3

I built a low dropout regulator following experiences building a discrete linear regulator. The circuit is simple like this:

schematic

simulate this circuit – Schematic created using CircuitLab

I would like to know how to test the maximum voltage inout and output, current carrying and heat dissipation ratings of this circuit with parts as specified in circuit, as well as its minimum voltage dropout.

By supplying a regulated 5.5V into this circuit I can get the maximum of 5.4V at the output under 30mA load. Does this spell a minimum 0.1V dropout at 30mA?

I also got a minimum of 2.5V voltage output without load. How to modify this regulator so that it can shut the load down by turning the pot one way all to the end?

Also is the op-amp getting power from the unregulated input a problem?

--- EDIT ---

Since this circuit is not protected, how to add overcurrent protectection? I am considering using LM358 instead of LM741 as the op-amp chip, and how to construct the current sensing and overcurrent protection circuit using the extra op-amp from LM358, a shunt resistor and a few more MOSFETs? (I have 2N7000, BS250, IRF4905 and IRF540 spare parts lying around in bulk)

--- EDIT 2 ---

Rounding up suggestions, does this seem like a better circuit?

schematic

simulate this circuit

The reference is modified per @SpehroPefhany's suggestion to use a TL431 powered through a 680Ω resistor. The old 741 is swapped out with OPA2134 dual rail-to-rail op-amp (expensive!) per @andyaka's suggestion. Some frequency compensation is attempted using capacitor C3 at the gate of pass transistor M1.

My intention of the overcurrent protection is like this:

The second op amp in the OPA2134 package is wired into a differential amplifier, monitoring the voltage difference between current shunt resistor R4. When the current approaches 4A the voltage drop across R4 increases to the point that the voltage output of the differential amplifier OA1b approaches the threshold voltage of 2N7000 and start to push it on, pulling the voltage at the inverting input of error amplifier OA1a lower, pushing the gate voltage of M1 higher and start to shut M1 down.

--- EDIT 3 ---

Rounding up suggestions again, would this be better?

schematic

simulate this circuit

The op amp is no longer driving the pass MOSFET directly, and current shunt voltage is directly matched against the threshold voltage of a MOSFET PNP BJT. This should be able to eliminate the need for a RRIO op amp. I am still using a relatively modern LM358 but is 741 suitable here now?

Community
  • 1
Maxthon Chan
  • 2,853
  • 1
  • 16
  • 33

3 Answers3

4

I suggest you do not use a 3.3V zener diode as a reference- you'll get horrible line regulation (and ripple rejection), especially with a resistor as the current source, as well as bad temperature stability.

At least use a TL431 (almost as cheap as a zener in volume) which (if you give it >1mA) will maintain a very steady voltage (nominally 2.495V) and has quite reasonable temperature stability. Your 1uF in parallel should result in unconditional stability.

An LM358 should work okay with a sufficiently low pullup resistor to allow the output to get close to the positive rail so the MOSFET can turn off. The LM358 is good for 32V, your MOSFET gate is probably not rated for 32V so that limits your maximum input unless you improve your circuit.

You've not made any attempt to deal with (frequency) compensation. At some point (probably very soon) you will find out why LDOs have problems in this area when it turns into an oscillator. You may not easily see the oscillation at the output because of the huge capacitor, so look at the op-amp output to see if the circuit is stable.

Anyway: How to test

  1. Line and load regulation test for the rated input range with different loads (minimum to maximum and a few inbetween). Measure the output voltage for each.

  2. Temperature stability- repeat tests at different temperatures from minimum to maximum.

  3. Stability- change the load from maximum to minimum with different input voltages and observe the output behavior-looking for droop or overshoot when the load is increased or decreased suddenly.

  4. Ripple rejection- apply some ripple on the input at the desired frequency and observe how much gets through to the output.

  5. Drop-out- observe the output behavior as the input is increased from zero to a voltage a few volts above the output, with various loads.

You can measure quiescent current Iq as well, if you like. If you pullup resistor is very low (like 1K) you may see a significant increase in current as the op-amp rails for a high set output voltage and input slightly too low to regulate.

Spehro Pefhany
  • 376,485
  • 21
  • 320
  • 842
  • Changed the design to include some crude frequency compensation and overcurrent protection. Is that enough? – Maxthon Chan Dec 21 '14 at 23:01
  • I don't think Andy suggested an OPA2134- it seems extraordinarily ill-suited to the tasks you've put it to, as well as being expensive. – Spehro Pefhany Dec 21 '14 at 23:13
  • Andy suggested a modern, rail-to-rail and my search on Taobao (Chinese equivalent of eBay) turned up this OPA2134. Other rail-to-rail parts are even more expensive. – Maxthon Chan Dec 22 '14 at 00:42
  • Thing is, it may be relatively modern (last 20 years), but it's not rail-to-rail input or output. It's even worse than the 741. It's actually a great audio amplifier (low noise, super-low distortion, fast slew) but has mediocre DC accuracy. You need RR input for the current limit (or at least the positive rail) and to work within 2.495V less tolerance, so maybe 2.47 of the negative rail for the reference and this part is only good to (V-)+2.5V (just misses). And the output only swings to within 1.2V of the positive rail (probably good enough for that particular MOSFET, but not rail-to-rail). – Spehro Pefhany Dec 22 '14 at 01:33
  • So my design calls for a real dual RR op amp chip? Please suggest some possible jellybean parts. – Maxthon Chan Dec 22 '14 at 12:19
  • @MaxthonChan Unfortunately, I don't think there is such a thing as a jellybean RRIO op-amp capable of high supply voltage. ADA4096-2 is RRIO and good for 30V, but it's not cheap. OPA2192 is not available yet, and probably expensive too. Better to change your design! – Spehro Pefhany Dec 22 '14 at 12:36
  • Changed the design again, please check if it removed the need of an expensive op amp. – Maxthon Chan Dec 22 '14 at 13:18
  • @MaxthonChan I like it (the current limit transistor would actually have to be a p-channel), but use a PNP BJT rather than a MOSFET for the current limit- the Vbe drop is better controlled than Vgs(th). – Spehro Pefhany Dec 22 '14 at 13:25
  • Updated the Edit 3 schematic to use your suggested PNP BJT to measure the current shunt. Current rating in my head is 3.5-4A here. – Maxthon Chan Dec 22 '14 at 13:44
  • 1
    Rolled one on breadboard following the Edit 3 schematic, and it seemed good enough and is able to turn off completely. No oscillation heard on the gate of pass MOSFET's gate. – Maxthon Chan Dec 26 '14 at 16:48
3

Here's a fairly simple, well-behaved LDO that simulates nicely:enter image description here

+V is 24VDC, the voltage into the load is >23.9V with R3 at 5V, and <100mV R3 at 0V.

I'll clean it up a little, maybe switch out the LT1007 for a RR out opamp to get really low LDO, and post a link to the LTspice file tomorrow.

EM Fields
  • 17,377
  • 2
  • 19
  • 23
  • +1 Nice simple obvious circuit. || Vgsmax of FET is 20V - and most FETs will be in danger at 24VDC supply used here. Adding a U1 output resistor and a Vgs clamp to FET helps (You know that :-) ). | Real world stability may require some playing. But overall it's the sort of circuit that immediately comes to mind. | Over current protection not present of course. – Russell McMahon May 13 '20 at 03:38
1

This circuit just will not work properly on low power loads. The 741 maximum output voltage is at least 3 volts under the supply rail voltage and this means the P channel MOSFET will always be conducting some current - gate-source voltage threshold can be as low as -2 volts and given the 741 probably can't get to within 3 volts of the positive supply rail you are going to have big problems at some point or another.

This also explains why you cannot properly shut down the regulated output voltage.

Try using an op-amp that wasn't designed in the middle of the last century. hint - rail-to rail output types are needed to cure the above. Please also note that you may encounter large instability problems with newer op-amps due to their increased GBW product. on the olde worlde 741 they are likely to slow to cause problems in your circuit.

Andy aka
  • 434,556
  • 28
  • 351
  • 777
  • Is LM358 or LM324, both op amp chips I stock in bulk, good op-amps for this purpose? Or please recommend some jellybean parts cheap enough for me to stock in bulk. – Maxthon Chan Dec 21 '14 at 20:08
  • @MaxthonChan my recommendation is for you to choose an op-amp that has rail-to-rail output capability. At the minimum it should be able to output a voltage that is within 0.5 volts of the positive supply rail - read the data sheets - it's all in there unless you want me to read them for you and report back dutifully LOL. – Andy aka Dec 21 '14 at 20:12
  • Answering the edit section of your question is not relevant until you have the main design covered. I'll also add that there are many examples on the net of over-current protection. Having a stock of certain devices should not restrict anyone on designing an adequate circuit - that's my philosophy anyway. – Andy aka Dec 21 '14 at 20:15
  • Something called OPA2134 came up to me, claiming a Vcc-0.5V maximum output voltage and 8MHz speed. More than $1 each (LM324 costs 1 cent each) and is that seem okay to you? I will order a few samples to try. – Maxthon Chan Dec 21 '14 at 20:21
  • 1
    It's not easy to get low drop out regulators to work because of the tons of extra gain due to the pass transistor's drain/collector being connected towards the load. I wouldn't encourage anyone to spend past their budget on anything but if you are intent on getting it going, be prepared to fight instability (downright massive oscillation) on the output with a firm hand. Like I say, LDO regulators are a little bit unstable on the best of days. – Andy aka Dec 21 '14 at 20:31