regulator protection

[r00n]

My next project is this one.



I am wondering whether to protect the 7815 and 7915 regulators by putting protection diodes around them.  I recall reading somewhere that the regs don't like to have the forward voltage > input voltage. Could I put place 1n4007 diodes from the outputs or the 7815 and 7915 back to the input to protect when and if that occurs? What would that look like? Would the 7915 go from input to output or vice versa?

[Whales]

> I recall reading somewhere that the regs don't like to have the forward voltage > input voltage

What do you mean by forward voltage?

Input voltage = voltage at the input terminal of the reg (compared to ground terminal)
Output voltage = voltage at the output terminal of the reg (compared to ground terminal)
Voltage drop = input voltage - output voltage

All 3 of those numbers will have max specs in the datasheet.  Additionally: you have to calculate power loss in the regulator (voltage drop * current) to make sure your regulator won't overheat.

There is also a thing called "dropout voltage" which is a minimum voltage drop.  Below this the regular won't work correctly.

[rdl]

Protection diodes are a good idea. They hurt nothing and cost very little. They are often left off for some reason. And yes, the diode on the 7915 will be opposite the one on the 7815. The cathode will connect to the output.

[Whales]

Protection diodes are a good idea. They hurt nothing and cost very little. They are often left off for some reason. And yes, the diode on the 7915 will be opposite the one on the 7815. The cathode will connect to the output.

What do you mean by protection diodes?   Reverse voltage shorting?  Overvoltage zener protection?  ESD protection?  Diode in reverse-parallel to regulator (going from output to input) for when voltage is applied (externally) to the output but not the input?

[rdl]

They're in most data sheets. They go across the regulator to prevent excess charge on the output from going backwards through the regulator. In a controlled situation, maybe they're not necessary. But in something like a bench supply where who knows what might be connected, it's not a bad idea to have them

[wraper]

You should put diodes between +/-15V rails and GND if there are circuits powered between both power rails (like opamps). Because otherwise during powering, output voltage may become negative for 7815 or positive for 7915 (depending on which side pulls stronger) and vreg or powered IC may latch-up because of that. As of diodes across vreg in-out, generally they are not needed unless there is large capacitance connected to the output that discharges slower than caps before vregs.

[golden_labels]

I believe r00n means V_OUT > V_IN. That’s the situation earlier explained in a thread on LM317 protection diode.

[Whales]

You should put diodes between +/-15V rails and GND if there are circuits powered between both power rails (like opamps). Because otherwise during powering, output voltage may become negative for 7815 or positive for 7915 (depending on which side pulls stronger) and vreg or powered IC may latch-up because of that.

What process would cause the power rail caps to charge negative to enough of a voltage to cause latchup?  Or is this a short transient case? (very fast MHz+ ESD, too fast for the electrolytic caps?)

As of diodes across vreg in-out, generally they are not needed unless there is large capacitance connected to the output that discharges slower than caps before vregs.

Also circuits that can potentially be backpowered, but I don't know the OP's use case.

[wraper]

You should put diodes between +/-15V rails and GND if there are circuits powered between both power rails (like opamps). Because otherwise during powering, output voltage may become negative for 7815 or positive for 7915 (depending on which side pulls stronger) and vreg or powered IC may latch-up because of that.

What process would cause the power rail caps to charge negative to enough of a voltage to cause latchup?  Or is this a short transient case? (very fast MHz+ ESD, too fast for the electrolytic caps?)

When power is applied, output voltages/currents can rise at different speeds. And depending on various factors one vreg can overpower another for a short moment and pull its output to opposite polarity (beyond GND). Of course if positive and negative supplies do not actually pass any current to each other, there is no need for such protection.

[floobydust]

I recommend using four protection diodes. Two for reverse-polarity, and two for backfeed from output to input. Just to make it reliable.
Pic is taken from Elliott Sound Products Power Supply for Preamplifiers

Also, single point (star) grounding at least for the main filter capacitors keeps hum and ripple low.

[Konkedout]

You should put diodes between +/-15V rails and GND if there are circuits powered between both power rails (like opamps). Because otherwise during powering, output voltage may become negative for 7815 or positive for 7915 (depending on which side pulls stronger) and vreg or powered IC may latch-up because of that. As of diodes across vreg in-out, generally they are not needed unless there is large capacitance connected to the output that discharges slower than caps before vregs.

YES! If you have a load between +15 to -15 then you definitely want to clamp the +15 from going below ground.   If the output of MANY positive regulators goes below ground, the IC will go brain dead until that voltage is removed.  I have been bitten by this more than once.   This is because most ICs use junction isolation which works only when the substrate is negative relative to the circuitry.  USE A SCHOTTKY DIODE (such as 30V or 40V rating; maybe a 1A type) because it has lower Vf than the junction isolation in the 15 regulator. 

[ArdWar]

What process would cause the power rail caps to charge negative to enough of a voltage to cause latchup?  Or is this a short transient case? (very fast MHz+ ESD, too fast for the electrolytic caps?)

This generally happen during power up. Regulators misbehaving due to "wrong" sequencing is a rather well known problem in multiple rail system where each rails are not independent of each others.

[Whales]

Thankyou all, I was not aware of that one.  So (for example) the -15V feeds through a part (like an opamp) and into the +15V rail whilst it's still starting up, dragging the nascent +15V rail slightly below 0V.  Huh.

More diodes to the rescue!

[r00n]

Thanks for sharing your knowledge everyone. Most kind.