Op-amp for power supply

There are several problems with your design. They require understanding how a transistor works. On the net you will mainly find the outcome of others in the same situation, so I wouldn't rely on this. First, that diode will have a huge diffusion capacitance when conducting, it is only dependent on the forward current and the reverse recovery time, because when it is supposed to shut off, the charge carriers have to be removed. This is a multiple of the depletion capacitance and can be several nF. Luckily the very low ohmic junction resistance is parallel to it. The same is true for the transistor. But the opamp is *much* faster and sees this big capacitance. It will start oscillating and turning the diode+transistor on again. So for the voltage control output this is bad. Better to use a resistor instead of the diode. Slow down the response time with a capacitor from the Vreg opamp O/P to its neg. input, maybe 100p. This will greatly improve stability. What you were doing was the opposite and aggravates the problem.

Now this "pull-up" resistor will draw a lot of current when the voltage is low and not enough to pull the darlington up, since the opamp with the diode cannot source current. So use a current source made from a resistor and current mirror between the rails and set it to say 5mA when output is zero. The new resistor on the voltage regulator has now to pull the darlington base down to 0.9V or less and has to sink those 5mA, so maybe 0.7V/5mA=

120R. If that is not enough put a high current diode from output gnd to the negative rail to improve this margin.

Now the current limiter can not work any more, since when it kicks in, the voltage opamp output goes high and delivers another 20mA until the current limit. There is a possibility: you pull down its *reference voltage* through a Schottky diode and also a 120R resistor. But now the regulation goes thru

2 opamps so the current one needs to have an even slower response, maybe 1n in the same position as before with the voltage opamp. Now your voltage opamp will never be out of regulation. Don't forget a resistor on the DAC voltage O/P since it gets pulled down by the current limiter.

--
ciao Ban
Apricale, Italy

Where is the current sensor located. If you are sensing current delivered by the NPN, say in its collector, then this will be a relatively wideband loop with greater chances of oscillation. Your current limiting OA gain is amplified by the NPN gm and converted back into a voltage by the sensor. It is a sure bet 20dB of additional loop gain- possibly as high as 60dB, and a phase shift that can be close to

90o depending on the conduction state of the diode. If that schematic is supposed to be a Darlington, then things are exponentially worse from a stability perspective, very large gm.

Almost any op-amp can be adapted to perform in this design. Old does not mean bad, the early designs are brilliant. And what is that odd looking varactor type of thing you have hung on the output?

I have not included everything on the schematic for complexity sake. But, the current sensor is a 0.1 ohm resistor, on the transistor emitter. That differential is amplified 10X with an Analog Devices AMP04. This way I get a nice 1V/1A output. The output transistor is supposed to be a Darlington. How do I stabilize the current sensor? I can put a capacitor from the OA output to non-inv intput, this I know. I'm not sure if this is the proper way though.

That odd looking thing is supposed to be an electrolytic. I agree it looks strange. Is the diode design I have here really that bad? I am using Schottkys, BTW.

How is this done? If I replace the diode with a resistor, then you have a different circuit alltogether. The diode scheme allows either the current OA or voltage OA alone to control the output. Whether this is a good design, however, is questionable.

I believe I've tried this. It does help some.

The pull-up isn't much of a problem. Actually, I have a 47K on there and it has no problem supplying enough current to the transistor. And the OA has no problem sinking this current.

Ok, so I think you're saying to build a plain old voltage regulator with an OA and output transistor, like I have here, minus the diodes. But, pull down the voltage reference into the op-amp with my current limiting OA. I still end up using diodes.

I'm confused now.

Look with a scope at the O/P of the voltage opamp. Whenever you give an increase in voltage by your D/A, it will also loose regulation. It goes high and the wimpy 47k resistor has to charge up the base capacity, which is

*very * high when conducting, maybe 0.3uF. When a resistor instead of the diode is there, the opamp doesn't loose regulation, but will help giving more current, at least as long as you do not reach the upper voltage limit. Now you can take a Rail-to-Rail opamp and increase performance. And get rid of this capacitor at the voltage divider, It will *never* allow a decent performance, but always let the opamp overreact. Look into your textbook about loop stabilisation.

Really? Have you thought how much current you need on the base to drive the darlington? With 5A the hFE is hardly more than 5000, maybe when it's hot you will have 10000. You would need 25V across the resistor to deliver that current. Ah now I see why this resistor is so big. Because otherwise when there is a short circuit, you would burn out the transistor, because the current limiter is also saturated and not regulating at all. Well at least use a current mirror with those 0.5mA, so you can also deliver some current when the voltage is maximal.

Only one diode for the current limiter at the voltage D/A output. Better would be another additional transistor as an emergency current limiter, which is driven by a small sense resistor, you could take your current shunt for that, to pull down the base of your darlington.

'm confused now.

Also before?

ciao Ban Apricale, Italy