Forums

series input parallel output SMPS

Started by bitrex March 29, 2018
On Friday, March 30, 2018 at 4:13:35 PM UTC-4, Jon Elson wrote:
> George Herold wrote: > > > On Thursday, March 29, 2018 at 3:04:18 PM UTC-4, Jon Elson wrote: > >> bitrex wrote: > >> > >> > I have a bunch of small 12 volt input ATX SMPS power modules each rated > >> > for 120 watts. about 14 bux each, surplus. Fanless 200 watt modules are > >> > harder to come by and significantly more expensive. > >> > > >> > I'd like to slave two together to get ~200 watts out, with the inputs > >> > in series. 24 volt 8 amp power bricks are pretty common. > >> > > >> > I found this whitepaper on voltage balancing SMPS inputs in series but > >> > it seems more applicable to from-scratch designs. > >> > > >> Nope, won't work. Paralleled outputs will likely not current share, one > >> will put out limit current, the other will pick up the remaining current > >> demand. So, with delivered power not being shared equally, the voltage > >> drop across the supplies will be different, but in series the input > >> current MUST > >> be the same. So, the one carrying less output current will have less > >> voltage drop, and it will increase until the other unit is starved for > >> voltage. It will be totally unstable, and the supplies will be pulsing > >> on and off. > >> > >> Jon > > > > I've done it with bench power supplies. (individual adjustments > > for voltage and current.) I never thought much about it but it's > > worked fine the few times I've needed more current. I see bitrex > > is asking about DC-DC converters, which may behave differently. > > > > Sure, paralleling the OUTPUTS of well-behaved supplies should work fine. > Paralleling the inputs to DC-DC supplies should also work. The OP wants to > parallel the outputs and SERIES the inputs! As far as I can see, that can > NEVER work in the real world. > > Jon
Thanks Jon, sorry to be an idiot sometimes. As is often the case I didn't read the post that well... Well if it's going to oscillate maybe it can be used as an inverter. :^) George H.
On 30.3.18 23:13, Jon Elson wrote:
> George Herold wrote: > >> On Thursday, March 29, 2018 at 3:04:18 PM UTC-4, Jon Elson wrote: >>> bitrex wrote: >>> >>>> I have a bunch of small 12 volt input ATX SMPS power modules each rated >>>> for 120 watts. about 14 bux each, surplus. Fanless 200 watt modules are >>>> harder to come by and significantly more expensive. >>>> >>>> I'd like to slave two together to get ~200 watts out, with the inputs >>>> in series. 24 volt 8 amp power bricks are pretty common. >>>> >>>> I found this whitepaper on voltage balancing SMPS inputs in series but >>>> it seems more applicable to from-scratch designs. >>>> >>> Nope, won't work. Paralleled outputs will likely not current share, one >>> will put out limit current, the other will pick up the remaining current >>> demand. So, with delivered power not being shared equally, the voltage >>> drop across the supplies will be different, but in series the input >>> current MUST >>> be the same. So, the one carrying less output current will have less >>> voltage drop, and it will increase until the other unit is starved for >>> voltage. It will be totally unstable, and the supplies will be pulsing >>> on and off. >>> >>> Jon >> >> I've done it with bench power supplies. (individual adjustments >> for voltage and current.) I never thought much about it but it's >> worked fine the few times I've needed more current. I see bitrex >> is asking about DC-DC converters, which may behave differently. >> > > Sure, paralleling the OUTPUTS of well-behaved supplies should work fine. > Paralleling the inputs to DC-DC supplies should also work. The OP wants to > parallel the outputs and SERIES the inputs! As far as I can see, that can > NEVER work in the real world. > > Jon
A good switcher behaves as a negative resistance load to the input. Putting two of them in series will create interesting effects, for sure. It will not work, even with isolated supplies. -- -TV
>" The OP wants to parallel the outputs and SERIES the inputs! As far as I can see, that can NEVER work in the real world. "
Sure it can. All you need is to have a team study the problem, come up with a modification that makes it possible and then design a circuit that will make it possible. Piece of cake. Then pay the fabrication costs for the new PC boards, after designing them of course, and then assembling this kludge in house depending on the designer of the SMPSes hoping they are deigned right and used good reliable components. And only $ 14 each, oops, $28 each. What could possibly go wrong ?
On 03/31/2018 04:55 AM, jurb6006@gmail.com wrote:
>> " The OP wants to parallel the outputs and SERIES the inputs! As far as I can see, that can NEVER work in the real world." > > Sure it can. All you need is to have a team study the problem, come up with a modification that makes it possible and then design a circuit that will make it possible. Piece of cake. > > Then pay the fabrication costs for the new PC boards, after designing them of course, and then assembling this kludge in house depending on the designer of the SMPSes hoping they are deigned right and used good reliable components. And only $ 14 each, oops, $28 each. > > What could possibly go wrong ? >
Relax, relax, you've made your case. Whenever there's a task that needs not-doing I assure you, you're gonna be my go-to guy.
On Sat, 31 Mar 2018 11:38:07 +0300, Tauno Voipio
<tauno.voipio@notused.fi.invalid> wrote:

>On 30.3.18 23:13, Jon Elson wrote: >> George Herold wrote: >> >>> On Thursday, March 29, 2018 at 3:04:18 PM UTC-4, Jon Elson wrote: >>>> bitrex wrote: >>>> >>>>> I have a bunch of small 12 volt input ATX SMPS power modules each rated >>>>> for 120 watts. about 14 bux each, surplus. Fanless 200 watt modules are >>>>> harder to come by and significantly more expensive. >>>>> >>>>> I'd like to slave two together to get ~200 watts out, with the inputs >>>>> in series. 24 volt 8 amp power bricks are pretty common. >>>>> >>>>> I found this whitepaper on voltage balancing SMPS inputs in series but >>>>> it seems more applicable to from-scratch designs. >>>>> >>>> Nope, won't work. Paralleled outputs will likely not current share, one >>>> will put out limit current, the other will pick up the remaining current >>>> demand. So, with delivered power not being shared equally, the voltage >>>> drop across the supplies will be different, but in series the input >>>> current MUST >>>> be the same. So, the one carrying less output current will have less >>>> voltage drop, and it will increase until the other unit is starved for >>>> voltage. It will be totally unstable, and the supplies will be pulsing >>>> on and off. >>>> >>>> Jon >>> >>> I've done it with bench power supplies. (individual adjustments >>> for voltage and current.) I never thought much about it but it's >>> worked fine the few times I've needed more current. I see bitrex >>> is asking about DC-DC converters, which may behave differently. >>> >> >> Sure, paralleling the OUTPUTS of well-behaved supplies should work fine. >> Paralleling the inputs to DC-DC supplies should also work. The OP wants to >> parallel the outputs and SERIES the inputs! As far as I can see, that can >> NEVER work in the real world. >> >> Jon > > > >A good switcher behaves as a negative resistance load to the input. >Putting two of them in series will create interesting effects, for sure. >It will not work, even with isolated supplies.
There is no problem as long as you have adequate (input) capacitance between them. We do it all the time (by the millions). Galvanic isolation has nothing to do with it.
On Saturday, 31 March 2018 15:41:27 UTC+1, k...@notreal.com  wrote:
> On Sat, 31 Mar 2018 11:38:07 +0300, Tauno Voipio > <tauno.voipio@notused.fi.invalid> wrote: > >On 30.3.18 23:13, Jon Elson wrote: > >> George Herold wrote: > >>> On Thursday, March 29, 2018 at 3:04:18 PM UTC-4, Jon Elson wrote: > >>>> bitrex wrote: > >>>> > >>>>> I have a bunch of small 12 volt input ATX SMPS power modules each rated > >>>>> for 120 watts. about 14 bux each, surplus. Fanless 200 watt modules are > >>>>> harder to come by and significantly more expensive. > >>>>> > >>>>> I'd like to slave two together to get ~200 watts out, with the inputs > >>>>> in series. 24 volt 8 amp power bricks are pretty common. > >>>>> > >>>>> I found this whitepaper on voltage balancing SMPS inputs in series but > >>>>> it seems more applicable to from-scratch designs. > >>>>> > >>>> Nope, won't work. Paralleled outputs will likely not current share, one > >>>> will put out limit current, the other will pick up the remaining current > >>>> demand. So, with delivered power not being shared equally, the voltage > >>>> drop across the supplies will be different, but in series the input > >>>> current MUST > >>>> be the same. So, the one carrying less output current will have less > >>>> voltage drop, and it will increase until the other unit is starved for > >>>> voltage. It will be totally unstable, and the supplies will be pulsing > >>>> on and off. > >>>> > >>>> Jon > >>> > >>> I've done it with bench power supplies. (individual adjustments > >>> for voltage and current.) I never thought much about it but it's > >>> worked fine the few times I've needed more current. I see bitrex > >>> is asking about DC-DC converters, which may behave differently. > >>> > >> > >> Sure, paralleling the OUTPUTS of well-behaved supplies should work fine. > >> Paralleling the inputs to DC-DC supplies should also work. The OP wants to > >> parallel the outputs and SERIES the inputs! As far as I can see, that can > >> NEVER work in the real world. > >> > >> Jon > > > > > > > >A good switcher behaves as a negative resistance load to the input. > >Putting two of them in series will create interesting effects, for sure. > >It will not work, even with isolated supplies. > > There is no problem as long as you have adequate (input) capacitance > between them. We do it all the time (by the millions). Galvanic > isolation has nothing to do with it.
I sure am curious what mass market appliance contains switchers in series NT
On 31.3.18 17:41, krw@notreal.com wrote:
> On Sat, 31 Mar 2018 11:38:07 +0300, Tauno Voipio > <tauno.voipio@notused.fi.invalid> wrote: > >> On 30.3.18 23:13, Jon Elson wrote: >>> George Herold wrote: >>> >>>> On Thursday, March 29, 2018 at 3:04:18 PM UTC-4, Jon Elson wrote: >>>>> bitrex wrote: >>>>> >>>>>> I have a bunch of small 12 volt input ATX SMPS power modules each rated >>>>>> for 120 watts. about 14 bux each, surplus. Fanless 200 watt modules are >>>>>> harder to come by and significantly more expensive. >>>>>> >>>>>> I'd like to slave two together to get ~200 watts out, with the inputs >>>>>> in series. 24 volt 8 amp power bricks are pretty common. >>>>>> >>>>>> I found this whitepaper on voltage balancing SMPS inputs in series but >>>>>> it seems more applicable to from-scratch designs. >>>>>> >>>>> Nope, won't work. Paralleled outputs will likely not current share, one >>>>> will put out limit current, the other will pick up the remaining current >>>>> demand. So, with delivered power not being shared equally, the voltage >>>>> drop across the supplies will be different, but in series the input >>>>> current MUST >>>>> be the same. So, the one carrying less output current will have less >>>>> voltage drop, and it will increase until the other unit is starved for >>>>> voltage. It will be totally unstable, and the supplies will be pulsing >>>>> on and off. >>>>> >>>>> Jon >>>> >>>> I've done it with bench power supplies. (individual adjustments >>>> for voltage and current.) I never thought much about it but it's >>>> worked fine the few times I've needed more current. I see bitrex >>>> is asking about DC-DC converters, which may behave differently. >>>> >>> >>> Sure, paralleling the OUTPUTS of well-behaved supplies should work fine. >>> Paralleling the inputs to DC-DC supplies should also work. The OP wants to >>> parallel the outputs and SERIES the inputs! As far as I can see, that can >>> NEVER work in the real world. >>> >>> Jon >> >> >> >> A good switcher behaves as a negative resistance load to the input. >> Putting two of them in series will create interesting effects, for sure. >> It will not work, even with isolated supplies. > > There is no problem as long as you have adequate (input) capacitance > between them. We do it all the time (by the millions). Galvanic > isolation has nothing to do with it. >
If you connect inputs in series and outputs in parallel and the inputs and outputs are not isolated from each other, you'll have a sturdy short-circuit somewhere. -- -TV
>"Relax, relax, you've made your case. Whenever there's a task that needs not-doing I assure you, you're gonna be my go-to guy. "
LOL. But you do get the point. With all this todo maybe those transformers in the units you have can handle more and you can just build your own. Really, the limiting factor is probably the chopper transistor. Put in a better one and made sure it switches right, ba da bing ba da boom. Well not boom... You can probably use most of the original. the transformer is nothing but a coil of wire so between the transistor and rectifiers, it might not be so bad to just upgrade them.
On 03/31/2018 03:38 PM, jurb6006@gmail.com wrote:
>> "Relax, relax, you've made your case. Whenever there's a task that needs not-doing I assure you, you're gonna be my go-to guy." > > LOL. But you do get the point. With all this todo maybe those transformers in the units you have can handle more and you can just build your own. Really, the limiting factor is probably the > chopper transistor. Put in a better one and made sure it switches right, ba da bing ba da boom. Well not boom... > You can probably use most of the original. the transformer is nothing but a coil of wire so between the transistor and rectifiers, it might not be so bad to just upgrade them. >
That's a good point, I did take the time to check the part IDs on all the ICs and switches on board. The driver IC for the high current rails e.g. +3.3 and +12 can drive beefier devices than are used on the board, synchronous rectifiers so don't need to replace any diodes. The other rails, +5 and -12 don't get as much demand in modern PCs so could likely leave those as is. I don't believe -12 is used for much of anything in fact but it's part of the spec so they have to include it. Might just need to do a quick inductor/MOSFET/filter cap swap to push up to 200, the filter caps and inductors are thru-hole so that part's easy enough.
On 03/31/2018 03:56 PM, bitrex wrote:
> On 03/31/2018 03:38 PM, jurb6006@gmail.com wrote: >>> "Relax, relax, you've made your case. Whenever there's a task that >>> needs not-doing I assure you, you're gonna be my go-to guy." >> >> LOL. But you do get the point. With all this todo maybe those >> transformers in the units you have can handle more and you can just >> build your own. Really, the limiting factor is probably the > chopper >> transistor. Put in a better one and made sure it switches right, ba da >> bing ba da boom. Well not boom... >> You can probably use most of the original. the transformer is nothing >> but a coil of wire so between the transistor and rectifiers, it might >> not be so bad to just upgrade them. >> > That's a good point, I did take the time to check the part IDs on all > the ICs and switches on board. The driver IC for the high current rails > e.g. +3.3 and +12 can drive beefier devices than are used on the board, > synchronous rectifiers so don't need to replace any diodes. > > The other rails, +5 and -12 don't get as much demand in modern PCs so > could likely leave those as is. I don't believe -12 is used for much of > anything in fact but it's part of the spec so they have to include it. > > Might just need to do a quick inductor/MOSFET/filter cap swap to push up > to 200, the filter caps and inductors are thru-hole so that part's easy > enough.
I think I'm going to give it a try in any case, it'll take the better part of a month for a higher power unit I need to arrive from China and if these can be easily modded for higher output power that'll be a good thing to know