Current limiting with a mosfet?...

[whit3rd]

On Saturday, June 17, 2023 at 11:34:38 AM UTC-7, John Larkin wrote:

On Sat, 17 Jun 2023 18:03:02 +0100, \"Commander Kinsey\"
C...@nospam.com> wrote:

On Sat, 17 Jun 2023 01:00:46 +0100, John Larkin <jla...@highlandsnipmetechnology.com> wrote:

On Sat, 17 Jun 2023 00:21:35 +0100, \"Commander Kinsey\"
C...@nospam.com> wrote:

On Fri, 16 Jun 2023 22:45:39 +0100, John Larkin <jla...@highlandsnipmetechnology.com> wrote:

On Fri, 16 Jun 2023 21:16:29 +0100, \"Commander Kinsey\"
C...@nospam.com> wrote:

On Fri, 16 Jun 2023 18:43:59 +0100, John Larkin <jla...@highlandsnipmetechnology.com> wrote:

On Fri, 16 Jun 2023 17:58:05 +0100, \"Commander Kinsey\"
C...@nospam.com> wrote:

On Fri, 16 Jun 2023 04:16:02 +0100, John Larkin <jla...@highlandsnipmetechnology.com> wrote:

You mean the dpaks that run at 200 amps?

Uh oh....
I got these: https://www.vishay.com/docs/77646/sum40014m.pdf

375 watts is absurd for that toy fet. Even 125 is crazy.

I don\'t like the package. I like them screwed firmly to the heatsink like TO-247. Those are the size I see inside 1-3kW supplies.

Phil is referring to IR posting absurd specs for their fets, like
ignoring the leads, or submerging them in boiling coolants to get the
ratings. So, everybody else had to do it to be competitive.

It does say \"at 25C\".

Solder it to half the universe made of silver.

I could use a liquid nitrogen setup like some crazy folk do with their CPUs to gain a few GHz. Or fill the whole computer with oil.

Oil doesn\'t conduct heat very well.

It\'s better than air, and computers dislike being filled with water.
But don\'t cool your TO247 mosfets with oil or air. They need a metal
heat sink first.

The needed \'metal heat sink\' is the tab of the TO247; you can cool it with liquid air, fluorinert, refrigerant,
moving air, moving oil... removing heat takes more than some metal conduction.

IBM had some hot computers that used helium and springloaded plungers.
<https://www.ibm.com/ibm/history/exhibits/vintage/vintage_4506VV2137.html>>

 

[John Larkin]

On Sat, 17 Jun 2023 19:05:48 -0700 (PDT), whit3rd <whit3rd@gmail.com>
wrote:

On Saturday, June 17, 2023 at 11:34:38?AM UTC-7, John Larkin wrote:
On Sat, 17 Jun 2023 18:03:02 +0100, \"Commander Kinsey\"
C...@nospam.com> wrote:

On Sat, 17 Jun 2023 01:00:46 +0100, John Larkin <jla...@highlandsnipmetechnology.com> wrote:

On Sat, 17 Jun 2023 00:21:35 +0100, \"Commander Kinsey\"
C...@nospam.com> wrote:

On Fri, 16 Jun 2023 22:45:39 +0100, John Larkin <jla...@highlandsnipmetechnology.com> wrote:

On Fri, 16 Jun 2023 21:16:29 +0100, \"Commander Kinsey\"
C...@nospam.com> wrote:

On Fri, 16 Jun 2023 18:43:59 +0100, John Larkin <jla...@highlandsnipmetechnology.com> wrote:

On Fri, 16 Jun 2023 17:58:05 +0100, \"Commander Kinsey\"
C...@nospam.com> wrote:

On Fri, 16 Jun 2023 04:16:02 +0100, John Larkin <jla...@highlandsnipmetechnology.com> wrote:

You mean the dpaks that run at 200 amps?

Uh oh....
I got these: https://www.vishay.com/docs/77646/sum40014m.pdf

375 watts is absurd for that toy fet. Even 125 is crazy.

I don\'t like the package. I like them screwed firmly to the heatsink like TO-247. Those are the size I see inside 1-3kW supplies.

Phil is referring to IR posting absurd specs for their fets, like
ignoring the leads, or submerging them in boiling coolants to get the
ratings. So, everybody else had to do it to be competitive.

It does say \"at 25C\".

Solder it to half the universe made of silver.

I could use a liquid nitrogen setup like some crazy folk do with their CPUs to gain a few GHz. Or fill the whole computer with oil.

Oil doesn\'t conduct heat very well.

It\'s better than air, and computers dislike being filled with water.
But don\'t cool your TO247 mosfets with oil or air. They need a metal
heat sink first.

The needed \'metal heat sink\' is the tab of the TO247; you can cool it with liquid air, fluorinert, refrigerant,
moving air, moving oil... removing heat takes more than some metal conduction.

One needs to spread the heat out, with a high thermal conductivity
metal, before transferring to a wimpy thermal conductor like air.
Namely increase the hot surface area.

There is a reason people bolt TO247 transistors to an expensive heat
sink.

https://www.dropbox.com/s/nmyb9pz0qpma2xa/Amp.jpg?dl=0

That has 32 giant mosfets clamped to copper bars to do the initial
heat spreading, into a huge air-cooled aluminum heat sink. All 32
drains (16 p-channel, 16 n-channel) clamp to the heat sink with no
insulators. The heat sink is the amp output.

IBM had some hot computers that used helium and springloaded plungers.
https://www.ibm.com/ibm/history/exhibits/vintage/vintage_4506VV2137.html

The chips were in contact with metal plungers that fit closely into
holes in a water-cooled block.

 

[whit3rd]

On Saturday, June 17, 2023 at 8:44:48 PM UTC-7, John Larkin wrote:

On Sat, 17 Jun 2023 19:05:48 -0700 (PDT), whit3rd <whi...@gmail.com
wrote:
On Saturday, June 17, 2023 at 11:34:38?AM UTC-7, John Larkin wrote:
On Sat, 17 Jun 2023 18:03:02 +0100, \"Commander Kinsey\"
C...@nospam.com> wrote:

On Sat, 17 Jun 2023 01:00:46 +0100, John Larkin <jla...@highlandsnipmetechnology.com> wrote:

On Sat, 17 Jun 2023 00:21:35 +0100, \"Commander Kinsey\"
C...@nospam.com> wrote:

On Fri, 16 Jun 2023 22:45:39 +0100, John Larkin <jla...@highlandsnipmetechnology.com> wrote:

On Fri, 16 Jun 2023 21:16:29 +0100, \"Commander Kinsey\"
C...@nospam.com> wrote:

On Fri, 16 Jun 2023 18:43:59 +0100, John Larkin <jla...@highlandsnipmetechnology.com> wrote:

On Fri, 16 Jun 2023 17:58:05 +0100, \"Commander Kinsey\"
C...@nospam.com> wrote:

On Fri, 16 Jun 2023 04:16:02 +0100, John Larkin <jla...@highlandsnipmetechnology.com> wrote:

You mean the dpaks that run at 200 amps?

Uh oh....
I got these: https://www.vishay.com/docs/77646/sum40014m.pdf

375 watts is absurd for that toy fet. Even 125 is crazy.

I don\'t like the package. I like them screwed firmly to the heatsink like TO-247. Those are the size I see inside 1-3kW supplies.

Phil is referring to IR posting absurd specs for their fets, like
ignoring the leads, or submerging them in boiling coolants to get the
ratings. So, everybody else had to do it to be competitive.

It does say \"at 25C\".

Solder it to half the universe made of silver.

I could use a liquid nitrogen setup like some crazy folk do with their CPUs to gain a few GHz. Or fill the whole computer with oil.

Oil doesn\'t conduct heat very well.

It\'s better than air, and computers dislike being filled with water.
But don\'t cool your TO247 mosfets with oil or air. They need a metal
heat sink first.

The needed \'metal heat sink\' is the tab of the TO247; you can cool it with liquid air, fluorinert, refrigerant,
moving air, moving oil... removing heat takes more than some metal conduction.
One needs to spread the heat out, with a high thermal conductivity
metal, before transferring to a wimpy thermal conductor like air.
Namely increase the hot surface area.

There is a reason people bolt TO247 transistors to an expensive heat
sink.

What if they\'re for switchmode use? The first watt or two doesn\'t require any
heatsink other than the tab. Other uses, maybe the bolt-to item is
required. It ain\'t for the transistor, it\'s for the waste heat, and
you have to know how much before buying a place to dump heat.

IBM had some hot computers that used helium and springloaded plungers.
https://www.ibm.com/ibm/history/exhibits/vintage/vintage_4506VV2137.html
The chips were in contact with metal plungers that fit closely into
holes in a water-cooled block.

.... and the heat transfer fluid from chip to plunger was helium.

 

[John Larkin]

On Sat, 17 Jun 2023 21:02:56 -0700 (PDT), whit3rd <whit3rd@gmail.com>
wrote:

On Saturday, June 17, 2023 at 8:44:48?PM UTC-7, John Larkin wrote:
On Sat, 17 Jun 2023 19:05:48 -0700 (PDT), whit3rd <whi...@gmail.com
wrote:
On Saturday, June 17, 2023 at 11:34:38?AM UTC-7, John Larkin wrote:
On Sat, 17 Jun 2023 18:03:02 +0100, \"Commander Kinsey\"
C...@nospam.com> wrote:

On Sat, 17 Jun 2023 01:00:46 +0100, John Larkin <jla...@highlandsnipmetechnology.com> wrote:

On Sat, 17 Jun 2023 00:21:35 +0100, \"Commander Kinsey\"
C...@nospam.com> wrote:

On Fri, 16 Jun 2023 22:45:39 +0100, John Larkin <jla...@highlandsnipmetechnology.com> wrote:

On Fri, 16 Jun 2023 21:16:29 +0100, \"Commander Kinsey\"
C...@nospam.com> wrote:

On Fri, 16 Jun 2023 18:43:59 +0100, John Larkin <jla...@highlandsnipmetechnology.com> wrote:

On Fri, 16 Jun 2023 17:58:05 +0100, \"Commander Kinsey\"
C...@nospam.com> wrote:

On Fri, 16 Jun 2023 04:16:02 +0100, John Larkin <jla...@highlandsnipmetechnology.com> wrote:

You mean the dpaks that run at 200 amps?

Uh oh....
I got these: https://www.vishay.com/docs/77646/sum40014m.pdf

375 watts is absurd for that toy fet. Even 125 is crazy.

I don\'t like the package. I like them screwed firmly to the heatsink like TO-247. Those are the size I see inside 1-3kW supplies.

Phil is referring to IR posting absurd specs for their fets, like
ignoring the leads, or submerging them in boiling coolants to get the
ratings. So, everybody else had to do it to be competitive.

It does say \"at 25C\".

Solder it to half the universe made of silver.

I could use a liquid nitrogen setup like some crazy folk do with their CPUs to gain a few GHz. Or fill the whole computer with oil.

Oil doesn\'t conduct heat very well.

It\'s better than air, and computers dislike being filled with water.
But don\'t cool your TO247 mosfets with oil or air. They need a metal
heat sink first.

The needed \'metal heat sink\' is the tab of the TO247; you can cool it with liquid air, fluorinert, refrigerant,
moving air, moving oil... removing heat takes more than some metal conduction.
One needs to spread the heat out, with a high thermal conductivity
metal, before transferring to a wimpy thermal conductor like air.
Namely increase the hot surface area.

There is a reason people bolt TO247 transistors to an expensive heat
sink.

What if they\'re for switchmode use? The first watt or two doesn\'t require any
heatsink other than the tab. Other uses, maybe the bolt-to item is
required. It ain\'t for the transistor, it\'s for the waste heat, and
you have to know how much before buying a place to dump heat.

There\'s no reason to use giant TO247 fets to dissipate a couple of
watts. You can do that with a dpak on a PC board. PCB copper and vias
can be your heat spreader at the few-watts level, but most power
transistors need to connect to metal first to slurp the heat away.

IBM had some hot computers that used helium and springloaded plungers.
https://www.ibm.com/ibm/history/exhibits/vintage/vintage_4506VV2137.html
The chips were in contact with metal plungers that fit closely into
holes in a water-cooled block.

... and the heat transfer fluid from chip to plunger was helium.

No, the plunger contacted the chip directly, with a bunch of force.
The helium helped transfer heat from the plungers into their bores in
the water-cooled block.

 

[whit3rd]

On Saturday, June 17, 2023 at 9:33:11 PM UTC-7, John Larkin wrote:

On Sat, 17 Jun 2023 21:02:56 -0700 (PDT), whit3rd <whi...@gmail.com
wrote:
On Saturday, June 17, 2023 at 8:44:48?PM UTC-7, John Larkin wrote:

IBM had some hot computers that used helium and springloaded plungers..
https://www.ibm.com/ibm/history/exhibits/vintage/vintage_4506VV2137.html
The chips were in contact with metal plungers that fit closely into
holes in a water-cooled block.

... and the heat transfer fluid from chip to plunger was helium.

No, the plunger contacted the chip directly, with a bunch of force.
The helium helped transfer heat from the plungers into their bores in
the water-cooled block.

What does the \'no\' mean? Helium was a thermal transfer medium,
both from chip to plunger and plunger to refrigerated parts.
Contact conduction happens too, but is imperfect.

 

[Phil Hobbs]

whit3rd <whit3rd@gmail.com> wrote:

On Saturday, June 17, 2023 at 9:33:11 PM UTC-7, John Larkin wrote:
On Sat, 17 Jun 2023 21:02:56 -0700 (PDT), whit3rd <whi...@gmail.com
wrote:
On Saturday, June 17, 2023 at 8:44:48?PM UTC-7, John Larkin wrote:

IBM had some hot computers that used helium and springloaded plungers.
https://www.ibm.com/ibm/history/exhibits/vintage/vintage_4506VV2137.html
The chips were in contact with metal plungers that fit closely into
holes in a water-cooled block.

... and the heat transfer fluid from chip to plunger was helium.

No, the plunger contacted the chip directly, with a bunch of force.
The helium helped transfer heat from the plungers into their bores in
the water-cooled block.

What does the \'no\' mean? Helium was a thermal transfer medium,
both from chip to plunger and plunger to refrigerated parts.
Contact conduction happens too, but is imperfect.

Helium-filled TCMs only lasted one generation. By the 3090 period
(1985ish), TCMS were oil-filled, and stayed that way.

Cheers

Phil Hobbs

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

 

[Chris Jones]

On 17/06/2023 4:03 am, John Larkin wrote:

On Fri, 16 Jun 2023 10:56:05 -0700 (PDT), whit3rd <whit3rd@gmail.com
wrote:

On Friday, June 16, 2023 at 9:49:31?AM UTC-7, Commander Kinsey wrote:
On Thu, 15 Jun 2023 23:03:50 +0100, John Larkin <jla...@highlandsnipmetechnology.com> wrote:

On Thu, 15 Jun 2023 21:57:37 +0100, \"Commander Kinsey\"
C...@nospam.com> wrote:

On Thu, 15 Jun 2023 17:17:42 +0100, John Larkin <jla...@highlandsnipmetechnology.com> wrote:


Why not use resistors or, as someone has suggested, wire?

Because I want to be able to tweak it. A variable resistor of that power output, and at fractions of an ohm, is impossible to obtain.

Measure the power suppies and pick the resistor or wire length.

I\'ve never seen milliohm huge variable resistors.

They\'re not called resistors, rather are \'shunts\'.
https://m.media-amazon.com/images/I/41-myRflg9L.jpg

They\'re variable by grinding material off the edges...

... and with a bit of bronze shim stock, or (at lower power dissipation) wire,
they\'re nearly trivial to build. Brazing is better than soldering, though, and
attaching wires is not easy; tapered plugs fitted to conical holes DOES make a
good mechanical and electrical connection for welders.

There are giant wirewound resistors with slider contacts. ebay may
have some.

I saw in an old book they used to fill a glass pipe with mercury (maybe
an inch diameter and a couple of feet long), with an eletrode at the top
and bottom of the mercury column. To adjust the resistance you stick a
glass rod part way down into the pipe, to reduce the cross-section area
of part of the mercury column. I presume there would be a bowl at the
top so the mercury didn\'t spill on the floor. I guess you\'d need some
sort of clamp to stop the glass rod floating out when you let go.
Probably best not to get it too hot either in case it evaporates too fast.

 

[Chris Jones]

On 18/06/2023 6:34 am, John Larkin wrote:

On Sat, 17 Jun 2023 21:00:13 +0100, \"Commander Kinsey\"
CK1@nospam.com> wrote:

On Sat, 17 Jun 2023 19:34:12 +0100, John Larkin <jlarkin@highlandsnipmetechnology.com> wrote:

On Sat, 17 Jun 2023 18:03:02 +0100, \"Commander Kinsey\"
CK1@nospam.com> wrote:

On Sat, 17 Jun 2023 01:00:46 +0100, John Larkin <jlarkin@highlandsnipmetechnology.com> wrote:

On Sat, 17 Jun 2023 00:21:35 +0100, \"Commander Kinsey\"
CK1@nospam.com> wrote:

On Fri, 16 Jun 2023 22:45:39 +0100, John Larkin <jlarkin@highlandsnipmetechnology.com> wrote:

On Fri, 16 Jun 2023 21:16:29 +0100, \"Commander Kinsey\"
CK1@nospam.com> wrote:

On Fri, 16 Jun 2023 18:43:59 +0100, John Larkin <jlarkin@highlandsnipmetechnology.com> wrote:

On Fri, 16 Jun 2023 17:58:05 +0100, \"Commander Kinsey\"
CK1@nospam.com> wrote:

On Fri, 16 Jun 2023 04:16:02 +0100, John Larkin <jlarkin@highlandsnipmetechnology.com> wrote:

You mean the dpaks that run at 200 amps?

Uh oh....
I got these: https://www.vishay.com/docs/77646/sum40014m.pdf

375 watts is absurd for that toy fet. Even 125 is crazy.

I don\'t like the package. I like them screwed firmly to the heatsink like TO-247. Those are the size I see inside 1-3kW supplies.

Phil is referring to IR posting absurd specs for their fets, like
ignoring the leads, or submerging them in boiling coolants to get the
ratings. So, everybody else had to do it to be competitive.

It does say \"at 25C\".

Solder it to half the universe made of silver.

I could use a liquid nitrogen setup like some crazy folk do with their CPUs to gain a few GHz. Or fill the whole computer with oil.

Oil doesn\'t conduct heat very well.

It\'s better than air, and computers dislike being filled with water.

But don\'t cool your TO247 mosfets with oil or air. They need a metal
heat sink first.


Don\'t they cool transformers with oil?

Cool and insulate. On big ones, the oil is pumped around because
convection isn\'t enough.

Oil and even air will cool things pretty well if you move a lot of it
across the surfaces.

Presumably oil is way better than air at conducting.

Ballpark 5:1, but you can move air a lot faster than you can move oil.
So it sort of cancels out.

Air is a lot nicer to work with than oil, too.


And since in a computer, the heatsinks stay and oil replaces air, it should improve. However.... I wonder if they change the fans to something capable of shifting oil? Which conducts better, almost stationary oil (I assume you can\'t pump it very fast) or fast moving air?

Cray submerged their logic in refrigerated, pumped liquid freon.

Are any computers oil cooled? I think giant server farms are air or
water cooled.

https://dug.com/dug-cool/

I wonder what it does to the microstrip impedances.

 

[John Larkin]

On Sun, 18 Jun 2023 22:35:52 +1000, Chris Jones
<lugnut808@spam.yahoo.com> wrote:

On 18/06/2023 6:34 am, John Larkin wrote:
On Sat, 17 Jun 2023 21:00:13 +0100, \"Commander Kinsey\"
CK1@nospam.com> wrote:

On Sat, 17 Jun 2023 19:34:12 +0100, John Larkin <jlarkin@highlandsnipmetechnology.com> wrote:

On Sat, 17 Jun 2023 18:03:02 +0100, \"Commander Kinsey\"
CK1@nospam.com> wrote:

On Sat, 17 Jun 2023 01:00:46 +0100, John Larkin <jlarkin@highlandsnipmetechnology.com> wrote:

On Sat, 17 Jun 2023 00:21:35 +0100, \"Commander Kinsey\"
CK1@nospam.com> wrote:

On Fri, 16 Jun 2023 22:45:39 +0100, John Larkin <jlarkin@highlandsnipmetechnology.com> wrote:

On Fri, 16 Jun 2023 21:16:29 +0100, \"Commander Kinsey\"
CK1@nospam.com> wrote:

On Fri, 16 Jun 2023 18:43:59 +0100, John Larkin <jlarkin@highlandsnipmetechnology.com> wrote:

On Fri, 16 Jun 2023 17:58:05 +0100, \"Commander Kinsey\"
CK1@nospam.com> wrote:

On Fri, 16 Jun 2023 04:16:02 +0100, John Larkin <jlarkin@highlandsnipmetechnology.com> wrote:

You mean the dpaks that run at 200 amps?

Uh oh....
I got these: https://www.vishay.com/docs/77646/sum40014m.pdf

375 watts is absurd for that toy fet. Even 125 is crazy.

I don\'t like the package. I like them screwed firmly to the heatsink like TO-247. Those are the size I see inside 1-3kW supplies.

Phil is referring to IR posting absurd specs for their fets, like
ignoring the leads, or submerging them in boiling coolants to get the
ratings. So, everybody else had to do it to be competitive.

It does say \"at 25C\".

Solder it to half the universe made of silver.

I could use a liquid nitrogen setup like some crazy folk do with their CPUs to gain a few GHz. Or fill the whole computer with oil.

Oil doesn\'t conduct heat very well.

It\'s better than air, and computers dislike being filled with water.

But don\'t cool your TO247 mosfets with oil or air. They need a metal
heat sink first.


Don\'t they cool transformers with oil?

Cool and insulate. On big ones, the oil is pumped around because
convection isn\'t enough.

Oil and even air will cool things pretty well if you move a lot of it
across the surfaces.

Presumably oil is way better than air at conducting.

Ballpark 5:1, but you can move air a lot faster than you can move oil.
So it sort of cancels out.

Air is a lot nicer to work with than oil, too.


And since in a computer, the heatsinks stay and oil replaces air, it should improve. However.... I wonder if they change the fans to something capable of shifting oil? Which conducts better, almost stationary oil (I assume you can\'t pump it very fast) or fast moving air?

Cray submerged their logic in refrigerated, pumped liquid freon.

Are any computers oil cooled? I think giant server farms are air or
water cooled.

https://dug.com/dug-cool/

Sounds messy. At least water evaporates.

I wonder what it does to the microstrip impedances.

Dielectric constant is around 2, so it will pull down impedance a bit.
That probably won\'t affect a computer board.

 

[John Larkin]

On Sat, 17 Jun 2023 21:48:23 -0700 (PDT), whit3rd <whit3rd@gmail.com>
wrote:

On Saturday, June 17, 2023 at 9:33:11?PM UTC-7, John Larkin wrote:
On Sat, 17 Jun 2023 21:02:56 -0700 (PDT), whit3rd <whi...@gmail.com
wrote:
On Saturday, June 17, 2023 at 8:44:48?PM UTC-7, John Larkin wrote:

IBM had some hot computers that used helium and springloaded plungers.
https://www.ibm.com/ibm/history/exhibits/vintage/vintage_4506VV2137.html
The chips were in contact with metal plungers that fit closely into
holes in a water-cooled block.

... and the heat transfer fluid from chip to plunger was helium.

No, the plunger contacted the chip directly, with a bunch of force.
The helium helped transfer heat from the plungers into their bores in
the water-cooled block.

What does the \'no\' mean? Helium was a thermal transfer medium,
both from chip to plunger and plunger to refrigerated parts.
Contact conduction happens too, but is imperfect.

Do you ever design electronics where thermals matter? Show us.

 

[John Walliker]

On Thursday, 15 June 2023 at 13:48:24 UTC+1, Commander Kinsey wrote:

On Thu, 15 Jun 2023 13:18:28 +0100, Chris Jones <lugn...@spam.yahoo.com> wrote:

On 15/06/2023 4:37 am, Commander Kinsey wrote:
On Wed, 14 Jun 2023 18:31:09 +0100, boB <b...@k7iq.com> wrote:

On Wed, 14 Jun 2023 18:07:05 +0100, \"Commander Kinsey\"
C...@nospam.com> wrote:

On Wed, 14 Jun 2023 17:59:32 +0100, Commander Kinsey <C...@nospam.com
wrote:

Is it easy to make a simple circuit to limit a 12V DC current to 50A
using a mosfet?

Can I just give it a variable voltage to turn it on a certain
amount? Or is it not that easy?

If it\'s simpler, a potentiometer to vary the voltage drop across the
mosfet would do.

I\'m trying to balance a few power supplies all running in parallel,
so they do the same work each. I don\'t mind tweaking a potentiometer
on each one to make the current about equal on each.


You could make a circuit that shuts the FET OFF when you hit 51
amps. If you want to limit current by lowering the output voltage
using that FET, then you are asking for blown up FET I think.

You\'d need a huge amount of dissipation to make that work and maybe
you don\'t want to reduce the output voltage.

But just shutting off could work fairly easily.

Shutting it off would worsen the problem - the other supplies would then
be likely to hit their limits.

Surely the TO-247 MOSFETs can take a high current? Or is that only when
turned fully on? I guess it would be complicated to make them pulse,
and probably upset the SMPS it\'s adjusting.

At the moment I\'m just going to sort out the biggest imbalance - two of
the supplies being 12.85V, two being 12.35V, and one adjustable. The
two high voltage ones I can put big TO-220 schottkys on to drop half a
volt.


Simplest method:
Open up the power supplies and find the resistive divider in the voltage
sensing circuit that sets the output voltage.
I thought of that, but they\'re tightly packed complex circuits, and I don\'t even have a pinout. The power output was obvious, the huge connectors. I had to find the power on pin by trial and error.
Put a trimpot with series
resistor across one of the resistors, to make it slightly adjustable,
but with hardly any more range than needed to reach the desired voltage..
Then adjust each of the SMPS to the same output voltage. Then, rather
than paralleling all of the SMPS with fat wire and running longer wire
from there to your load, instead run separate, fairly long (maybe longer
than physically necessary) (and equally long for each PSU) wires, no
thicker than necessary, from each SMPS separately to the load. This will
provide a small ballast resistor between each SMPS and the load. You can
still twist the positive and negative output wires of each SMPS, to
minimise the loop area between them and thereby minimise inductance. If
the inductance of the long wires is still a problem, you could add lots
more electrolytic capacitors at the load end.

Proper method:
Buy SMPS designed for parallel operation.
They are, but I don\'t have the specs sheet. They\'re from an HP blade server, and the end user is supposed to just plug them in and not know what the pins do.

At £14 for 2600W at 12V, I wasn\'t going to buy anything else!

Now that you have told us (approximately) what power supplies you are using
I can probably help. I have a few of these. I have found the part number of the
high current edge connector and bought a few. I even have some data sheets!
Just give me a day to dig out the information from my \"filing system\"...
If you can\'t wait, the clue is that they are made by Murata who DO publish data.
They are designed to be operated in parallel with no external components (as long
as all the parallel PSUs are the same model and power rating).
They have an internal trimpot that can be used to tweak the output voltage slightly.

John

 

[whit3rd]

On Sunday, June 18, 2023 at 8:19:05 AM UTC-7, John Larkin wrote:

On Sat, 17 Jun 2023 21:48:23 -0700 (PDT), whit3rd <whi...@gmail.com
wrote:
On Saturday, June 17, 2023 at 9:33:11?PM UTC-7, John Larkin wrote:

No, the plunger contacted the chip directly, with a bunch of force.
The helium helped transfer heat from the plungers into their bores in
the water-cooled block.

What does the \'no\' mean? Helium was a thermal transfer medium,
both from chip to plunger and plunger to refrigerated parts.
Contact conduction happens too, but is imperfect.

Do you ever design electronics where thermals matter? Show us.

I already did; I did some stepper control with TO220 transistors, just
sticking up in air, because they didn\'t need any attached heatsinking.
That\'s a word picture...

 

[John Larkin]

On Sun, 18 Jun 2023 18:46:29 -0700 (PDT), whit3rd <whit3rd@gmail.com>
wrote:

On Sunday, June 18, 2023 at 8:19:05?AM UTC-7, John Larkin wrote:
On Sat, 17 Jun 2023 21:48:23 -0700 (PDT), whit3rd <whi...@gmail.com
wrote:
On Saturday, June 17, 2023 at 9:33:11?PM UTC-7, John Larkin wrote:

No, the plunger contacted the chip directly, with a bunch of force.
The helium helped transfer heat from the plungers into their bores in
the water-cooled block.

What does the \'no\' mean? Helium was a thermal transfer medium,
both from chip to plunger and plunger to refrigerated parts.
Contact conduction happens too, but is imperfect.

Do you ever design electronics where thermals matter? Show us.

I already did; I did some stepper control with TO220 transistors, just
sticking up in air, because they didn\'t need any attached heatsinking.
That\'s a word picture...

How much current per transistor? A stepper driver may not need heat
sinking.

In the OP\'s case, he may drop 0.7 volts at 50 amps, 35 watts per fet.
That will need heat sinking.

 

[John Walliker]

On Sunday, 18 June 2023 at 21:30:48 UTC+1, John Walliker wrote:

On Thursday, 15 June 2023 at 13:48:24 UTC+1, Commander Kinsey wrote:
On Thu, 15 Jun 2023 13:18:28 +0100, Chris Jones <lugn...@spam.yahoo.com> wrote:

On 15/06/2023 4:37 am, Commander Kinsey wrote:
On Wed, 14 Jun 2023 18:31:09 +0100, boB <b...@k7iq.com> wrote:

On Wed, 14 Jun 2023 18:07:05 +0100, \"Commander Kinsey\"
C...@nospam.com> wrote:

On Wed, 14 Jun 2023 17:59:32 +0100, Commander Kinsey <C...@nospam.com
wrote:

Is it easy to make a simple circuit to limit a 12V DC current to 50A
using a mosfet?

Can I just give it a variable voltage to turn it on a certain
amount? Or is it not that easy?

If it\'s simpler, a potentiometer to vary the voltage drop across the
mosfet would do.

I\'m trying to balance a few power supplies all running in parallel,
so they do the same work each. I don\'t mind tweaking a potentiometer
on each one to make the current about equal on each.


You could make a circuit that shuts the FET OFF when you hit 51
amps. If you want to limit current by lowering the output voltage
using that FET, then you are asking for blown up FET I think.

You\'d need a huge amount of dissipation to make that work and maybe
you don\'t want to reduce the output voltage.

But just shutting off could work fairly easily.

Shutting it off would worsen the problem - the other supplies would then
be likely to hit their limits.

Surely the TO-247 MOSFETs can take a high current? Or is that only when
turned fully on? I guess it would be complicated to make them pulse,
and probably upset the SMPS it\'s adjusting.

At the moment I\'m just going to sort out the biggest imbalance - two of
the supplies being 12.85V, two being 12.35V, and one adjustable. The
two high voltage ones I can put big TO-220 schottkys on to drop half a
volt.


Simplest method:
Open up the power supplies and find the resistive divider in the voltage
sensing circuit that sets the output voltage.
I thought of that, but they\'re tightly packed complex circuits, and I don\'t even have a pinout. The power output was obvious, the huge connectors. I had to find the power on pin by trial and error.
Put a trimpot with series
resistor across one of the resistors, to make it slightly adjustable,
but with hardly any more range than needed to reach the desired voltage.
Then adjust each of the SMPS to the same output voltage. Then, rather
than paralleling all of the SMPS with fat wire and running longer wire
from there to your load, instead run separate, fairly long (maybe longer
than physically necessary) (and equally long for each PSU) wires, no
thicker than necessary, from each SMPS separately to the load. This will
provide a small ballast resistor between each SMPS and the load. You can
still twist the positive and negative output wires of each SMPS, to
minimise the loop area between them and thereby minimise inductance. If
the inductance of the long wires is still a problem, you could add lots
more electrolytic capacitors at the load end.

Proper method:
Buy SMPS designed for parallel operation.
They are, but I don\'t have the specs sheet. They\'re from an HP blade server, and the end user is supposed to just plug them in and not know what the pins do.

At £14 for 2600W at 12V, I wasn\'t going to buy anything else!

Now that you have told us (approximately) what power supplies you are using
I can probably help. I have a few of these. I have found the part number of the
high current edge connector and bought a few. I even have some data sheets!
Just give me a day to dig out the information from my \"filing system\"...
If you can\'t wait, the clue is that they are made by Murata who DO publish data.
They are designed to be operated in parallel with no external components (as long
as all the parallel PSUs are the same model and power rating).
They have an internal trimpot that can be used to tweak the output voltage slightly.

John

The connectors I used are FCI/Amphenol 10046971-001LF
They are available from multiple distributors.

Here is a data sheet for an equivalent Murata power supply with details
about how to connect them for parallel operation.
https://www.murata.com/products/productdata/8807026982942/d1u86g-w-460-12-hxx.pdf
\"Current share is achieved using the droop method. Nominal output voltage (12.30V) is achieved
at 1A load and output voltage drops at a rate of 7.83mv per amp increase.\"
There is a current monitoring output where the output voltage is proportional
to the load current. The sensitivity is scaled according to the power supply rating.
There is also an i2c interface which gives more data.
If you want more information, search for \"HP common slot hacking\"

John