7812 stuck on 0.7V

[kamtar]

Hello,
I have a problem which occurs like once a few days, when I'm switching my board on, sometimes happens that my 12V power rail is stuck on 0.7V

I have 12V and -12V rails regulated by 7812 and 7912 from 23V. -12V rail is okay, the problem happens only on the 12V. To solve it I just need to switch it off and on a few times.
Everything is properly decoupled and I have 100uF tantalum bulk caps.

The regulator isn't heating up and my PTC fuse isn't tripping.

When the 12V rail is running on 12V it's perfectly fine and clean.
I didn't have the chance to look with an oscilloscope when it's stuck on those 0.7V yet.

My question would be can you tell something from that peculiar value of 0.7V?
It's possible it's due to some inrush current?

I'm not sharing schematics at this moment, I'm mainly curious if you can deduce something specific from that behavior.

[Jwillis]

0.7V is a typical forward voltage of a diode.

[Mp3]

Is there a voltage divider somewhere before the regulator? Is it the proper resistance still?

[TimFox]

Re latch-up:
Latch-up occurs in ICs when any pin goes negative with respect to the most-negative rail on the device, which is the substrate, by at least a P-N voltage drop.  (Similarly for positive excess.)  If there is a path between the negative supply (7912 output) and the 7812 output, and the negative power supply comes on before the positive supply, the 7812 output can go negative and cause the latch-up.  Before the 7812 powers up, the output impedance is high, and it does not take much current to take the pin negative.  The usual cure for this type of latch-up is to put a Schottky diode (with less voltage drop) across the relevant pins (reverse biased), to prevent the parasitic SCR from turning on.  In this case, put the diode from the 7812 output to ground.  Latch-up may destroy a device if the current supplying the input is sufficient to exceed its power rating, but if there be insufficient current available, the device may survive until the power is cycled.

[kamtar]

Thanks, I will try out Schottky on the output.

[David Hess]

If the schottky diode across the output does not solve the problem, then there is another possibility.

When power is first applied, if the load on the regulated positive supply is great enough at low output voltage, like with current drawn to a negative supply, then the regulator may latch off due to foldback current limiting and a high input to output voltage difference.

This really should not be happening with an input supply voltage of only 23 volts but it might if the regulator was a counterfeit.