Problems with 12V and 5V lines on a PC ATX supply

[Commander Kinsey]

Why do (cheap? expensive ones may be better) PC ATX power supplies need current drawn from the 5V line to make the 12V line work correctly?

I have a PC with 3 graphics cards running scientific applications. I acquired three old graphics cards that take about 300W each, and have loads of cheap (CIT) PSUs that are rated at 650W on the 12V line, which is what those cards use. So I run each card off its own supply. But the 12V line at no load, or even at 300W, is only giving out 10 to 10.5V. If I attach a small dummy load of an amp or so to the 5V line, the 12V line suddenly becomes 12V.

Why are the two lines related in any way?

Sorry for the crosspost, I'm not sure which of these groups are active.

[RobH]

Why don't you buy decent, not cheap, PC power supplies.

[Commander Kinsey]

I could do, but I have several of these lying around and hate wasting stuff. I like to put old equipment to good use. They work fine with the 5V under load. Just wondering why I have to do that. I've currently got the two offending supplies connected to the dip and full beam of a car headlamp on their 5V lines, which works fine.

[default]

\
There's often a good reason for it. The 5 volt supply is regulated
and the others are not (generally speaking). The feedback path is
from the 5VDC output back to the mains side of the controller.

The others get line (but not load) regulation via the 5V supply
because they share a common transformer. They are also switching
supplies that work at a high frequency so the transformers have fewer
turns of wire and more volts per turn which results in excellent
transformer "regulation."

[Phil Hobbs]

A lot of cheap supplies regulate only one output, and rely on
cross-regulation via the transformer to control the others. If the
regulated output isn't loaded, it rises out of spec and so do the others.

Cheers

Phil Hobbs

[Commander Kinsey]

So on a cheap shit supply, the 5V is guaranteed to be very close to 5V, but the 12V will drop under heavy load?
And on a decent supply like Corsair, they must regulate both seperately?
I still don't understand why the regulation goes to pot when under 1.5A is taken from 5V. It still regulates that 5V perfectly with no load, but the 12V goes wildly wrong. Why does the regulation need current to be flowing through 5V?

[Commander Kinsey]

Strangely, with no 5V load, I get 5.2V and 10.5V. A small rise and a large drop.

I can't understand why the following happens: No 5V load, 12V is out. Small (2A) 5V load, 12V is ok. Yet if I draw 30A from 5V, the 12V is still ok? How can zero load upset it, but 2A or 30A (big difference) both be ok?

[default]

On Wed, 19 Feb 2020 18:55:27 -0000, "Commander Kinsey"

I suppose there may be some that regulate the 12 for both line and
load but I haven't seen it. Not having separate regulation for the
coarse 12 supply isn't necessarily a design flaw. The 12V is less
critical,

Some (particularly older) designs do require some minimal loading of
the 5V to keep it from drifting above 5, or above 5 AND causing the
crowbar over-voltage protection from kicking in and shorting it to
~1V.

5V is more critical since a lot of components start smoking at 5.2V or
higher and those are the logic components.

The 12V runs fans, platter motors in CDROM, disk drive motors, etc.
and may be used for RS232 protocols, and it can be regulated on the
board close to the load if the load is critical. The -12V is often
pretty wimpy power-wise and it's function is for bias and/or
communications protocols.

The PS should work for the purpose intended, but if you are trying to
use it as a general bench power supply you may encounter issues with
poor regulation.

[Commander Kinsey]

If you look at the 12V line on a Corsair supply under any load, it will always be between 11.8 and 12. A cheap PSU like CIT, with the 5V loaded normally, the 12V can be 11V to 12V depending on its own load. Buy an Alpine supply and it will literally go bang if you exceed 50% load for more than an hour. I went through 10 Alpine supplies, costing the shop a fortune, before I told them enough was enough and got my money back.

> Some (particularly older) designs do require some minimal loading of
> the 5V to keep it from drifting above 5, or above 5 AND causing the
> crowbar over-voltage protection from kicking in and shorting it to
> ~1V.

The supplies I'm having bother with are not that old, probably 5 years. But they were the second cheapest. They also lie on their specs. They're sold as 850W supplies, but you can only draw 650W of that on the 12V line, which is where 99% of the power goes in a modern PC.

> 5V is more critical since a lot of components start smoking at 5.2V or
> higher and those are the logic components.

Nowadays, aren't all the chips running at about 1V and powered by their own VRMs, fed off the 12V line?

> The 12V runs fans, platter motors in CDROM, disk drive motors, etc.
> and may be used for RS232 protocols, and it can be regulated on the
> board close to the load if the load is critical. The -12V is often
> pretty wimpy power-wise and it's function is for bias and/or
> communications protocols.
>
> The PS should work for the purpose intended, but if you are trying to
> use it as a general bench power supply you may encounter issues with
> poor regulation.

I am using it for PC components, but only for one component, the graphics card, which only has a 12V input.

[default]

On Wed, 19 Feb 2020 19:43:11 -0000, "Commander Kinsey"

You mean to say "you get what you pay for?"

>
>> Some (particularly older) designs do require some minimal loading of
>> the 5V to keep it from drifting above 5, or above 5 AND causing the
>> crowbar over-voltage protection from kicking in and shorting it to
>> ~1V.
>
>The supplies I'm having bother with are not that old, probably 5 years. But they were the second cheapest. They also lie on their specs. They're sold as 850W supplies, but you can only draw 650W of that on the 12V line, which is where 99% of the power goes in a modern PC.

You could just get a nice bench supply if you only need 12V,

>
>> 5V is more critical since a lot of components start smoking at 5.2V or
>> higher and those are the logic components.
>
>Nowadays, aren't all the chips running at about 1V and powered by their own VRMs, fed off the 12V line?

Yeah, a lot of the newer stuff particularly those processors made for
smart phones, minimalist single board computers, TV box, Tablets, "Eee
box" style ones, etc., do use 3.3 and 1.2 volt supplies on the board.

Most of my desktop computers are "off-lease" second hand ones.

I notice that my super-compact desktops use a single ~20V supply but
that's probably just to keep the wire gauge small between the supply
and computer.

>
>> The 12V runs fans, platter motors in CDROM, disk drive motors, etc.
>> and may be used for RS232 protocols, and it can be regulated on the
>> board close to the load if the load is critical. The -12V is often
>> pretty wimpy power-wise and it's function is for bias and/or
>> communications protocols.
>>
>> The PS should work for the purpose intended, but if you are trying to
>> use it as a general bench power supply you may encounter issues with
>> poor regulation.
>
>I am using it for PC components, but only for one component, the graphics card, which only has a 12V input.

How many graphics cards are you running from this supply at the same
time? Just get a dedicated 12V supply(s). I'm using some ebay ones
in projects, and have had good results. They even have a pot that
lets you adjust the voltage +/- 10%. They go from 1 amp to 60 amp
sizes.

[Commander Kinsey]

I can get more reliable supplies using PC supplies from Corsair, for not much more money. Plus they have the extra voltage lines if I need some.

>>> 5V is more critical since a lot of components start smoking at 5.2V or
>>> higher and those are the logic components.
>>
>> Nowadays, aren't all the chips running at about 1V and powered by their own VRMs, fed off the 12V line?
>
> Yeah, a lot of the newer stuff particularly those processors made for
> smart phones, minimalist single board computers, TV box, Tablets, "Eee
> box" style ones, etc., do use 3.3 and 1.2 volt supplies on the board.
>
> Most of my desktop computers are "off-lease" second hand ones.
>
> I notice that my super-compact desktops use a single ~20V supply but
> that's probably just to keep the wire gauge small between the supply
> and computer.

Laptops do the same. Trouble is you end up with a thin wire which the user always breaks. Surely a thicker wire wouldn't cost much compared with the whole laptop? Maybe it also means thinner wires inside the laptop on the motherboard, less tracks, less space.

I can understand them wanting the voltage to be higher than that of the battery to make charging easier, but 20V is a bit much.

>>> The 12V runs fans, platter motors in CDROM, disk drive motors, etc.
>>> and may be used for RS232 protocols, and it can be regulated on the
>>> board close to the load if the load is critical. The -12V is often
>>> pretty wimpy power-wise and it's function is for bias and/or
>>> communications protocols.
>>>
>>> The PS should work for the purpose intended, but if you are trying to
>>> use it as a general bench power supply you may encounter issues with
>>> poor regulation.
>>
>> I am using it for PC components, but only for one component, the graphics card, which only has a 12V input.
>
> How many graphics cards are you running from this supply at the same
> time?

One each. The supplies are CIT 850W (650W on the 12V rail). The graphics cards are AMD Radeon R9 280X (250W TDP). I can actually run two per supply, but since I've been having bother, I'm letting them have one each to be sure, since I have a surplus of those supplies sat on a shelf. If I get short, I'll double them up.

> Just get a dedicated 12V supply(s).

I was thinking of car batteries with a charger or three. But that would provide 13.8V....

> I'm using some ebay ones
> in projects, and have had good results. They even have a pot that
> lets you adjust the voltage +/- 10%. They go from 1 amp to 60 amp
> sizes.

I might try that next time I need to buy one. Last time I looked they would have cost more than PC supplies for anything over about 200W. But that was a few years ago.

[Jasen Betts]

On 2020-02-19, Commander Kinsey <CFKi...@military.org.jp> wrote:
> Why do (cheap? expensive ones may be better) PC ATX power supplies need current drawn from the 5V line to make the 12V line work correctly?
>
> I have a PC with 3 graphics cards running scientific applications. I acquired three old graphics cards that take about 300W each, and have loads of cheap (CIT) PSUs that are rated at 650W on the 12V line, which is what those cards use. So I run each card off its own supply. But the 12V line at no load, or even at 300W, is only giving out 10 to 10.5V. If I attach a small dummy load of an amp or so to the 5V line, the 12V line suddenly becomes 12V.
>
> Why are the two lines related in any way?

because all the output voltages come from taps on the same transformer
and the voltage regulation is applied to the input to that transformer
and the voltage regulation only watches the 5V line.

--
Jasen.

[Jasen Betts]

On 2020-02-19, Commander Kinsey <CFKi...@military.org.jp> wrote:

The difference between nothing and 2A is a factor of infinity
The difference between 2A and 20A is only a factor of 10

Actually the fisrt is more like the difference betweem 2mA and 2A
because the internal feedback takes avout 2mA to run the LM431 and the
optocoupler. So going to 2A loads the 5V output by 1000 times more, better
than infinity. but not by much.

Diode voltage drop is logarythmic vs current so the voltage on the transformer
needed to make 5V on the output is less with a 2mA load than it is with
a 2A load.


--
Jasen.

[Commander Kinsey]

Ok, but why does current need to be taken from 5V to make the voltage monitor work?

[Commander Kinsey]

If it's going to be so shit, they could have added a dummy load inside the PSU to make it work properly. They could even have it shut off if there was enough external load.

[Jasen Betts]

because if there's enough voltage on the 5V no effort is made to
supply the 12V

--
Jasen.

[Jasen Betts]

They could, but it was already working propperly with their existing
external loads.

--
Jasen.

[default]

It is designed to be in a computer, and there's always some load on
the 5 V line. It probably didn't seem terribly important to worry
about a high voltage condition where none should ever exist.

Some power supplies sit and oscillate if they don't have a load on the
5 volt line...

I notice my desktop has 5v present on the USB connector even when it
is turned off, turned on, or just in standby. I suspect it may have a
small independent supply to run the USB connectors for power, and
perhaps that also supplies the CMOS memory so the clock and settings
don't drain the battery.

[Jasen Betts]

It does, theres a separate 2A(typ.) 5v "standby" supply on the same circuit
board as the main supply.

--
Jasen.

[default]

On Thu, 20 Feb 2020 10:08:06 -0000 (UTC), Jasen Betts

Yup, here are two ATX supply schematics, one calls it "second power
supply," and the other "flyback converter for bias and 5V SB"

Concise, well-written Theory of Operation too.

http://www.pavouk.org/hw/en_atxps.html
When power supply is connected to the line voltage, then at first are
charged capacitors C5 and C6 together for about 300V. Then take a run
secondary power supply controlled by transistor Q12 and on his output
will be voltage. Behind the voltage regulator IC3 will be voltage 5V,
which goes in to the motherboard and it is necessary for turn-on logic
and for "Wake on something" functions.

http://www.smpspowersupply.com/atx-power-supply.html#bias
An auxiliary flyback converter with power switch Q6 and isolation
transformer T2 provides 5V standby and bias for control circuitry.

[Commander Kinsey]

On Thu, 20 Feb 2020 03:43:58 -0000, default <def...@defaulter.net> wrote:

> On Thu, 20 Feb 2020 00:32:05 -0000, "Commander Kinsey"
> <CFKi...@military.org.jp> wrote:
>
>> On Wed, 19 Feb 2020 23:31:01 -0000, Jasen Betts <ja...@xnet.co.nz> wrote:
>>
>>> On 2020-02-19, Commander Kinsey <CFKi...@military.org.jp> wrote:
>>>> Why do (cheap? expensive ones may be better) PC ATX power supplies need current drawn from the 5V line to make the 12V line work correctly?
>>>>
>>>> I have a PC with 3 graphics cards running scientific applications. I acquired three old graphics cards that take about 300W each, and have loads of cheap (CIT) PSUs that are rated at 650W on the 12V line, which is what those cards use. So I run each card off its own supply. But the 12V line at no load, or even at 300W, is only giving out 10 to 10.5V. If I attach a small dummy load of an amp or so to the 5V line, the 12V line suddenly becomes 12V.
>>>>
>>>> Why are the two lines related in any way?
>>>
>>> because all the output voltages come from taps on the same transformer
>>> and the voltage regulation is applied to the input to that transformer
>>> and the voltage regulation only watches the 5V line.
>>
>> Ok, but why does current need to be taken from 5V to make the voltage monitor work?
>
> It is designed to be in a computer, and there's always some load on
> the 5 V line. It probably didn't seem terribly important to worry
> about a high voltage condition where none should ever exist.

What load would that be? In my newest computer for example (which has a decent supply which doesn't need the 5V load), it uses an SSD which draws fuck all power. The CPU and graphics card and memory all operate from 12V with their own regulator modules. I can't think of anything drawing much 5V, and the dodgy supplies I have need approaching 2 amps! I tried 1 amp and that didn't completely stabilize the 12V line.

[default]

On Thu, 20 Feb 2020 20:07:47 -0000, "Commander Kinsey"

[Commander Kinsey]

[default]

On Thu, 20 Feb 2020 20:07:47 -0000, "Commander Kinsey"

<CFKi...@military.org.jp> wrote:

>On Thu, 20 Feb 2020 03:43:58 -0000, default <def...@defaulter.net> wrote:
>
>> On Thu, 20 Feb 2020 00:32:05 -0000, "Commander Kinsey"
>> <CFKi...@military.org.jp> wrote:
>>
>>> On Wed, 19 Feb 2020 23:31:01 -0000, Jasen Betts <ja...@xnet.co.nz> wrote:
>>>
>>>> On 2020-02-19, Commander Kinsey <CFKi...@military.org.jp> wrote:
>>>>> Why do (cheap? expensive ones may be better) PC ATX power supplies need current drawn from the 5V line to make the 12V line work correctly?
>>>>>
>>>>> I have a PC with 3 graphics cards running scientific applications. I acquired three old graphics cards that take about 300W each, and have loads of cheap (CIT) PSUs that are rated at 650W on the 12V line, which is what those cards use. So I run each card off its own supply. But the 12V line at no load, or even at 300W, is only giving out 10 to 10.5V. If I attach a small dummy load of an amp or so to the 5V line, the 12V line suddenly becomes 12V.
>>>>>
>>>>> Why are the two lines related in any way?
>>>>
>>>> because all the output voltages come from taps on the same transformer
>>>> and the voltage regulation is applied to the input to that transformer
>>>> and the voltage regulation only watches the 5V line.
>>>
>>> Ok, but why does current need to be taken from 5V to make the voltage monitor work?
>>
>> It is designed to be in a computer, and there's always some load on
>> the 5 V line. It probably didn't seem terribly important to worry
>> about a high voltage condition where none should ever exist.
>
>What load would that be? In my newest computer for example (which has a decent supply which doesn't need the 5V load), it uses an SSD which draws fuck all power. The CPU and graphics card and memory all operate from 12V with their own regulator modules. I can't think of anything drawing much 5V, and the dodgy supplies I have need approaching 2 amps! I tried 1 amp and that didn't completely stabilize the 12V line.

The whole idea of taking a supply designed for a specific purpose,
using it for a different purpose is "dodgy," to use your words. Your
expectations may not be realistic.

If I'm designing something to be used in an industrial application I
can make it damn near foolproof and bullet proof, because the
application justifies it. If it's a military contract cost is
secondary to reliability and ruggedness. If I'm designing for a mass
market commercial application cost is very important...

Engineering is all about compromise. There's a lot of different ways
to do things. Sure the PS could be made better, but if that means you
price yourself out of the market what did you achieve?

You should also consider that ATX isn't much of a standard in the
sense that it wasn't carved in stone and handed down from the
mountain, never to deviate. There have been many iterations of the
basic ATX since it was introduced. Requirements change, obsolete
parts get supplanted with newer ones, etc..

The only thing you can count on is change. (and humans are involved -
lower your expectations or face disappointment)

[Commander Kinsey]

Sounds like a horrid design making way too many assumptions.

[Commander Kinsey]

They assume all computers take a couple of amps at 5V. With modern machines that may not be true. Almost everything takes 12V now.

[Commander Kinsey]

On Thu, 20 Feb 2020 21:33:42 -0000, default <def...@defaulter.net> wrote:

> On Thu, 20 Feb 2020 20:07:47 -0000, "Commander Kinsey"
> <CFKi...@military.org.jp> wrote:
>
>> On Thu, 20 Feb 2020 03:43:58 -0000, default <def...@defaulter.net> wrote:
>>
>>> On Thu, 20 Feb 2020 00:32:05 -0000, "Commander Kinsey"
>>> <CFKi...@military.org.jp> wrote:
>>>
>>>> On Wed, 19 Feb 2020 23:31:01 -0000, Jasen Betts <ja...@xnet.co.nz> wrote:
>>>>
>>>>> On 2020-02-19, Commander Kinsey <CFKi...@military.org.jp> wrote:
>>>>>> Why do (cheap? expensive ones may be better) PC ATX power supplies need current drawn from the 5V line to make the 12V line work correctly?
>>>>>>
>>>>>> I have a PC with 3 graphics cards running scientific applications. I acquired three old graphics cards that take about 300W each, and have loads of cheap (CIT) PSUs that are rated at 650W on the 12V line, which is what those cards use. So I run each card off its own supply. But the 12V line at no load, or even at 300W, is only giving out 10 to 10.5V. If I attach a small dummy load of an amp or so to the 5V line, the 12V line suddenly becomes 12V.
>>>>>>
>>>>>> Why are the two lines related in any way?
>>>>>
>>>>> because all the output voltages come from taps on the same transformer
>>>>> and the voltage regulation is applied to the input to that transformer
>>>>> and the voltage regulation only watches the 5V line.
>>>>
>>>> Ok, but why does current need to be taken from 5V to make the voltage monitor work?
>>>
>>> It is designed to be in a computer, and there's always some load on
>>> the 5 V line. It probably didn't seem terribly important to worry
>>> about a high voltage condition where none should ever exist.
>>
>> What load would that be? In my newest computer for example (which has a decent supply which doesn't need the 5V load), it uses an SSD which draws fuck all power. The CPU and graphics card and memory all operate from 12V with their own regulator modules. I can't think of anything drawing much 5V, and the dodgy supplies I have need approaching 2 amps! I tried 1 amp and that didn't completely stabilize the 12V line.
>
> The whole idea of taking a supply designed for a specific purpose,
> using it for a different purpose is "dodgy," to use your words. Your
> expectations may not be realistic.

I didn't. I'm powering computer parts. Just a few less than it was designed for. For something to be upset because it's doing less work is crazy.

> If I'm designing something to be used in an industrial application I
> can make it damn near foolproof and bullet proof, because the
> application justifies it. If it's a military contract cost is
> secondary to reliability and ruggedness. If I'm designing for a mass
> market commercial application cost is very important...
>
> Engineering is all about compromise. There's a lot of different ways
> to do things. Sure the PS could be made better, but if that means you
> price yourself out of the market what did you achieve?

Corsair make supplies that don't need a 5V current. Everything is rock solid at the correct voltage no matter what. That's why they sell so many, because their units just work.

> You should also consider that ATX isn't much of a standard in the
> sense that it wasn't carved in stone and handed down from the
> mountain, never to deviate. There have been many iterations of the
> basic ATX since it was introduced. Requirements change, obsolete
> parts get supplanted with newer ones, etc..
>
> The only thing you can count on is change. (and humans are involved -
> lower your expectations or face disappointment)

It states on the label the maximum current I can draw. It does not state any minimum. Shortfalls should be clearly advertised.