Best design option for over voltage monitoring?

[rfengg]

Need your advice on implementation of  an over voltage monitoring circuit  and cutoff on a LM317 based power supply.

The LM317 drives several expensive ($45 per piece) MMIC's which have a normal supply voltage of 4V and  a max voltage rating on 5V. The current requirement for the MMIC is  around 0.5A.
Was wondering what would be the best option to pursue  in order to make sure that the supply voltage does not exceed 4.5V in case the LM317 gets shorted or the trimmer resistor goes open which causes the LM317 to swing to the rail voltage?
Should it be a dedicated voltage supervisor IC or just a zener in combination with a BJT  that switches off a NC relay at the output of the power supply?
I was envisaging an NC relay at the output leads of the power supply that switches to open when an over voltage is detected.
Would it be better to use a relay as the output "switch" or a MOSFET?
Would love to hear your ideas......

thanks

[capt bullshot]

Driving an NC relay to open the supply might fail in case it's needed, and opening the contact might take too long to protect the circuit.

For this kind of protection, one usually employs a crowbar circuit: https://en.wikipedia.org/wiki/Crowbar_(circuit)
That is a voltage threshold detector driving an element that shorts out the supply voltage of concern.

A crowbar implementation can be as simple as an SCR, a resistor from its gate to ground and a suitable zener diode driving the gate from the supply - one might even find a very simple solution called "TRISIL ( https:// www.st.com/resource/en/application_note/cd00004007.pdf ). Maybe even a simple TVS (or "Transil") is sufficient - one would have to check their datasheet against the capabilities of the power source in case of a failure.

[bdunham7]

Need your advice on implementation of  an over voltage monitoring circuit  and cutoff on a LM317 based power supply.

The LM317 drives several expensive ($45 per piece) MMIC's which have a normal supply voltage of 4V and  a max voltage rating on 5V. The current requirement for the MMIC is  around 0.5A.
Was wondering what would be the best option to pursue  in order to make sure that the supply voltage does not exceed 4.5V in case the LM317 gets shorted or the trimmer resistor goes open which causes the LM317 to swing to the rail voltage?
Should it be a dedicated voltage supervisor IC or just a zener in combination with a BJT  that switches off a NC relay at the output of the power supply?
I was envisaging an NC relay at the output leads of the power supply that switches to open when an over voltage is detected.
Would it be better to use a relay as the output "switch" or a MOSFET?
Would love to hear your ideas......

Assuming you are setting the LM317 at exactly 4 volts, I would suggest a simple, mostly passive approach.  First, get the divider resistor set up so that you are close to the 4 volts using only fixed resistors, then if you want a more precise setting, add a higher resistance trimmer in parallel to one of the resistors to make a fine adjustment only.  Depending on which resistor you trim this way, you can design the system to fail either on the high or low side of the setting if the trimmer goes open, but it probably doesn't matter much if the difference is small.

The LM317 is pretty unlikely to otherwise fail in a way to overvoltage your part and I don't think I'd go to much more trouble to protect a $45 part.  However, if you do, I'd use simple, passive methods and not complex, active circuits.  A simple 1N5336B (selected 1N5337B might be better) zener across the output along with a ~1A fuse or an appropriate power resistor on the input would do the trick.  The power resistor, if selected appropriately, would both let your LM317 run cooler and allow the circuit to continue to operate properly even if the regulator fails shorted.  I'd need to know the input power supply voltage/current and so forth to determine exact numbers.

[rfengg]

Thanks bdunham7.

The input to the LM317 is around 12 volts and the output set with a trimmer is 4V exactly with a load current of 1A.
I forgot to mention that there are 10 MMIC's ($450 at stake) all driven by this supply, so that's why I am a bit concerned in case the voltage exceeds 5V.

Regarding the zener implementation, are you advocating the use of a zener at the output of the supply in parallel with the load and a 1A fuse just directly at the output of the 317, such that the zener crowbars the fuse if the voltage of the zener is exceeded?

You mentioned that "The power resistor, if selected appropriately, would both let your LM317 run cooler and allow the circuit to continue to operate properly even if the regulator fails shorted. "
I agree that a resistor would allow the 317 to run cooler , but do not understand why the circuit would continue to operate properly even if the regulator fails shorted?
How can I ensure that the voltage output does not exceed 5V if the 317 indeed gets shorted?

[bdunham7]

The input to the LM317 is around 12 volts and the output set with a trimmer is 4V exactly with a load current of 1A.
I forgot to mention that there are 10 MMIC's ($450 at stake) all driven by this supply, so that's why I am a bit concerned in case the voltage exceeds 5V.

I was just giving you an idea, and of course every detail affects the design.  So is the load 0.5A for 10 MMICs in parallel?  And does that vary up and down?  What is the absolute maximum peak load?

Regarding the zener implementation, are you advocating the use of a zener at the output of the supply in parallel with the load and a 1A fuse just directly at the output of the 317, such that the zener crowbars the fuse if the voltage of the zener is exceeded?

Fuse on the input side of the LM317, although it would work either way if you went the fuse route.  And a zener is not quite the same as a crowbar.

You mentioned that "The power resistor, if selected appropriately, would both let your LM317 run cooler and allow the circuit to continue to operate properly even if the regulator fails shorted. "
I agree that a resistor would allow the 317 to run cooler , but do not understand why the circuit would continue to operate properly even if the regulator fails shorted?
How can I ensure that the voltage output does not exceed 5V if the 317 indeed gets shorted?

By selecting the appropriate resistor and zener.  And that depends on what you need for a maximum current.  If you can live with a maximum current of 0.5A, then you would select a 10 ohm power resistor and one of the 5W zeners I mentioned (there's a small issue here selecting these and I haven't figured out the datasheet yet).  At 0.5A, there would be a 5 volt drop across the resistor, leaving 7 volts on the input of the LM317 which is adequate for a 4V output, but has a low power dissipation in the regulator of 1.5W.  If the regulator goes rogue, then the zener will step in and start conducting at ~4.5V, which will result in the other 7.5 volts being dropped across the resistor and the total current to be 750mA.  If the MMICs draw 0.5A, the current through the zener is 250mA and the power is ~1.2W.  If the MMICs draw no current, then the zener current is 750mW and the power is ~3.6W.  Toasty, but not fried if the circuit board is designed well. 

If you need higher currents, you'll need to adjust those values and you might need to fuse it instead.  I don't know how big a deal having the whole thing shut off would be.  Or, perhaps use two 0.5A regulator circuits.  But really, I think if you use fixed resistors with a large-value parallel trimmer, you have eliminated the most likely cause of an LM317 going rogue and beyond that they are pretty reliable and will shut down rather than short out.

[pqass]

 CuriousMarc just did a video about a crowbar.  He implements an hp design that they use on their 9825 computers; just 4 parts + fuse.
See 8:40 for the schematic here:

[Doctorandus_P]

A crowbar circuit would also be my first thought, and with a TL431 it can be adjusted quite accurately.

A crowbar does need a fuse or some other method to stop the current flow.

[John B]

Yep I've used a TL431 + TRIAC crowbar circuit. Pretty easy to get +/- 10mV or tighter trigger voltage, also you can RC filter the trigger input depending on how much noise there is on the input and how strict / fast acting you need the clamping to be.

[ledtester]

If you're interested in what's used in cell phones, laptops, chargers, etc. do a search for "over voltage load/protection switch".

For instance the KTS1662EUW:

https://www.mouser.com/ProductDetail/Kinetic-Technologies/KTS1662EUW-TR?qs=lc2O%252BfHJPVZkAuNkM71H7w%3D%3D

This particular device has an adjustable overvoltage threshold which can go down to 4V, an integrated MOSFET and -- most importantly -- it's in stock.

[David Hess]

Should it be a dedicated voltage supervisor IC or just a zener in combination with a BJT  that switches off a NC relay at the output of the power supply?
I was envisaging an NC relay at the output leads of the power supply that switches to open when an over voltage is detected.

Relays are awfully slow for an application protecting semiconductors.  Usually an SCR or TRIAC controlled by some sort of trigger circuit is used as a crowbar to short the output with the current limited by a fuse.

A TRIAC instead of SCR is required in John B's suggestion because an SCR can only be triggered from the negative side and the TL431 can only trigger from the positive side because it is referenced from the negative side.  An LM385 could be used with an SCR, or a transistor could be added to reverse the trigger signal to use an SCR with an TL431.

Would it be better to use a relay as the output "switch" or a MOSFET?

Integrated load protection switches commonly use a MOSFET in series with the load.  I have seen one design which used an SCR or TRIAC to latch-on and shut off the MOSFET.  If the MOSFET was driven from a comparator, then it could potentially switch off in 10s of nanoseconds.

[rfengg]

Thanks David.

You mentioned "A TRIAC instead of SCR is required in John B's suggestion because an SCR can only be triggered from the negative side and the TL431 can only trigger from the positive side because it is referenced from the negative side.  An LM385 could be used with an SCR, or a transistor could be added to reverse the trigger signal to use an SCR with an TL431."

I am not sure I understood this....if I put the SCR across the rail voltage, with the anode on the positive side and the cathode on the negative side, then I should be able to trigger the SCR with a (+) voltage greater than the cathode terminal, right?

I am inclined to go with a p channel mosfet in series with the load which driven by a comparator so that the mosfet is switched off when the voltage exceeds the threhold set by a zener/TL431.

[David Hess]

You mentioned "A TRIAC instead of SCR is required in John B's suggestion because an SCR can only be triggered from the negative side and the TL431 can only trigger from the positive side because it is referenced from the negative side.  An LM385 could be used with an SCR, or a transistor could be added to reverse the trigger signal to use an SCR with an TL431."

I am not sure I understood this....if I put the SCR across the rail voltage, with the anode on the positive side and the cathode on the negative side, then I should be able to trigger the SCR with a (+) voltage greater than the cathode terminal, right?

That is right.  The problem is that if you do this with the ubiquitous TL431 as the reference/comparator/trigger, then the TL431 cannot directly trigger the SCR because its reference is on the negative side and it can only sink current to turn something on.  An LM385 has its reference on the positive side, so it sources a current which could trigger an SCR on the negative side.  A TRIAC solves this problem because it can be triggered either way.

Update: Why not always use a TRIAC?  SCRs are more powerful in the same package.

I am inclined to go with a p channel mosfet in series with the load which driven by a comparator so that the mosfet is switched off when the voltage exceeds the threhold set by a zener/TL431.

I think that is a fine solution, and it is even faster than using an SCR or TRIAC crowbar circuit.

[Doctorandus_P]

If you use a MOS fet, you have to add something to keep it off after it has been "triggered".

With a quad comparator, you can make an under & overvoltage lockout, use a third comparator as a flipflop to remember it has been triggered, and the forth mabye to drive a blinking led or buzzer to indicate a fault.

[cdev]

The time delay may be a problem. The time to open a relay is going to be short, but possibly significant. You could measure it fairly easily.

Need your advice on implementation of  an over voltage monitoring circuit  and cutoff on a LM317 based power supply.

The LM317 drives several expensive ($45 per piece) MMIC's which have a normal supply voltage of 4V and  a max voltage rating on 5V. The current requirement for the MMIC is  around 0.5A.
Was wondering what would be the best option to pursue  in order to make sure that the supply voltage does not exceed 4.5V in case the LM317 gets shorted or the trimmer resistor goes open which causes the LM317 to swing to the rail voltage?
Should it be a dedicated voltage supervisor IC or just a zener in combination with a BJT  that switches off a NC relay at the output of the power supply?
I was envisaging an NC relay at the output leads of the power supply that switches to open when an over voltage is detected.
Would it be better to use a relay as the output "switch" or a MOSFET?
Would love to hear your ideas......

thanks

[Terry Bites]

see LM74502 from TI