Implementing a simple current limit for a basic follower

[Shay]

Hello,
Consider this voltage "regulator" output stage. it is a part of a low noise regulator that is more like a reference instead of a regulator, as there is no error amplifier feedback system.



Quick brief description of the circuit:

V1 is the input supply. V2 is what sets the output voltage minus Vbe of Q1. R3 is a minimum load, and C2 is output capacitance for noise and stability.

Q1, Q2, Q3 form a szilaki pair with a pnp darlington pair, a hybrid of both. R4+R7 is for current balancing between Q2 and Q4.



R8, R5 is the basis of current limiting. if the voltage difference across R7 is enough to make Q5 start conducting, it means there is an overcurrent.



I am stuck on the next steps. How can I utilize this to reduce the voltage?

I tried attaching the collector to a N-MOSFET gate, and the drain to Q1 base, and source to GND. but this caused a lot of oscillations, which I am not sure how I can fix.



Would like to hear your opinions.

Thanks.

[mikerj]

There is no circuit visible in your post.  You need to attach the image to your post rather than linking to an external website.

[Shay]

There is no circuit visible in your post.  You need to attach the image to your post rather than linking to an external website.

Thanks. I think I fixed it

[CountChocula]

I can see your circuit, but I agree that you should attach it here and not use an external service, or a lot of people won't be able to see it because of their browser or network settings (plus, the external service may just disappear or be unavailable).

You should do your current sensing after your pass elements—add a separate resistor between the emitter of Q1 and R3, then use Q5 in the same configuration to turn on an NPN transistor that steals current away from the base of Q1. Try this simulation of the circuit I attached below, for example (I tried to keep the same designators as in your original).


—CC

[Shay]

I can see your circuit, but I agree that you should attach it here and not use an external service, or a lot of people won't be able to see it because of their browser or network settings (plus, the external service may just disappear or be unavailable).

You should do your current sensing after your pass elements—add a separate resistor between the emitter of Q1 and R3, then use Q5 in the same configuration to turn on an NPN transistor that steals current away from the base of Q1. Try this simulation of the circuit I attached below, for example (I tried to keep the same designators as in your original).


—CC

Thank you for you response.
I am trying to stay away from putting any additional resistance on the output side. As this circuit doesn't have any voltage correction (e.g feedback) it will cause additional voltage drop under load, which contributes to more ripple. This is why the resistors are on the input side to the regulator.

[CountChocula]

Well, you can move the current limiter back where you had it (see attached), but I don't think it's going to work that well, because you'll be sensing off of a single transistor. Out of curiosity, why not add feedback to the circuit so that you can have good regulation?

[Shay]

Well, you can move the current limiter back where you had it (see attached), but I don't think it's going to work that well, because you'll be sensing off of a single transistor. Out of curiosity, why not add feedback to the circuit so that you can have good regulation?

I've got something similar to your solution, but I am facing the same issue: oscillations.
The reason the circuit does not have feedback is because of customer-specific requirements. A bunch of bullshit, really, but it is what it is.

[wasedadoc]

The circuit DOES have feedback.  If the output voltage falls, Q1 conducts more which increases the base current to the pass transistors.

You CAN introduce current sensing on the output side.  On end of resistor at the join of Q2, Q3 and Q4 collectors. Other end of resistor to output and emitter of Q1.  (I'm not giving an opinion if this is a good or bad approach, just refuting the view that it is not possible.)

[Shay]

The circuit DOES have feedback.  If the output voltage falls, Q1 conducts more which increases the base current to the pass transistors.

You CAN introduce current sensing on the output side.  On end of resistor at the join of Q2, Q3 and Q4 collectors. Other end of resistor to output and emitter of Q1.  (I'm not giving an opinion if this is a good or bad approach, just refuting the view that it is not possible.)

Ah! This is good indeed. You're correct about the feedback. However, even in this approach, I get oscillations.
I've tried placing a capacitor from Q3 base to ground. It slowed down the feedback and stopped the oscillations, but the dynamic response is awful, it takes a lot of time to react to changes. This is because the capacitors slows down the feedback, thus reducing the gain(?) (I think so, not sure about the exact control theory stuff).

[Terry Bites]

Achieving proper compensation will always be a compromise beween stability and response time. You'd usually aim for critical damping.
So the question is where in the circuit do you add compensation?
Identify the feedback loop and work from there.

I'm wondering about the nature of the oscillations. What frequncy are they, kHz, MHz?
Try some adding some series resistance in the base leads of Q1, Q3. Say 100R.
Chuck a ferrite bead on those base leads as well.
You can also experiment with emitter degeneration on these transistors. Again in the 100s of Ohms range.

[David Hess]

I am stuck on the next steps. How can I utilize this to reduce the voltage?

The usual solution is to move Q5 to the output and then use it to rob base drive from Q1.

In this case, connect the collector of Q5 to the base of Q3 to rob base drive from and shut down Q2 through Q4, which leaves Q1 supplying the output, but if the base drive to Q1 is limited, then "beta current limiting" may be sufficient.

Classic discrete transistor regulators sometimes used a differential pair in place of Q5, so current limiting was much more precise, and used the differential pair to rob current or otherwise control Q1.

[edavid]

Why are you using common emitter PNP pass transistors instead of NPN emitter followers?  That makes your stability problem much worse.