Need help for LM337 current limiting

[ElectronSurf]

How do I detect voltage drop on Node1 and pull down Node2 to limit the current without using an op-amp?

[Ian.M]

You don't!

With two OPAMPs introducing phase shift + whatever lag the current sensor adds in the current control loop, the result will be gross instability.

There are two approaches to current limiting that can be made to work:

• Limiting the current available to the LM337's Vin pin.  You can do that with another LM337 or LM317 + a power resistor, wired as a current source.  Unfortunately that requires a lot of headroom and the limiting current isn't easily adjustable.
• Pulling up the Adj pin to approx +1.3V when the current limit is reached.  This can be done with a NPN long tailed pair sensing the voltage across RS via two potential dividers (to allow adjustment of the threshold + sensing down to zero current), driving a potential divider feeding the base of a PNP transistor with its emitter at +1.5V and its collector to the Adj pin.

[ElectronSurf]

• Pulling up the Adj pin to approx +1.3V when the current limit is reached.  This can be done with a NPN long tailed pair sensing the voltage across RS via two potential dividers (to allow adjustment of the threshold + sensing down to zero current), driving a potential divider feeding the base of a PNP transistor with its emitter at +1.5V and its collector to the Adj pin.

That sounds more complicated than using an op amp.

I don't want a constant current, just want to define a maximum current for LM337, 200mA for example.

I'm looking for an approach like this:

[Ian.M]

In that case, if you can stand approx 0.7V drop across RS, you can reduce the long-tailed pair to a single NPN transistor.  Put its base-emitter junction across RS (with a resistor in series with the base so you don't blow the junction during transient surges), and use its collector to drive the potential divider driving the base of the PNP Adj pullup transistor.  Try RS of 3.3R, 2W  initially, and see what it limits at.  Fine adjustments can be made by tweaking the potential divider feeding the PNP adj pullup, but coarse adjustments will need a different RS.  N.B. as it relies on the NPN's Vbe to set the limit, it will decrease with temperature.

If you can guarantee the load will never draw more than 200mA at -1.5V output, you can eliminate the +1.5V bias rail for the pullup transistor and pull Adj up to ground instead.  That has the disadvantage that the output will be at approx -1.4V during current limiting, and if its shorted, only the LM337 internal limiting will be effective, and RS will burn up.

[ElectronSurf]

and use its collector to drive the potential divider driving the base of the PNP Adj pullup transistor.

The problem is the current at ADJ pin can be tens of mA (because it also is connected to OA1 output), I wanted to pull down the reference voltage so the current consumption can be in uA region.

Is that possible?

Edit: Did you mean something like this? Link to simulator

[Ian.M]

The problem with pulling down an external reference is there's too much phase shift in the OPAMPs + LM337 internal circuit.  Doing it that way tends to make a power oscillator, unless you severely limit the current loop bandwidth, which will result in a transient of the LM337's max output current, possibly over 3A, before it settles to your 200mA limit.

[ElectronSurf]

So it seems that having another LM337 in series as a current limiter is the only choice, can you think of any alternatives?

[Zero999]

So it seems that having another LM337 in series as a current limiter is the only choice, can you think of any alternatives?

That'll work, but the drop-out voltage (the minimum input-output differential required to maintain regulation) will be huge: double the usual 3V of the LM337 plus 1.25V, towards the current at which the limit kicks in.

Just use one op-amp, to pull the ADJ pin below 0V, when the current through the sense resistor is exceeded. There are a couple of ways to do that:

1) Connect the reference, with respect to the input voltage.

2) Use Howland topology.

I don't have time to post any schematics, at the moment. If you're interested, you'll have to wait a few days.

[David Hess]

If you want a variable current limit, then it takes two operational amplifiers.  One level shifts the ground reference control voltage to the negative supply, and then an error amplifier makes a direct comparison at the negative supply with the voltage across RS and overdrives the adjustment pin of the LM337 overriding OA1.  Diodes are used to combine the voltage and current control signals.  Only one diode is required if OA1 supplies the bias current to drive the LM337.

[Zero999]

If you want a variable current limit, then it takes two operational amplifiers.  One level shifts the ground reference control voltage to the negative supply, and then an error amplifier makes a direct comparison at the negative supply with the voltage across RS and overdrives the adjustment pin of the LM337 overriding OA1.  Diodes are used to combine the voltage and current control signals.  Only one diode is required if OA1 supplies the bias current to drive the LM337.

Only one op-amp is required, in Howland configuration. Of course some frequency compensation capacitor(s) may still be necessary, but I haven't included them in the schematic. If the op-amp is slow, such as the old '741, it's more likely you'll get away without it.

The maximum voltage across R3, thus the maximum current, is determined by V3 multiplied by the gain = R5/R6 = 2*100k/1M = 0.2. I = V/R 0.2/1 = 0.2A. R5 = R7 & R4 = R6

R1 draws a current from the output of U1, which will be subtracted from the load current, so in reality the current limit will be around 10mA, less the calculation above.

[David Hess]

If you want a variable current limit, then it takes two operational amplifiers.  One level shifts the ground reference control voltage to the negative supply, and then an error amplifier makes a direct comparison at the negative supply with the voltage across RS and overdrives the adjustment pin of the LM337 overriding OA1.  Diodes are used to combine the voltage and current control signals.  Only one diode is required if OA1 supplies the bias current to drive the LM337.

Only one op-amp is required, in Howland configuration.

I wonder why I never see designs doing it that way.  Common mode rejection is worse but I doubt that would matter in most cases.