I've seen the following precision voltage reference circuit in schematic diagrams of several vintage instruments:

enter image description here

I understand how this circuit works in general: it is basically two constant current sources back-to-back, providing the current to the Zener diode Z1 that acts as a voltage reference.

Current through Q1 will increase until voltage drop on R1 will be approximately equal to voltage on Z2, at which point the transistor starts to close. D1 provides some temperature compensation. If Ube changes due to ambient temperature then forward voltage on D1 will change for a similar amount and the current through Q1 will not change.

Q2 works in a similar fashion to Q1 and provides a constant current to Z2, with the current defined by Z1 and R4.

R5 is there to provide some initial current through Z2 and ensures that the circuit starts up. Without it the circuit can end up in a stable state with both transistors closed.

I'm puzzled by the inclusion of R6. It seems to have a destabilizing effect: Consider if current through R1 increases by some small amount. Then voltage on Q1 emitter will decrease, less current will flow through R6. This in term lowers voltage on Q2 emitter. Q2 passes more current, increasing the voltage drop on Z2. This decreases the voltage on Q1 base and causing it to open more, increasing the current through Q1 even further.

However in practice, this positive feedback effect is minuscule because of the large value of R6 compared to R4.

Considering the function of this circuit I'm guessing R6 is there because it somehow improves the stability. In simulation the circuit works fine without it.

My best guesses are that R6 is there either to improve the temperature stability or to prevent some kind of oscillations (perhaps due to C1?), but I fail to see the mechanism behind it.

  • \$\begingroup\$ Why not use a simulator and test your theory? \$\endgroup\$
    – Andy aka
    Commented Jun 18 at 10:01
  • \$\begingroup\$ I've never really bothered to understand the effects of finite collector output impedance or Early voltage too deeply. Does it compensate for those? \$\endgroup\$
    – Neil_UK
    Commented Jun 18 at 10:01
  • \$\begingroup\$ @Andyaka can you suggest how to test any of that in simulation? As I mention in my question, I tried simulating the circuit. It does not oscillate without R6. Does that disprove the theory about oscillations or prove that the simulation is not accurate in some aspect? \$\endgroup\$
    – avian
    Commented Jun 18 at 10:21
  • \$\begingroup\$ Why not show your results that indicate R6 destabilizes the circuit? You should also make a citation for the schematic image in your question. \$\endgroup\$
    – Andy aka
    Commented Jun 18 at 10:45
  • \$\begingroup\$ @Andyaka The image was drawn by me for the purpose of asking this question. If you mean the original author of the circuit, I would cite them if I knew. The fact that this exact same circuit appears in different designs of the time suggests that perhaps it was published in some text book, article or application note. Despite my best effort I was unable to find the original source. \$\endgroup\$
    – avian
    Commented Jun 18 at 11:37

3 Answers 3


My best guesses are that R6 is there either to improve the temperature stability or to prevent some kind of oscillations ...

No, it does not improve the temperature stability.

R6 modify the output a "little".
When you vary the voltage supply, it makes Vo very "horizontal".
The "best" value should be ~ 360 kOhm.

enter image description here

But these curves depend on the temperature ...

enter image description here


I think @Neil_UK's intuition is correct and it's compensation for Early voltage in Q1 and as such is dependent on Q1's characteristics (and, of course, any loading on the output). In other words, it increases the output resistance seen at the collector of Q1 with the 6.1V applied and the Zener disconnected.

It has no significant effect on the temperature coefficient. There is only a small effect on supply rejection.

A simulation suggests the compensation is a bit heavy (like 2:1) for modern transistors, or perhaps they are assuming a resistive load on the output of 1MΩ or so. It would be interesting to compare Va on the actual transistors used vs. the 100V-ish Va (116V) I assumed for Q1. Perhaps the OP can provide a part number.

Aha, okay looking at the schematic OP linked in the comments, the transistor used in one case was a 2N5086. The 2N5086 is a lower gain bin of the 2N5087 which has a Va of 45.7V according to the model I have, so just about a perfect match.


When Vcc increases the current thru zeners rises a bit and the zener voltages also. The Q1 emitter voltage rises also, but R6 now pushes some little extra current to R4 what causes Q2 collector current decrease back to normal. So the Z2 is approximately at the same current again as with Vcc=15V, so Z1 also.

It isn’t precise because zener VI is knee but resistor linear.

I’m not hundred % sure.


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.