Paupers prefered reference

[IconicPCB]

For a while I have been looking for a low cost reasonable performer Zener amplifier solution.

Conrad Hoffman's idea appealed to me how ever it was not quite the minimalist approach.

Here is something for Your consideration and critique:


A single opamp with a single resistor and a ( compensated )Zener diode.
The op amp is an LM 3900 a Norton op amp.
Norton op amp relies on a current differencing current mirror input stage. Connecting non inverting input to ground disables parts of the input stage and produces lowest noise outcomes.

The input pin will happily work with 200 uA max input current and when driven within the input range will develop typically 0,55V on the inverting input.

The current through the zener diode is set by the resistor such that I zener  = 0.55/ R typically.
The other limit on zener current is the aximum drive capability of the op amp ... typically 10mA.
The implication is that the resistor should not be any less than 56 ohm and probably no more than 1Kohm.
 
I have built a prototype consisting of four sources using 6.2V zener and following resistor 56ohm, 100ohm, 220 ohm and 470ohm,
Unfortunately all i have is a Fluke 8050 4.5 digit instrument and my measurements are inconclusive to the extent that i see a variation of a couple of millivolts as the temperature in my workshop fluctuates from 18 C to 24 C over a 24 hour cycle. and the four references drift by unequal amounts.

The nice thing about the LM3900 package is there are four opamps.

I am yet t o build the temperture controller using one of them while the other three are used as zener amplifiers.

A very simple bang bang controller based on a comparator can be built with an LED status indicator. The LED should flash on and off at an increased frquency when set point is achieved and this would be a cue that the source has stabilised thermally.

As usual comments invited.

[cellularmitosis]

Does the Vbe drop inside the opamp create a temperature compensate zener + diode pair?

[zhtoor]

Does the Vbe drop inside the opamp create a temperature compensate zener + diode pair?

my thoughts exactly, maybe at the current level selected inside the norton amp, *may* somewhat compensate the zener.

regards.

[IconicPCB]

On further reflection, I do not think there is much in the way of compensation that can be attributed to the BE junction.
In the so far considered voltage op amps the compensating effects of the diode junction were in the form of a matched inverse temp co slope added to the overall ref voltage.
In this instance the diode current needs to be sampled and and fed back in series with the zener voltage.
The simple solution is to view the voltage drop across the current limiting resistor as being the same as the VBE compensating drop.
Now consider a resistor in series with the Zener in the feedback path.
Since the diode and the two resistor effectively carry the same current then adjusting the second resistor ( R2 = a * R1) will inject a * VBE volts as compensating signal in series with the diode.
Careful measurement of thermal effects will point to the ratio of R2 to R1.

Cheap..Cheerful and probably acceptable. In my case with only 4.5digits to play with.. i think i am about to fill my boots. 

EDIT2: The business of averaging the multiple reference... just a handful of low value resistors. The answer to a Young maidens prayer.

[IconicPCB]

The PCB is fabricated and assembled.
I shall set up a data logger  ( Advantech USB-4716  16 bit logger )  for some overnight collection.
 
Hope to have the first bunch of data for discussion soon.

[IconicPCB]

A few days down the road and a change to the temperature controller followed by some improvements.

Inital configuration retained with a change to the temperature control circuit. In the original posting I slapped together an idea for discussion. It turns out that the thermistor should be in the non inverting leg of the comparator.

So rehashed that. And yes the oven was regulating but due to a variety of reasons with tooo much overshoot. Net result was a ripple up to 3 millivolts directly attributable to localised heating on the PCB.

Let me expand on that... Two zeners were fitted without compensation, the third zener was fitted with a 1N4148 in series.
The notionally uncompensated zeners were showing a ripple of approx positive 1mV. The compensated zener was showing a ripple of approximately negative 3 mV. Overcompensated?

The current was set to approximately 2.4mA.

Second bite of the cherry....  Temperature control modified.. replace the BJT with a Power mosfet. Add a .22uF cap integrator cap to the comparator. Change power supply from 9V to nominally 10V.
Temperature fluctuations less pronounced.
Ripple due to temperature variations lower.
Change the zener operating current to closer to 10mA.( probably limiting case of LM3900)

Have noted inversion in polarities of temperature related ripple. See the spreadsheet.
Note data collected using a 16 bit data logger.

Logger specs absolute accuracy  0.03% of full scale ( on a 10V range.... measly 3mV) so data has to be truncated since the professional who wrote the logger software made it work with double precision.

Well so far so good...The idea seems to have merit and i may separate the temperature control function onto another bit of silicone and simply run the LM3900 as four references with a fifth averaged output and adjust current for minimal temperature sensitivity without additional temperature compensation.

 

[Kleinstein]

The circuit does not look sensible at all. The Norton amplifier is not at all made for precision applications and it looks like it is not even used as it is supposed to be. With the non inv. input to GND the current there is zero and thus it is just the internal bias the is used as a reference current.  To a certain degree the input transistor in the LM3900 is already used as a temperature compensation.

There is not much sense in reinventing the wheel and make it square.

A true OP is way more sensible and not that much more expensive (an LM324 is likely cheaper and still more useful than the LM3900) - though it usually takes 3 resistors, or the old style circuit the 2 zeners and 2 resistors.

[IconicPCB]

Further simulations suggest a fairly broad band of temperature compensation is possible.

Simulation suggests a temperature coefficient of entire assembly under 10 ppm/C over a +/-15C range centred on 40 C.

Hope to be able to assemble second version by end of day today. Will see how the second version behaves.
If everything looks OK the Australian Club will get to judge it .