Zener Diodes as voltage references

[lk]

Greetings,

I have been fooling around with Voltage references , specifically the LM431, but also “pure” zener diodes that are used as a reference, and no long before I discover what is probably obvious to everyone else, and that is the fact that these are only stable if their current is not changing.

Before I discovered this I build this SMPS,http://www.romanblack.com/smps/smps.htm
 but there I noticed that the output voltage , followed the changes in the input voltage, this must be due to the changes in the current flowing through the zener,  and in my case that I usually tinker things that are powered from 12 volt solar systems, I have to live with voltages between 11-14.

I guess that the resistor that is used to limit current to the zener diode, needs to be replaced with at constant current source instead? To increase the voltage stability of the zener.  I imagine that I would need to build current limiting circuitry , similar to what is in the Voltage standard from EEV Blog 210, to supply the diode in there.
http://www.eevblog.com/2011/10/24/eevblog-210-krohn-hite-dc-voltage-standard-teardown-calibration/


But what I worry about is if I end up in a chicken and egg, situation where the current is not stable either, because my voltage is not stable (11-14v)


-lk

[BravoV]

All you need is to build a constant current circuit and stop worrying about the voltage that is supplying it.

As the name implied "constant" current, it will give you a stable current no matter what voltage its working on, of course the voltage range must be in the certain limit for it to work properly.

Your 11-14 voltage range is not an issue for the CC circuit.

[lk]

Thank you BravoV,

I might need to "just" build a constant current source, but so far all the solutions i have found seem to rely on the fact the the input voltage is stable, but this is not really the case for my setup.  I realise that I'm most likely have missed something



-lk

[Rerouter]

the constant current source used in the HV dc supply is actually the one reccomended in a few old motorola power supply handbooks, 2 transistors, and a  bais resistor from memory, hmm, wonder where that book went, as it went a few steps furthur in actually reccomending trasistors for thermal offset cancellation, with a few pairs being down in the uV fer degree C,

here is the most similar image that i can find for the time being http://1.bp.blogspot.com/_1KIwZrxSy6A/S3TqbPeuyYI/AAAAAAAAAWw/EaQmFg6ELdk/s320/ConstantCurrent.gif,

but instead of an led, you would use your zener

[BravoV]

The CC circuit is very simple, made this just under 5 minutes at ltspice.

See how the voltage from "Source" with the gigantic ripple at 10 Volt peak to peak swinging from 10 volt to 20 volt  was tamed, and compared to the output of the flat line/stabilized voltage (V(ref)) generated from the zener that is supplied with the constant current.



The "BufferLoad" is for simulating the op-amp buffer input at 100uA.

[lk]

Perfect!

Thank you both, this is exactly what i was looking for, i apparently spend hours looking in the wrong places, i look forward to get home and experiment with this

-lk

[Rerouter]

also while playing around, this seems to me like the most stable form of zener reference i can find, the odd values would essentially be calibration

the 408K is to suppress the ripple,
the 26K is almost like a fine adjust for the current,
and the 83 ohms is like a course adjust for the current,

and the 100uF was simply because any less and it seemed to difficult to suppress down to the lower uV of ripple, (yes i know its technically cheating)
also having one on the output seemed to have no overall effect,

yes there are probably other ways to tweak it, expecially tweaking the 540 ohm current limiting resistor, however this design held up to even 8V of ripple superimposed on 13V

http://pastebin.com/x9we75qn

[Mechatrommer]

thank you as well for example... but i got a funny thinking, what if we replace the zener in constant current circuit, with... a simple resistor, can it be made as voltage reference as well? Vref = IR?

[amspire]

You do need a constant current source, but the best way to do that is to use the Zener voltage to regulate the current in the current source. So if the zener voltage is stable, the current source is stable.

All you need is a circuit like this:



Now the zener can be either a zener alone, or a slightly higher voltage zener in series with a forward biased diode junction with the opposite coefficient.

Start off by assuming the voltage across D1 will be constant - that is the aim after all. The voltage on the opamp output will be a constant voltage (about double the zener voltage) and so the voltage across R1 will be constant. Since the voltage across R1 is constant, the current through R1 (ie the current through the zener ) is constant.

The secret to stability is the zener temperature coefficient changes with current. Typically as the current increases, the temperature coefficient decreases. So if you start with a zener that is very close to zero temperature coefficient, then by testing the coefficient, and then adjusting the current, you can find the current with the minimum coefficient. Increase the current and the coefficient goes negative. Reduce the current and the coefficient goes positive. Somewhere in between, it is very close to zero. That is why in Dave's reference box, they included the calibration number. When the current is adjusted to give that voltage from the zener - it is at the point that the manufacturer measured the lowest coefficient. It is a number that only works with the current zener, and if the zener had to be replaced, the whole unit would probably have to go back to the manufacturer to determine the best operating point for the new zener.

In the circuit above, you could vary R1 slightly to set the current.

The resistor coefficients come into it, but if you calibrate for zero coefficient using this circuit with all parts in place, then you can pick the zener current that gives lowest net coefficient. So if the resistor and opamp coefficients end up giving a slight positive coefficient to the current source, then you pick a current where the zener has a slight negative coefficient to compensate.

Put the whole circuit in an oven, and the thermal stability improves a couple of orders of magnitude, and you have the big advantage in that you only have to find the best coefficient for one single temperature operating point.

Also the other key to a zener reference is the voltage at which the zener is most stable is not important - what is important is that the voltage is extremely stable.  You then calibrate elsewhere.

Can you make a stable reference with cheap standard zeners? You could be lucky, but the real reference zeners are often pre-aged and often have a construction that minimizes the effects of the surface conditions on the silicon die on the zener voltage. Many of the best zener references use buried zener diodes. If you did use a cheap zener you selected because it had a zero coefficient between 5 and 10mA, I would probably keep it running for a few years, and then recalibrate it for zero temperature coefficient. At that point, it will be much more stable then when you built the reference.

Richard.

[free_electron]

Us an lm199 zener diode. Comes with built in heater and temperature controller..

[SeanB]

What also works well is a second, higher voltage zener as a pre regulator. The first diode provides a more stable source of voltage for the resistor driving the primary reference. This just requires the first diode is run at a current higher than the second, and it may have to have a higher power rating, or you can use 2 lower voltage devices in series instead.

I used this to get a very stable voltage out of a 723 regulator, using a 60VAC rail that was half wave rectified, then regulated with a 1W 68V zener, and then a 33V 1W fed by a further resistor. The 723 was fed by this and provided a 7V reference voltage. Very stable, did not drift much with time and temperature and was used to provide a suppressed zero meter used to show output on a variac, worked well to give a 60 - 360VAC display on a moving coil meter built into the case. Variac provides 0 - 340VAC from a 220VAC input @ 10A.

[jimmc]

For completeness I'll mention the 'Ring of Two' see p227 of this extract
http://books.google.co.uk/books?id=EEcemABAU44C&pg=PA226&lpg=PA226&dq=zener+%22ring+of+two%22&source=bl&ots=ywpZ-NGFfp&sig=pCYHBTnWY26hsGIkC0MgHsybU28&hl=en&sa=X&ei=qx2tT9jKA5LT8QP6uP3hCg&sqi=2&ved=0CE0Q6AEwAA#v=onepage&q=zener%20%22ring%20of%20two%22&f=false

However it is not suitable for your requirement as it requires more then 11v.

A point to note is that neither the 'Ring of Two' nor the Op-Amp circuit mentioned previously are necessarily self starting and and can sit with 0v across the zener.
For the 'Ring of Two' a high value resistor between the bases will set it off, for the Op- Amp, a positive input offset will do the trick.

(As shown the OP-Amp circuit will not work at 11v, reducing R2 (and R1) sort this  out.)

Of course if you want an easy way to get a stable reference voltage, use a three terminal reference with the voltage stabilisation built in. An extreme example is the LT1019 at 0.5ppm/V
http://cds.linear.com/docs/Datasheet/1019fd.pdf