How did I burn them?

[akis]

I have a breadboard with a single op-amp amplifying a sinewave. The breadboard gets fed +/-21V and I use an LM7818 and LM7918 to being this down to +/-18V. Next to the breadboard is a power-amp. As soon as I connect the output of the op-amp to the power amp, the LM7918 blows. I have so far blown half a dozen.

I am puzzled as to what has caused this. I even placed 10R resistors before the LM7918 but it still blew. They blew instantly, no heat, no smoke, nothing. Just ceased working, half of them then letting the input voltage through (very naughty that), and the others regulating to -3V (rather than -18V).

The LM7818 has not suffered the same fate.

So, what can you do to burn an LM7918? It can accept up to -35V. My highest rail was -21V. I know that you can burn it by applying at the output a voltage higher than the input. Maybe that is what happened through the op-amp . Some high frequency component from the power-amp (under construction and heavy harmonics at 4MHz) made it back through the op-amp's output, into the op-amp's negative supply pin, and into the negative rail.

Truth to be told, I had not placed capacitors anywhere around the LM7918 or around the op-amp. Because I had thought they should be more robust than that.

The only way for these oscillations to have killed them must have been through the op-amp, yet the op-amp (LM6172) survives and we have half a dozen dead LM7918s.

I am still puzzled as to what happened. Currently it is working with a 100nF cap at the output and a 100uF at the input. The power-amp does not oscillate anymore. Any of these may have cured it.

[w2aew]

It would be helpful if you could provide a diagram of ALL of the connections, including any instrumentation like a scope, signal generator, etc. that you might also be using.  Most common fault - connecting the ground lead of a scope probe to a non-grounded node in the circuit...

[wraper]

Maybe amplifier manages to short +/- rails somehow. 7818 overpowers 7918 current limit and manages to pull voltage on it's output above 0V what destroys IC. I have an idea, attach diode between 7918 output and GND so it clamps voltage if it goes above 0V on negative rail.

[T3sl4co1l]

Regulators require capacitors for dynamic stability, regardless of load.  It's not a matter of energy storage.

Tim

[Andreas]

Regulators require capacitors for dynamic stability, regardless of load.  It's not a matter of energy storage.

Especially negative regulators need even more capacyity.

If using 2 voltage rails you also need 2 diodes parallel to the outputs to prevent reverse polarity.
https://www.fairchildsemi.com/datasheets/LM/LM7815.pdf
See figure 16 in the datasheet.

With best regards

Andreas

[XFDDesign]

79xx series of regulators do not connect the same way as 78xx series regulators. With no Schematic, I'm not sure the assumption was that they were wired the same (vin, gnd, vout), but it's a common error.

[akis]

I did not wire it the wrong way or anything like that, even though it has different pin out to the LM78XX.

I have added two protection diodes, from ground to positive rail and from negative rail to ground. LM317/337 designs also show a protection diode crossing over the IC to protect against "back EMF" type of faults, but I have not added those. So far I have not blown any more LM7918s so fingers crossed.

[macboy]

you need those capacitors. They are not nearly as optional as the datasheets sometimes imply.

Also, the 79xx regualtors are built with NPN pass transistors since that is cheaper, but this makes them similar to a low-dropout regulator in design and operation. In other words, fickle, picky, and prone to oscillation. It is especially important with these to use input and output caps, and the output caps should generally not be ultra-low ESR types. A cheap small value mini electrolytic usually works fine since its ESR works do dampen the circuit. Pay attention to the datasheet applications information section, especially for 79xx.

[akis]

Thank you all for the comments. On the final board I will definitely place protection diodes and caps to be sure.

[Mr Simpleton]

This is really tricky when not seeing the schematics... Do you have DC-coupling between preamp and amp or a capacitor in the signal path??

Now for blowing 7918.. if there is a serious power consuption usually it gets pretty warm, and so if there is oscillations! One of the few times I managed to instant kill a 78xx/79xx is to feed high voltage in reverse, i.e. having current go the wrong way.  A diode from output to input will protect for this.  Just make sure you do have it the correct way not shunting the stabilizer when all is normal.