[No TL;DR provided]

I am an undergraduate electronics engineering student and I am working on designing a linear bench top power supply with foldback current limiting. Inspired by designs all over the internet (30V-3A types mostly) I've already came up with a complete design and with the help of LTspice I have resolved most of the oscillation issues (to the extent I am concerned, of course).

For voltage and current adjustment, I have used a zener regulator with an opamp to obtain various reference voltages and I observed that the input power supply ripples and transients are affecting my reference voltages which causes oscillation.

My solution was to change the zener configuration, as I've seen the issue with a trivial design. Since my whole design is irrelevant, I'm posting the zener voltage regulator part:

Various zener with opamp reference voltage schematics

Trivial design was #3. Suffering from supply ripple, changed to #4 but than realized that better(?) alternatives are #1 and #2. They turned out to be great.

I know why #1 and #2 are superior to #3 and #4, basically power supply rejection . Simulations also verified this. Small signal analysis on ripple source with only gain plots (ngspice) are as shown below:

AC Analysis: Gain plot of various zener opamp voltage references

It is obvious that #1 and #2 are doing great, but my first question is, which is better? Is there any aspect of why one should be preferred over the other?

And my second question: Is it possible to make design #4 perform better by playing with capacitor-resistor values while still having a reasonable settling time?

Thanks in advance.

  • \$\begingroup\$ If you are wanting to regulate down to 0V you are going to have to add in negative supply rail. If you want a low voltage of 1.2V or so then there are more choices. Look inside a LM723 for ideas, this has all been done before. Use modern reference as in the one answer to benefit from technology. - google.com/search?q=lm723+internal+circuit&tbm=isch \$\endgroup\$
    – KalleMP
    Feb 1, 2017 at 6:43
  • \$\begingroup\$ Designs #1 and #2 can both latch with output at ground, depending on the offsest voltage on the op amp. Zero equals zero is a balanced input condition. \$\endgroup\$
    – Whit3rd
    Feb 1, 2017 at 8:28
  • \$\begingroup\$ @KalleMP This regulators are used to produce fixed voltage references, so no need for adjustment (output well above 1.2V). And I am aware of LM723, as it was in my "inspiration list". My upperside zener configuration comes from LM723 and my question is about why to use this instead of other? What could be a theory/intuition backed reasoning? \$\endgroup\$ Feb 1, 2017 at 10:29
  • \$\begingroup\$ @Whit3rd you are indeed right but adding a diode drop to negative rail (of opamp only) will solve that. Even more, If opamp had a negative rail below something like -1V, another stable point will be the case with diode in forward region but these are still solvable. \$\endgroup\$ Feb 1, 2017 at 10:33
  • 1
    \$\begingroup\$ @KalleMP no problem, I should've clarified it beforehands. Thanks for the suggestions about padded cell, by the way. \$\endgroup\$ Feb 2, 2017 at 10:53

2 Answers 2


While I agree with Marcus that option 1 seems the most viable, I am surprised that you would still consider a Zener voltage reference acceptable in a power supply today.

For many years now there have been good shunt voltage references, and for the last couple of years the AS431 and TL431 can viably replace a whole range of zener diodes with much better characteristics.
Internally they are not so different to your circuit, with only 2 resistors required to set any reference voltage from 2.5 V up to about 36 V.

enter image description here

enter image description here

  • \$\begingroup\$ Thanks for the suggestion but as I said this is not a modern power supply project. Think of it as an "Analog IQ Test" question (in some sense). If I were to go with modern options I would've used specialized adjustable power supply chips (LT seems to have a lot of options like LT3081) and even a Swithing+LDO Linear hybrid approach with MCU-set references. \$\endgroup\$ Feb 1, 2017 at 10:23

With the precision voltage reference being directly referenced to ground, which you probably want for the output signal, too, I'd personally go for 1.

Unless, of course, you need to recover fast from short circuits/high loads. In that case, having the resistive part of the voltage divider on the low side might make it easier to drive the diode back into saturation. But that's pure speculation. Since both options are easy to simulate or build and actually test, I'd go for that.

  • \$\begingroup\$ could you eleborate on "(...) drive the diode back into saturation"? As you said trial and error is easy but I would like to get the idea/intuition behind? \$\endgroup\$ Feb 1, 2017 at 1:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.