Voltage limiter/clamp, the easy way 1

[MacGyverS2000]

I would like to limit/clamp a voltage to a specific value, let's say +15V. I don't need to regulate the output voltage to +15V, per se, only insure that it doesn't go over. Of course, minimal voltage loss is hoped for as I wish to generate as little heat as possible with this piece

Would a typical +15V regulator, such as the MC7815, work?

The only graph I've been able to find is for a 5V LDO regulator I've been using, which shows a nice, linear drop in output versus input until around 2V. I want to know if this behavior is typical of regulators (I'm sure it is, just checking), or specific to this package.


Of course, other inexpensive design ideas are welcome, I just figured this was a cheap way out...

 

[Skogsgurra]

I don't know how much output power you need and not how precisly the output signal needs to follow your "pre-limiter" voltage. An operational amplifier with limiting input(s) is the normal technique, but then you need a power supply for the amplifier so I guess that it cannot be used in your application.

Zeners cannot be used if you want to minimize losses. But what about a "switching zener"? All you need is an inductor and a transistor connected between output and ground. Control the transistor with a comparator with some hysterisis so that the transistor shorts the output to ground momentarily when the output gets too high. If the voltage gets higher, the duty cycle increases and keeps the voltage (mean value on the limit).

Some filtering may be needed if your load is sensitive to the PWM waveform. And, of course, the inductor and the transistor must be chosen to take care of your worst case input voltage.

 

[MacGyverS2000]

1.5-2A is probably the outer limit of what I'll need, closer to 1A is possibly more realistic. The closer I can get the input and output to follow, the happier I am... not because I truly need one to track the other, but simply because that means less dissipated power. This is going in a faily compact place (so parts count/size is an issue), and heat retention is a concern.

The 5V LDO I'm using for another project dissipates around 100mW dropping 15V down to 5V at 10mA, and you can barely feel it (SOT-89 package) with the back of your finger. I'm sure 500mW wouldn't be too bad in the designed-for space, maybe 1W with some testing.

 

[felixc]

Operating a regulator in this manner does not make it regulate anything. But you just want something that limits the voltage at the upper end, right?

Then all the electronics in the regulator is wasted, and you get a high drop. But the basic idea is good, so let's strip it down. If you want the smallest possible voltage drop, you can use the classical zener/NPN transistor configuration. Vin at the collector, load at the emitter, a resistor between the collector and base, and a zener between the base and the ground. (maybe a small cap in parallel) As long as Vin does not reach Vz, the output follows the input minus a VCEsat drop, above Vz, the output stays at Vz minus 0.65V. Your voltage is limited, and you will not blow any fuse if you exceed your Vin max. Like the regulator, you have to evaluate the power dissipation when Vin gets high, but at lower voltages you will get less dissipation than the regulator.

 

[MacGyverS2000]

felix, this is starting to sound like a better idea... it may take up more board space than a regulator package, but I can move the individually smaller parts around as necessary.

I wonder if I can redesign the basic circuit to use a MOSFET instead to take advantage of the low on resistance. That would certainly reduce wasted power.

Maybe some more info on my application would help. It's an automotive application, so I would probably lock Vz in at around 15V. Let's put an upper limit of 2A on this (realistically it will be just over 1A). I wish to protect the powered circuitry from the usual bad things that show up in the automotive world... load dumps, failing alternators giving way more than 15V, etc.

A true load dump is not very common, and I would expect a failing alternator to output less voltage rather than more the majority of the time... and I want the Tooth Fairy to give me a winning lottery ticket Bad stuff will happen, but when it does, I don't want my products giving up the ghost.

The circuit must be reasonably small and inexpensive (let's say under $2 in quantity and takes up no more than about 1" square of board space). Best case scenario is whatever circuit I use will essentially be self-healing (self-resetting fuses? No experience with them... any hints/tips?) Next best would be the circuit blows and the customer has to send it back for repair, but no ill effects shopuld come of any circuit further down the line (a firewall, of sorts)... I replace a component or two and ship it back... not ideal, but workable if the circuit was sturdy enough to handle typical day-to-day abuse.

 

[melone]

Why won't a transorb work? I know for a FACT that these devices are used in automotive applications SPECIFICALLY for load dump.

 

[MacGyverS2000]

No reason, other than the fact I currently don't know what they are or how they work. I'll go look now...

 

[felixc]

The tranzorbs for automotive use that can support load dumps are in the 5kw range. These are no small devices.
For an automotive application then you just need something that cuts off your circuit over a certain voltage, as everything else in the car will fry over 16 volts. Unless it is a piece of equipment that has to do with driver security.
Then all you need is a high-side switch. Look for the IPS511G from International rectifier for example. You need the zener detect circuit to turn the switch "off" for whatever exceeds your 15-16 volts. A tranzorb of a smaller power can be used after the switch, to absorb the peaks before the zener cuts the switch off.
Security standards may require that you fuse your circuit, and an inductor at the input will help for noise susceptibility. If your device is a user-installed gadget, I also suggest a series schottky (Vf at 0.3V) diode to prevent a reverse connection.
Felix

 

[MacGyverS2000]

Below was an initial thought, although it won't work ... (hoping the Code tags work in this forum)

Vin--+---------+ +------+
     |         |_|      |
     R1         |       |
     |          |       |
     +----R3----+       Rload
     |          |       |
     R2         Z       |
     |          |       |
Gnd--+----------+-------+

R1 = 10k
R2 = 4.7k
R3 = 4.7k
Z = 5V zener

R1 and R2 make up a voltage divider network. The initial thought process was for Vin=14.7V, potential between R1/R2 would be 4.7V, not enough to trip the zener, but enough to enable the FET and allow current to pass on to the load with minimal power dissipation in the FET. When Vin begins to swing higher, the zener conducts and turns off the FET.

Hey, it was an initial thought, even if it obviously wouldn't work... getting the creative juices flowing

However, what if the voltage divider (with adjusted values) was connected to an SCR. An overvoltage would trip the circuit, and it would stay tripped until the circuit was disconnected from power (or at the very least Vin was dropped significantly). Something like this...

Vin--+----+----+ +------+
     |    |    |_|      |
     R1   R3    |       |
     |    |     |       |
     |    +-----+       |
     |          |       |
     +----------S       Rload
     |          |       |
     R2         |       |
     |          |       |
Gnd--+----------+-------+

R1 and R2 again provide a voltage divider network that attaches to the gate of an SCR. R3 provides a small current path to charge the FET gate and enable it. If Vin rises too high, the SCR fires and drags the FET's gate to ground, essentially turning it off.

I've never worked with SCRs before, and I even had to look up their operation to make sure it was what I wanted... but I think this is a workable circuit.

What do you guys think?

 

[MacGyverS2000]

A bit of searching on the net gave me a slightly better idea... replace R1 with a Zener rated for the voltage I wish to trip at (in this case, 15V). If the voltage cranks above 15V, current flows through R2 creating a voltage high enough to fire the SCR.

 

[ScottyUK]

Hi MacGyver,

The thyristor circuit you are describing is a well-known last-ditch defense against overvoltage. Try searching Google for 'crowbar circuit' or similar. You *must* include a fuse in the supply which will blow when the thyristor fires.



------------------------------

If we learn from our mistakes,
I'm getting a great education!

 

[MacGyverS2000]

This isn't a typical crowbar circuit in that it doesn't connect Vin directly to ground. Instead, when the circuit fires, it DISconnects Vin from the rest of the circuit... the SCR disables the FET on overvoltages. In this case, there's no need for a fuse as no major currents will flow in the case of an overvoltage.

 

[ScottyUK]

Sorry, misinterpreted your diagram. Foot in mouth!

If you replace R1 with a zener, when the overvoltage occurs you will drive a lot of current through the gate of the thyristor down to the negative supply. Could you use a FET in place of the thyristor with a resistor in series with the gate and a clamp diode across the gate of the FET, or do you need a latching condition?

Isn't it a shame we can't post 'proper' diagrams unless we upload them to our own websites.





------------------------------

If we learn from our mistakes,
I'm getting a great education!

 

[MacGyverS2000]

Ah, I kinda missed that one when I replaced R1 with a zener... need to put a resistor back in between the R2/zener junction and the SCR to keep the current levels decent.

Or, as you said, just use a FET in place of the SCR. I wonder if I could get away with no resistor on the second FET's gate. True, there would be an inrush of current once the zener breaks down, but no more current would hit the FET's gate than is rushing through the zener. Using the FET-with-no-gate-resistor may be acceptable in this application. At least it would keep my component count down to 5, and since I'll already have FETs in the circuit, it's one less component to order.