Current limiter

[electronics man]

I have a current limiter circuit which consists of a diferential opamp across a current shunt resistor, the output of that opamp goes to a comparator whith a 10 turn pot on the other input, so when the comparator senses that both inputs are more than or equal to one another is turns on and switches on a MOSFET to pull the adjust of an lm317 low to tun the output off and stopp the current, the problem is it oscillates as I expected. This is because as the output is turned off the current drops so the MOSFET switches off, this then means the current increase so the circuit triggers again, my question is how can I make the circuit stay low until the over current stops.

Daniel

[tszaboo]

step 1: post schematic.

[electronics man]

sorry its not that great it is just a section from a larger schematic. the output of that opamp just connects to the mosfet

[dannyf]

You don't want it to be a limiter; you want it to be a latching switch.

[electronics man]

yeah i suppose but not one that will stay on for too long so not a relay

[miceuz]

I've seen an excellent post on the topic somewhere in electronic.stackexchange.com. Comparator is your problem. I'd suggest putting another opamp instead of comparator and setting a finite gain for that opamp.

[T3sl4co1l]

Sounds like a bad recipe for a good oscillator.  (What happens when it exceeds current?  Turns off.  So current is zero?  Let's turn it on again!  Repeat at probably 100kHz or so, or until circuits fry...)

Tim

[electronics man]

When dave made this circuit for his usupply he put a unity gain opamp between the differential amp and the high side of the current shunt.

[codeboy2k]

Some DC-DC controllers go into a hiccup mode during over current shutdown to avoid this very oscillation problem; granted its still an oscillation, but its very slow

Basically, the output switch controller is gated via the hiccup controller. When the hiccup controller turns on the gate and allows the output switch to turn on, and another over-current event is detected , it resets the hiccup timer and the gate turns off. More sophisticated controllers even increase the timeout each time to some max, and reset the timeout to zero after its working again (the cause of the overload is removed).

[electronics man]

could someone sugest a diferent soluution mabe a single chip would be nice but it must be variable whith an external pot.

[dannyf]

A potential solution was suggested to you earlier: use a latching switch.

A schmitt trigger that can never be reset for example.

[ConKbot]

Sounds like a bad recipe for a good oscillator.  (What happens when it exceeds current?  Turns off.  So current is zero?  Let's turn it on again!  Repeat at probably 100kHz or so, or until circuits fry...)

Tim

Or put an inductor in series and make a buck converter that only turns on above the current limit, and make the oscillation useful

[electronics man]

Sounds like a bad recipe for a good oscillator.  (What happens when it exceeds current?  Turns off.  So current is zero?  Let's turn it on again!  Repeat at probably 100kHz or so, or until circuits fry...)

Tim

Or put an inductor in series and make a buck converter that only turns on above the current limit, and make the oscillation useful

what value would i need.

[matkar]

If you're searching for an integrated solution you could find something useful browsing web pages from different chip producers.
Depending on your current / voltage needs you might find something on TI site:
http://www.ti.com/lsds/ti/power-management/power-protection-distribution-monitoring-products.page
These two might do the trick:
TPS2421-1
TPS2421-2
Look at Maxim site for suitable solutions as well: http://www.maximintegrated.com/products/power/power-switching/
Try other producers too.

[dzseki]

Once I inherited a DIY amplifier and there were some pcbs in series with every voltage lines, as it turned out they were electronic current limiters, I reverse engineered the schematics then, now I loked it up for you, I think it is a very clever circuit, it can be used in both polarity just keep in mind the current flow direction.

[electronics man]

Once I inherited a DIY amplifier and there were some pcbs in series with every voltage lines, as it turned out they were electronic current limiters, I reverse engineered the schematics then, now I loked it up for you, I think it is a very clever circuit, it can be used in both polarity just keep in mind the current flow direction.

thanks for that very helpfull.

[dannyf]

I reverse engineered the schematics then

BD243/BD740 form a PNP; Together with BD749, they form a SCR. BC237 is normally not conducting (that 250p capacitor is very important, ). As such, the SCR is conducting hard.

As the voltage drop over the scr goes up, the voltage across BC237's b-e junction goes up. Once BC237 starts to conduct, it diverts current away from the "SCR", thus turning it off. Once the scr is turned off, all voltage is drop'd on the whole device, further locking it up.

aka, this is a latching switch.

Because of that, it can be implemented a variety of ways, using SCR, mosfet, solid state relays, etc.

[dzseki]

I reverse engineered the schematics then

BD243/BD740 form a PNP; Together with BD749, they form a SCR. BC237 is normally not conducting (that 250p capacitor is very important, ). As such, the SCR is conducting hard.

As the voltage drop over the scr goes up, the voltage across BC237's b-e junction goes up. Once BC237 starts to conduct, it diverts current away from the "SCR", thus turning it off. Once the scr is turned off, all voltage is drop'd on the whole device, further locking it up.

aka, this is a latching switch.

Because of that, it can be implemented a variety of ways, using SCR, mosfet, solid state relays, etc.

Thanks for the clear explanation, I also figured out that the little capacitor has a great importance here, but never knew exactly why is this working at all -that's why I drawed the schematics then... And yes, it is latching. Once it goes out you have to remove the power completely to switch on again (I guess shoring in and out would do the trick as well, but this approach isn't very wise if you use it as a fuse)

[ohmineer]

If there aren't special requirements, the simplest current limiter that I recall is based on 2 PNP transistors (or 1 PNP + 1 PMOS).
PMOS (Or right PNP) is used as pass transistor.
Current demand can increase until top left PNP BE junction is forward biased.
Therefore, VBE voltage drop over R1 resistor (Sense resistor) set maximum allowed current.

[SeanB]

True that limits current, but the pass element will get mighty hot when it is limiting and dropping most of the supply voltage. the earlier one at least latches into a low dissipation mode when it trips out until the supply is interrupted.

[electronics man]

that is probably the best one i have seen so far, the pass element could be heatsinked, as they are on my BOM already and there will be a fan to keep everything nice and cool.

[ohmineer]

Beware! I have just realised that there is a mistake in the screenshot I posted. Iout is not Vcc/R1. The correct expression would be Iout=Vbe/R1.

The drawback of this approach is the change on Vbe with ambient temperature. In any case, I think that there are solutions to compensate for this, although sometimes it does not worth it.

[electronic_eel]

If you don't want oscillation, you want
a) latching
    a1) forever until poweroff
    a2) automatic retry after a few milliseconds to seconds
b) foldback

The latching ones completely shut the output off when tripped. The foldback ones always let a little bit of current through. If the current drawn drops below a threshold, the limiter is reset and the full current is available again.

Here is a discussion with some schematics:
http://www.electronics-related.com/sci.electronics.design/thread/196166/electronic-fuse.php

[dannyf]

VBE voltage drop over R1 resistor (Sense resistor) set maximum allowed current.

Replace Q1 with some leds and they act as over-current indicators.