How to improve or make this old design more robust?

[XaviPacheco]

The circuit attached is a power supply based on zener diodes and a LM317 regulator. It's a real application circuit built on a PCB. This is how the circuit is expected to work:

a) The ground is defined as the point between the two output capacitors C3 and C4, then D11 establishes the most negative voltage as -13V.

b) D10 then puts the ADJ pin of the regulator at +12V, which means that the regulator output is 12 + 1.25 = 13.25V. This feeds another low power circuit I'm not showing.

c) D7 regulates the gate of M1 at 43V above the negative rail, for an absolute voltage of +30V relative to ground. Note that M1 is wired as a current limiter.

d) If I measure +27V at the input of the regulator, then there's a difference of 3V between that point and the gate of M1. Most of this is the threshold voltage of M1, but there will also be a voltage drop across R2 + R3 that depends on the load current.

e) D8 limits that voltage to 10V at most, which means that the maximum current that can flow through R2 + R3 is about 7/515 = 13.6 mA — at which point, the input voltage to the regulator will have dropped to about 20 V. This range of current is not enough to activate the coil L4.

d) When the SCR is triggered, then the current is limited only by R2, which means that about 1.37 A will flow through the coil. When the coil activates, it shutdowns the power line and the whole circuit is turned off.

I simulated the circuit and that's how it works.

Additional details:

1. The input voltage ranges from 100 to 480 VAC. I'm simulating with 480 VAC line-to-line so it's the worst operating point. But the circuit has to handle it.
2. The coil has a resistance about 290 ohms
3. The capacitor after the rectifier is 22nF/1000V. This mean the output ripple is high. This is the capacitor used.
4. D13, D14, and D15 are three TVS diodes in series. I think the designer chose this combination to increase high current surge protection. Is there a way to reduce the number of diodes?
5. The three line inductors L1, L2, L3  datasheet is this: RL-1294-680 . I want to replace them with these as they seem more robust.

Main problem I'm trying to address: The fuse blows up from time to time.

The fuse being used is a time-delay (slow blow) fuse rated at 125 mA, 250V. I want to increase this rating because it seems like the actual fuse doesn't handle "normal transients" maybe?

What other general improvements would you make? Suggestions? Change the capacitor?, Add more capacitance? The varistor here is protecting against DC transients as it's placed after the rectifier.

In my personal opinion, this is not a good design.

[XaviPacheco]

D12 and D9 are the same part: 1N4007.
The TVS diodes are 1.5KE300A

[magic]

Not sure why you started a new thread about this PSU.

Anyway, maybe something is triggering the SCR? Otherwise, current limit through M1 was less than 20mA IIRC. You should be able to put a short from D9 to -13V and the fuse shouldn't blow. (Perhaps try with resistors first instead of a direct short if you want to test it and avoid more smoke than necessary).

Maybe M1 is shorted and current limiting doesn't work.

[XaviPacheco]

Maybe M1 is shorted and current limiting doesn't work.

All the components remain healthy except for the fuse. I don't think the fuse is being damaged by a short circuit.

[duak]

What is the gate circuit of the SCR?  If there is not a low enough impedance from the gate to the cathode, the SCR can be triggered by leakage current.  The current can come from the external circuit or from the SCR itself at high temperatures.  SCRs are also sensitive to rapid increases in anode voltage relative to the cathode.  This is shown in the data sheet as the dV/dt spec.

Maybe adding a 10n to 100n cap in parallel with the SCR?  Probably need to add a 10R resistor in series with the cap to limit current when the SCR is triggered.

I wonder if under some conditions, the MOSFET M1 could be oscillating.