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

[John Larkin]

On Fri, 21 Feb 2020 11:33:51 -0500, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-02-20 17:55, Commander Kinsey wrote:
On Thu, 20 Feb 2020 22:27:48 -0000, Lasse Langwadt Christensen
langwadt@fonz.dk> wrote:

torsdag den 20. februar 2020 kl. 23.16.57 UTC+1 skrev 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?

they are all on the same transformer so they are related, and the
feedback is a combination of 12V and 5V

Not sure how a shared transformer can work.  Since I can draw for
example 5 times as much current from one line so the other would be
adversely affected surely?  They have to adjust each individually or it
would go completely out of whack.  I take more 5V current, so the whole
thing gets boosted, now 12V becomes 14V.


Nope. The coupling between windings is high, so they track pretty
accurately--drag one output down and the others follow.
Cross-regulating like that is quite common. In the limit of tight
coupling (k=1), low resistance, and zero diode drop, the
cross-regulation becomes perfect.

Cheers

Phil Hobbs

Doesn't cross-reg go negative on a typical flyback switcher? If you
load the winding that makes the feedback, the drive goes up to the
others.

Do PC supplies do some sort of averaged or compromise or nonlinear
feedback?

--

John Larkin Highland Technology, Inc
picosecond timing precision measurement

jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com

 

[Phil Hobbs]

On 2020-02-21 14:12, John Larkin wrote:

On Fri, 21 Feb 2020 11:33:51 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-02-20 17:55, Commander Kinsey wrote:
On Thu, 20 Feb 2020 22:27:48 -0000, Lasse Langwadt Christensen
langwadt@fonz.dk> wrote:

torsdag den 20. februar 2020 kl. 23.16.57 UTC+1 skrev 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?

they are all on the same transformer so they are related, and the
feedback is a combination of 12V and 5V

Not sure how a shared transformer can work.  Since I can draw for
example 5 times as much current from one line so the other would be
adversely affected surely?  They have to adjust each individually or it
would go completely out of whack.  I take more 5V current, so the whole
thing gets boosted, now 12V becomes 14V.


Nope. The coupling between windings is high, so they track pretty
accurately--drag one output down and the others follow.
Cross-regulating like that is quite common. In the limit of tight
coupling (k=1), low resistance, and zero diode drop, the
cross-regulation becomes perfect.

Cheers

Phil Hobbs

Doesn't cross-reg go negative on a typical flyback switcher? If you
load the winding that makes the feedback, the drive goes up to the
others.

Flybacks are a bit anomalous in that they won't work without a fair bit
of leakage inductance--that's where the energy gets stored. With poor
coupling, there's room for loading to affect different windings
differently, sure.

Do PC supplies do some sort of averaged or compromise or nonlinear
feedback?

Dunno. I only use them in PCs.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[John Larkin]

On Fri, 21 Feb 2020 15:16:23 -0500, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-02-21 14:12, John Larkin wrote:
On Fri, 21 Feb 2020 11:33:51 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-02-20 17:55, Commander Kinsey wrote:
On Thu, 20 Feb 2020 22:27:48 -0000, Lasse Langwadt Christensen
langwadt@fonz.dk> wrote:

torsdag den 20. februar 2020 kl. 23.16.57 UTC+1 skrev 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?

they are all on the same transformer so they are related, and the
feedback is a combination of 12V and 5V

Not sure how a shared transformer can work.  Since I can draw for
example 5 times as much current from one line so the other would be
adversely affected surely?  They have to adjust each individually or it
would go completely out of whack.  I take more 5V current, so the whole
thing gets boosted, now 12V becomes 14V.


Nope. The coupling between windings is high, so they track pretty
accurately--drag one output down and the others follow.
Cross-regulating like that is quite common. In the limit of tight
coupling (k=1), low resistance, and zero diode drop, the
cross-regulation becomes perfect.

Cheers

Phil Hobbs

Doesn't cross-reg go negative on a typical flyback switcher? If you
load the winding that makes the feedback, the drive goes up to the
others.

Flybacks are a bit anomalous in that they won't work without a fair bit
of leakage inductance--that's where the energy gets stored.

In the leakage? Not magnetizing inductance?

--

John Larkin Highland Technology, Inc
picosecond timing precision measurement

jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com

 

[Commander Kinsey]

On Fri, 21 Feb 2020 20:16:23 -0000, Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-02-21 14:12, John Larkin wrote:
On Fri, 21 Feb 2020 11:33:51 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-02-20 17:55, Commander Kinsey wrote:
On Thu, 20 Feb 2020 22:27:48 -0000, Lasse Langwadt Christensen
langwadt@fonz.dk> wrote:

torsdag den 20. februar 2020 kl. 23.16.57 UTC+1 skrev 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?

they are all on the same transformer so they are related, and the
feedback is a combination of 12V and 5V

Not sure how a shared transformer can work. Since I can draw for
example 5 times as much current from one line so the other would be
adversely affected surely? They have to adjust each individually or it
would go completely out of whack. I take more 5V current, so the whole
thing gets boosted, now 12V becomes 14V.


Nope. The coupling between windings is high, so they track pretty
accurately--drag one output down and the others follow.
Cross-regulating like that is quite common. In the limit of tight
coupling (k=1), low resistance, and zero diode drop, the
cross-regulation becomes perfect.

Cheers

Phil Hobbs

Doesn't cross-reg go negative on a typical flyback switcher? If you
load the winding that makes the feedback, the drive goes up to the
others.

Flybacks are a bit anomalous in that they won't work without a fair bit
of leakage inductance--that's where the energy gets stored. With poor
coupling, there's room for loading to affect different windings
differently, sure.


Do PC supplies do some sort of averaged or compromise or nonlinear
feedback?

Dunno. I only use them in PCs.

Mine is next to a PC, running one PC component. Which is why I sometimes feel like defenestrating it.

 

[Commander Kinsey]

On Fri, 21 Feb 2020 17:09:01 -0000, Grant Taylor <gtaylor@tnetconsulting..net> wrote:

On 2/20/20 10:57 PM, whit3rd wrote:
Nonsense; it's not a fault to find a way to repurpose an otherwise
obsolescent part for such a use, it's a talent.

I largely agree.

However, I've found that salvage parts often have a down side that makes
them sub-optimal for re-use. It's a matter of is the down side a
blocker for a specific re-use.

In this case, it sounds like there must be current on a different
voltage rail. Depending on the power levels, this doesn't sound like
too big of a down side to me.

Well even with the dummy current on the 5V line, it has other problems. The wire guage is too thin, the 12V is not as stable as some of the graphics cards like. One of them keeps dropping the clock speed, and another occasionally stops functioning. I may have to replace the three of them with:

1 decent Corsair for the computer and one graphics card.
1 1kW 12V LED supply from China for the other two graphics cards.

I'd prefer to buy it in the UK for speed of receiving it and more likely to get something that actually works, but there aren't any here. There are some in Germany but they've all sold out.

If anyone knows anywhere other than China that I can buy a 12V regulated 1kW supply for a decent price, please let me know (the Chinese ones are about Ł38).

 

[Phil Hobbs]

On 2020-02-21 15:36, John Larkin wrote:

On Fri, 21 Feb 2020 15:16:23 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-02-21 14:12, John Larkin wrote:
On Fri, 21 Feb 2020 11:33:51 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-02-20 17:55, Commander Kinsey wrote:
On Thu, 20 Feb 2020 22:27:48 -0000, Lasse Langwadt Christensen
langwadt@fonz.dk> wrote:

torsdag den 20. februar 2020 kl. 23.16.57 UTC+1 skrev 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?

they are all on the same transformer so they are related, and the
feedback is a combination of 12V and 5V

Not sure how a shared transformer can work.  Since I can draw for
example 5 times as much current from one line so the other would be
adversely affected surely?  They have to adjust each individually or it
would go completely out of whack.  I take more 5V current, so the whole
thing gets boosted, now 12V becomes 14V.


Nope. The coupling between windings is high, so they track pretty
accurately--drag one output down and the others follow.
Cross-regulating like that is quite common. In the limit of tight
coupling (k=1), low resistance, and zero diode drop, the
cross-regulation becomes perfect.

Cheers

Phil Hobbs

Doesn't cross-reg go negative on a typical flyback switcher? If you
load the winding that makes the feedback, the drive goes up to the
others.

Flybacks are a bit anomalous in that they won't work without a fair bit
of leakage inductance--that's where the energy gets stored.

In the leakage? Not magnetizing inductance?

AIUI the energy is basically stored in the air gap, which is the source
of the leakage inductance. But I've never designed a flyback, so I
wouldn't want to be too dogmatic about that.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Commander Kinsey]

On Fri, 21 Feb 2020 04:42:58 -0000, Rick C <gnuarm.deletethisbit@gmail.com> wrote:

On Thursday, February 20, 2020 at 11:03:37 PM UTC-5, bitrex wrote:
On 2/20/20 8:46 PM, DecadentLinuxUserNumeroUno@decadence.org wrote:
bitrex <user@example.net> wrote in
news:CnF3G.207815$MX1.148708@fx43.iad:

On 2/20/20 7:15 PM, whit3rd wrote:
On Thursday, February 20, 2020 at 2:55:54 PM UTC-8, Commander
Kinsey wrote:
On Thu, 20 Feb 2020 22:27:48 -0000, Lasse Langwadt Christensen
langwadt@fonz.dk> wrote:

torsdag den 20. februar 2020 kl. 23.16.57 UTC+1 skrev 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 l... 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?

they are all on the same transformer so they are related, and
the feedback is a combination of 12V and 5V

Not sure how a shared transformer can work.

The transformer puts proportional (AC) voltages on its windings,
so the DC outputs are always proportional as long as resistive
and rectifier losses are similar to the test condtions. If you
draw zero current through one of the rectifiers, but normal
current through another, there's a discrepancy to be expected.
Usually, 12V power has a high tolerance, it's often the one to be
left out of the regulation feedback loop.


In modern PCs just about everything runs of the 12V but almost
nothing runs off it directly, it's a bulk supply to feed other
converters on the mobo.

LEDs, fans, hard drive spindles get it raw I guess.



Hard drive spindle motor drive circuits get it raw, but the spindle
motor itself gets a quite controlled feed from that circuit.


I think some PCIe cards like GPUs still use 3.3 off the PCI bus for
something. The really honking ones have a connector elsewhere on the PCB
for a higher current connection to the 12.

Pretty much all PC CPUs and GPUs need a direct 12 volt connection for power. It's been that way for over 10 years now. They used to use 5 volts for the CPUs, but the power levels got too high and 5 volt connections had too much current flowing. So they switched to 12 volts for the CPU on board PSU. These days GPUs use pretty much the same power levels.

A lot more in fact. The average decent CPU is 90W. The average decent GPU is 250W. That would be why they have 3 fans instead of 1.

 

[sea moss]

AIUI the energy is basically stored in the air gap, which is the source
of the leakage inductance. But I've never designed a flyback, so I
wouldn't want to be too dogmatic about that.

Cheers

Phil Hobbs

In a flyback the energy is stored in the magnetizing inductance. You gap the core to decrease the value of magnetizing inductance vs. the large, unregulated value the magnetic material gives you. For a given FET ON time, V = L*dI/dt means that peak current increases proportional to the decrease in L. But since energy is 1/2*L*I^2, a smaller inductance means more stored energy.

The leakage inductance is the non-coupled inductance of each winding. For modeling the effect of leakage on transformers with multiple windings, Erickson/Maksimovic have a really good paper:
https://ecee.colorado.edu/~rwe/references/PESC98.pdf

 

[Phil Hobbs]

On 2020-02-21 20:49, sea moss wrote:

AIUI the energy is basically stored in the air gap, which is the source
of the leakage inductance. But I've never designed a flyback, so I
wouldn't want to be too dogmatic about that.

Cheers

Phil Hobbs


In a flyback the energy is stored in the magnetizing inductance. You gap the core to decrease the value of magnetizing inductance vs. the large, unregulated value the magnetic material gives you. For a given FET ON time, V = L*dI/dt means that peak current increases proportional to the decrease in L. But since energy is 1/2*L*I^2, a smaller inductance means more stored energy.

The leakage inductance is the non-coupled inductance of each winding. For modeling the effect of leakage on transformers with multiple windings, Erickson/Maksimovic have a really good paper:
https://ecee.colorado.edu/~rwe/references/PESC98.pdf

Sure. But almost all the energy is stored in the gap, because the
magnetic energy density is proportional to B*H. Perpendicular B is
continuous across the interface, so H is 1/mu_R times higher in the gap.
Since mu is in the thousands, the origin of the energy storage is the
same as that of the leakage inductance.

Cheers

Phil Hobbs


--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Phil Hobbs]

On 2020-02-21 20:58, Phil Hobbs wrote:

On 2020-02-21 20:49, sea moss wrote:
AIUI the energy is basically stored in the air gap, which is the source
of the leakage inductance.  But I've never designed a flyback, so I
wouldn't want to be too dogmatic about that.

Cheers

Phil Hobbs


In a flyback the energy is stored in the magnetizing inductance.  You
gap the core to decrease the value of magnetizing inductance vs. the
large, unregulated value the magnetic material gives you.  For a given
FET ON time, V = L*dI/dt means that peak current increases
proportional to the decrease in L.  But since energy is 1/2*L*I^2, a
smaller inductance means more stored energy.

The leakage inductance is the non-coupled inductance of each winding.
For modeling the effect of leakage on transformers with multiple
windings, Erickson/Maksimovic have a really good paper:
https://ecee.colorado.edu/~rwe/references/PESC98.pdf

Sure.  But almost all the energy is stored in the gap, because the
magnetic energy density is proportional to B*H.  Perpendicular B is
continuous across the interface, so H is 1/mu_R times higher in the gap.
 Since mu is in the thousands, the origin of the energy storage is the
same as that of the leakage inductance.

mu_R times higher.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Phil Hobbs]

On 2020-02-21 21:06, Phil Hobbs wrote:

On 2020-02-21 20:58, Phil Hobbs wrote:
On 2020-02-21 20:49, sea moss wrote:
AIUI the energy is basically stored in the air gap, which is the source
of the leakage inductance.  But I've never designed a flyback, so I
wouldn't want to be too dogmatic about that.

Cheers

Phil Hobbs


In a flyback the energy is stored in the magnetizing inductance.  You
gap the core to decrease the value of magnetizing inductance vs. the
large, unregulated value the magnetic material gives you.  For a
given FET ON time, V = L*dI/dt means that peak current increases
proportional to the decrease in L.  But since energy is 1/2*L*I^2, a
smaller inductance means more stored energy.

The leakage inductance is the non-coupled inductance of each winding.
For modeling the effect of leakage on transformers with multiple
windings, Erickson/Maksimovic have a really good paper:
https://ecee.colorado.edu/~rwe/references/PESC98.pdf

Sure.  But almost all the energy is stored in the gap, because the
magnetic energy density is proportional to B*H.  Perpendicular B is
continuous across the interface, so H is 1/mu_R times higher in the
gap.   Since mu is in the thousands, the origin of the energy storage
is the same as that of the leakage inductance.

mu_R times higher.

*sigh*. Shouldn't post after having two glasses of wine. Magnetic
energy density is |B|**2/mu, so since perpendicular B is continuous
across the boundary, the energy density is mu_r times higher in the gap.

Cheers

Phil Hobbs
(Still on vacation in a beach house on Anna Maria Island)

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Commander Kinsey]

On Fri, 21 Feb 2020 04:03:29 -0000, bitrex <user@example.net> wrote:

On 2/20/20 8:46 PM, DecadentLinuxUserNumeroUno@decadence.org wrote:
bitrex <user@example.net> wrote in
news:CnF3G.207815$MX1.148708@fx43.iad:

On 2/20/20 7:15 PM, whit3rd wrote:
On Thursday, February 20, 2020 at 2:55:54 PM UTC-8, Commander
Kinsey wrote:
On Thu, 20 Feb 2020 22:27:48 -0000, Lasse Langwadt Christensen
langwadt@fonz.dk> wrote:

torsdag den 20. februar 2020 kl. 23.16.57 UTC+1 skrev 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 l... 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?

they are all on the same transformer so they are related, and
the feedback is a combination of 12V and 5V

Not sure how a shared transformer can work.

The transformer puts proportional (AC) voltages on its windings,
so the DC outputs are always proportional as long as resistive
and rectifier losses are similar to the test condtions. If you
draw zero current through one of the rectifiers, but normal
current through another, there's a discrepancy to be expected.
Usually, 12V power has a high tolerance, it's often the one to be
left out of the regulation feedback loop.


In modern PCs just about everything runs of the 12V but almost
nothing runs off it directly, it's a bulk supply to feed other
converters on the mobo.

LEDs, fans, hard drive spindles get it raw I guess.



Hard drive spindle motor drive circuits get it raw, but the spindle
motor itself gets a quite controlled feed from that circuit.


I think some PCIe cards like GPUs still use 3.3 off the PCI bus for
something. The really honking ones have a connector elsewhere on the PCB
for a higher current connection to the 12.

I run many GPUs off a 4 way multiplexer card. There's nothing but 12V feeding those cards. There might be a small amount of current available through the USB cables to the cards, but it ain't for power. Maybe just communication.

 

[bitrex]

On 2/21/20 10:22 PM, bitrex wrote:

On 2/21/20 6:18 PM, Phil Hobbs wrote:

Doesn't cross-reg go negative on a typical flyback switcher? If you
load the winding that makes the feedback, the drive goes up to the
others.

Flybacks are a bit anomalous in that they won't work without a fair bit
of leakage inductance--that's where the energy gets stored.

In the leakage? Not magnetizing inductance?


AIUI the energy is basically stored in the air gap, which is the
source of the leakage inductance.  But I've never designed a flyback,
so I wouldn't want to be too dogmatic about that.

Cheers

Phil Hobbs


The stock flyback is an irritating topology anyway and scales poorly. I
don't think most beefy modern PC supplies use them anyway they use
two-switch forward converters.

But the same statements wrt cross-regulation apply.

 

[bitrex]

On 2/21/20 6:18 PM, Phil Hobbs wrote:

Doesn't cross-reg go negative on a typical flyback switcher? If you
load the winding that makes the feedback, the drive goes up to the
others.

Flybacks are a bit anomalous in that they won't work without a fair bit
of leakage inductance--that's where the energy gets stored.

In the leakage? Not magnetizing inductance?


AIUI the energy is basically stored in the air gap, which is the source
of the leakage inductance.  But I've never designed a flyback, so I
wouldn't want to be too dogmatic about that.

Cheers

Phil Hobbs

The stock flyback is an irritating topology anyway and scales poorly. I
don't think most beefy modern PC supplies use them anyway they use
two-switch forward converters.

 

[sea moss]

*sigh*. Shouldn't post after having two glasses of wine. Magnetic
energy density is |B|**2/mu, so since perpendicular B is continuous
across the boundary, the energy density is mu_r times higher in the gap.

Cheers

Phil Hobbs
(Still on vacation in a beach house on Anna Maria Island)

Yeah the energy is mostly in the gap. The goal with the flyback transformer design is to have as much of this energy be transferred from one winding to another; by definition this is the mutual inductance.

Have a nice vacation. For warm weather I like to mix red wine with sprite and ice. Cheap wine of course...

 

[George Herold]

On Friday, February 21, 2020 at 9:14:38 PM UTC-5, Phil Hobbs wrote:

On 2020-02-21 21:06, Phil Hobbs wrote:
On 2020-02-21 20:58, Phil Hobbs wrote:
On 2020-02-21 20:49, sea moss wrote:
AIUI the energy is basically stored in the air gap, which is the source
of the leakage inductance.  But I've never designed a flyback, so I
wouldn't want to be too dogmatic about that.

Cheers

Phil Hobbs


In a flyback the energy is stored in the magnetizing inductance.  You
gap the core to decrease the value of magnetizing inductance vs. the
large, unregulated value the magnetic material gives you.  For a
given FET ON time, V = L*dI/dt means that peak current increases
proportional to the decrease in L.  But since energy is 1/2*L*I^2, a
smaller inductance means more stored energy.

The leakage inductance is the non-coupled inductance of each winding.
For modeling the effect of leakage on transformers with multiple
windings, Erickson/Maksimovic have a really good paper:
https://ecee.colorado.edu/~rwe/references/PESC98.pdf

Sure.  But almost all the energy is stored in the gap, because the
magnetic energy density is proportional to B*H.  Perpendicular B is
continuous across the interface, so H is 1/mu_R times higher in the
gap.   Since mu is in the thousands, the origin of the energy storage
is the same as that of the leakage inductance.

mu_R times higher.

*sigh*. Shouldn't post after having two glasses of wine. Magnetic
energy density is |B|**2/mu, so since perpendicular B is continuous
across the boundary, the energy density is mu_r times higher in the gap.

Cheers

Phil Hobbs
(Still on vacation in a beach house on Anna Maria Island)
Don't drink and derive. :^)

I'm going to pull out Erickson's book on SMPS's.. it got a long
section on magnetics that I only skimmed.
(who as a physics type, sees energy stored in the B-field... which
is mostly in the gap..)

George H.

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Phil Hobbs]

On 2020-02-21 22:12, sea moss wrote:

*sigh*. Shouldn't post after having two glasses of wine. Magnetic
energy density is |B|**2/mu, so since perpendicular B is continuous
across the boundary, the energy density is mu_r times higher in the gap.

Cheers

Phil Hobbs
(Still on vacation in a beach house on Anna Maria Island)

Yeah the energy is mostly in the gap. The goal with the flyback transformer design is to have as much of this energy be transferred from one winding to another; by definition this is the mutual inductance.

Have a nice vacation. For warm weather I like to mix red wine with sprite and ice. Cheap wine of course...

Thanks. When I have my druthers it's a G&T with Inverroche Verdant or
Junipero gin and Fever Tree tonic, with a twist of lime holding a little
umbrella.

(My 15 minutes of fame arrived some years back when my flight-attendant
daughter posted a photo of me in a Delta Sky Club with a little umbrella
in my drink. The in-flight folks loved it.)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

On Sat, 22 Feb 2020 15:18:53 -0500, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-02-21 22:12, sea moss wrote:
*sigh*. Shouldn't post after having two glasses of wine. Magnetic
energy density is |B|**2/mu, so since perpendicular B is continuous
across the boundary, the energy density is mu_r times higher in the gap.

Cheers

Phil Hobbs
(Still on vacation in a beach house on Anna Maria Island)

Yeah the energy is mostly in the gap. The goal with the flyback transformer design is to have as much of this energy be transferred from one winding to another; by definition this is the mutual inductance.

Have a nice vacation. For warm weather I like to mix red wine with sprite and ice. Cheap wine of course...


Thanks. When I have my druthers it's a G&T with Inverroche Verdant or
Junipero gin and Fever Tree tonic, with a twist of lime holding a little
umbrella.

My MD suggested that quinine can prevent cramps, but she said that I'd
have to drink a lot of tonic water to get much. A few bananas a week
seems like a more pleasant therapy.



--

John Larkin Highland Technology, Inc

The cork popped merrily, and Lord Peter rose to his feet.
"Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"

 

[Commander Kinsey]

On Fri, 21 Feb 2020 01:50:39 -0000, <DecadentLinuxUserNumeroUno@decadence.org> wrote:

"Commander Kinsey" <CFKinsey@military.org.jp> wrote in
newsp.0gbbg6ebwdg98l@glass:

On Fri, 21 Feb 2020 00:17:00 -0000, bitrex <user@example.net
wrote:

On 2/20/20 7:11 PM, Commander Kinsey wrote:
On Thu, 20 Feb 2020 23:41:11 -0000, bitrex <user@example.net
wrote:

On 2/20/20 5:55 PM, Commander Kinsey wrote:
On Thu, 20 Feb 2020 22:27:48 -0000, Lasse Langwadt Christensen
langwadt@fonz.dk> wrote:

torsdag den 20. februar 2020 kl. 23.16.57 UTC+1 skrev
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?

they are all on the same transformer so they are related, and
the feedback is a combination of 12V and 5V

Not sure how a shared transformer can work. Since I can draw
for example 5 times as much current from one line so the other
would be adversely affected surely? They have to adjust each
individually or it would go completely out of whack. I take
more 5V current, so the whole thing gets boosted, now 12V
becomes 14V.


Time-averaged output voltage(s) of an e.g. ideal flyback
converter with either single or multiple secondaries in
continuous conduction mode is a function of the input voltage,
duty cycle, and turns ratio only it isn't directly dependent on
any currents.

That is to say in the non-ideal case with a heavy load on the
5V and a light load on the 12 the 12 will not go out of
regulation nearly that bad in a well-designed converter; the
duty cycle is not supposed to be swinging wildly over light
load to full load in CCM (until maybe it drops into DCM at very
light load.)

Sounds like a clever design which fails under certain
circumstances. Plenty folk have a computer with two or more
graphics cards, so they need a 2nd power supply. Nowhere does
it state the supply has to be loaded on the 5V line.

Yes it "fails" when they cheap out on everything else in the
supply, surely. How much you pay for them again?

Not regulating the 12V properly is cheaping out, and from what
I've read in these groups a lot of them are like that.

They're CIT supplies. They don't explode like Alpine when you
dare to use more than 50% of the rated current, but they're not as
stable as a proper Corsair.


Best PC PSUs I have used are by EVGA.

Those guys know how to make circuit boards and propely manage what
is on them and where the juice is moving and what parts are producing
heat. Their MOBOs are pretty good too, thought the gamingtard mobo
makers have won out now due to so many gaming dopes building their
own machines. Once you get all that money you can start doing better
and better engineering in your products. Look at AMD. They *might*
beat out Intel for a bit, but don't count out the big boys that got
us all here yet.

I haven't seen much from EVGA. I got a graphics card from them once and it was fine. They make other stuff then?

 

"Commander Kinsey" <CFKinsey@military.org.jp> wrote in
newsp.0gim98w6wdg98l@glass:

On Fri, 21 Feb 2020 01:50:39 -0000,
DecadentLinuxUserNumeroUno@decadence.org> wrote:

"Commander Kinsey" <CFKinsey@military.org.jp> wrote in
newsp.0gbbg6ebwdg98l@glass:

On Fri, 21 Feb 2020 00:17:00 -0000, bitrex <user@example.net
wrote:

On 2/20/20 7:11 PM, Commander Kinsey wrote:
On Thu, 20 Feb 2020 23:41:11 -0000, bitrex <user@example.net
wrote:

On 2/20/20 5:55 PM, Commander Kinsey wrote:
On Thu, 20 Feb 2020 22:27:48 -0000, Lasse Langwadt
Christensen <langwadt@fonz.dk> wrote:

torsdag den 20. februar 2020 kl. 23.16.57 UTC+1 skrev
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?

they are all on the same transformer so they are related,
and the feedback is a combination of 12V and 5V

Not sure how a shared transformer can work. Since I can
draw for example 5 times as much current from one line so
the other would be adversely affected surely? They have to
adjust each individually or it would go completely out of
whack. I take more 5V current, so the whole thing gets
boosted, now 12V becomes 14V.


Time-averaged output voltage(s) of an e.g. ideal flyback
converter with either single or multiple secondaries in
continuous conduction mode is a function of the input
voltage, duty cycle, and turns ratio only it isn't directly
dependent on any currents.

That is to say in the non-ideal case with a heavy load on the
5V and a light load on the 12 the 12 will not go out of
regulation nearly that bad in a well-designed converter; the
duty cycle is not supposed to be swinging wildly over light
load to full load in CCM (until maybe it drops into DCM at
very light load.)

Sounds like a clever design which fails under certain
circumstances. Plenty folk have a computer with two or more
graphics cards, so they need a 2nd power supply. Nowhere does
it state the supply has to be loaded on the 5V line.

Yes it "fails" when they cheap out on everything else in the
supply, surely. How much you pay for them again?

Not regulating the 12V properly is cheaping out, and from what
I've read in these groups a lot of them are like that.

They're CIT supplies. They don't explode like Alpine when you
dare to use more than 50% of the rated current, but they're not
as stable as a proper Corsair.


Best PC PSUs I have used are by EVGA.

Those guys know how to make circuit boards and propely manage
what
is on them and where the juice is moving and what parts are
producing heat. Their MOBOs are pretty good too, thought the
gamingtard mobo makers have won out now due to so many gaming
dopes building their own machines. Once you get all that money
you can start doing better and better engineering in your
products. Look at AMD. They *might* beat out Intel for a bit,
but don't count out the big boys that got us all here yet.

I haven't seen much from EVGA. I got a graphics card from them
once and it was fine. They make other stuff then?

Motherboards, Graphics adapters, PSUs and a few other things.
They actually took Nvidia and did some cards which were fully
compliant wher other only had some of the elements. Look though their
cards, they had some of the best and the other guys followed, and now
all have high end cards with huge memoray arrays on them. EVGA was
among the earliest to make high end Nvidia offerings.