**Short version:**

I've come up with a current limiting system with low dropout voltage, intended primarily for output short-circuit protection. I think it hits a sweet spot in terms of balancing accuracy/repeatability vs cost and complexity. Current flowing through D1, D2, and R-Bias sets a reasonably stable voltage difference between supply voltage and the base of Q3. Current through Q3 can increase until Q3's Vbe plus the voltage drop across R-Sense add up to that supply to base voltage difference. So, the *approximate* current limit can be found with this formula:

Current limit = (Vf(D1 + D2) – Vbe(Q3)) / R-Sense

I'm interested in hearing comments and criticisms:

**Long version:**

I've recently become interested in simple, compact, cheap solutions for current limiting circuits. These don't necessarily have to be perfectly accurate, not what you'd call current sources, just something to protect the outputs of a board from short circuits. I found an intriguing circuit that just uses a couple extra resistors and is surprisingly effective, but it's pretty finicky about pass transistor Beta, supply voltage variations, etc. and it's not very straightforward to calculate values for. This circuit is what got me curious about other circuit options:

I came up with a system using two common diodes to create a moderately stable reference voltage for the transistor's base (instead of the voltage divider in the circuit above.) This circuit delivers a much more stable current limit when facing varying supply voltage or Beta values. It's also much easier (at least for me) to calculate the values for, although it's still not the most straightforward. I like that I could incorporate it into different designs with less fear that changing transistor component choice or having minor inaccuracies in supply voltage regulation would dramatically change the current limit:

I also looked at a few other known circuits for comparison. There's a circuit shared by @crutschow which is more stable and easier to calculate, basically better in every way, except that it's significantly more complex, so if you had a board with 20 outputs to protect and absolute precision wasn't that important, you might want something cheaper and simpler. Here's that circuit:

The other known circuit I looked at was the LM317 configured as a current limit, which provides great accuracy and stability, but at the cost of high dropout voltage:

Finally, as I was looking at the various influences that would push the current limit on my own design around, I thought perhaps a more stable reference would help, so I tried a version with a TL431, instead of two signal diodes, providing the reference voltage. This did yield more stable results, but at the cost of much higher dropout voltage (although still not as bad as the LM317.)

Some key parameters I looked at when comparing these circuits were:

- dropout voltage
- stability of current limit with varying input voltage
- stability of current limit with varying transistor Beta
- stability of current limit with varying transistor Vbe

Next is the effect of varying input voltage. The graphs below represent 5V +/-0.5V supply voltage. You can see that the two circuits with ICs deliver very consistent current limits, @crutschow's circuit shows very little variation in current limit, my circuit (EBS) shows a few mA of variation, and the Doughty circuit that originally got me curious about all this swings up or down nearly 15mA based on these 0.5V supply variations:

Next let's look at that effect of transistor Beta on current limit set point. I ran these sims with the default 200 Beta setting, as well as Beta of 100 and 300. The LM317 circuit doesn't require an external transistor, and the TL341 and Crutschow circuits appear to be immune to Beta variations. My circuit shows a little variation, maybe a few mA, while the Doughty circuit shows a little over 10mA of total variation as a function of Beta.

Finally, let's look at the effect of transistor Vbe on current limit set point. I ran these sims with varying IS parameters, corresponding to Vbe variations from around 0.65 to around 0.9V (rough estimates on those numbers.) Since all of these circuits except the LM317 rely heavily on Vbe, most of the graphs below show large variations. The Tl431 showed the least variation, Crutschow's circuit and mine were quite similar on this, and Doughty's was the worst, albeit not by terribly much.

Basically, the result of all these sims seems to be that the Doughty circuit works well enough if you have a well regulated supply voltage and you have a pretty good idea what Vbe and Beta numbers to expect. If not, Crutschow's circuit provides the best performance, and I'd like to think maybe mine finds the sweet spot in terms of reasonable performance with a simpler circuit. The other two, while interesting for comparison, don't meet my needs in terms of dropout voltage. If there are other things I should be looking at, like if my circuit is especially vulnerable to temperature effects or something else that I've overlooked, I'd love to know. (@OBW0549, you've given me some good tutoring on other current limiting circuits, a modified current-mirror circuit in particular, so I'd love to hear your thoughts on this arrangement.)