If you want an even lower dropout voltage than the above circuits provide you can get it, albeit at some extra complexity and cost. The following 5V, 50mA limiter has a dropout voltage of only 50 mV and is also stable with temperature since it doesn't rely on a BJT's Vbe to set the current.

View attachment 201138

Voltage divider R1/R2 determines the setpoint for the voltage across current sense resistor R5. Op amp U1 provides the control function, and MUST be a RRIO op amp. C1 and R4 discourage instability and oscillations.

U1 and Q1 are a lot costlier than a pair of jellybean BJTs; but if you absolutely must have the lowest possible dropout voltage the above circuit is one way to get it.

That's the opposite direction of what I was searching for with my cheap, simple experiment at the moment, but looks like a good circuit to keep in mind for other uses in the future. I'll be sure to save it for later.

There only thing I see as a potential weakness, depending on how picky you want to be, is that the current limit is proportionally dependent on supply voltage, since the reference comes from a voltage divider.

If you wanted a current limit that was more immune to supply voltage variation or inaccuracy, presumably you could set something up with a Zener, TL431, etc. to provide a more stable reference.

 

After staring at the above, I concluded that the next time I need to design a current limiting circuit the first I would consider is an adjustable (or not) constant current source with enough allowance for a certain variation of the input voltage.

I'm curious what you have in mind when you mention constant current sources. Are you referring to a different class of sub circuits, or to ICs?

It seems like most of the IC current source solutions I've seen have pretty high dropout voltage, although if there are good ones I'm overlooking, I'd love to know more about them.

 

That's the opposite direction of what I was searching for with my cheap, simple experiment at the moment, but looks like a good circuit to keep in mind for other uses in the future. I'll be sure to save it for later.

I understand; I posted that just to illustrate one method of achieving a rock-bottom, minimum dropout voltage if that were the overriding consideration. For your purpose it's not a practical solution.

There only thing I see as a potential weakness, depending on how picky you want to be, is that the current limit is proportionally dependent on supply voltage, since the reference comes from a voltage divider.

If you wanted a current limit that was more immune to supply voltage variation or inaccuracy, presumably you could set something up with a Zener, TL431, etc. to provide a more stable reference.

That's exactly what you'd do if you wanted to eliminate the dependence on supply voltage. My choice would be to use a TL431 to get a reference point 2.5 volts below Vcc, then divide down from that.

It seems like most of the IC current source solutions I've seen have pretty high dropout voltage, although if there are good ones I'm overlooking, I'd love to know more about them.

I don't make much of a distinction between constant current source circuits and current limiting circuits; to me, they're pretty much the same thing except possibly what characteristics they're optimized for, such as precision control over current, programmability via an external control voltage, etc.

 

I'm curious what you have in mind when you mention constant current sources. Are you referring to a different class of sub circuits, or to ICs?

It seems like most of the IC current source solutions I've seen have pretty high dropout voltage, although if there are good ones I'm overlooking, I'd love to know more about them.

@ebeowulf17

It seems that I am derailing your thread after all. Sorry for that.

To your question: when posting my last comment, in fact, I had in mind not your interest on low dropout (my bad) but many different current limiting circuits I've built in the past plus a series of constant current sources where I could adjust the voltage controlling the current. After all, a current-limited PSU may be used as a constant current source if required.

I don't make much of a distinction between constant current source circuits and current limiting circuits; to me, they're pretty much the same thing except possibly what characteristics they're optimized for, such as precision control over current, programmability via an external control voltage, etc.

There only thing I see as a potential weakness, depending on how picky you want to be, is that the current limit is proportionally dependent on supply voltage, since the reference comes from a voltage divider

If you wanted a current limit that was more immune to supply voltage variation or inaccuracy, presumably you could set something up with a Zener, TL431, etc. to provide a more stable reference.

In all my designs I've always used some voltage reference. The LM 236 (2,5V) was a recurrent component, many times.

 

It seems that I am derailing your thread after all. Sorry for that.

No need to apologize! Although I posted this thread specifically about one circuit in a certain situation, I'm interested in all of these variations too.

 

The difference being that in post #11 R_Sense is missing.

'R.sense' was removed in my (#11) post because it wasn't necessary. It was actually more controllable adjusting the resistor at the bottom left. In truth, the circuit works by oscillation between the two transistors to maintain a 'stable' limited current output. I didn't come up with the circuit either, I just through I'd post it as it works pretty well.

 

Here's an interesting and fairly simple current limit circuit with a very low drop that uses a variation of a current mirror to sense the current across the shunt resistor.
The circuit has the interesting property that the Vbe difference between the two transistors is about 6mV times the operating current difference between the two.
So for a factor of 10 difference (such as R3 = 100kΩ and R4+U1 = 10K) the voltage difference would be 60mV.
Thus the current limit would be approximately 60Vm / R1 = 1.2A.
Note that it's the current ratio that determines the sense voltage, so this voltage is relatively insensitive to the supply voltage.

The advantages of this circuit are that it has a low sense voltage so the value of the shunt sense resistor R1 can be small, minimizing its power dissipation and voltage drop, and the current limit can be easily adjusted by pot U1.
Of course if adjustability is not needed, a fixed resistor can be used in place of the pot.
For best performance (low offset and good stability), the inexpensive DMMT3906W-7-F Dual Matched Transistors (or similar) can be used for the current mirror transistors.

 

Here's an interesting and fairly simple current limit circuit with a very low drop that uses a variation of a current mirror to sense the current across the shunt resistor.
The circuit has the interesting property that the Vbe difference between the two transistors is about 6mV times the operating current difference between the two.
So for a factor of 10 difference (such as R3 = 100kΩ and R4+U1 = 10K) the voltage difference would be 60mV.
Thus the current limit would be approximately 60Vm / R1 = 1.2A.
Note that it's the current ratio that determines the sense voltage, so this voltage is relatively insensitive to the supply voltage.

The advantages of this circuit are that it has a low sense voltage so the value of the shunt sense resistor R1 can be small, minimizing its power dissipation and voltage drop,and the current limit can be easily adjusted by pot U1.
Of course if adjustability is not needed, a fixed resistor can be used in place of the pot.
For best performance and stability, the inexpensive DMMT3906W-7-F Dual Transistor can be used for the current mirror transistors.

View attachment 201155

That's really interesting and cool. Thanks for sharing!

I must confess that it's a bit of a stretch for me. If I stare at it long enough, I can start to understand it, and as soon as I get distracted, it's gone again! Guess I'll have to keep staring some more until it really sticks.

 

I must confess that it's a bit of a stretch for me. If I stare at it long enough, I can start to understand it, and as soon as I get distracted, it's gone again!

With no output current through R1, the lower collector resistance of Q1 as compared to Q4 causes its collector voltage to be low from the current-mirror action (the collector of Q1 being essentially equal to that of Q4), turning on the P-MOSFET (Vgs is negative).
As the output current increases the increased IR voltage on the left side of R1 causes Q1's Veb to increase as compared to Q1's Veb.
This causes Q1's collector current to increase and the collector voltage to rise, reducing the MOSFET Vgs.
When that voltage gets high enough, the MOSFET will start to turn off, limiting the current.
At that point the differences in Veb between the two transistors is about 6mV times their current ratio (difference).

 

With no output current through R1, the lower collector resistance of Q1 as compared to Q4 causes its collector voltage to be low from the current-mirror action (the collector of Q1 being essentially equal to that of Q4), turning on the P-MOSFET (Vgs is negative).
As the output current increases the increased IR voltage on the left side of R1 causes Q1's Veb to increase as compared to Q1's Veb.
This causes Q1's collector current to increase and the collector voltage to rise, reducing the MOSFET Vgs.
When that voltage gets high enough, the MOSFET will start to turn off, limiting the current.
At that point the differences in Veb between the two transistors is about 6mV times their current ratio (difference).

Am I right in thinking that the 6mV factor depends on specific MOSFET specs, specifically its Vgs specs? In other words, would different MOSFET choices deliver different factors other than 6?

 

Am I right in thinking that the 6mV factor depends on specific MOSFET specs, specifically its Vgs specs? In other words, would different MOSFET choices deliver different factors other than 6?

No, the 6mV is determined by the two BJT's, which varies very little between transistors.
The MOSFET Vgs is inside the closed feedback loop so has only a small effect on the actual current limit point.

 

No, the 6mV is determined by the two BJT's, which varies very little between transistors.
The MOSFET Vgs has a small effect on the actual current limit point.

Ok, cool. Thanks for the clarification. I think I'm going to have to play around with it in simulation and monitor current and voltage in a few spots to really start understanding it.

 

After staring at the above, I concluded that the next time I need to design a current limiting circuit the first I would consider is an adjustable (or not) constant current source with enough allowance for a certain variation of the input voltage.

Perhaps of some interest.
https://www.electronics-notes.com/a...s/current-limiter-circuit.php#google_vignette

 

There is an error in my equation of current limit vs. current-mirror current ratios.
It's not a linear function of the difference in ratio.
I'll return when I sort that out.

 

There is an error in my equation of current limit vs. current-mirror current ratios.
It's not a linear function of the difference in ratio.
I'll return when I sort that out.

I'm glad you said something, cause it wasn't matching up when I ran the sims, and I was starting to fear that I was crazy (or to be less melodramatic, that I had somehow miscopied the circuit!)

I tried a Log function, just on a hunch, and this appears to be a pretty good match, although I certainly couldn't tell you why:

Log10( IQ1 / IQ2 ) * 60mV / 0.05ohm

 

I'm glad you said something, cause it wasn't matching up when I ran the sims, and I was starting to fear that I was crazy (or to be less melodramatic, that I had somehow miscopied the circuit!)

I tried a Log function, just on a hunch, and this appears to be a pretty good match, although I certainly couldn't tell you why:

Log10( IQ1 / IQ2 ) * 60mV / 0.05ohm

View attachment 201260

@ebeowulf17

Yesterday I spent some time simulating but could not understand what I was seeing. Definitely well behind you.

Actually, cannot say I understand how it works. Rusty me, to say the least.

BTW, do you mind posting the pot model and accessories so I can include it in my LTSpice bag? Gracias.

 

@ebeowulf17

Yesterday I spent some time simulating but could not understand what I was seeing. Definitely well behind you.

Actually, cannot say I understand how it works. Rusty me, to say the least.

BTW, do you mind posting the pot model and accessories so I can include it in my LTSpice bag? Gracias.

Sorry, I don't have a pot model either. I'm sort of a cave-man when it comes to LTspice. I work on a bunch of different computers and I got frustrated trying to keep all of my parts library file sets in sync, so I finally just gave up. With very rare exceptions, I almost always work exclusively with the standard parts that come with LTspice.

In the case of this circuit, the pot is being used as a variable resistor (rheostat?) more so than a full potentiometer - I may not be describing this quite right, but basically I just mean that since the wiper is tied to one of the other legs, you're effectively only using two of the three terminals, with the result being a single variable resistance. As such, I just used a step param variable on a standard resistor model for my version of the sim. I've attached the file in case it helps.

 

I'm glad you said something, cause it wasn't matching up when I ran the sims, and I was starting to fear that I was crazy (or to be less melodramatic, that I had somehow miscopied the circuit!)

I tried a Log function, just on a hunch, and this appears to be a pretty good match, although I certainly couldn't tell you why:

Log10( IQ1 / IQ2 ) * 60mV / 0.05ohm

View attachment 201260

The formula that I made a lucky guess on above now makes a lot more sense to me. It's hard to read the datasheet with enough detail to get this right, but at least in LTspice simulation, it's a pretty consistent ~18mV per doubling in current across the useful range of the 2N3906:


Since Log10(2) = 0.301, that means that for each doubling in current my formula above would predict a Vbe change of 0.301 * 60mV = 18.06mV. That simple formula approximation is pretty darn close to the 18.18mV I measured above when doubling from 1mA to 2mA. The simulation results aren't exactly the same across that whole slope, but they stay pretty close to 18mV until around 6mA or so of collector current.

 

So it would seem the current limit of my circuit is then:
ILim ≈ [60mV * log(IQ1/IQ2)] / R1.
The revised circuit is below showing the current limit versus pot rotation, indicating the log nature of the limit versus linear pot rotation (which can be advantageous as it allows for finer adjustment of the limit at the lower current levels).

I also noticed that R2 was redundant, so it was removed.

 

Here's a variation of the circuit that has a more linear current-limit adjustment versus pot position (less curve):