T. Ment wrote:
On Mon, 03 Jun 2019 20:29:07 -0400, Flasherly wrote:

Get your spare PS, or whatever else, all on your good money
on your good time on your best sale price.

My spare PS has a bad 16v 3300uf capacitor. I have a replacement part,
but it's waiting for a day with nothing better to do.

Paul said 5v minimum spec is 4.75v. But there must be some wiggle room
outside the spec. The question is, how much?

That's all I want to know.



That's the spec. 5%. That's all the margin you're allowed.

It's in all three of the ATX specs.

Now, pull a TTL datasheet. What is the spec in there ?
It's 5%. (Page 2, "Recommended Operating Conditions")
I realize this is an old part, and this is purely to
illustrate the situation we might fall into.

https://www.futurlec.com/Datasheet/74ls/74LS02.pdf

Design works with specs. There's a source and a sink
for that spec. Notice in this case, we've got no margin
for distribution loss.

So what we can conclude from this, is the circuit
actually runs with less than 4.75V. It will probably
run down to around 3.6V or so. However, you don't
want "things that cannot afford to make mistakes",
running at 3.6V! It's OK to say they run down to
3.6V, as part of a bar bet.

We had something in the lab one day, with a feedthru problem.
Two modules (boxes full of logic chips), one had power,
the other lost power, and there was so much leakage through
some logic signals, that the second unpowered module "siphoned"
power out of the other unit, and its logic was still running!
Needless to say, we were shocked, and needed to
go to the nearest bar for a drink :-) (Or something)

Now, on good power supplies, the typical variability
is 3%. But that's not how specsmanship works. I hope you
can see that something is a bit broken in this picture,
in terms of conservative design.

I'm still not concerned. But I'm also not that happy.
There are some brands of ATX supplies, that push crossload
right to the spec limit on purpose, like evil school boys.
And they really should not do that.

Paul