Whats the chances of the mains "hiccuping".

[cupoftea]

Hi,
Customer sent us 20W offline flyback SMPS (60-265vac). C(in) is 47uF. Its one of a number of power supplies in a unit.
The input fuse on this board is a 7mm long piece of PCB track which is 0.25mm wide, and is of 2 oz copper...Saturn PCB says it would blow after 1sec at 3.5A.
But if the mains hiccups at say 0.01Hz, then the SMPS will inrush repeatedly through a hot NTC, and so do you think the fuse Track could blow?
Cust doesnt want to change the fuse track for a normal fuse.

 

[Easy peasy]

In India - a fairly common occurrence

 

[dick_freebird]

I'm sure this is "local flavor" but re your inrush current scenario, the
key question is not the rep rate (which here should not be a cyclic
thing without some close-in bad actor) but the dropout duration,
which then determines how much input filter charge must be
returned in quick time. If your sag doesn't hit some UVLO then
it's a nothingburger, right?

 

[prairiedog]

Never let the "customer" dictate a safety component, such as a fuse. That is for you, the responsible member to specify- unless you are not liable in your work per your contract. Even then, you have a professional ethics obligation to not make unsafe product. Congratulations on saving pennies there.

PCB fuse traces do not give coverage like a fuse. They do arc and burn, what did you expect, esp. if not done correctly. They are a last resort against calamity. Your SMPS going through certification will need something acting as a fuse sans flames instead of the IC blowing up or NTC burning up etc.

Mains can flicker, such as breaker reclosing (3 strikes) or when something fails in the transmission/distribution system. It's uncommon in North America for multiple dropouts seconds apart.

 

[dick_freebird]

In this context you would like to know per "instance" what the
worst case temp rise in the critical conductor is, for whatever
you consider the range of "dropout" behavior. Probably something
like the Joule energy held in the upstream output filters, or the
fraction that makes it to the conductor anyhow (Does second
DC-DC stage get cut out, or process it through?).

Then you can work with temp-rise-per-pulse and observed
decay time to determine whether or under what conditions,
a PCB track temp rise would exceed your reliability limits.
Like if you have 10 degC per dropout-restart cycle, and a
100ms thermal time constant then you'd say fuggedaboudit
for a once-a-minute dropout but probably look closer if it's
10Hz as now you could "staircase" up.

You might have some idea of PCB allowable DC and pulsed
current density vs width per plane as another cut at reliability.

How you'd instrument up that bit of wire to get time-domain
temperature data, I can't recommend exactly although I have
worked with some IR camera setups which had at least ability
to capture at standard frame rate, so you could manually pick
off points from a frame-series if you were that ambitious /
curious and get to time domain attribues for self-heating, self-
cooling.