. Yes a rectifier diode at full current will drop 1 Volt that’s why I preferred welder diodes; you don’t get anywhere near full current. But Nigel’s idea with 3.3 + 5 +5 will work fine without any diodes.

Ante :roll: