"Commander Kinsey" wrote in message
newsp.0f8x12x7wdg98l@glass...
On Wed, 19 Feb 2020 17:39:22 -0000, Rod Speed
wrote:



"Commander Kinsey" wrote in message
newsp.0f8u8x2fwdg98l@glass...
On Wed, 19 Feb 2020 17:02:33 -0000, Rod Speed
wrote:

Commander Kinsey 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?

Depends on what you mean by work correctly.

If I don't load the 5V at all, the 12V becomes 10.5V. Connecting a car
headlamp to the 5V rail (which draws about 1.75A) seems to make it
behave.

By definition, when whatever it is that decides
what mark space ratio is used with a switch
mode power supply, it has to measure the
voltage somewhere and since the spec on
the 5V rail is a lot tighter than on the 12V
rails, it makes sense to measure the 5V rail.

That means that unless the 5V rail has some
load on it, the regulation isnt great and so
you can get some odd effects on the 12V rail.

Its obvious possible to have a separate switch mode
power supply for each rail but even the more expensive
PC PSUs don't do it like that because the 5V and 3.3V
rails are normally well loaded in a PC,

I don't understand.

That's obvious.

Surely the mark space ratio has to be completely different for the 12V
and
5V rails

Nope. The high frequency chopped DC goes
into the high frequency transformer and the 5V
and 12V rails are just different output windings
on that. Look at a PC PSU circuit diagram.
http://www.pavouk.org/hw/en_atxps.html

because any computer at any given moment could be drawing more from one
than the other, then differently several seconds from now.

The current on the 5V rail whose voltage is used to vary
the mark space ratio doesn't actually vary all that much.

Incorrect,

Nope.

this PSU is happy with me taking 1.75A from 5V. It's rated at up to 35A
on 5V. So clearly it will be ok anywhere from 1.75A to 35A on 5V.
Explain how that is possible to vary what I take from 5V without affecting
the 12V regulation, yet 0A on 5V upsets it.

The explanation is obvious, the mark space
ratio doesn't vary with the current, stupid.

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.

You'd be better off with dedicated 12V switch mode power
supplys. Surprisingly cheap now even at the higher powers.

Link me to an example,

https://www.ebay.co.uk/itm/Transform...DPmF 8NegnqEg

That's just the one at the top of the hit list.

because I looked at them before, and at the high powers I need (300W per
card), it's massively cheaper to buy even the top of the range PC
supplies
from Corsair etc.,

Not true with that one.

I'll stick to Corsair PC PSUs.

More fool you.

That one you linked to is Chinese noname rubbish,

Its just the first hit on the list. Plenty of other branded ones.

I find those tend to overheat and break if you load them over 50%.

Mine doesn't,.

Yours is £21 for 360W. I can get a Corsair of the same power rating for
£30 that I know will work. Or 550W for £35.50.

And you have to fart around loading the 5V rail. Yes, you are that stupid.