• Save 15% on ALL SFF Network merch, until Dec 31st! Use code SFF2024 at checkout. Click here!

Power Supply A Guide to 12V PSU

Thehack

Spatial Philosopher
Original poster
Creator
Mar 6, 2016
2,813
3,670
J-hackcompany.com
Hey guys, I'm working on a little PCB to make 12V builds a little easier / intuitive. The intent is that you'll be able to solder an XT90 connector onto the board to bring power in, solder on a 4pin mini fit junior to power up a pico psu, a 4 pin cpu jack you'll plug the Pico's cpu connector to in order to switch a MOSFET on, and space for two 8 pin PCIe power jacks, and one 4/8 pin cpu power connector.

https://www.mouser.com/ProductDetail/Infineon-Technologies/IRF250P225?qs=sGAEpiMZZMshyDBzk1/Wi/D7Em5shE8qLkc5hPl0smFrCq3qpcjopQ==
This mosfet can do up to 69A with power dissipation of 347W, would I be correct in assuming that I should divide 347 by 12 to find the maximum amperage this can handle with 12V? How much wattage could I viably push through this thing without a heatsink? And considering 8 awg only does 50A, how can the tiny little leads on this package possibly hold up to such high amperage?

I'd like to use an XT90 connector. It's pretty big and bulky, but it can handle a ton of amperage for high powered builds. Do you think using an XT90 would make a board like this less helpful for the general community, or would most 12V builders be willing to throw XT90 connectors on their power brick of choice if it means they can use a little distribution board to minimize loose components / awkward connections?

Finally, how much temperature rise should I allow on the board? The more temperature rise I allow, the smaller I can potentially make the board.

1. XT90 as a connector doesn't make sense. You should use screw terminals. Simpler, inexpensive, reliable. 400W builds is what require a pcb, below that you are fine without the pcb. XT requires soldering and is about 5x the cost of cheaper screw terminals.

2. That mosfet is specced too high, and should be an SMD type.

3. You will need a controller for the mosfet.

4. Tiny leads can push a lot of current because the distance is fairly short. Current also depends on surface area so properly soldered it has plenty of surface area.

5. I would allow for 10C since the ambient temperature would be fairly warm inside a case.

You are essential designing a solid state relay. I don't think we have an electrical engineer on the forums so I recommend you check an EE forum to properly design it.

I also intend to design a solution but it is not on my radar until next year.

I have already designed a board with more features than this called the Distro 400. It is our forum and has all the specs one would need + more.
 
Last edited:
  • Like
Reactions: teisysadmin

teisysadmin

Caliper Novice
Oct 15, 2018
26
20
1. XT90 as a connector doesn't make sense. You should use screw terminals. Simpler, inexpensive, reliable. 400W builds is what require a pcb, below that you are fine without the pcb. XT requires soldering and is about 5x the cost of cheaper screw terminals.

2. That mosfet is specced too high, and should be an SMD type.

3. You will need a controller for the mosfet.

4. Tiny leads can push a lot of current because the distance is fairly short. Current also depends on surface area so properly soldered it has plenty of surface area.

5. I would allow for 10C since the ambient temperature would be fairly warm inside a case.

You are essential designing a solid state relay. I don't think we have an electrical engineer on the forums so I recommend you check an EE forum to properly design it.

I also intend to design a solution but it is not on my radar until next year.

I have already designed a board with more features than this called the Distro 400. It is our forum and has all the specs one would need + more.

1. XT90 allows for very high amperages, rapid connection / disconnection, and looks a little sharper on the back of a case than screw terminals. With SFF builds typically placing an emphasis on portability, I think the ability to set up your rig quickly without a screwdriver would go a long way.

2. Noted.

3. Why? I was under the impression that if you put a suitable current / voltage on the base of a mosfet current will flow freely through the other two terminals. Controller in this case should pretty much just be a resistor, no? constant 12V supply on the pico, pico powers on and supplies 12v to its integrated 4pin cpu connector which comes back, through a resistor, and to the base of the transistors on the board and 12V power is passed through to cpu / gpu.

4. Noted.

5. Annoying, but wise.
 

Thehack

Spatial Philosopher
Original poster
Creator
Mar 6, 2016
2,813
3,670
J-hackcompany.com
1. XT90 allows for very high amperages, rapid connection / disconnection, and looks a little sharper on the back of a case than screw terminals. With SFF builds typically placing an emphasis on portability, I think the ability to set up your rig quickly without a screwdriver would go a long way.

2. Noted.

3. Why? I was under the impression that if you put a suitable current / voltage on the base of a mosfet current will flow freely through the other two terminals. Controller in this case should pretty much just be a resistor, no? constant 12V supply on the pico, pico powers on and supplies 12v to its integrated 4pin cpu connector which comes back, through a resistor, and to the base of the transistors on the board and 12V power is passed through to cpu / gpu.

4. Noted.

5. Annoying, but wise.

1. No other unit I know of uses XT90. I know of 12V external units using 6 pin Mini fit. Since one would have to create a matching connector, it makes more sense to just use mini fit through out. Internal units use other connectors like screw terminals which I thought you were referring to. If you're talking about modded bricks then it kinda makes sense. Also XT90 is not known to be the safest due to slight physical alignment causing a bad connection. I would look for better industrial designs that are safer if we're talking modding.

3. Additional features. Just applying 12V is quite crude. You also don't want to just pump it 12V from the rail. If you don't use logic and the mosfet fail, it may short. Well, you can also use resistors as well. Mine detects polarity and warns if it is reversed, it also has an LED if power is on.

Note you can use thru hole type as well. It may be cheaper just to use 2 cheap mosfet instead of one big one. Idk, I only send the specs and an EE designed mine with me.
 

teisysadmin

Caliper Novice
Oct 15, 2018
26
20
1. No other unit I know of uses XT90. I know of 12V external units using 6 pin Mini fit. Since one would have to create a matching connector, it makes more sense to just use mini fit through out. Internal units use other connectors like screw terminals which I thought you were referring to. If you're talking about modded bricks then it kinda makes sense. Also XT90 is not known to be the safest due to slight physical alignment causing a bad connection. I would look for better industrial designs that are safer if we're talking modding.

3. Additional features. Just applying 12V is quite crude. You also don't want to just pump it 12V from the rail. If you don't use logic and the mosfet fail, it may short. Well, you can also use resistors as well. Mine detects polarity and warns if it is reversed, it also has an LED if power is on.

Note you can use thru hole type as well. It may be cheaper just to use 2 cheap mosfet instead of one big one. Idk, I only send the specs and an EE designed mine with me.

I'm concerned about the maximum amperage minifit can stand up to. I have a G-Unique kit which uses XT90 connectors, I like the look and feel of them so I figured they wouldn't be a bad option to use in a kit. What other connectors out there can viably handle 50A+?
 

Thehack

Spatial Philosopher
Original poster
Creator
Mar 6, 2016
2,813
3,670
J-hackcompany.com
I'm concerned about the maximum amperage minifit can stand up to. I have a G-Unique kit which uses XT90 connectors, I like the look and feel of them so I figured they wouldn't be a bad option to use in a kit. What other connectors out there can viably handle 50A+?


Note we're talking about wire to PCB here. I think you're confusing power entry with what is actually gonna be soldered onto the PCB.

Brick - - - wire - - XT90>>XT90>----PCB

How are you going to connect the XT90 from the case to the PCB? Soldered?

For power entry you can use whatever terminals match up with your external unit. Do note that XT90 is only on modded units. Rest of industry uses mini fit so far as I've seen. Rates 9A per contact.
 

teisysadmin

Caliper Novice
Oct 15, 2018
26
20
Note we're talking about wire to PCB here. I think you're confusing power entry with what is actually gonna be soldered onto the PCB.

Brick - - - wire - - XT90>>XT90>----PCB

How are you going to connect the XT90 from the case to the PCB? Soldered?

For power entry you can use whatever terminals match up with your external unit. Do note that XT90 is only on modded units. Rest of industry uses mini fit so far as I've seen. Rates 9A per contact.

9A per contact is a little dicey for high power applications. Even with an 8 pin connector, you've got 4 x +12V and 4 x Neutral connectors, comes out to 36A * 12V = 432W. Nothing to sneeze at, but you're not leaving very much headroom.

The XT90 would be soldered directly onto the PCB. The PCB would have holes on either side of the XT90 for standoffs / bolts to mount the whole assembly onto the case.

My hope is to build a simple distro board with a cheap BOM for new builders to implement high wattage 12V systems. I think that 12V systems have way more potential, and can be much more affordable than 19V silliness, but they can feel intimidating / dicey. I don't want to build a commercial product, just a simple accessory that makes cobbling together a 12V system a little less intimidating, so the fact that XT90 is only modded units is okay if people would be willing to make that mod.
 

Thehack

Spatial Philosopher
Original poster
Creator
Mar 6, 2016
2,813
3,670
J-hackcompany.com
9A per contact is a little dicey for high power applications. Even with an 8 pin connector, you've got 4 x +12V and 4 x Neutral connectors, comes out to 36A * 12V = 432W. Nothing to sneeze at, but you're not leaving very much headroom.

The XT90 would be soldered directly onto the PCB. The PCB would have holes on either side of the XT90 for standoffs / bolts to mount the whole assembly onto the case.

My hope is to build a simple distro board with a cheap BOM for new builders to implement high wattage 12V systems. I think that 12V systems have way more potential, and can be much more affordable than 19V silliness, but they can feel intimidating / dicey. I don't want to build a commercial product, just a simple accessory that makes cobbling together a 12V system a little less intimidating, so the fact that XT90 is only modded units is okay if people would be willing to make that mod.

The 6 pin mini fit is what companies use for stock units.

Now you say you want direct solder to pcb XT90...

1. Where are you going to place your board components? Where are you going to place the connectors for 8 pin PCIE, EPS, etc? You could use a low profile right angle headers, but I'm pretty sure there is no XT90 right angle headers, I've only seen wire to wire.


2. If it's direct soldered then it needs to be mounted to the chassis and then there's limited area around the chassis. So now you have to determine which chassis it fits in, and then make sure it matches up with the chassis, unless you intend to only make it compatible with the S4M.

It would make more sense just use it as a power entry module and put the pcb somewhere else.
 

smitty2k1

King of Cable Management
Dec 3, 2016
978
500
@Thehack do you think a PicoPSU 160 powered by the meanwell EPP 200 can power a Ryzen 2400g + 1050ti? No other accessories besides a single SSD and no overclocking.
 

Legion

Airflow Optimizer
Nov 22, 2017
364
402
@Thehack do you think a PicoPSU 160 powered by the meanwell EPP 200 can power a Ryzen 2400g + 1050ti? No other accessories besides a single SSD and no overclocking.
I ran a Xeon E3 (95W) with a fulltime all core turbo and 1050ti on a 160w pico and 192w brick, avg gaming draw was 130-150w never saw any spikes that put it into overdraw. The 2400g is much more efficient than a Xeon E3.
 
  • Like
Reactions: smitty2k1

smitty2k1

King of Cable Management
Dec 3, 2016
978
500
I ran a Xeon E3 (95W) with a fulltime all core turbo and 1050ti on a 160w pico and 192w brick, avg gaming draw was 130-150w never saw any spikes that put it into overdraw. The 2400g is much more efficient than a Xeon E3.

Yessssssssss
 

arataisozaki

Chassis Packer
Oct 20, 2018
15
3
Hey guys, just discovered this forum. Great source of knowledge and ideas.
I'm building a SFF system and when looking for small PSUs were astonished how little options there are, especially if you want to go brickless.
Now I came across this EPP-200-12 option which is looking nearly a little bit too good to be true. I have a 65tdp CPU and want to use a LP gtx 1050 ti. So i guess wattage wise I should be good.
Now I'm wondering how well this actually will work since the meanwell probably wasn't build to be used with computer hardware.
I mean the first thing you learn when building a computer is not to go with the cheap PSUs as they are supposed to be "ticking timebombs".
So what do you think? Will the EPP-200-12 be a "ticking timebomb" as well? Will it damage or reduce the lifespan of my hardware in the long run by not delivering a clean power source?
How does it compare in that terms to the HDPLEX 160W AC-DC "internal brick"? (I know the HDPLEX is 19 V but it serves a similar purpose I guess)
Hope to hear some opinions about that.
Cheers!
 

Thehack

Spatial Philosopher
Original poster
Creator
Mar 6, 2016
2,813
3,670
J-hackcompany.com
Hey guys, just discovered this forum. Great source of knowledge and ideas.
I'm building a SFF system and when looking for small PSUs were astonished how little options there are, especially if you want to go brickless.
Now I came across this EPP-200-12 option which is looking nearly a little bit too good to be true. I have a 65tdp CPU and want to use a LP gtx 1050 ti. So i guess wattage wise I should be good.
Now I'm wondering how well this actually will work since the meanwell probably wasn't build to be used with computer hardware.
I mean the first thing you learn when building a computer is not to go with the cheap PSUs as they are supposed to be "ticking timebombs".
So what do you think? Will the EPP-200-12 be a "ticking timebomb" as well? Will it damage or reduce the lifespan of my hardware in the long run by not delivering a clean power source?
How does it compare in that terms to the HDPLEX 160W AC-DC "internal brick"? (I know the HDPLEX is 19 V but it serves a similar purpose I guess)
Hope to hear some opinions about that.
Cheers!

This and the HDPLEX 160 AC-DC is very similar in performance. The HDPLEX version is fully enclosued/passive if that is what you need.

The EPP-200-12, now RPS-200-12, is an industrial high quality unit. They are warrantied for 3 years. I am not worried if I used it. Power is power. As long as it meets the required power specifications, there is no fancy black magic to worry about.
 

arataisozaki

Chassis Packer
Oct 20, 2018
15
3
Thanks for the reply!
I was thinking for example about the dynamic current load ripple which is not quite that good. Also the ripple limits for the 5, 3.3, -12 and 5VSB lines are lower at 50mV. I guess a DC-DC board will be responsible to reach that, but did anyone have a chance to test that? I'm just really curious on how good in quality a PSU this combination actually can be.
Did the issue with the PS-ON get solved? wouldn't it be as "easy" as just bringing the 5 VSB of the MW to the DCDB and the PWR_ON from the DCDB through an inverter to the PS-ON?
 

Thehack

Spatial Philosopher
Original poster
Creator
Mar 6, 2016
2,813
3,670
J-hackcompany.com
Thanks for the reply!
I was thinking for example about the dynamic current load ripple which is not quite that good. Also the ripple limits for the 5, 3.3, -12 and 5VSB lines are lower at 50mV. I guess a DC-DC board will be responsible to reach that, but did anyone have a chance to test that? I'm just really curious on how good in quality a PSU this combination actually can be.
Did the issue with the PS-ON get solved? wouldn't it be as "easy" as just bringing the 5 VSB of the MW to the DCDB and the PWR_ON from the DCDB through an inverter to the PS-ON?

The board uses 12V for input to power the microcontroller.

You can certainly use 5VB standby + a power_on signal but that would require a plug in dc board to be designed from the ground up to support said features. It may also increase cost. A current board without these features is already $35-45.

The meanwell logic to power on is also reverse logic of atx.
 

arataisozaki

Chassis Packer
Oct 20, 2018
15
3
Hm, I think I might give it a try if I get my hands on an EPP-300-12. Modding a DCDB shouldn't be too hard. Inverting the signal should be possible with a simple 74HC04 in between, or not?
For my project I will use a EPP-200-12 so I will have to design a a bit more sophisticated circuit with an extra 5v AcDc unit. Probably similar to the one used by aquelito.
 

Thehack

Spatial Philosopher
Original poster
Creator
Mar 6, 2016
2,813
3,670
J-hackcompany.com
Hm, I think I might give it a try if I get my hands on an EPP-300-12. Modding a DCDB shouldn't be too hard. Inverting the signal should be possible with a simple 74HC04 in between, or not?
For my project I will use a EPP-200-12 so I will have to design a a bit more sophisticated circuit with an extra 5v AcDc unit. Probably similar to the one used by aquelito.

*PDCB* for plug in dc board.

The issue is that when the PSU is off, there is no 12V supply for the PDCB. You will require a way to supply 12V, probably through a booster circuit. Then you need a way to turn that booster circuit off, or prevent circuit flow when the main 12V turns on. Once you figure that out, you can either use the micro-controller logic and program to output the logic to the MW on, or piggy back off the PS_on signal use an NOT gate.

A simpler way is to just design a solid state relay, since you need a terminal block or something similar to move from 2x M4 round terminals to 6/8 pin PCIe. Which... may or may not be a planned product. I can't say.
 
  • Like
Reactions: NinoPecorino

arataisozaki

Chassis Packer
Oct 20, 2018
15
3
I just did some testing and my old mainboard was fine with just ground and the 5 VSB to pull down on PS_ON when i push the power button. So I don't see why I would need the 12 V to be on for this circuit to work. Do you know if V- is connected to DC_COM on the MW? Is V- ground even when the psu is in standby mode?
Thanks for the super fast responses. I'm really excited to try this out. Still waiting for my EPP-200-12 to arrive :)
 

Thehack

Spatial Philosopher
Original poster
Creator
Mar 6, 2016
2,813
3,670
J-hackcompany.com
I just did some testing and my old mainboard was fine with just ground and the 5 VSB to pull down on PS_ON when i push the power button. So I don't see why I would need the 12 V to be on for this circuit to work. Do you know if V- is connected to DC_COM on the MW? Is V- ground even when the psu is in standby mode?
Thanks for the super fast responses. I'm really excited to try this out. Still waiting for my EPP-200-12 to arrive :)

The plugin board uses 12V. How will you power the plugin board when the MW is off.
 

arataisozaki

Chassis Packer
Oct 20, 2018
15
3
I don't. I only pass through the 5 VSB and ground from the MW directly to the mainboard. That is enough for the mainboard to pull down on PS_ON which I connect to the mw PS_ON through an inverter IC. In that regard the EPP-400 is already nearly a full fledged PC PSU except for not having the other power rails like 3.3 V and 5 V. It even has a PG (power good) output that could be connected to the PWR_OK pin on the mainboard as well (has to be inverted as well i think).
Oh, and also the 5 VSB from the MW has 120mV p-p ripple which isn't in spec unfortunately :/
 

Thehack

Spatial Philosopher
Original poster
Creator
Mar 6, 2016
2,813
3,670
J-hackcompany.com
I don't. I only pass through the 5 VSB and ground from the MW directly to the mainboard. That is enough for the mainboard to pull down on PS_ON which I connect to the mw PS_ON through an inverter IC. In that regard the EPP-400 is already nearly a full fledged PC PSU except for not having the other power rails like 3.3 V and 5 V. It even has a PG (power good) output that could be connected to the PWR_OK pin on the mainboard as well (has to be inverted as well i think).
Oh, and also the 5 VSB from the MW has 120mV p-p ripple which isn't in spec unfortunately :/

I understand now.

However, the PDCB will produce it's own +5VSB. That means you need to modify the +5VSB PDCB pin by desoldering it, and wiring something different. If you do a straight tap, then I'm unsure of what would happen if you have two separate +5VSB fighting each other. It would be a difficult job to remove the solder and isolate one of the 24 pin terminals from the board.

I guess a better solution is to use a 24 pin extension and splice it from there.

Theoretically, it sounds like it'll work.

Have thought about designing one that fully integrates with the meanwell, but the other issue is you'll also need a way to safely connect 8pin PCIe to the round terminals. I concluded a solid state relay would just be simpler overall and ensures more compatibility with other types of AC-DC units. I will probably revisit this issue when I get more opportunities.
 
Last edited: